04-24101. HUD's Housing Goals for the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac) for the Years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac

AGENCY:

Office of the Assistant Secretary for Housing—Federal Housing Commissioner, HUD.

ACTION:

Final rule.

SUMMARY:

Through this final rule, the Department of Housing and Urban Development establishes new housing goal levels for the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac) (collectively, the government sponsored enterprises, or GSEs) for calendar years 2005 through 2008. The new housing goal levels are established in accordance with the Federal Housing Enterprises Financial Safety and Soundness Act of 1992 (FHEFSSA) and govern the purchase by Fannie Mae and Freddie Mac of mortgages financing low- and moderate-income housing, special affordable housing, and housing in central cities, rural areas and other underserved areas. This rule also establishes new subgoals for the GSEs' acquisitions of home purchase loans that qualify for each of the housing goals. The final rule also establishes a new regulatory section relating to GSE data integrity, amends and adds certain definitions, provides a method for imputing the distribution of GSE-purchased mortgages that lack income data, prohibits goals credit for purchases of loans in transactions with an option to dissolve the purchase in less than one year, and makes a technical change to the counting rules to clarify HUD's rules on double counting of loans.

EFFECTIVE DATE:

January 1, 2005.

FOR FURTHER INFORMATION CONTACT:

Sandra Fostek, Director, Office of Government Sponsored Enterprises, Office of Housing, Room 3150, telephone [phone]. For questions on data or methodology, contact John L. Gardner, Director, Financial Institutions Regulation Division, Office of Policy Development and Research, Room 8212, telephone [phone]. For legal questions, contact Paul S. Ceja, Deputy Assistant General Counsel for Government Sponsored Enterprises/RESPA, Office of the General Counsel, Room 9262, telephone [phone]. The address for all of these persons is Department of Housing and Urban Development, 451 Seventh Street, SW., Washington, DC, 20410. Persons with hearing and speech impairments may access the phone numbers via TTY by calling the Federal Information Relay Service at [phone].

SUPPLEMENTARY INFORMATION:

I. General

A. Authority

HUD's authority to regulate the GSEs is established under:

(1) The Federal National Mortgage Association Charter Act (“Fannie Mae Charter Act”), which is Title III of the National Housing Act, section 301 et seq. (12 U.S.C. 1716 et seq.);

(2) The Federal Home Loan Mortgage Corporation Act (“Freddie Mac Act”), which is Title III of the Emergency Home Finance Act of 1970, section 301 et seq. (12 U.S.C. 1451 et seq.);

(3) FHEFSSA, enacted as Title XIII of the Housing and Community Development Act of 1992 (Pub. L. 102-550, approved October 28, 1992) (12 U.S.C. 4501-4641); and

(4) Section 7(d) of the Department of Housing and Urban Development Act (42 U.S.C. 3535(d)).

B. Background: Fannie Mae and Freddie Mac

Fannie Mae and Freddie Mac were chartered by the Congress as GSEs. Pursuant to section 301 of the Fannie Mae Charter Act (12 U.S.C. 1716) and section 301(b) of the Freddie Mac Act (12 U.S.C. 1451), the GSEs were chartered expressly to:

(1) Provide stability in the secondary market for residential mortgages;

(2) Respond appropriately to the private capital market;

(3) Provide ongoing assistance to the secondary market for residential mortgages (including activities relating to mortgages on housing for low- and moderate-income families involving a reasonable economic return that may be less than the return earned on other activities) by increasing the liquidity of mortgage investments and improving the distribution of investment capital available for residential mortgage financing; and

(4) Promote access to mortgage credit throughout the nation (including central cities, rural areas, and other underserved areas) by increasing the liquidity of mortgage investments and improving the distribution of investment capital available for residential mortgage financing.

Fannie Mae and Freddie Mac engage in two principal businesses: (1) Purchasing and otherwise investing in residential mortgages, and (2) guaranteeing securities backed by residential mortgages. As a result of their status as GSEs, Fannie Mae and Freddie Mac receive significant explicit benefits that are not enjoyed by fully private shareholder-owned corporations in the mortgage market. These benefits include:

• Conditional access to a $2.25 billion line of credit from the U.S. Treasury (see section 306(c)(2) of the Freddie Mac Act and section 304(c) of the Fannie Mae Charter Act);
• Exemption from the securities registration requirements of the U.S. Securities and Exchange Commission and the State securities regulatory agencies (see section 306(g) of the Freddie Mac Act and section 304(d) of the Fannie Mae Charter Act); ^[1] and

• Exemption from all State and local taxes except property taxes (see section 303(e) of the Freddie Mac Act and section 309(c)(2) of the Fannie Mae Charter Act).

While the securities that the GSEs guarantee, and the debt instruments they issue, are explicitly not backed by the full faith and credit of the United States, and nothing in this rule should be construed otherwise, such securities and instruments trade at yields only a few basis points over those of U.S. Treasury securities with comparable terms. These securities also offer yields lower than those for securities issued by fully private firms that are more highly capitalized but otherwise comparable. In addition, the market does not require that individual GSE securities be rated by a national rating agency. Consequently, the GSEs are able to fund their operations at lower cost than other private firms with similar financial characteristics. In a recent report, the Congressional Budget Office (CBO) estimated that this funding advantage for the year 2003 resulted in a $19.6 billion annual combined subsidy for both GSEs. Of this amount, CBO estimated that the GSEs retained about $6.2 billion, or approximately one-third of the subsidy, for their officers and shareholders, while the remainder accrued to borrowers.^[2]

In return for the public benefits they receive, Congress has mandated in the GSEs' Charter Acts that the GSEs carry out public purposes not required of other private sector entities in the housing finance industry. These statutory mandates obligate the GSEs to work to ensure that everyone in the nation has a reasonable opportunity to enjoy access to the mortgage financing benefits resulting from the activities of these enterprises.

With respect to these public purposes, Congress does not simply expect the GSEs to strive toward achievement of these purposes but rather to “lead the mortgage finance industry” and to “ensure that citizens throughout the country enjoy access to the public benefits provided by these federally related entities.” (See S. Rep. No. 102-282, at 34 (1992).)

C. Statutory and Regulatory Background

The statutory and regulatory background applicable to the chartering of Fannie Mae and Freddie Mac and HUD's regulatory authority over these two GSEs were set out in detail in the preamble to HUD's proposed rule published on May 3, 2004 (69 FR 24228). Therefore, this background information is not repeated here in the preamble to this final rule. Interested members of the public should refer to Section I.A. of the preamble to the proposed rule at pages 69 FR 24228 through 69 FR 24230 for this information.

D. The Proposed Rule

On May 3, 2004, HUD published a proposed rule setting forth new housing goal levels for Fannie Mae and Freddie Mac. (See 69 FR 24228.) HUD's rule proposed to increase the level of the housing goals (“Housing Goals”) for the purchase by Fannie Mae and Freddie Mac of mortgages financing low- and moderate-income housing, special affordable housing, and housing in central cities, rural areas, and other underserved areas. The rule also proposed to establish new subgoals for the GSEs' acquisitions of home purchase loans that qualify for each of the housing goals.

In addition to soliciting public comments on the proposed goal levels and new subgoals, the rule solicited public comments on several other issues related to the housing goals, including: (1) Provisions relating to GSE data integrity, such as verification, certification, treatment of errors, omissions or discrepancies, and other enforcement authority; (2) amended definitions of “underserved area,” “metropolitan area” and “minority,” and a new definition of the term “home purchase mortgage”; (3) a method for imputing the distribution of GSE-purchased mortgages that lack income data; and (4) other changes related to the GSEs' bulk purchases of seasoned loans. More detailed information about HUD's proposals can be found in the preamble to HUD's May 3, 2004, proposed rule.

E. This Final Rule—Overview

Under this 2004 rulemaking, the Department is setting new, higher levels for the Housing Goals, accompanied by subgoals under each of the Housing Goals for purchases of home purchase mortgages (i.e., excluding refinance mortgages) on owner-occupied properties in metropolitan areas. (The subgoals are referred to in this rule as the “Home Purchase Subgoals.”)

The Department's purpose in setting higher Housing Goals and in establishing new Home Purchase Subgoals in this final rule is to encourage the GSEs to facilitate greater financing and homeownership opportunities for families and neighborhoods targeted by the Housing Goals. The final rule establishes levels of the Housing Goals that will bring the GSEs to a position of market leadership in a range of foreseeable economic circumstances related to the future course of interest rates and consequent fluctuations in origination rates on home purchase and refinance mortgages—both multifamily and single-family.

For each goal, HUD has projected goal-qualifying percentages of mortgage originations in terms of ranges that cover a variety of economic scenarios. The objective of HUD's Housing Goals is to bring the GSEs' performance to the upper end of HUD's market range estimate for each goal, consistent with the requirement in FHEFSSA that HUD should consider the GSEs' ability to lead the market for each goal.

To enable the GSEs to achieve this leadership, the Department has established staged increases in Housing Goal levels for 2005, which will increase further, year-by-year through 2008, to achieve the ultimate objective for the GSEs to lead the market under a range of foreseeable economic circumstances by 2008.

The staged increases established by this rule, are consistent with the statutory requirement that HUD consider the past performance of the GSEs in setting the Housing Goals. Staged annual increases in the Goals will provide the GSEs with the opportunity to adjust their business models, so as to meet the required 2008 levels without compromising other business objectives and requirements.

The Department believes that the Home Purchase Subgoals established by this final rule are necessary and warranted. Increasing homeownership is a national priority. The past average performance of the GSEs in the home purchase market has been below market levels. As further discussed below, the GSEs must apply greater efforts to increasing homeownership for low- and moderate-income families, families living in underserved areas, and very-low income families and low-income families living in low-income areas. The addition of Home Purchase Subgoals to the regulatory structure will serve to better focus the GSEs' efforts in a clear and transparent manner. The Home Purchase Subgoals will better allow the government and public alike to monitor the GSEs' efforts in meeting the nation's homeownership needs. The increases in the levels of the Housing Goals, and the addition of the new Home Purchase Subgoals, are predicated upon the Department's recognition that the GSEs not only have the ability to achieve these Housing Goals and Subgoals but, also, that they are fully consistent with the statutory factors established under FHEFSSA. In addition, this rule is supported by the Department's comprehensive analyses of the size of the mortgage market, the opportunities available to the GSEs, America's unmet housing needs, and identified credit gaps.

In addition to the establishment of higher Housing Goals for the years 2005 through 2008, and the establishment of Home Purchase Subgoals, specific changes included in the final rule from the provisions included in the May 3, 2004, proposed rule are as follows:

(1) The final rule expands the existing provisions to permit the GSEs to impute incomes or rents when data are missing for some purchases, addressing the market's expanding use of low documentation mortgages;

(2) The final rule provides that goals credit is available for purchases of loans in transactions involving seller dissolution options, such as repurchase agreements, only when the option provides for a minimum one-year lockout period;

(3) The final rule clarifies the proposed provisions regarding HUD's procedures for correcting errors, omissions and discrepancies in current year-end data and in remedying material overstatements of housing goals performance for prior years;

(4) The final rule changes the scope of the proposed certification statement that the GSEs must provide to make it closer to the certification used by the Office of Federal Housing Enterprise Oversight (OFHEO), the GSEs' financial safety and soundness regulator; and

(5) The final rule makes a technical correction to the special counting rules prohibiting double counting of GSE purchases of seasoned mortgages toward the housing goals.

In developing these regulations, the Department was guided by, and re-affirms, the following principles established in the Housing Goals 1995 final rule (published on December 1, 1995 at 60 FR 1846):

(1) The GSEs should fulfill FHEFSSA's intent that they lead the industry in ensuring that access to mortgage credit is made available for very low-, low- and moderate-income families and residents of underserved areas. HUD recognizes that, to lead the mortgage industry over time, the GSEs will have to stretch to reach certain Housing Goals and to close gaps between the secondary mortgage market and the primary mortgage market for various categories of loans. This recognition is consistent with the Congressional directive that “the enterprises will need to stretch their efforts to achieve” the goals. (See S. Rep. No. 102-282, at 35 (1992).)

(2) The Department's role as a regulator is to set broad performance standards for the GSEs through the Housing Goals, but not to dictate the specific products or delivery mechanisms the GSEs will use to achieve a Housing Goal. Regulating two exceedingly large financial enterprises in a dynamic market requires that HUD provide the GSEs with sufficient latitude to use their innovative capacities to determine how best to develop products to carry out their respective missions. HUD's regulations are intended to allow the GSEs the flexibility to respond quickly to market opportunities. At the same time, the Department must ensure that the GSEs' strategies address national credit needs, especially as they relate to housing for low- and moderate-income families and housing located in underserved geographical areas. The addition of Home Purchase Subgoals to the regulatory structure provides an additional means of encouraging the GSEs' affordable housing activities to address identified, persistent credit needs while leaving to the GSEs the specific approaches used to meet these needs.

(3) Discrimination in lending continues to limit access to credit for purchasing homes by racial and ethnic minorities. Troublesome gaps in homeownership remain for minorities even after record growth in affordable lending and homeownership during the nineties. Studies indicate that, over the next few years, minorities will account for a growing share of the families seeking to buy their first home. HUD's analyses indicate, however, that Fannie Mae and Freddie Mac account for a disproportionately small share of the minority first-time homebuyer market. The GSEs have a responsibility to promote access to capital for minorities and others who are seeking their first homes, and to demonstrate the benefits of such lending to industry and borrowers alike. The GSEs also have an integral role in eliminating predatory mortgage lending practices.

(4) In addition to the GSEs' purchases of single-family home mortgages, the GSEs also must continue to assist in the creation of an active secondary market for mortgages on multifamily rental housing. Affordable rental housing is essential for those families who cannot afford to become, or who choose not to become, homeowners. For this reason, the GSEs must assist in making capital available to assure the continued development of single-family and multifamily rental housing.

II. Discussion of Public Comments

A. Overview of Public Comments

At the close of the public comment period on July 16, 2004, which was extended an additional two weeks beyond the original public comment deadline of July 2, 2004, HUD had received 302 comments, which are in HUD's docket file for this rule. In addition to the public comments received on the rule, during the public comment period, HUD met with representatives of several organizations, including Fannie Mae and Freddie Mac, to accommodate oral presentation of concerns about the rule. HUD's docket file for this rule contains information on the dates of these meetings, the attendees, and the subject discussed.

Of the public comments received on the proposed rule, the most detailed comments were those submitted by the two directly affected GSEs, Fannie Mae and Freddie Mac. Neither GSE was supportive of the higher goal levels proposed for 2005-2008, nor did either support the creation of HUD's proposed Home Purchase Subgoals. The GSEs stated, among other comments that they made on the rule, that the effect of many goals and subgoals would be micromanagement of the GSEs. With their comments, the GSEs provided several appendices that provided alternative analyses of data and questioned the Department's methodology in determining market share for the three affordable housing goals, a key component for establishing the appropriate level of the housing goals and the subgoals.

The GSEs did not object to HUD's special affordable multifamily subgoal levels for 2005-2008, but other commenters (mostly public advocacy groups) recommended that HUD increase the levels of these subgoals.

In addition to the GSEs, the commenters included national and regional housing industry organizations, nonprofit organizations, alliances, councils, and advocacy organizations involved in housing or housing issues, lenders, academic researchers, Members of Congress, state and local government officials, and two individuals.

In large measure, except for several nonprofit organizations and public advocacy groups that favored higher goals, the majority of commenters were not supportive of HUD's proposed goals, especially in the outer years when the goal levels would reach their highest levels. A particular concern cited by a number of commenters was the potential for adverse impact on middle-income borrowers, particularly higher interest rates and fees. Another concern raised by the commenters was the possibility of unintended consequences for the industry. Many commenters, including the GSEs, urged HUD to exclude all single-family refinances from the calculation of the goals.

The Department received fewer comments that addressed other proposals in the rule, such as those regarding data integrity, large-scale transactions involving seasoned loans, the treatment of missing income data, and modifying the definition of rural underserved areas. For those commenters who submitted comments on these proposals, the reactions were generally mixed.

With respect to HUD's proposals for new data integrity provisions, the majority of those who commented on the new data integrity proposals were generally supportive of the concept and acknowledged the need for some sort of data verification process. However, two industry-related commenters expressed concern about the potential for HUD's proposals to result in increased reporting burdens for lenders. The GSEs' comments also reflected several concerns about the data integrity provisions, mainly with respect to definitions, procedures, and enforcement.

The GSEs favored generous proxy provisions for the treatment of missing income data and submitted several suggestions. The majority of commenters on this issue, consisting chiefly of nonprofit and advocacy organizations, opposed using proxies, and several favored an outright ban on purchasing “no income” subprime mortgages.

With regard to large-scale transactions involving seasoned loans, the GSEs commented that they should receive housing goals credit and that no change in HUD's current definition of “mortgage purchase” was warranted. However, a group of industry-related organizations opposed providing goals credit for seasoned loans, as did several advocacy organizations. Commenters offered no alternative definitions for “mortgage purchase” in HUD's regulations.

All but one commenter who addressed the issue of HUD's rural underserved area definition favored changing this definition to one that is census tract-based, rather than county-based. Those commenters favoring conversion to a tract-based definition believed that county-level data do not show disparities in service that the GSEs should address. The dissenting commenter felt that lenders serving rural areas would face operational difficulties and expenses in shifting to a tract-based orientation.

In addition to comments on its proposals related to housing goals, HUD received other comments on subjects pertaining to HUD's regulatory authority over the GSEs but which were not related to the rule's proposals on housing goals (for example, comments on new program authority, monitoring and reporting procedures, and public access to GSE mortgage data). Because these comments raised issues outside the scope of the May 3, 2004, proposed rule, they are not addressed in this final rule.

A discussion of the general and specific comments on the rule, as well as HUD's responses to these comments, follows in subsequent sections in this preamble, as well as in the Appendices to this Final Rule. While comments are summarized, not all the comments are addressed explicitly in this preamble. HUD is appreciative of the full range of public comments received and acknowledges the value of all of the comments submitted in response to the proposed rule.

B. Subpart A—General

In the May 3, 2004, rule, HUD proposed to add a definition of “home purchase mortgage” in connection with its proposal to specify Home Purchase Subgoals under each of the three Housing Goals, to revise the definitions of “metropolitan area” and “minority” to conform HUD's regulations to changes in data collection practices made by the Office of Management and Budget (OMB), and to modify the current definition of “underserved area” with respect to the delineation of underserved portions of non-metropolitan areas.

1. Home Purchase Mortgage

HUD proposed to insert a definition of “home purchase mortgage” for purposes of specifying the Home Purchase Mortgage Subgoals. Since no comments bearing directly on this definition were received and the Department has retained the subgoal concept in this final rule, the definition is adopted.

2. Metropolitan Area

HUD proposed to alter the definition of “metropolitan area” to reflect a change in the definition of “metropolitan area” recently promulgated by OMB, in which the concept of “Primary Metropolitan Statistical Area” was removed. No comments were received on this proposed change; accordingly, it is adopted.

3. Minority

HUD proposed to alter the definition of “minority” to reflect changes in standards for the classification of federal data on race and ethnicity previously promulgated by OMB and implemented in the 2000 census and in data collection under the Home Mortgage Disclosure Act in 2004. No comments were received on this proposed change; accordingly, it is adopted.

4. Underserved Area

HUD proposed to alter the definition of “underserved area” to provide for the specification of underserved areas outside of metropolitan areas at the census tract level rather than at the county level.

For properties in non-metropolitan (rural) areas, mortgage purchases have counted toward the Underserved Areas Housing Goal where such purchases finance properties that are located in underserved counties. This final rule incorporates a determination that mortgage purchases will count toward the Underserved Areas Housing Goal where such purchases finance properties that are located in underserved census tracts. These are defined as census tracts where either: (1) the median income in the tract does not exceed 95 percent of the greater of the median income for the non-metropolitan portions of the state or the median income of the non-metropolitan portions of the nation as a whole; or (2) minorities comprise at least 30 percent of the residents and the median income in the tract does not exceed 120 percent of the greater of the median incomes for the non-metropolitan portions of the state or of the nation as a whole.

HUD originally adopted its current county-based definition for targeting GSE purchases to underserved non-metropolitan areas primarily based on information that rural lenders did not perceive their market areas in terms of census tracts, but rather, in terms of counties. A further concern was an apparent lack of reliability of geocoding software applied to non-metropolitan areas.

Thirteen commenters endorsed HUD's proposed change in definition, observing that the change would produce more precise targeting and improved service toward underserved segments of the market within counties. One banking trade association advocated continuation of a county-based definition, stating that because the business perspective of community banks in rural areas is geared toward entire counties, there would be costs associated with monitoring the tract location of loans, and therefore, marketing toward borrowers at the tract level would be difficult.

Recent research summarized in Appendix B to this rule indicates that a tract-based system will improve the extent to which the underserved area definition distinguishes areas by key socioeconomic and demographic characteristics such as median family income, poverty, unemployment, school dropout rates, and minority populations. Under a tract-based definition underserved areas stand out more as areas of lower income and low economic activity and as having somewhat larger minority population proportions. A tract-based definition will also improve the targeting of the goal to areas with relatively greater housing needs. Based on these findings, which are detailed in Appendix B to this rule, HUD is adopting a re-specification of underserved areas within non-metropolitan (rural) areas to be based on census tracts rather than counties.

C. Subpart B—Housing Goals

1. Overview

A substantial majority of the comments received criticized HUD's proposed levels of the housing goals on the basis that they would be difficult for the GSEs to achieve, particularly in periods of high refinance activity when higher-income borrowers comprise a relatively high proportion of mortgage borrowers. Several types of adverse consequences of such high goals were forecast, including diminution of availability of mortgage credit to some sectors of the mortgage market, unfavorable effects on neighborhood housing quality, and other adverse effects discussed below. This section of the final rule reviews the statutory factors the Department must consider in setting the level of the housing goals and the Department's determinations with regard to the levels of each of the housing goals as well as the proposed Home Purchase Subgoals.

2. Statutorily Required Factors in Setting the Levels of the Housing Goals and Subgoals

The Housing Goals and Home Purchase Subgoals being implemented by this final rule were established following consideration of the six factors required by statute to be considered in establishing goal levels and establishing subgoals. A summary of HUD's findings relative to each of the six statutory factors follows. More detailed discussion of these points is included in Appendices A, B, and C to this rule.

a. Demographic, Economic, and Housing Conditions

(i) Demographic Trends

Changing population demographics will result in a need for the primary and secondary mortgage markets to meet nontraditional credit needs, respond to diverse housing preferences and overcome information and other barriers that many immigrants and minorities currently face.

The U.S. Census Bureau has projected that the U.S. population will grow by an average of 2.5 million persons per year between 2000 and 2025, resulting in about 1.2 million new households per year. The aging of the baby-boom generation and the entry of the baby-bust generation into prime home-buying age will have a dampening effect on housing demand. Growing housing demand from minorities, immigrants and non-traditional homebuyers will help offset declines in the demand for housing caused by the aging of the population.

The continued influx of immigrants will increase the demand for rental housing, while those who immigrated during the 1980s and 1990s will be in the market for homeownership. Immigrants and minorities—who accounted for nearly 40 percent of the growth in the nation's homeownership rate over the past five years—will be responsible for almost two-thirds of the growth in the number of new households over the next 10 years.

Non-traditional households have become more important, as overall household formation rates have slowed. With later marriages, divorce, and non-traditional living arrangements, the fastest growing household groups are single-parent and single-person households. By 2025, non-family households will make up one-third of all households. The role of traditional 25-to-34 year-old married, first-time homebuyers in the housing market will be smaller in the current decade due to the aging of the population. Between 2000 and 2025, the Census Bureau projects that the largest growth in households will occur among householders who are age 65 and older.

As these demographic factors play out, the overall effect on housing demand will likely be continued growth and an increasingly diverse household population from which to draw new renters and homeowners. A greater diversity in the housing market will, in turn, require greater adaptation by the primary and secondary mortgage markets.

(ii) Economic and Housing Conditions

While most other sectors of the economy were weak or declining during 2001 and 2002, the housing sector showed remarkable strength. The housing market continued at a record pace during 2003.

In 2002, the U.S. economy moved into recovery, with real Gross Domestic Product (GDP) growing 2.2 percent, although measures of unemployment continued to rise before declining again in 2003. In October 2002, the average 30-year home mortgage interest rate slipped below 6 percent for the first time since the mid-1960s. Favorable financing conditions and solid increases in house prices were the key supports to record housing markets during both 2002 and 2003. By the end of 2003, the industry had set new records in single-family home permits, new home sales, existing home sales, low interest rates, and rates of homeownership. Other indicators—total permits, starts, completions, and affordability—reached levels that were among the highest in the past two decades.

The Administration's forecast for real GDP growth is 3.7 percent for 2005 and 3.1-3.4 percent in 2006-2009, while CBO projects that real GDP will grow at an average rate of 4.1 percent in 2005 and annual rates of 2.9-3.2 percent in 2006 through 2009.^[3] The Administration projects the 10-year Treasury rate to average 5.1 percent in 2005 and 5.4-5.8 percent between 2006 and 2009 compared to its average of 4.6 percent in 2002 and 4.0 percent in 2003.

Standard & Poor's expects housing starts to average 1.8 million units in 2004-2005. Fannie Mae projects existing home sales for 2004 at 6.1 million units, and for 2005 at 5.8 million, compared to their record level in 2003 of 6 million units.

(iii) Mortgage Market Conditions

Low interest rates and record levels of refinancing caused mortgage originations to soar from $2.0 trillion in 2001 to $2.6 trillion in 2002 and around $3.8 trillion in 2003. The Mortgage Bankers Association projects that mortgage originations will drop to $2.7 trillion in 2004 and $1.8 trillion in 2005, as refinancing returns to more normal levels.^[4]

The volume of home purchase mortgages was $910 billion to $1.1 trillion between 1999 and 2001 before jumping to $1.2 trillion in 2002 and $1.3 trillion in 2003. As with housing starts, the home purchase origination market is expected to exhibit sustained growth.

b. National Housing Needs

(i) Affordability Problems

Data from the 2000 Census and the American Housing Survey demonstrate that there are substantial housing needs among low- and moderate-income families. Many of these households are burdened by high homeownership costs or rent payments and, consequently, are facing serious housing affordability problems. There is evidence of persistent housing problems for Americans with the lowest incomes. Since 1977 the percentage of U.S. households with worst case needs has hovered around five percent, with the worst year being 1983 (6.03 percent) and the best being 1999 (4.72 percent). The proportion in 2001 was 4.77 percent, which is not significantly different from the 1999 figure. HUD's analysis of American Housing Survey data reveals that, in 2001, 5.1 million unassisted very-low-income renter households had “worst case” housing needs, defined as housing costs greater than 50 percent of household income or severely inadequate housing. Among these households, 90 percent had a severe rent burden, 6 percent lived in severely inadequate housing, and 4 percent suffered from both problems. Among the 34 million renters in all income categories, 6.3 million (19 percent) had a severe rent burden and over one million renters (3 percent) lived in housing that was severely inadequate.

(ii) Disparities in Housing and Mortgage Markets

Despite the strong growth in affordable lending over the past ten years, there are families who are not being adequately served by the nation's housing and mortgage markets. Serious racial and income disparities remain. The homeownership rate for minorities is 25 percentage points below that for whites. A major HUD-funded study of discrimination in the sales and rental markets found that discrimination still persists in both rental and sales markets of large metropolitan areas nationwide, although its incidence has generally declined since 1989. The most prevalent form of discrimination observed in the study against Hispanic and African-American home seekers was Hispanics and African Americans being told that housing units were unavailable when non-Hispanic whites found them to be available. Levels of consistent adverse treatment experienced by the nation's largest minority groups when they inquire about a unit advertised for sale in metropolitan areas nationwide in 2000-2001 were: African Americans 16.8 percent, Hispanics 18.3 percent, and Asians and Pacific Islanders 20.4 percent.

The study also found other worrisome trends of discrimination in metropolitan housing markets that persisted in 2000. Examples include geographical steering experienced by African-American homebuyers, and real estate agents who provided less assistance in obtaining financing for Hispanic homebuyers than for non-Hispanic whites.^[5] Racial disparities in mortgage lending are also well documented. HUD-sponsored studies of the pre-qualification process conclude that African Americans and Hispanics risk unequal treatment when they visit mainstream mortgage lenders. Studies reveal higher mortgage denial rates for African Americans and Hispanics, even after controlling for applicant income and a host of underwriting characteristics, such as the credit record of the applicant.^[6] However, substantial progress has been made since 1989.

The existence of substantial neighborhood disparities in homeownership and mortgage credit is also well documented for metropolitan areas. HUD's analysis of Home Mortgage Disclosure Act (HMDA) data shows that mortgage credit flows in metropolitan areas are substantially lower in high-minority and low-income neighborhoods and that mortgage denial rates are much higher for residents of these neighborhoods. Studies have also documented that mainstream lenders often do not operate in inner-city minority neighborhoods, leaving their residents with only high-cost lenders as options. Too often, residents of these same neighborhoods have been subjected to the abusive practices of predatory lenders.

These troublesome disparities mostly affect those families (minorities and immigrants) who are projected to account for almost two-thirds of the growth in the number of new households over the next 10 years.

(iii) Single-Family Market: Trends in Affordable Lending and Homeownership

Many younger, minority and lower-income families did not become homeowners during the 1980s due to slow growth of some earnings, high real interest rates, lower inflation, and continued increases in housing prices. Over the past 10 years, economic expansion, accompanied by low interest rates and increased outreach on the part of the mortgage industry, has improved affordability conditions for these families.

As this preamble and the appendices note, there has been a “revolution in affordable lending” that has extended homeownership opportunities to historically underserved households. The mortgage industry, including the GSEs, has offered more customized mortgage products, more flexible underwriting, and expanded outreach to low-income and minority borrowers.

HMDA data suggest that the industry and GSE initiatives are increasing the flow of credit to underserved borrowers. Between 1993 and 2002, conventional loans to low-income and minority families increased at much faster rates than loans to upper-income and non-minority families. Conventional home purchase originations to African Americans more than doubled between 1993 and 2002, and those to Hispanic borrowers more than tripled during this period. Home loans to low-income borrowers and to low-income and high-minority census tracts also more than doubled during this period.

Thus, the 1990s and the early part of the current decade have seen the development of a strong affordable lending market. Homeownership statistics show similar trends. After declining during the 1980s, the homeownership rate has increased every year since 1994, reaching a record mark of 69.2 percent in the second quarter of 2004.

The number of households owning their own home increased by 13.3 million between 1994 and 2003. Gains in homeownership rates during the period of 1994 to 2003 have been widespread, with the homeownership rate for African-American households increasing from 42.5 percent to 48.8 percent, for Hispanic households from 41.2 percent to 46.7 percent, for non-Hispanic white households from 70.0 percent to 75.4 percent, and for central city residents from 48.5 percent to 52.3 percent.

Despite the record gains in homeownership since 1994, a gap of approximately 25 percent in the homeownership rate prevails for African-American and Hispanic households as compared to white non-Hispanic households. Studies show that these lower homeownership rates are only partly accounted for by differences in income, age, and other socioeconomic factors.

In addition to low income, barriers to homeownership that disproportionately affect minorities and immigrants include: lack of capital for downpayment and closing costs; poor credit history; lack of access to mainstream lenders; little understanding of the home buying process; a limited supply of modestly priced homes in locations where these populations reside; and continued discrimination in housing markets and mortgage lending. These barriers are discussed in Appendix A to this rule.

(iv) Single-Family Market: Potential Homeowners

As already noted, the potential homeowner population over the next decade will be highly diverse, as growing housing demand from immigrants (both those who are already in this country and those who are projected to arrive), minorities, and non-traditional homebuyers will help to offset declines in the demand for housing caused by the aging of the population.

Studies cited in Appendix A to this rule reveal that increased immigration during the 1990s directly accounted for 35 percent of the nation's rise in population during that decade, as a result of which the foreign-born population of the United States was 31.1 million in 2000. These trends do not depend on the future inflow of new immigrants, as immigrants do not, on average, enter the home purchase market until they have been in this country for eleven years. Fannie Mae staff have noted that there are enough immigrants already in this country to keep housing demand strong for several years.

Thus, the need for the GSEs and other industry participants to meet nontraditional credit needs, respond to diverse housing preferences, and to overcome the information barriers that many immigrants face will take on added importance. A new or recent immigrant may have no credit history or, at least, may not have a credit history that can be documented by traditional methods. In order to address these needs, the GSEs and the mortgage industry have been developing innovative products and seeking to extend their outreach efforts to attract these homebuyers, as discussed in Appendix A to this rule.

In addition, the current low homeownership rates in inner cities (as compared with the suburbs) also suggest that urban areas may be a potential growth market for lenders. As explained in Appendix A to this rule, lenders are beginning to recognize that urban borrowers and properties have different needs than suburban borrowers and properties.

Surveys indicate that these demographic trends will be reinforced by the fact that most Americans desire, and plan, to become homeowners. According to Fannie Mae's 2002 National Housing Survey, Americans rate homeownership as the best investment they can make, far ahead of 401(k) plans, other retirement accounts, and stocks. Forty-two percent of African-American families reported that they were “very or fairly likely” to buy a home in the next three years, up from 38 percent in 1998 and 25 percent in 1997. Among Hispanics and Hispanic immigrants, the numbers reached 37 percent and 34 percent, respectively. The survey also reported that more than half of Hispanic renters cite homeownership as being “one of their top priorities.”

Despite these trends, potential minority and immigrant homebuyers see more obstacles to buying a home than does the general public. Typically, the primary barriers to homeownership are credit issues and a lack of funds for a downpayment and closing costs. However, other barriers also exist, such as a lack of affordable housing, little understanding of the home buying process, and language barriers. Thus, the new group of potential homeowners will have unique needs.

The GSEs can play an important role in tapping this potential homeowner population. Along with others in the industry, they can address these needs on several fronts, such as expanding education and outreach efforts, introducing new products, and adjusting current underwriting standards to better reflect the special circumstances of these new households. These efforts are necessary for achieving the Administration's goal of expanding minority homeownership by 5.5 million families by the end of the decade.

The single-family mortgage market has been very dynamic over the past few years, experiencing volatile swings in originations (with the 1998 and 2001-2003 refinancing waves), witnessing the rapid growth in new types of lending (such as subprime lending), incorporating new technologies (such as automated underwriting systems), and facing serious challenges (such as predatory lending). Fannie Mae and Freddie Mac have played a major role in the ongoing changes in the single-family market and in helping the industry address the problems and challenges that have arisen.

The appendices to this final rule discuss the various roles that Fannie Mae and Freddie Mac have played in the single-family market. A wide range of topics is examined, including the GSEs' automated underwriting technology used throughout the industry, their many affordable lending partnerships and underwriting initiatives aimed at extending credit to underserved borrowers, their development of new targeted low-downpayment products, their entry into new markets such as the subprime market, and their attempts to reduce predatory lending. As that discussion emphasizes, the GSEs have the ability to bring increased efficiencies to a market and to attract mainstream lenders into markets. (Readers are referred to Appendices A, B, and C to this rule for further discussion of the GSEs' role in different segments of the single-family mortgage market.)

(v) Multifamily Mortgage Market

The market for financing of multifamily apartments has reached record volume. The favorable long-term prospects for apartments, combined with record low interest rates, have kept investor demand for apartments strong and have supported property prices despite recently high vacancy rates.

Fannie Mae and Freddie Mac have been among those boosting their volumes of multifamily financing and both have introduced new programs to serve the multifamily market. Fannie Mae and, especially (considering its earlier withdrawal from the market), Freddie Mac have rapidly expanded their presence in the multifamily mortgage market under the Housing Goals.

Freddie Mac has successfully rebuilt its multifamily acquisition program, as reflected by the increase in its purchases of multifamily mortgages: from $27 million in 1992 to $3.9 billion in 1998 and then rising to $9.5 billion in 2001, $10.7 billion in 2002, and $21.5 billion in 2003. Multifamily units accounted for 9.0 percent of all dwelling units (both owner and rental) financed by Freddie Mac between 1999 and 2003. Concerns regarding multifamily capabilities no longer constrain Freddie Mac's performance with regard to the Housing Goals.

Although Fannie Mae never withdrew from the multifamily market, it has stepped up its activities in this area substantially, with multifamily purchases rising from $3.0 billion in 1992 to $10.0 billion in 1999, and $19.1 billion in 2001, then declining slightly to $16.6 billion in 2002, and then rising markedly to $30.9 billion in 2003. Multifamily units accounted for 8.8 percent of all dwelling units (both owner and rental) financed by Fannie Mae between 1999 and 2003.

The increased role of Fannie Mae and Freddie Mac in the multifamily market has major implications for the Low- and Moderate-Income Housing and Special Affordable Housing Goals, since high percentages of multifamily units have affordable-level rents and can count toward one or both of these Housing Goals. However, the potential of the GSEs to lead the multifamily mortgage industry has not been fully developed. The GSEs' purchases between 1999 and 2002 accounted for less than 40 percent of the multifamily units that received financing during this period. Certainly there are ample opportunities and room for expansion of the GSEs' share of the multifamily mortgage market.

The GSEs' size and market position between loan originators and mortgage investors make them the logical institutions to identify and promote needed innovations and to establish standards that will improve market efficiency. As their role in the multifamily market continues to grow, the GSEs will have the knowledge and market presence to push simultaneously for standardization and for programmatic flexibility to meet special needs and circumstances, with the ultimate goal of increasing the availability and reducing the cost of financing for affordable and other multifamily rental properties.

The long-term outlook for the multifamily rental market is sustained, moderate growth, based on favorable demographics. The minority population, especially Hispanics, provides a growing source of demand for affordable rental housing. “Lifestyle renters” (older, middle-income households) are also a fast-growing segment of the rental population.

At the same time, the provision of affordable housing units will continue to challenge suppliers of multifamily rental housing as well as policy makers at all levels of government. Low incomes, combined with high housing expenses, define the difficult situation of millions of renter households. Housing cost reductions are constrained by high land prices and construction costs in many markets. Regulatory barriers at the state and local level have an enormous impact on the development of affordable rental housing. Government action—through land use regulation, building codes, and occupancy standards—is a major contributor to high housing costs.

Since the early 1990s, the multifamily mortgage market has become more closely interconnected with global capital markets, although not to the same degree as the single-family mortgage market. Loans on multifamily properties are still viewed as riskier by some than are mortgages on single-family properties, despite delinquency rates that in recent quarters have been lower than those on single-family mortgages.

There is a need for an ongoing GSE presence in the multifamily secondary market, both to increase liquidity and to advance affordable housing efforts. The potential for an increased GSE presence is enhanced by the fact that an increasing proportion of multifamily mortgages are now originated in accordance with secondary market standards. Small multifamily properties, and multifamily properties with significant rehabilitation needs, have historically experienced difficulty gaining access to mortgage financing, and the flow of capital into multifamily housing for seniors has been historically characterized by volatility. The GSEs can play a role in promoting liquidity for multifamily mortgages and increasing the availability of long-term, fixed-rate financing for these properties.

c. GSEs' Past Performance and Effort Toward Achieving the Housing Goals

Since the enactment of FHEFSSA and HUD's establishment in 1993 of the Housing Goals, both Fannie Mae and Freddie Mac have improved their affordable housing loan performance. However, the GSEs' mortgage purchases have generally lagged, and not led, the overall primary market in providing financing for affordable housing to low- and moderate-income families and underserved borrowers and their neighborhoods, indicating that there is more that the GSEs can do to improve their performance.

(i) Performance on the Housing Goals

The year 2001 was the first year under the higher levels of the Housing Goals established in the Housing Goals 2000 final rule. Fannie Mae met all three Housing Goals in 2001, 2002, and 2003. Freddie Mac met all three Housing Goals in 2001 and 2003. However, in 2002 HUD discovered that Freddie Mac had counted 22,371 housing units towards the Low- and Moderate Income Goal even though it had previously counted these same housing units towards the same goal in 2001. Freddie Mac also counted 22,424 housing units towards the Underserved Area Goal even though these units had also been credited towards the same goal in 2001. HUD's regulations prohibit double counting. To correct for these double-counting errors, the Department has adjusted its official performance results for Freddie Mac in 2002 by deducting the double-counted housing units, including all bonus point credit that had been awarded for most of these units, from the official performance results it had previously reported publicly. As a result of these adjustments, Freddie Mac continued to meet the Low- and Moderate-Income Goal in 2002. However, Freddie Mac fell short of the 31 percent target for the Underserved Areas goal by 90 units or 0.002 percent. Freddie Mac's 2002 goal performance results are described more fully in Tables 4, 6 and 8 in this preamble, including the accompanying footnotes.

(ii) The GSEs' Efforts in the Home Purchase Mortgage Market

The Appendices to this final rule include a comprehensive analysis of each GSE's performance in funding home purchase mortgages for borrowers and neighborhoods targeted by the three Housing Goals—special affordable and low- and moderate-income borrowers and underserved areas. The GSEs' role in the first-time homebuyer market is also analyzed. Because homeownership opportunities are integrally tied to the ready availability of affordable home purchase loans, the main findings from that analysis are provided below.

• Both Fannie Mae and Freddie Mac have increased their purchases of affordable home purchase mortgages since the Housing Goals were put into effect, as indicated by the increasing share of their business going to the three goals-qualifying categories. Between 1992 and 2003, the special affordable share of Fannie Mae's business almost tripled, rising from 6.3 percent to 17.1 percent, while the underserved areas share increased more modestly, from 18.3 percent to 26.8 percent. The figures for Freddie Mac are similar. The special affordable share of Freddie Mac's business rose from 6.5 percent to 15.6 percent, while the underserved areas share also increased but more modestly, from 18.6 percent to 24.0 percent.
• While both GSEs improved their performance, they have historically lagged the primary market in providing affordable home purchase mortgage loans to low-income borrowers and underserved neighborhoods. Freddie Mac's average performance, in particular, fell far short of market performance during the 1990s. Fannie Mae's average performance was better than Freddie Mac's during the 1993-2003 period as well as during the 1996-2003 period, which covers the period under HUD's currently-defined Housing Goals. Between 1993 and 2003, 12.2 percent of Freddie Mac's mortgage purchases were for special affordable borrowers, compared with 13.3 percent of Fannie Mae's purchases, 15.4 percent of loans originated by depositories, and 15.5 percent of loans originated in the conventional conforming market (without estimated B&C subprime loans).^[7]

• Between 2001 and 2003, both Fannie Mae and Freddie Mac fell significantly below the market in funding affordable home purchase mortgage loans. During this period, special affordable loans accounted for 15.1 percent of Fannie Mae's purchases, 14.7 percent of Freddie Mac's purchases, and 16.2 percent of loans originated in the market; thus, the “Fannie-Mae-to-market” ratio was 0.93 and the “Freddie-Mac-to-market” ratio was also 0.91. During the same period, underserved area loans accounted for 24.7 percent of Fannie Mae's purchases, 23.1 percent of Freddie Mac's purchases, and 26.2 percent of loans originated in the market; the “Fannie-Mae-to-market” ratio was 0.94 and the “Freddie-Mac-to-market” ratio was only 0.88.
• While Freddie Mac has improved its affordable lending performance in the past two years, it has continued to lag the conventional conforming market in funding affordable home purchase loans for special affordable and low- and moderate-income borrowers and underserved neighborhoods targeted by the Housing Goals. In 2003, Freddie Mac's performance on the underserved areas goal was particularly low relative to both the performances of Fannie Mae and the market; in that year, underserved area loans accounted for only 24.0 percent of Freddie Mac's purchases compared with 26.8 percent of Fannie Mae's purchases and 27.6 percent of market originations.
• As noted above, Fannie Mae's average performance during past periods (e.g., 1993-2003, 1996-2003, 1999-2003) has been below market levels. However, it is encouraging that Fannie Mae markedly improved its affordable lending performance relative to the market during 2001, 2002, and 2003, the first three years under the higher housing goal targets that HUD established in the GSE Final Rule dated October 2000. Over this three-year period, Fannie Mae led the primary market in funding special affordable and low- and moderate-income home purchase mortgage loans but lagged the market in funding underserved areas home purchase loans. In 2003, Fannie Mae's increased performance placed it significantly above the special affordable market (a 17.1 percent share for Fannie Mae compared with a 15.9 percent share for the market) and the low-mod market (a 47.0 percent share for Fannie Mae compared with a 44.6 percent share for the market). However, Fannie Mae continued to lag the underserved areas market in 2003 (a 26.8 percent share for Fannie Mae compared with a 27.6 percent share for the market). These data are based on the “purchase year” approach, that is, Fannie Mae's performance is based on comparing its purchases of all home purchase loans (both seasoned loans and newly-originated mortgages) during a particular year with loans originated in the market in that year. When Fannie Mae's performance is measured on an “origination year” basis (that is, allocating Fannie Mae's purchases in a particular year to the year that the purchased loan was originated), Fannie Mae also led the 2003 market in funding special affordable and low- and moderate-income loans, and lagged the market in funding underserved area loans.
• Appendix A compares the GSEs' funding of first-time homebuyers with that of primary lenders in the conventional conforming market. Both Fannie Mae and Freddie Mac lag the market in funding first-time homebuyers, and by a rather wide margin. Between 1999 and 2002, first-time homebuyers accounted for 27 percent of each GSE's purchases of home purchase mortgages, compared with 38 percent for home purchase mortgages originated in the conventional conforming market. For minority first-time homebuyers, the GSE ratio was 6.2 percent, compared to a market originations ratio of 10.6 percent. For African-American and Hispanic first-time homebuyers, the GSE ratio was 3.8 percent, compared to a market originations ratio of 6.9 percent. For first-time homebuyers, particularly first-time minority homebuyers, both GSEs substantially lag the private conventional conforming market.
• The GSEs account for a small share of the market for important groups such as minority first-time homebuyers. Considering all mortgage originations (both government and conventional) between 1999 and 2001, it is estimated that the GSEs purchased only 14 percent of all loans originated for African-American and Hispanic first-time homebuyers, or one-third of their share (42 percent) of all home purchase loans originated during that period. Considering conventional conforming originations during the same time period, it is estimated that the GSEs purchased only 31 percent of loans for African-American and Hispanic first-time homebuyers, or about one-half of their share (57 percent) of all home purchase loans in that market. A large percentage of the lower-income loans purchased by the GSEs had relatively low loan-to-value ratios and consequently high downpayments, which may explain the GSEs' limited role in the first-time homebuyer market.

Appendix A to this rule provides evidence that there is a significant population of potential homebuyers who are likely to respond well to increased homeownership opportunities produced by increased GSE purchases in this area. Immigrants and minorities, in particular, are expected to be a major source of future homebuyers.

d. Size of the Mortgage Market That Qualifies for the Housing Goals

The Department has estimated the size of the conventional, conforming market for loans that would qualify under each Housing Goal category based on 2000 Census data and geography. These estimates, which are changed slightly from estimates reported in the proposed rule, are as follows:

• 51-56 percent for the Low- and Moderate-Income Housing Goal
• 23-27 percent for the Special Affordable Housing Goal
• 35-39 percent for the Underserved Areas Housing Goal

These market estimates exclude the B&C (i.e., subprime loans that are not A-minus grade) portion of the subprime market. The estimates, expressed as ranges, allow for economic and market affordability conditions that are more adverse than recent conditions. The market estimates are based on several mortgage market databases such as HMDA and American Housing Survey data. The Department's estimates of the size of the conventional mortgage market for each Housing Goal are discussed in detail in Appendix D to this rule.

The GSEs have room for growth in serving the affordable housing mortgage market. The Department estimates that the two GSEs' mortgage purchases accounted for 55 percent of the total (single-family and multifamily) conventional, conforming mortgage market between 1999 and 2002. In contrast, GSE purchases comprised 48 percent of the low- and moderate-income market, 48 percent of the underserved areas market, and a still smaller 41 percent of the special affordable market. Thus, the remaining 52-59 percent of the Goals-qualifying markets have not yet been touched by the GSEs.

The GSEs' presence in mortgage markets for rental properties, where much of the nation's affordable housing is concentrated, is below that in the single-family-owner market. The GSEs' share of the total rental market (including both single-family and multifamily) was also less than 40 percent between 1999 and 2002. Obviously, there is room for the GSEs to increase their presence in the single-family rental and multifamily rental markets.

Table 1 summarizes the Department's findings regarding GSE performance relative to market projections for 2005-2008 and the Housing Goal levels for 2005-2008.

The analysis for 2005 and later reflected in Table 1 is based on 2000 Census data on area median incomes and minority concentrations, using the metropolitan area boundaries specified by OMB in June 2003. This affects the market percentages for all three Housing Goals, as well as the figures on area median incomes and minority percentage figures that will be used to measure GSE performance on the Housing Goals beginning in 2005. The greatest effect of the updated data is on the Underserved Areas Housing Goal. Expressing this goal in terms of 2000 Census data adds approximately 5 percentage points to the Housing Goal and market levels, compared with analysis using 1990 Census data with Metropolitan Statistical Areas (MSAs) as defined prior to 2000.

The GSEs' baseline performance figures in Table 1 exclude the effects of the bonus points for small multifamily and single-family two-to-four unit owner-occupied properties and the Temporary Adjustment Factor (TAF) for Freddie Mac that were applied in official scoring toward the Housing Goals in 2001-2003. The Department did not extend these adjustments beyond 2003.

Table 1 reveals several features of HUD's Housing Goals. First, it is evident from this table that the 2005 level (22 percent) for the Special Affordable Housing Goal is below the low end (23 percent) of HUD's projected market range for 2005-2008. The 2005 level (52 percent) of the Low- and Moderate-Income Housing Goal is slightly above the low-end (51 percent) of HUD's market estimate range.

Second, the 2005 Underserved Areas Housing Goal level (37 percent) is consistent with the market range (35-39 percent) now projected by HUD for the Housing Goals using 2000 Census data.

Third, the GSEs' performance on all of the Housing Goals was significantly below the market averages for 1999-2002. Appendix D to this rule provides market estimates for the years 1999-2002 under different assumptions about the multifamily mix (i.e., newly-mortgaged multifamily units as a share of all financed dwelling units). The estimates differ between the two home purchase years (1999 and 2000) and the heavy refinance years (2001 and 2002). For the low-mod goal, the estimates average approximately 56 percent for the two home purchase years and 52 percent for the two heavy refinance years, with an overall 1999-2002 low-mod average of 54 percent (five percentage points above Fannie Mae's performance and seven percentage points above Freddie Mac's performance). The market estimates for the underserved areas goal average slightly over 37 percent (38 percent during the two home purchase years and 36 percent during the two heavy refinance years), or approximately 2-4 percentage points above the GSEs' performance (see Table 1). The higher Housing Goals are intended to move the GSEs closer to or within the market range for 2005, and to the upper end of the market range projection by 2008.

An analysis of the GSEs' mortgage purchases by property type shows that they have had much less presence in the “goals-rich” rental segments of the market, as compared with the “less-goals-rich” owner segment of the market. As shown in Figure 1, GSE mortgage purchases represented 37 percent of single-family and multifamily rental units financed between 1999 and 2002. This figure is much lower than their 61 percent market share for single-family owner-occupied properties. (Figure 2 provides unit-level detail comparing the GSEs' purchases with originations in the conventional conforming market.)

Typically, about 90 percent of rental units in single-family rental and multifamily properties qualify for the Low- and Moderate-Income Housing Goal, compared with about 44 percent of owner units. Corresponding figures for the Special Affordable Housing Goal are almost 60 percent of rental units and 16.4 percent of owner units. Thus, one reason that the GSEs' performance under the Low- and Moderate-Income Housing and Special Affordable Housing Goals has fallen short of HUD's market estimates is that the GSEs have had a relatively small presence in the two rental market segments, notwithstanding that these market segments are important sources of affordable housing and important components in HUD's market estimates.

In the overall conventional conforming mortgage market, rental units in single-family properties and in multifamily properties represented approximately 25 percent of the overall mortgage market between 1999 and 2002, 42 percent of the units that collateralize mortgages qualifying for the Low- and Moderate-Income Housing Goal, and 56 percent of the units that collateralize mortgages qualifying for the Special Affordable Housing Goal. Yet between 1999 and 2002, units in such properties accounted for only 17 percent of the GSEs' overall purchases, 32 percent of the GSEs' purchases meeting the Low- and Moderate-Income Housing Goal, and 44 percent of the GSEs' purchases meeting the Special Affordable Housing Goal.^[8] Continuing weakness in GSE purchases of mortgages on single-family rental and multifamily properties has been a significant factor underlying the shortfall between GSE performance and that of the primary mortgage market.

e. Ability of the GSEs To Lead the Industry

An important factor in determining the overall Housing Goal level is the ability of the GSEs to lead the industry in making mortgage credit available for Housing Goals—qualifying populations and areas.

The legislative history of FHEFSSA reflects Congress's strong concern that the GSEs need to do more to benefit low- and moderate-income families and residents of underserved areas that lack access to credit. (See, e.g., S. Rep. No. 102-282, at 34.) The Senate Report on FHEFSSA emphasized that the GSEs should “lead the mortgage finance industry in making mortgage credit available for low- and moderate-income families.” (Id.)

Thus, FHEFSSA specifically requires that HUD consider the ability of the GSEs to lead the industry in establishing the level of the Housing Goals. FHEFSSA also clarified the GSEs' responsibility to complement the requirements of the CRA (see section 1335(a)(3)(B) of FHEFSSA, 12 U.S.C. 4565(a)(3)(B)), and fair lending laws (see section 1325 of FHEFSSA, 12 U.S.C. 4545) in order to expand access to capital to those historically underserved by the housing finance market.

While leadership may be exhibited through the GSEs' introduction of innovative products, technology, and processes, and through their establishment of partnerships and alliances with local communities and community groups, leadership must always involve increasing the availability of financing for homeownership and affordable rental housing. Thus, the GSEs' obligation to “lead the industry” entails leadership in facilitating access to affordable credit in the primary market for borrowers at different income levels, and with different housing needs, as well as in underserved urban and rural areas.

Because the GSEs' market presence varies significantly by property type, the Department examined whether the GSEs have led the industry in three different market sectors served by the GSEs: single-family-owner, single-family rental (those with at least one rental unit and no more than four units in total), and multifamily rental.

The GSEs' purchases between 1999 and 2002 financed almost 61 percent of the approximately 36 million owner-occupied units financed in the conventional conforming market during that period. The GSEs' state-of-the-art technology, staff resources, share of the total conventional conforming market, and financial strength strongly suggest that they have the ability to lead the industry in making home purchase credit available for low-income families and underserved neighborhoods. From the analysis in Appendices A-D to this rule, it is clear that the GSEs are able to improve their performance and lead the primary market in financing Housing Goals—qualifying home purchase mortgages. In fact, Fannie Mae's improved performance in 2003 is evidence of this potential, as it led the market in funding home purchase loans for special affordable and low- and moderate-income families.

As discussed in Appendix A to this rule, there are a wide variety of quantitative and qualitative indicators that demonstrate that the GSEs have ample, indeed robust, financial strength to improve their affordable lending performance. For example, the combined net income of the GSEs has risen steadily over the last 15 years, from $888 million in 1988 to $12.7 billion in 2003. This financial strength provides the GSEs with the resources to lead the industry in making mortgage financing available for families and neighborhoods targeted by the Housing Goals.^[9]

As noted above, the GSEs have been much less active in providing financing for the rental housing market. Between 1999 and 2002, the GSEs financed 4.5 million rental dwelling units, which represented 37 percent of the 12 million single-family and multifamily rental dwelling units that were financed in the conventional market during this period. Thus, the GSEs' share of the rental mortgage market was just three-fifths of their share of the market for mortgages on single-family owner-occupied properties.

Clearly there is room for the GSEs to increase their presence in the single-family rental and multifamily rental markets. As explained above, these markets are an important source of low- and moderate-income housing since these units qualify for the Housing Goals in a greater proportion than do single-family owner-occupied properties. Thus, Fannie Mae and Freddie Mac can improve their performance on each of the three Housing Goals if they increase their purchases of mortgages on rental properties.

As discussed below in Section II.C.4 of this preamble with respect to the Home Purchase Subgoals, both GSEs should be able to lead the market for single-family owner-occupied properties in all three housing goal categories—special affordable, low- and moderate-income, and underserved areas. The GSEs are already dominant players in this market, which, unlike the rental markets, is their main business activity. However, as already discussed, research studies conducted by HUD and academic researchers conclude that except for Fannie Mae's recent performance on the special affordable and low- and moderate-income categories, the GSEs have not led the primary market in financing owner-occupied housing for low-income families, for first-time homebuyers, or for properties located in underserved areas.

As discussed above, the Housing Goals established by this rule are quantitative measures of how well the GSEs are serving low- and moderate-income homebuyers. HUD received comments on this factor from Freddie Mac and one other commenter. The commenter stated that, in addition to measuring leadership through the purchase of goal-qualifying mortgages, Fannie Mae and Freddie Mac's leadership should be measured in more qualitative ways such as their “development of products and technologies that the private sector may not be willing or able to do as well.” This commenter asserted that through the qualitative leadership of the GSEs, homeownership opportunities are expanded and costs lowered for all potential purchasers, including those in more affordable markets.

With respect to the issue of leadership, Freddie Mac contended in its comments on the proposed rule that HUD misinterpreted the “leading the industry” statutory factor and asserted that “[t]here is no intimation in the Act or its legislative history that Congress intended industry leadership to be determined based on the enterprises purchases of goal-qualifying mortgages.” Moreover, Freddie Mac commented that the GSEs are statutorily mandated to “facilitate the financing of affordable housing for low- and moderate-income families in a manner consistent with their overall public purposes.” Freddie Mac stated that the overall public purpose of the GSEs is to facilitate the operation of, and provide ongoing assistance to, the secondary market for residential mortgages. To the extent that the proposed goals inhibit or endanger Freddie Mac's ability to accomplish its general purpose of bringing liquidity and stability to the residential mortgage market, Freddie Mac contended that its ability to “lead the market” is in jeopardy. While the Department recognizes the degree of qualitative leadership provided by the GSEs, the Department also believes that their expertise and substantial financial resources allow them to lead quantitatively as well.

f. Need To Maintain the Sound Financial Condition of the GSEs

Based on HUD's economic analysis prepared for this final rule (Economic Analysis) and review by OFHEO, the Department has concluded that the Housing Goals in this final rule will not adversely affect the sound financial condition of the GSEs. Further discussion of this issue is found in the Economic Analysis.

3. Determinations Regarding the Levels of the Housing Goals

There are several reasons why the Department, having considered all the statutory factors as well as the comments on the May 3, 2004, proposed rule, is increasing the levels of the Housing Goals. The following sections describe these reasons and discuss and respond to comments received by HUD regarding the levels of the housing goals.

a. HUD's Market Analysis

Summary of Comments and HUD's Determination. As part of the process of establishing goals, HUD estimates the size of the conventional conforming mortgage market. In this process, HUD separately analyzes the markets for several different categories of mortgage loans: single-family owner-occupied housing units, rental units in two-to-four unit properties where the owner occupies one unit, rental units in one-to-four-unit investor-owned properties, and rental units in multifamily (five or more units) properties. This categorization is necessary because the data sources differ for the various categories, and it is also desirable because goals-qualifying shares of units vary markedly by category. HUD described its methodology for analyzing each category in Appendix D to the proposed rule, and the GSEs commented on that analysis. Other commenters expressed concern about the magnitude of the goals, but did not discuss the analysis on which the goals calculations were based.

(i) Multifamily Share of the Mortgage Market

An important component of HUD's calculation process is estimating the number of multifamily units financed each year as a percentage share of the total number of dwelling units financed (often referred to as the “multifamily mix”); this is important because of the high proportions of multifamily units, which qualify for credit under all three goals. Section C of Appendix D to this Final Rule provides a detailed discussion of estimates of the size of the multifamily mortgage market, including estimates by HUD, the GSEs, and other researchers. As explained in Appendix D, comprehensive data on the annual volume of multifamily mortgage originations are much less available than similar data on single-family mortgage originations. This introduces a degree of uncertainty into the market sizing analysis and highlights the need for sensitivity analyses to show the effects of different multifamily mixes on the size of the goals-qualifying markets. As explained below, HUD's market analysis focused on multifamily mixes between 13.5 percent and 16.0 percent, with a baseline of 15 percent. This range and baseline is consistent with HUD's historical estimates of the multifamily mix reported in Table D.5b of Appendix D. For example, between 1995 and 2002, HUD estimated that the multifamily mix was in the 14-16 percent range.

In its comments, Fannie Mae estimated a multifamily mix of 12.3 percent, stating that HUD's range is too high for current conditions in the multifamily market. Fannie Mae cited the current high vacancy rates for multifamily properties and the fact that the population aged 20 to 34 will not begin to increase until after 2007; this age group tends to be predominantly renters. Fannie Mae also projected a low multifamily refinance volume, because of a recent peak in multifamily originations; these recent originations will not be able to refinance easily under their current contracts until 2008 or later.

At Freddie Mac's request, ICF Consulting also calculated the multifamily mix. In its best estimate, ICF projected an average of 14.2 percent over the 2005-2008 period, ranging between 13.7 percent and 14.7 percent in individual years, while recognizing that the actual outcomes may be higher or lower. ICF projected multifamily refinancings based on the number of units financed eight, nine, and ten years ago, because 10-year balloon mortgages are the most common multifamily mortgages, and prepayment possibilities are limited by yield maintenance agreements in their current mortgage contracts.

In Appendix D to this rule, HUD reviews the evidence provided by the GSEs in their comments. HUD notes that the 2001 Residential Finance Survey (RFS) has recently been published by the Census Bureau, and that the RFS provides higher estimates of the multifamily mix for 1999-2001 (the most recent years available) than either Fannie Mae or ICF. The RFS data and other data analyzed in Appendix D to this rule suggest that 15.0 percent is a reasonable baseline, particularly in a home purchase mortgage market environment, with a relatively small volume of refinanced mortgage originations. HUD also notes that the ICF average of 14.2 percent is fairly close to HUD's estimate of 15.0 percent. HUD therefore continues to use 15.0 percent as the best estimate of the projected share of multifamily mortgages over the 2005-2008 period. HUD reports the goals-qualifying shares of mortgage originations on the basis of this estimate in Appendix D to this rule. HUD also publishes sensitivity analyses using other estimates of the multifamily mix, including 12.3 percent (Fannie Mae estimate), 13.5 percent (low end of HUD's range), 14.2 percent (ICF's best estimate), and 16.0 percent (high end of HUD's range). Using this range of multifamily mix estimates, the estimate of the goals-qualifying share of mortgage originations varies by about 1.5 to 2.5 percentage points for the low-mod goal, by about 1.0 percentage point for the underserved areas goal, and by about 1.2 to 1.7 percentage points for the special affordable housing goal. The estimate varies depending on other market factors.

As also discussed in Appendix D to this final rule, the multifamily mix is even lower during heavy refinance environments, as single-family owner refinance loans dominate both the market and the GSEs' purchases. This makes it more difficult for the GSEs to meet specific Housing Goal targets. As discussed in section b below of this preamble, HUD is soliciting public comments on how to structure and implement a regulatory provision to take account of the effects of high volumes of refinance loans in some years on the GSEs' ability to achieve the Housing Goals.

(ii) Single-Family Rental Share of the Mortgage Market

HUD also estimated the distribution of mortgage originations for single-family properties, defined as structures with one-to-four units. In Appendix D to this rule, HUD disaggregates single-family mortgage originations into three categories: those on owner-occupied single-family homes, those on structures with two to four units having one unit owner-occupied, and those on structures with one to four rental units owned by investors. HUD bases this categorization on the fact that the rental units in the latter two categories qualify at much higher rates for the housing goals.

HUD uses two data sources in Appendix D to estimate the size of the investor category, the Residential Finance Survey (RFS) and the Home Mortgage Disclosure Act database (HMDA). HMDA provides data only on the investor category. The investor share of HMDA single-family loans averaged 7.8 percent over 1993-2003, and 8.3 percent over the recent period of 1999-2003. The share of investor loans has also been rising for home purchase loans; it was 9.6 percent over 1993-2003 and 11.2 percent over 1999-2003. The RFS for 2001 reported a larger share of investor loans than HMDA, 13.4 percent compared to 7.8 percent. The RFS also reported larger investor shares for 1999 and 2000.

In the proposed rule, HUD estimated the investor share of the single-family market at 10 percent, based on HMDA data and the 2001 RFS, which was then the most recent available. HUD also considered alternatives of 8 percent and 12 percent. Both GSEs and ICF commented that HUD should use HMDA data rather than RFS data, and should use a lower investor share in setting the goals. While they agreed with HUD that the RFS provides the most accurate estimate of the true investor share of the market, they stated that lender reporting of investor loans to the GSEs is conceptually closer to HMDA data, which are based on lender reports. They commented that the actual opportunities available to the GSEs in the single-family investor loan market are best measured by data that lenders report, based on actual loan applications.

Fannie Mae stated that HUD's two highest alternatives exceed the highest investor share ever reported in HMDA. Fannie Mae cited research indicating a reporting bias in HMDA, due to “hidden investors.” At the time of loan origination, a property may be owner-occupied or intended for owner-occupancy, but may become rental shortly after origination. Fannie Mae stated that the same bias exists in its own reporting. The hidden investors cannot be identified at the time of origination.

Freddie Mac stated that investors have an incentive to claim falsely that they are owner-occupants because investor properties are subject to higher underwriting standards and loans tend to carry higher interest rates. Freddie Mac concluded that HUD should measure the opportunities that are actually available in the market to the GSEs, which are best measured by lender-reported HMDA data.

In this rule, HUD has adopted HMDA data as the basis for its calculation of the investor share of single-family mortgage originations. The GSEs make a valid argument that lender-reported data at the time of origination measures the investor loans that are available for them to purchase; HMDA provides that data. As discussed in Appendix D to this rule, HUD projects the investor share to be 8.5 to 9.0 percent (based on HMDA) during the 2005-2008 home purchase environments, rather than 10 percent. HUD also reports sensitivity analyses for higher and lower investor shares of 8.0 and 9.5 percent. Using this range of single-family investor share estimates, the estimate of the goals-qualifying share of mortgage originations varies by about 1.5 percentage points for the low-mod goal, and by 0.5 percent or less for the other two goals. The estimate varies depending on other market factors.

In the proposed rule, HUD estimated that the share of the single-family market consisting of two-to-four units properties with one unit owner-occupied was 2.0 percent of all single-family mortgages. This category is reported only in the RFS. The 2001 RFS reports that this category comprised 1.5 percent of all single-family mortgages. Because the RFS calculates a higher share of investor mortgages in the single-family market (13.4 percent) than HUD employs in this rule (8.5 to 9.0 percent), it is necessary to adjust the 2001 RFS figure upward.

The RFS reports that 85.1 percent of all single-family mortgages were for owner-occupied homes. The estimated share of two-to-four units properties with one unit owner-occupied in the single-family market is calculated at 1.73 percent (i.e., 1.5 percent/[1.5 percent + 85.1 percent]). This figure lies between Fannie Mae's share of about 2.0 percent over 1999-2003 and Freddie Mac's share of about 1.5 percent. In this final rule, HUD uses a share of 1.6 percent. Sensitivity analyses for 2.0 percent are reported in Appendix D to this rule.

Similarly, the single-family owner-occupied share is adjusted upward to take account of the lower share of investor loans, from 85.1 percent to 89.9 percent.

The estimated market share range for each of the three goals categories is as follows: 51-56 percent for the Low- and Moderate-Income Goal, 35-39 percent for the Underserved Areas Goal, and 23-27 percent for the Special Affordable Goal. These estimates are one percentage point below the market ranges reported in the Proposed Rule, for the reasons discussed above and detailed in Sections F-H of Appendix D. The top ends of the market ranges were reduced as follows: from 57 percent to 56 percent for the low- and moderate-income market; from 40 percent to 39 percent for the underserved areas market; and from 28 percent 27 percent for the special affordable market. Accordingly, the 2008 goals were also reduced by one percentage point from those included in the Proposed Rule. In the Final Rule, the Low- and Moderate-Income Goal increases from 52 percent in 2005 to 56 percent in 2008, as compared with an increase of 52 percent to 57 percent in the Proposed Rule. In the Final Rule, the Underserved Areas Goal increases from 37 percent in 2005 to 39 percent in 2008, as compared with an increase of 38 percent to 40 percent in the Proposed Rule. In the Final Rule, the Special Affordable Goal increases from 22 percent in 2005 to 27 percent in 2008, as compared with an increase of 22 percent to 28 percent in the Proposed Rule.

b. Attainability of the Goals in a High Refinance Environment

Summary of Comments. A common theme of many of the public comments was concern about the volatility of the mortgage market and how such volatility makes setting Housing Goals a delicate and risky proposition.

These commenters indicated that the goals proposed by HUD would be unattainable, particularly in a high refinance environment when a large portion of the mortgage market is comprised of refinance loans rather than home purchase mortgages.

Fannie Mae and others suggested that including single-family refinance mortgages in goals calculations creates tension between liquidity goals and affordable housing goals by taking the emphasis away from increasing purchase money mortgages (and therefore homeownership) and placing the focus instead on meeting high goals.

Freddie Mac, several trade associations, a financial organization and consumer advocacy groups also expressed concern that inclusion of single-family refinances jeopardizes the GSEs' abilities to increase homeownership through acquisitions of purchase money mortgages because the focus would be on attaining goals rather than providing affordable home purchases for the target population.

One trade association, however, asserted that removing refinance mortgages from the goals calculations would only serve to encourage the GSEs to buy refinance loans instead of home purchase loans. By buying refinance loans, the GSEs could effectively ignore housing goals and both “jeopardize the safety and soundness of the GSEs due to the higher default rate of refinance loans and increase the minority housing gap due to the lower rate of minority borrowers for refinance loans.”

Other commenters suggested that the final rule should include mechanisms for making adjustments to the goals if there are changes in market conditions including a surge or drop in refinance volume. These commenters asserted that the GSEs' ability to successfully meet the goals should not be contingent upon interest rate stability. One suggestion that was offered for dealing with market mix fluctuations (i.e., between home purchase and refinance loans) was to remove from both the numerator and denominator “any mortgage activity in excess of the percentage of home refinance loans used by HUD for estimating the size of this market (i.e., above 35%).”

Another commenter stated that “HUD should simply set goals that require the GSEs to lead the market, whatever the market turns out to be.” This commenter explained that “if 50% of home purchase loans are to low-moderate income borrowers in 2005, then HUD should expect that a slightly higher percentage than this, say 51%, of Fannie's and Freddie's home purchase loans should fit in the purchase category of loans to low-moderate income borrowers.”

HUD's Determination. This final rule retains the approach of the May 3, 2004, proposed rule, in which the level of each Housing Goal will increase year-by-year so that by 2008 each goal will match the top of the market range established in section 2.d, above.

The last three years have shown unprecedented volumes of refinance activity. For both GSEs, refinance loans accounted for 64 percent of all loans on single-family owner-occupied properties in 2001.^[10] The refinance shares increased to 70 percent for Fannie Mae and 73 percent for Freddie Mac in 2002, and rose even further last year, to 79 percent for Fannie Mae and 82 percent for Freddie Mac. These unexpected record refinance rates made it more challenging for the GSEs to attain the housing goals in the past few years, as discussed elsewhere in this Preamble. The goals in HUD's proposed rule for the latter part of the 2005-2008 period would be even more challenging if (contrary to current expectations) very high refinance rates are experienced in those years.

HUD received a number of public comments seeking a regulatory solution to the issue of the ability of the GSEs to meet the housing goals during a period when refinances of home mortgages constitute an unusually large share of the mortgage market. HUD is not addressing the refinance issue as a regulatory change in this final rule. Elsewhere in today's Federal Register, HUD is publishing an Advance Notice of Proposed Rulemaking that advises the public of HUD's intention to consider by separate rulemaking a provision that recognizes and takes into consideration the impact of high volumes of refinance transactions on the GSEs' ability to achieve the housing goals in certain years, and solicits proposals on how such a provision should be structured and implemented. HUD believes that it would benefit from further consideration and additional public input on this issue. HUD also notes that FHEFSSA provides a mechanism by which HUD can take into consideration market and economic conditions that may make the achievement of housing goals infeasible in a given year. (See 12 U.S.C. 4566(b).)

c. Bonus Points

The Housing Goals 2000 final rule provided for the award of bonus points (double credit) toward the Housing Goals for both GSEs' mortgage purchases that financed single-family, owner-occupied two-to-four unit properties and 5-50 unit multifamily properties. The rule also established a temporary adjustment factor (TAF) that awarded Freddie Mac 1.2 units credit for each multifamily unit in properties over 50 units for calendar years 2001 through 2003. (Congress increased the level of the TAF to 1.35 per unit under section 1002 of Public Law 106-554.)

The Housing Goals 2000 final rule made clear that both of these measures were temporary, intended to encourage the GSEs to increase their efforts to meet financing needs that had not been well served. During the three years for which the temporary bonus points and TAF were established, HUD expected the GSEs to develop new, sustainable business relationships and purchasing strategies for the targeted needs. Data indicate that, because both GSEs did increase their financing of units targeted by the bonus points and the TAF, the original objectives were met. The Department determined at the end of the three years (2001-2003) not to extend the bonus points or the TAF.

Summary of Comments. A number of non-GSE commenters, including organizations representing affordable housing and consumer groups, trade associations, organizations representing racial and ethnic minorities, other organizations, and both Fannie Mae and Freddie Mac, recommended that the Department reinstate the award of bonus points for the GSEs which were established for 2001-2003 but which the Department did not continue after the end of 2003. Various non-GSE commenters, in addition to recommending reinstatement, also suggested that HUD develop new bonus point incentives for other unmet housing needs, such as manufactured housing, rural housing, or tax credit properties or for particular groups, e.g., Native Americans, other minority populations, or persons with disabilities.

Fannie Mae recommended that HUD provide bonuses for targeted business such as extremely low-income households, i.e., those with incomes less than 30% of area median income (AMI); first-time homebuyers; manufactured housing; rural areas; and small multifamily properties. Freddie Mac suggested that instead of purchase money subgoals, the Department could provide bonus point incentives for these mortgages. Freddie Mac stated that the bonus points for two-to-four unit and 5-50 unit properties provided an extremely effective incentive. Freddie Mac indicated that other markets that could be assisted by bonus points are rural and manufactured housing. Freddie Mac noted that the Department's concern that bonus points obfuscate the GSEs' actual goals-qualifying performance is easily remedied by having the GSEs report two numbers, one with and one without the bonuses.

HUD's Determination. The Department has fully considered the comments suggesting the re-introduction of bonus points, as well as other types of targeted incentives for the GSEs' mortgage purchases, and has determined not to reinstate the bonus points for the years covered by this rule. The position of the Department discussed in the preamble of the proposed rule (see 69 FR 24228, 24232) remains unchanged; that is, the continued use of the bonus points “would only result in misleading information about the extent to which the GSEs are, in fact, meeting the Housing Goals.” In addition, the Department reiterates that the “decision to increase the levels of the Housing Goals substantially in a staged manner * * * and, at the same time, not renew the bonus points or TAF, will ensure that the GSEs continue to address the areas formerly targeted by these measures'' (see 69 FR 24228, 24232).

d. Appropriate Levels of the Goals

In the May 3, 2004, proposed rule, HUD set the Goals to increase to levels at or near the high end of the estimated market range for each goal category by 2008. A large number of commenters expressed concern that these goal levels were set too high, and could have deleterious consequences for the mortgage market as a whole, or for specific sectors of the market.

Fannie Mae commented that a high allocation of affordable mortgage credit will take away from the broad middle class, especially in high-housing cost regions. For example, Fannie Mae asserted that if the special affordable housing goal had been set at 28 percent in 2003, then it would have needed to greatly curtail support to the overall market to meet that goal. Fannie Mae concluded that such manipulation does not promote stability and limits liquidity, and that it can shut out working middle class borrowers, contribute to higher interest rates and lower conforming loan limits.

Many commenters, including Freddie Mac, also claimed that setting the goals at a high percentage may lead to denominator management. They state that denominator management would occur if a GSE purposely abstained from buying mortgages in the markets that are not goals-eligible, rather than increasing its purchases in markets that are goals-eligible. Freddie Mac contended that this may be necessary if goals are set above the market percentage of available goal-qualifying loans. One financial institution observed that denominator management “will be exacerbated by the fact that the GSEs do not operate in the primary market and do not have any direct control over the origination strategies of their customers.”

In addition to the allocation problems discussed above, the GSEs stated that the liquidity requirements in their charters imply that they must stand ready to buy any and all conventional, conforming residential mortgages. They contend that denominator management is in direct conflict with these provisions, and goals set higher than market originations could force the GSEs to refuse to purchase mortgages that are not eligible. This, in turn, could reduce liquidity in the market. Knowing that the GSEs would no longer stand willing and able to purchase all conventional, conforming mortgages, other market participants might be less willing to hold these mortgages in their portfolios, and general liquidity would decline. The GSEs further asserted that changing market forces could cause swings in prices and trading volumes, and these temporary disturbances could create unstable markets, increase risk, and reduce the willingness of investors to invest in the sector. Thus, the GSEs maintained that denominator management decreases market stability.

The GSEs pointed to specific historical examples that describe their positive influence on stability. They maintained that during the 1990-1991 recession, the GSEs advised that they stood ready to purchase mortgages while many industry participants curtailed their purchase programs. Using historical trends in prices, the GSEs asserted that their presence in the mortgage market explains why mortgage-backed securities have a more stable price trend than commodity markets. They warned that because of denominator management resulting from unrealistic goals, they could not buy mortgage-backed securities and encourage stability in a financial crisis.

The GSEs further contended that if they reduce their willingness to buy non-goals eligible mortgages, it will be harder for borrowers whose incomes marginally exceed goals eligibility requirements to obtain financing since the two income-based Housing Goals compare the incomes of the borrower or resident to area median income. For example, the combined incomes in a working family may just disqualify that family's loan for eligibility under the low- and moderate-income goal even though each individual's income would not be considered to be affluent. The GSEs and other commenters provided examples of working families in the middle class, such as “teacher/fireman” households, that could encounter difficulties in financing a home.

Moreover, the commenters asserted that non-goal qualifying households may have higher costs associated with available financing since these mortgages would be less likely to be purchased by a GSE. Freddie Mac asserted that HUD did not take this into account in its cost/benefit analysis.

Furthermore, commenters claimed that denominator management may contribute to higher interest rates and, as a result, harm the precise borrowers that HUD is trying to help. These commenters stated that if denominator management reduces liquidity then the supply of mortgage funds will decline and interest rates will rise. The GSEs contended that if they are less willing to buy mortgages under all conditions, then investors will be less willing to provide funds to the market. As a result, the GSEs claimed that as investors seek out safer instruments, home mortgage interest rates will rise, and this rise in home mortgage rates will harm even those borrowers that are still goals-eligible.

Several commenters expressed concern about the effect of the goals on high cost markets. One commenter explained that while the goals are set with a national standard, a market level analysis “reveals a pronounced shortage of affordable mortgages in high cost housing markets.” Commenters stated that the GSEs' current loan purchasing patterns demonstrate that market affordability already has an impact on goals-related purchases. The commenters expressed concern that high cost markets could see even tighter credit if the proposed goals are enacted.

The GSEs note that under HUD's May 3, 2004, proposed rule, the goal levels rise to levels at the top of HUD's market range in 2008 and stabilize there. They state that the projected market range concept is one in which HUD projects market levels of loans generated within each goal category will fluctuate within the range, depending on relative volumes of single-family refinance loans relative to other loans, interest rates and other macroeconomic and housing market conditions. The GSEs express the concern that, in any particular year, they could be confronted with goal levels that are several percentage points higher than the market percentages of goal-qualifying loans, or goal levels that are at the market percentages. The GSEs state that if HUD's proposed Housing Goals are retained, they foresee years when the goal levels will be attainable only by means of “denominator management” in which they limit their purchases of loans that do not qualify for the goals.

HUD's Determination. Many of the comments expressed concern about the goal levels established for the last year or two of the period covered by this rule. In these years, the goals are set at the market levels estimated by HUD. Also, since they are the later years, market projections are necessarily more imprecise. In particular, the possibility of a decline in mortgage interest rates in those years raises the possibility of another boom in refinancing, and thus greater difficulty for the GSEs to meet the housing goals without denominator management. The comments relating to middle-income borrowers are predicated on the difficulty of foreseeing refinance volatility. Recent years have seen large unexpected home refinance rates. Since higher income homeowners disproportionately engage in refinancing, inclusion of refinance loans in the denominator increases the difficulty of GSE goals performance. A middle-income borrower just above the low/mod bracket would be less attractive to the GSEs in high refinance years. As noted in section II.C.3.b., HUD is considering in a separate rulemaking a provision that recognizes and takes into consideration the impact of high volumes of refinance transactions on the GSEs' ability to achieve the housing goals in certain years. HUD also notes that FHEFSSA provides a mechanism by which HUD can take into consideration market and economic conditions that may make the achievement of housing goals infeasible in a given year. (See 12 U.S.C. 4566(b).)

With regard to the effects of the goals on high-cost markets, HUD notes that the overall presence of the GSEs in these markets depends on the conforming loan limit, which has been established by Congress for all states, including states deemed to be “high-cost areas.” With regard to HUD's housing goals more specifically, the low- and moderate-income and special affordable goals are based on borrower income relative to area median income, thus a mortgage for a lower-income family in a high-income metropolitan area will count towards the goals in the same manner as a mortgage for a lower-income family in a low-income area. Underserved areas are defined in terms of median family income in a census tract relative for median income in the area; thus a mortgage for a family living in a lower-income tract in a high-income metropolitan area will count towards the goals in the same manner as a mortgage for a family living in a lower-income tract in a low-income area. Thus HUD concludes that its housing goals will have no adverse impact on borrowers or neighborhoods in areas with high housing costs.

e. Consequences of the Goals for FHA

Fannie Mae, Freddie Mac, several trade associations, two advocacy groups and two financial institutions expressed concern over the impact HUD's proposed goals would have on the future solvency of the FHA program. One trade association asserted that “excessive goals will push GSEs to expand into the least-risky part of the FHA market and put into question FHA's long-term viability.”

The aforementioned commenters reiterated this point by stating that unrealistically high goals would compel the GSEs to increase competition with FHA for higher credit quality borrowers and would therefore further undermine the FHA program in the long-run. One advocacy group asserted that not only will these goals encourage the GSEs to compete with FHA more in the single family sector but in the multifamily sector as well.

Freddie Mac and Fannie Mae agreed that they would be compelled to more aggressively compete with FHA in procuring top-quality borrowers. Freddie Mac stated that the GSEs would take as many as “1/3 of all FHA borrowers.” Freddie Mac and two trade associations further contended that such a loss to the FHA program would be seen in the increasing expenses to the remaining FHA borrowers. As the FHA program loses better quality loans to the GSEs, the result would be “higher fees to FHA borrowers or government subsidies to pay claims, effectively making FHA the lender of last resort,” said one trade association.

One financial institution stated that the so-called competition for goals-qualifying loans would not be between traditional conventional lenders vying for loans with a separate group of traditional FHA lenders, but rather an acceleration of product competition within a single group of existing lenders who originate for both markets. This commenter stated that 12 of the top 15 (by volume) FHA/VA lenders are also among the top 15 conventional lenders and indicated that the increased product competition would not result in a net increase in goals-qualifying loans, but in a shift from FHA to the GSEs of FHA's relatively lower risks.

HUD's Determination. The Department agrees with many of these commenters that improvements in technology, such as the widespread use of commercial credit scores, mortgage scores, and automated underwriting systems, have fundamentally changed the way lenders process loan applications in recent years. Where once rules-based underwriting distinctions between prime conventional and FHA loans were fairly clear, in recent years, with the new technology, these distinctions have become blurred. For example, loan applications with payment-to-income ratios above conventional market guidelines were once clearly candidates for FHA financing because FHA would accept applicants with higher payment-to-income ratios. However, today, the same application would be processed using an automated underwriting system (AUS) that scores the application based on the totality of the application's risk factors. What once may have been an unacceptable payment-to-income ratio for a prime conventional loan may now be acceptable if the application contains offsetting low risks in other key areas such as borrower cash reserves, loan-to-value ratio, or commercial credit scores.

In addition to these technological changes, FHA made several changes to its underwriting guidelines in FY 1995 in order to promote increased homeownership opportunities among low-income and minority homebuyers. By doing so, FHA modestly increased the risk characteristics of its post-1995 books of business, but it succeeded in raising FHA's proportion of first-time homebuyers from 60.9 percent in fiscal year 1994 to 73.0 percent in fiscal year 2003. During the same period (fiscal years 1994 to 2003), FHA's proportion of minority borrowers increased from 24.8 percent to 33.0 percent, and has since remained at this level, or higher.

The new technology may allow the conventional market to identify lower risk loan applications that historically have come to FHA. However, the ability to identify risks does not, in and of itself, equate to shifts in market share from FHA to conventional lenders. Better pricing for borrowers by the conventional market is required to lure lower risk borrowers from FHA. If conventional lenders use the new technology to not only evaluate risks but also to price according to risk, then there may be some shift from FHA to the conventional market. Such a shift can produce tangible benefits for borrowers in the form of lower cost mortgage financing.

The Department does not believe it is FHA's mission to compete with the private sector. Rather, FHA's mission is to complement the conventional market, using FHA's cost of capital advantage where it can have the most benefit in creating homeownership opportunities for those households who might not otherwise be served by the prime conventional market.

HUD gauges the soundness of FHA's insurance funds in several ways. The statutorily mandated annual independent actuarial review of FHA's principal single-family insurance fund, the Mutual Mortgage Insurance Fund (MMIF), provides the Department, and the public, with an outside expert's estimate of the capital ratio of the overall fund, and the economic value of new business coming into the fund. The capital ratio indicates whether the existing books of business (current portfolio) are financially sound, while the economic value estimates of new business show whether if the marginal impact of new loans insured is adding or detracting from the financial health of the fund.^[11] Specifically, the Fiscal Year 2003 actuarial review estimated the economic value of the MMIF at the end of Fiscal Year 2003 to be $22.7 billion and the fund's capital ratio to be 5.21 percent—the eighth full year this ratio has exceeded the Congressionally mandated minimum of 2.0 percent. The economic value of new loans endorsed for insurance during 2003 was estimated by FHA's independent actuary to be $2.8 billion, indicating new business coming into FHA is further contributing to FHA's reserves.

In comparison, the Fiscal Year 2002 actuarial review estimated the economic value and capital ratio of the MMIF at $22.6 billion and 4.52 percent, respectively. The increases in both measures for Fiscal Year 2003 were driven by the large positive economic value the actuary placed on a record dollar volume of new loans FHA insured in FY 2003 along with the rapid prepayment of older loans, keeping the end-of-year insurance-in-force (denominator of the capital ratio) down.

With regard to the GSEs taking multifamily business away from FHA, the Department notes that there are many differences between the types of multifamily mortgages FHA insures and those the GSEs purchase. For newly constructed multifamily properties, FHA insures the loan from the start of construction while GSE multifamily loan products generally do not. The GSEs do have forward commitment programs that can be used for new construction, but the purchase of the permanent loan by the GSEs generally requires the property to achieve minimum sustained occupancy levels, whereas FHA does not have this requirement. However, it is possible that the new goals will provide incentives for the GSEs to expand and refine their forward commitment products to be more attractive in the market for new multifamily housing. This could be a benefit to the market.

The greatest potential impact of the higher housing goals on FHA's multifamily business may come from a reduction in two of FHA's programs that address the purchase or refinance of existing properties. The first is the Section 223(f) program, which insures mortgages for the purchase or refinance of existing (over three year old) properties that are not currently financed with an FHA mortgage. This program accounted for about $0.8 billion in endorsements for FHA during Fiscal Year 2003, and is expected to produce about $0.5 billion in endorsements during Fiscal Year 2004. FHA's 223(f) business is estimated to be profitable to FHA—it is estimated to have a credit subsidy (net present value of all cash flows from the insurance contract at the time of endorsement) of negative 3.0%.^[12] The second is the Section 223(a)(7) program, which insures mortgages for FHA-to-FHA refinances—that is, the refinance of an existing FHA-insured mortgage. Section 223(a)(7) is used, for example, to refinance loans previously insured under FHA's most used programs—i.e., Section 221(d)(4) new construction/substantial rehabilitation, and Section 223(f). FHA endorsed over $2.1 billion in Section 223(a)(7) loans during Fiscal Year 2003, and is expected to endorse about $1.4 billion during Fiscal Year 2004. As with the Section 223(f) program, FHA's Section 223(a)(7) program is also profitable to FHA—operating with an estimated negative credit subsidy of 2.2%.

If FHA does lose some multifamily market share from its purchase or refinance programs for existing housing as a result of the goals, it would not likely have any significant impact on FHA overall.

f. Consequences of the Goals for the Multifamily Market

Summary of Comments. Several organizations commented on potential adverse consequences if the housing goals are set too high. Fannie Mae and Freddie Mac, among others, cited the recent high vacancy rates for multifamily rental housing as an example that increased lending by the GSEs at this time would encourage overbuilding.

Others stated that the multifamily market is already flush with capital and that inappropriate goals could promote overly aggressive bidding for loans and reckless lending.

One trade association stated that the increased presence of the two GSEs would promote a duopsony (a market with only two buyers) that would hinder competition in the multifamily mortgage market.

Other commenters suggested that increased loan purchases by the GSEs would skim the highest credit-quality loan from other mortgage lenders, and reduce the credit quality of multifamily loans remaining in the portfolios of pension funds or insured through FHA.

Another commenter stated that increased goals pressure on the GSEs would cause them to concentrate on large properties, where a single loan would contribute more toward goal attainment.

HUD's Determination. One of HUD's objectives in promulgating this final rule is to promote the availability of mortgage credit to affordable properties at the lowest possible cost. It is not the intent of this rule to promote the maximum flow of credit to this market, regardless of housing and mortgage market conditions.

Increased competition for business, as intended by the rule, should bring benefits to borrowers, and therefore renters, through lower interest rates and more attractive non-price terms. Such increased competition does not imply impaired credit quality or lax underwriting. As the GSEs compete more aggressively for multifamily business and gain market share, the market will not necessarily grow one-for-one with every additional loan purchased by the GSEs. It is likely that the market impacts will be more on the pricing of multifamily credit and less on the volume of credit supplied. Lower pricing of credit in and of itself does not promote overbuilding; its one unambiguous effect is to reduce the cost of supplying housing to consumers.

Demand for multifamily mortgages will be responsive to cyclical macroeconomic factors. Beyond these influences, demand for multifamily housing will be supported by favorable demographics. In its comments on the proposed rule, Fannie Mae highlighted the prospective growth in the number of people ages 20 through 34 in arguing that the demographics do not become clearly favorable to rental demand until late in this decade. But fewer than half of all renter households are headed by someone of this age, and more comprehensive estimates and projections suggest a steadier path of moderate growth in the demographic component of demand for multifamily housing.

Interest rates clearly will be important for the future path of mortgage lending, as noted by Fannie Mae, Freddie Mac, and other commenters. The historically low interest rates of recent years have spurred lending in both the multifamily and single-family markets. If interest rates should rise in the future, the volume of mortgage lending presumably would be lower than if rates were to remain at current levels. But the effect of higher rates on the GSEs' ability to achieve the housing goals is less clear. Because the goals are established in terms of shares of the GSEs' business, rather than levels, a key question is how higher interest rates would affect the relative demand for single-family and multifamily mortgage credit. Because of differences in prepayment provisions and other characteristics between single-family and multifamily mortgage lending, multifamily credit demand might drop off proportionally less than would single-family credit demand in response to higher rates.^[13] This in turn would make it easier to attain the goal levels if interest rates were to increase from current levels.

Regarding the market structure implications of increased GSE multifamily activity, HUD estimates that the GSEs purchased slightly less than one-third of the dollar volume of conventional multifamily loan originations during 2001-2003 (see Table D.2). There is room for increasing this market share without producing the duopsony alluded to in the previously cited comments. Furthermore, if the GSEs do increase their market penetration, it is because they are offering multifamily borrowers more attractive products or pricing than are their competitors, including the pension funds and FHA programs alluded to by some commenting organizations. The borrower and, ultimately, the rent-paying affordable housing resident benefit from these more attractive products and pricing.

In summary, the Department's determination is that the Housing Goal levels established by this rule are prudent and will improve the availability and pricing of credit for affordable multifamily properties. For the reasons stated above, it is the Department's view that the rule will not have the adverse consequences mentioned in some comments on the proposed rule.

g. Consequences of the Goals for the Single-Family Rental Market

Summary of Comments. Several community organizations raised concerns about encouraging the single-family rental market. They asserted that the goals should target families who want to live in the financed houses, as opposed to the investors who purchase these homes. In these commenters' view, investors take affordable housing stock off the market, which raises the price for low and moderate-income first-time homebuyers. They claimed that homeownership should be stressed because home equity is a large component of the disparity that exists in household net wealth between ethnic groups.

Some commenters cited studies that suggest homeownership has beneficial neighborhood effects relative to investor-owned properties. According to one cited study, absentee landlords are much more likely to let housing stock decline but homeowners are much more likely to invest in the upkeep of their homes. In the view of one of these organizations, the incentives that the GSEs receive for rental housing should be to promote multifamily developments, not single-family homes.

HUD's Determination. HUD considered many factors related to the single-family rental market. Single-family rentals are another source of affordable housing. Also, the capital provided by investors can help maintain demand for single-family homes in underserved neighborhoods. While some commenters complained that this raises the cost to first-time homebuyers, investors also help to maintain the liquidity and value of owner-occupied homes. Further, there are some investors who make it their business to renovate the housing stock and resell the properties. On balance, HUD found no compelling evidence that single-family rentals should be excluded from goals eligibility.

h. Consequences of the Goals for the Subprime Market

Summary of Comments. Both GSEs indicated that they would need to increase their purchase of subprime loans to meet the higher goals. Freddie Mac stated that the increased affordable housing goals created tension in its business practices between meeting the goals and conducting responsible lending practices.

In the past, Fannie Mae and Freddie Mac have voluntarily decided not to purchase subprime loans with features such as single-premium life insurance and prepayment penalty terms that exceed three years, or to purchase loans subject to the Home Ownership and Equity Protection Act (HOEPA). Freddie Mac indicated that the increased goals would limit its ability to influence subprime lending practices. More specifically, Freddie Mac claimed that, to meet the higher housing goals, it might not have the option in the future of turning away subprime loans that have less desirable loan terms than the subprime business it currently purchases.

Several commenters suggested that if the GSEs are pushed to serve more of the subprime market, they will skim a significant portion of the lower-risk borrowers from that market. The resulting smaller subprime market would include the neediest borrowers. The commenters stated that these higher risk borrowers would pay more because lower risk borrowers would not be present to subsidize them, and the market's high fixed costs would be distributed across fewer borrowers.

One industry group also suggested that a significantly smaller subprime market for private lenders would drive some lenders out of business and translate into less competition.

While some industry commenters welcomed the entrance of the GSEs into the subprime market because their presence would bring stability and standardize business practices, the commenters also expressed concern that unrealistically high goals could force the GSEs to jump into the market in a manner that negatively distorts underwriting and pricing. These commenters contended that the GSEs could bring capital and standards up, but that they must gradually and carefully enter the subprime market to have a positive effect. They strongly urged HUD to lower the goals to encourage the GSEs to expand their subprime activities at a measured pace.

Some commenters suggested that bonus points, or other incentives for the GSEs' purchases of certain nonprime loans, could foster more deliberate and prudent purchases by the GSEs of subprime loans. One lender also suggested that incentives could be granted to the GSEs for other underserved market segments, such as manufactured homes, minority first time buyers, and nonprime first-time buyers.

HUD's Determination. To date, the GSEs' involvement in the subprime market has benefited two types of borrowers: “A” risk and “near A” risk. The first group consists of borrowers with risk profiles similar to “A” borrowers, but receive mortgages from a subprime lender. The GSEs” outreach and education efforts increase the likelihood that “A” borrowers will use cheaper prime lenders for refinance mortgages, and reduce their reliance on subprime firms. The second group, borrowers who are near A credit risks, has growing access to mortgage products offered by the GSEs as these borrowers are increasingly served by GSE seller/servicers.

The GSEs have been prudent in their pursuit of subprime lending, focusing on the top part of the market, the “A-minus” and “Alt A” segments. A-minus mortgages are typically those where borrowers have less than perfect credit. Alt A mortgages are originated to borrowers who cannot document all of the underwriting information in the application but generally have FICO scores similar to those in the prime market. The GSEs' subprime products are integrated into their automated underwriting systems and are approved based on mortgage scoring models. These models have proven over the years to be an effective tool in limiting risk layering. The GSEs charge lenders higher fees for guaranteeing these loans. As a result these higher risk loans are priced above those offered to prime borrowers but below what subprime lenders would otherwise charge for these loans.

The GSEs' presence in the subprime mortgage market benefits many low-income and minority borrowers whose risk profiles differ markedly from borrowers who qualify for prime mortgage products. Millions of Americans with less than perfect credit or who cannot meet some of the tougher underwriting requirements of the prime market for reasons such as inadequate income documentation, limited downpayment or cash reserves, or the desire to take more cash out in a refinancing than conventional loans allow, rely on subprime lenders for access to mortgage financing. If the GSEs reach deeper into the subprime market, more borrowers will benefit from the advantages that greater stability and standardization create.

i. Consequences of the Goals for Mortgage Defaults; Neighborhood Impacts

Summary of Comments. HUD received several comments concerning the impact of mortgage default rates on neighborhoods. Comments from mortgage insurance companies highlighted that the higher goals will likely lead to more expanded affordable housing products as well as higher foreclosures. Affordable products present challenges to borrowers and lenders. For borrowers, qualifying for an affordable mortgage does not insure they have a clear understanding of the risks of homeownership. Where aggressive affordable products are aimed at qualifying borrowers for home loans rather than qualifying families for homeownership, lenders need to be cautious of products that test the limits of borrowers' credit capabilities. Affordable products that have been introduced into the market under favorable economic conditions can experience increasing defaults and foreclosures during periods of higher interest rates, higher unemployment and/or lower house price appreciation rates. One commenter indicated that 15 percent or more of borrowers in some affordable housing products could experience default in an economic downturn.

As defaults on affordable products rise, inner city neighborhoods can be especially hard hit. A large number of foreclosures in an area may lead to abandoned properties. While foreclosures devastate borrowers who lose their homes and damage borrowers' credit history, foreclosures also weaken the neighborhoods where the properties are located.

The potential for affordable lending products to result in higher foreclosure during a less prosperous economic environment was echoed in Freddie Mac's comments. Its comment discussed how too many defaults in one neighborhood can lead to serious blight and disinvestment in the community. One commenter recommended that HUD establish safeguards against aggressive affordable products. The commenter suggested that HUD deny Housing Goal credit for GSE mortgage purchases that experience early-term serious defaults (e.g., delinquent 90 days or longer within 12 months of the date of origination).

The GSEs and community groups cautioned that the struggle to meet high goals for low-income groups could cause the GSEs to relax underwriting standards and/or extend loans to people who are unprepared. For example, the commenters pointed out that FHA default rates are higher than the conventional conforming market. High goals would encourage the GSEs to enter markets served by FHA. This incentive to extend credit to unprepared low-income people would rise if unexpected refinances decreased the proportion of goals-eligible units produced in the market.

HUD's Determination. HUD carefully reviewed the comments regarding mortgage default rates. The Department believes that the GSEs' presence in underserved markets will be beneficial for neighborhoods. The GSEs have improved their underwriting methods to better identify risks in these markets, and also have instituted homebuyer education programs. An increased role for the GSEs' seller-services in inner-city neighborhoods will improve competition, reduce high-cost lending, and reduce predatory lending. As described in Appendix A, families living in inner-city, high-minority neighborhoods often have to rely on subprime lenders as their main source of mortgage credit. Studies indicate that many of these borrowers obtaining high-cost loans could qualify for lower-cost, prime mortgage credit. An active GSE effort in these neighborhoods will encourage traditional, mainstream lenders to increase their lending activities in these historically underserved areas. This will offer additional funding options for those lower-income and minority borrowers who today may have to take out a high-cost loan in order to purchase or renovate a home or to refinance an existing mortgage. Reductions in predatory lending reduce the costs of mortgages and the chances of default. As a result, the Department believes that GSE participation is a net benefit to lower income neighborhoods.

j. Consequences of the Goals for Residents of Puerto Rico

Summary of Comments. Several associations stated that HUD's proposed affordable housing goals could be disadvantageous to residents of Puerto Rico, alleging that less than 10 percent of loans that are originated in the Puerto Rican market would qualify for the goals. These commenters were concerned that the GSEs might be unable to buy loans from Puerto Rico, and urged HUD to take special measures to ensure that owner and rental housing production are not deleteriously affected by the demographic and economic differences that exist between the mainland markets and the Puerto Rico market.

HUD's Determination. Loans purchased by the GSEs for properties in Puerto Rico are counted in the same manner as loans purchased on properties in any other location. Since underserved areas are defined as low-income and/or high-minority census tracts in metropolitan areas or counties in non-metropolitan areas, the overwhelming majority of loans purchased by the GSEs on properties in Puerto Rico count toward that goal. In fact, in 2003, Fannie Mae reported that 95 percent of the units it financed in Puerto Rico qualified for the underserved areas goal; the corresponding figure for Freddie Mac was 98 percent.

Relatively few of the loans in Puerto Rico that are purchased by the GSEs qualify for the two income-based goals. Despite this, HUD does not believe that the final housing goals will adversely affect Puerto Rico. In 2003, Puerto Rico accounted for only 0.2 percent of all units financed by Fannie Mae and only 0.1 percent of all units financed by Freddie Mac. Thus overall performance on these broad national goals is not materially affected by the characteristics of loans purchased by the GSEs in Puerto Rico.

Apparently many lower-income families in Puerto Rico rely on consumer finance companies for financing their homes. Since such financing is typically more expensive to borrowers than traditional mortgages, this suggests that the GSEs could play an important role, working with mortgage originators, to better develop the mortgage market in Puerto Rico.

4. Determinations Regarding the Specification and Levels of the Home Purchase Subgoals

a. Overview

Given that the past average performance of the GSEs in the home purchase market has been below market levels, and the Administration's emphasis on increasing homeownership opportunities, including those for low- and moderate-income and minority borrowers, HUD proposed to set Home Purchase Subgoals for GSE mortgage purchase activities to increase financing opportunities for low- and moderate-income, underserved, and special affordable borrowers who are purchasing single-family homes.

Specifically, the Department proposed Home Purchase Subgoals for home purchase loans that qualify for the Housing Goals. The purpose of the Home Purchase Subgoals is to ensure that the GSEs focus on financing home purchases for the homeowners targeted by the Housing Goals. The Department believes that the establishment of Home Purchase Subgoals will place the GSEs in an important leadership position in the Housing Goals categories, while also facilitating homeownership. The GSEs have years of experience in providing secondary market financing for single-family properties and are fully capable of exerting such leadership.

The focus of these Subgoals on home purchase loans meeting the Housing Goals will also help address the racial and income disparities in homeownership that exist today. As noted earlier, although minority homeownership has grown, the homeownership rate for African-American and Hispanic families is still approximately 25 percentage points below that for non-Hispanic white families. The focus of the Subgoals on home purchase will also increase the GSEs' support of first-time homebuyers, a market segment where they have lagged primary lenders.

Summary of Comments. Fannie Mae claimed that the proposed Subgoals are not necessary and are, in fact, duplicative of the broader goals structure. Fannie Mae asserted that it is already a leader in financing home purchases, even in a period of aggressive refinancings. In addition, Fannie Mae stated that subgoals add complexity to the mortgage market and contribute to a loss of liquidity, and suggested that the proposed Subgoals do not reflect recent market experience because affordability may decline and HUD may mistreat missing data when formulating subgoals. Fannie Mae also stated that HUD improperly exercised its authority in proposing the Subgoals.

Specifically, Fannie Mae contended that a complex subgoal structure harms liquidity and that when Fannie Mae needs to stretch in one market to meet a goal, it may have to reduce its willingness to purchase mortgages in another market. Fannie Mae stated that conflicts between the goals arise because the goals are set as a percentage of business, and fulfilling the numerator of one goal adds to the denominator of the other goals. Fannie Mae asserted that the GSEs could be forced to abstain from buying non-goal eligible mortgages that would count in the denominator, but that would not benefit its calculation of goals performance in the numerator. In Fannie Mae's view, its own abstention from buying implies an illiquid market.

Other commenters affirmed Fannie Mae's comments and expressed concern that, given the market leadership of the GSEs, the manner in which home purchases are counted toward the Subgoals could distort the lending market.

In addition, both Fannie Mae and Freddie Mac asserted that FHEFSSA requires that HUD consider each of the six statutory factors set forth in sections 1332(b) and 1334(b) of the statute in setting the levels of any Subgoals within those Housing Goals. Freddie Mac objected to the home purchase Subgoals because it claimed these Subgoals would constitute micromanagement of the GSEs' business decisions. Freddie Mac also noted that, in the past, HUD has declined to implement subgoals for that very reason.

Several commenters expressed the view that HUD had overestimated available purchase money mortgages and noted that if Subgoals on these types of mortgages are set too high, adverse market distortions will occur.

Other commenters contended that, regardless of the level of the Subgoals, a subgoal that targets home purchase mortgages unfairly allocates credit toward home buying rather than mortgage refinances. These commenters asserted that this credit allocation is unfair in that it penalizes borrowers who want to lower mortgage costs or improve their homes. They also contended that credit allocation that promotes purchase mortgages could push refinance borrowers into high-cost loans rather than conforming, GSE-eligible mortgages. To combat such effects, one organization suggested separate subgoals for both purchase money mortgages and refinances, with the overall low- and moderate-income goal as the weighted average of the different subgoals.

Commenters also objected to mortgage purchase subgoals targeting only those loans originated in metropolitan areas because this geographic limitation allocates credit at the expense of residents of rural communities. The commenters stated that Congress charged the GSEs in their charters to “promote access to mortgage credit throughout the Nation (including central cities, rural areas, and underserved areas).” One commenter stated that the lack of detailed HMDA data in rural areas makes market size estimates difficult, but suggested that other data from private vendors could provide acceptable measures (without offering any specific sources).

HUD's Determination. Home purchase is a high national priority. The comments received and research reviewed document many studies revealing the desire of Americans to own their own home. HUD finds that the proposed home purchase subgoal furthers the statutory objectives of FHEFSSA. HUD set the level of the home purchase subgoal prudently. Details of HUD's methodology are found in Appendices A and D of this final rule and in chapter 3 of the Economic Analysis that accompanies the rule. Rather than distorting the market, the home purchase subgoal facilitates the desire of many Americans to use the market to acquire their own home.

Several commenters asked HUD to extend the counting for the home purchase subgoal to rural areas even though data for rural areas is sparse. HUD disagrees. Although HMDA data for rural areas has improved, it is still too incomplete to support extending the counting system. Alternative sources from private lenders are similarly flawed. While HMDA's reporting of non-metropolitan areas has improved over the years, it continues to be unreliable. In 2001, 3,757 (3,280 of which were small banks) of the 4,394 non-metropolitan-area banks did not report under HMDA. In that same year, 324 (246 of which were small thrifts) of the 458 non-metropolitan-area thrift institutions did not report under HMDA.

Except for Fannie Mae's recent performance in the Special Affordable and Low- and Moderate-Income categories, the GSEs have lagged the market in purchasing single-family, owner-occupied loans that qualify for the Housing Goals. In 2003, Fannie Mae continued to lag the market in financing properties located in underserved areas while Freddie Mac lagged the market in all three goals-qualifying categories. The Department's analysis reveals that there is ample room for both Fannie Mae and Freddie Mac to improve their performance in purchasing home loans that qualify for the Housing Goals, particularly in important market segments such as the minority, first-time homebuyer market.

Both GSEs' funding of mortgages for first-time homebuyers lags the market's provision of funding for these families, and the lag is particularly large for first-time minority homebuyers. Table 2 compares the GSEs' funding of mortgages for first-time homebuyers with market loan originations for first-time homebuyers. This table shows that first-time homebuyers represented 37.6 percent of market loan originations, compared with 26.5 percent of the GSEs' purchases; thus, the GSEs fell substantially short of the market originations ratio for first-time homebuyers, over the period 1999-2001.

For minority first-time homebuyers, the GSE ratio was 6.2 percent, compared to a market originations ratio of 10.6 percent. For African-American and Hispanic first-time homebuyers, the GSE ratio was 3.8 percent, compared to a market originations ratio of 6.9 percent. For first-time homebuyers, particularly first-time minority homebuyers, both GSEs substantially lag the private conventional conforming market.

As detailed in Appendix A to this rule, evidence suggests that there is a significant population of potential homebuyers who are likely to respond well to increased homeownership opportunities produced by increased GSE purchases in this area. Immigrants and minorities, in particular, are expected to be a major source of future homebuyers. Furthermore, studies indicate the existence of a large untapped pool of potential homeowners among the rental population. Indeed, the GSEs' recent experience with new outreach and affordable housing initiatives confirms the existence of this potential.

The Department therefore is establishing through this rule Subgoals for home purchase loans that qualify for the three Housing Goals to encourage the GSEs to take a leadership position in creating homeownership financing opportunities within the categories that Congress expressly targeted with the Housing Goals.

b. HUD's Determinations Regarding the Home Purchase Subgoals

Under FHEFSSA, HUD is authorized to establish nonenforceable Subgoals within the Low- and Moderate-Income Housing Goal and the Underserved Areas Housing Goal. HUD also is authorized under FHEFSSA to establish enforceable Subgoals within the Special Affordable Housing Goal. The Administration has proposed, as part of GSE regulatory reform, that Congress authorize HUD to establish a separate Home Purchase Goal that would include enforceable components. Pending the enactment of any such legislation, HUD is establishing the Home Purchase Subgoals described in this final rule under its current statutory authority.

HUD stated in the preamble to the proposed rule that in setting a subgoal, “[c]urrent law does not require that HUD consider the statutory factors set forth in FHEFSSA prior to establishing or setting the level of Subgoals.” (69 FR 24244.) HUD's interpretation of this portion of FHEFSSA is unchanged. Each of the subsections identifying the factors for consideration indicates that the factors are to be considered in setting each respective goal; no mention is made of the subgoals. However, despite the absence of any statutory requirement to consider the listed factors in setting the levels of the subgoals, HUD has nevertheless carefully considered each of these factors in setting the subgoal levels in this final rule.

The following sections provide an overview of HUD's reasons for establishing the Subgoals, which are detailed in the Appendices to this rule.

(i) The GSEs Have the Ability To Lead the Market

The GSEs have the ability to lead the primary market for mortgages on single-family owner-occupied properties, which are the GSEs' principal line of business. Both GSEs have long experience in the home purchase mortgage market, and therefore there is no issue of the degree to which they have penetrated this market. In addition, because the Subgoals focus on homeownership opportunities and, thus, do not include refinance loans, there is no issue regarding potentially large year-to-year changes in refinance mortgage volumes, which affect the magnitude of the denominator in calculating performance percentages under the Housing Goals, as experienced in the heavy refinance years of 1998 and 2001-2003.

Both GSEs have not only been operating in the single-family owner mortgage market for years, they have been the dominant players in that market, funding 57 percent of mortgages on single-family owner-occupied residences financed between 1999 and 2002. As discussed in Section G of Appendix A to this rule, their underwriting guidelines are industry standards and their AUS are widely used in the mortgage industry.

(ii) The GSEs' Performance Relative to the Market

Even though the GSEs have had the ability to lead the home purchase market, their past average performance (1993-2003, 1996-2003, and 1999-2003) has been below market levels. During 2002 and 2003, Fannie Mae improved its performance enough to lead the special affordable and low-mod markets for home purchase loans, but Fannie Mae continued to lag the primary market in funding homes in underserved areas. The subgoals will ensure that Fannie Mae maintains and further improves its above-market performance in the special affordable and low-mod markets, and also becomes a market leader in funding underserved areas. Freddie Mac, although it has also improved its recent performance, continues to lag behind the primary market on all housing goal categories. The subgoals will ensure that Freddie Mac erases its gaps with the market and takes a leadership position as well. The type of improvement needed for Freddie Mac to meet these new subgoals was demonstrated by Fannie Mae during 2001-2003. For example, Fannie Mae increased its low-mod purchases from 40.8 percent of its single-family-owner business in 2000 to 45.3 percent in 2002 to 47.0 percent in 2003.

(iii) Disparities in Homeownership and Credit Access Remain

HUD notes that there remain troublesome disparities in our housing and mortgage markets, even after the “revolution in affordable lending” and the growth in homeownership that has taken place since the mid-1990s. As noted previously in the discussion of the goals, the homeownership rate for African-American and Hispanic households remains 25 percentage points below that of white households. In 2002, the mortgage denial rate for African-American borrowers was over twice that for white borrowers, even after controlling for the income of the borrower.

HUD also notes that there is growing evidence that inner city neighborhoods are not always being adequately served by mainstream lenders. Some have concluded that a dual mortgage market has developed in our nation, with conventional mainstream lenders serving mainly white families living in the suburbs and FHA and subprime lenders serving minority families concentrated in inner city neighborhoods. In addition to the unavailability of mainstream lenders, families living in high-minority neighborhoods generally face many additional hurdles, such as lack of cash for a downpayment, credit problems, and discrimination.

Immigrants and minorities are projected to account for almost two-thirds of the growth in the number of new households over the next ten years. As emphasized throughout this preamble and the Appendices to this rule, changing population demographics will result in a need for the primary and secondary mortgage markets to meet nontraditional credit needs, respond to diverse housing preferences and overcome information and other barriers that many immigrants and minorities face. HUD finds that the GSEs must increase their efforts towards providing financing for these families.

(iv) There Are Ample Opportunities for the GSEs To Improve Their Performance in the Home Purchase Market

Home purchase loans that qualify for the Housing Goals are available for the GSEs to purchase, which means they can improve their performance and lead the primary market in purchasing loans for lower-income borrowers and properties in underserved areas. Three indicators of this have already been discussed.

First, the affordable lending market has shown an underlying strength over the past few years that is unlikely to vanish (without a significant increase in interest rates or a decline in the economy). Since 1999, the shares of the home purchase market accounted for by the three Housing Goal categories are as follows: 16.3 percent for special affordable, 31.4 percent for underserved areas, and 44.1 percent for low- and moderate-income.

Second, market share data reported in section G of Appendix A to this rule show that almost half of newly-originated loans that qualify for the Housing Goals are not purchased by the GSEs. As noted above, the situation is even more extreme for special sub-markets, such as the minority first-time homebuyer market where the GSEs have only a minimal presence. In terms of the overall mortgage market (both conventional and government), the GSEs funded only 24 percent of all first-time homebuyers and 17 percent of minority first-time homebuyers between 1999 and 2001. Similarly, during the same period, the GSEs funded only 40 percent of first-time homebuyers in the conventional conforming market, and only 33 percent of minority first-time homebuyers in that market.

Finally, the GSEs' purchases that can count toward the Subgoal are not limited to new mortgages that are originated in the current calendar year. The GSEs can purchase loans from the substantial, existing stock of affordable loans held in lenders' portfolios, after these loans have seasoned and the GSEs have had the opportunity to observe their payment performance. In fact, based on Fannie Mae's recent experience, the purchase of seasoned loans is at present one strategy employed for purchasing Housing Goals-qualifying loans and meeting the goals.

The current low homeownership rate of minorities and others living in inner cities suggests that there will be considerable growth in the origination of CRA loans in urban areas. For banks and thrifts, selling their CRA originations will free up capital to make new CRA loans. As a result, the CRA market segment provides an opportunity for the GSEs to expand their affordable lending programs. As explained in Appendix A to this rule, Fannie Mae and Freddie Mac have already started developing programs to purchase CRA-type loans on a flow basis as well as after they have seasoned.

While the GSEs can choose any strategy for leading the market, this leadership role can likely be accomplished by building on the many initiatives and programs that the enterprises have already started, including: (1) Their outreach to underserved markets and their partnership efforts that encourage mainstream lenders to move into these markets; (2) their incorporation of greater flexibility into their purchase and underwriting guidelines, (3) their development of new products for borrowers with little cash for a downpayment and for borrowers with credit blemishes or non-traditional credit histories; (4) their targeting of important markets where they have had only a limited presence in the past, such as the markets for minority first-time homebuyers; (5) their purchases of both newly-originated and seasoned CRA loans; and (6) their use of automated underwriting technology to qualify creditworthy borrowers that would have been deemed not creditworthy under traditional underwriting rules.

The experience of Fannie Mae and Freddie Mac in the subprime market indicates that they have the expertise and experience to develop technologies and new products that allow them to enter new markets in a prudent manner. Given the innovativeness of Fannie Mae and Freddie Mac, other strategies will be available as well. In fact, a wide variety of quantitative and qualitative indicators suggest that the GSEs have the expertise, resources and financial strength to improve their affordable lending performance enough to lead the home purchase market for special affordable, low- and moderate-income, and underserved areas loans. The recent improvement in the affordable lending performance of the GSEs, and particularly Fannie Mae, further demonstrates the GSEs' capacity to lead the home purchase market.

c. Structure and Levels of the Home Purchase Subgoals

Under this rule, performance on the Home Purchase Subgoals will be calculated as Housing Goal-qualifying percentages of the GSEs' total purchases of mortgages that finance purchases of single-family, owner-occupied properties located in metropolitan areas, based on the owner's income and the location of the property. Specifically, for each GSE the following Subgoals would apply. (A “home purchase mortgage” is defined as a residential mortgage for the purchase of an owner-occupied single-family property.)

• 45 percent of home purchase mortgages purchased by the GSE in metropolitan areas must qualify under the Low- and Moderate-Income Housing Goal in 2005, with this share rising to 46 percent in 2006 and 47 percent in both 2007 and 2008;
• 32 percent of home purchase mortgages purchased by the GSE in metropolitan areas must qualify under the Underserved Areas Housing Goal in 2005, with this share rising to 33 percent in both 2006 and 2007 and 34 percent in 2008; and
• 17 percent of home purchase mortgages purchased by the GSE in metropolitan areas must qualify under the Special Affordable Housing Goal in both 2005 and 2006, with this share rising to 18 percent in both 2007 and 2008.

Calculation of performance under the Home Purchase Subgoals will be in terms of numbers of mortgages, not numbers of units. This is consistent with the basis of reporting in HMDA data, which were HUD's point of reference in establishing the Home Purchase Subgoal levels. HMDA data are reported in terms of numbers of mortgages in metropolitan areas.

These Home Purchase Subgoals are shown in Table 3, along with information on what the GSEs' performance on the Subgoals would have been if they had been in effect for 1999-2003 (under the proposed counting rules for 2005-2008). Table 3 also presents HUD's estimates of the average shares of mortgages on owner-occupied single-family properties in metropolitan areas that were originated in 1999-2003 that would have qualified for these Home Purchase Subgoals.

d. Counting Mortgages Toward the Home Purchase Subgoals

The Department is amending 24 CFR 81.15 to add a new paragraph (i) that would clarify that the procedures in § 81.15 generally govern the counting of home purchase mortgages toward the Home Purchase Subgoals in §§ 81.12, 81.13 and 81.14. The new paragraph provides, however, that the numerator and denominator for purposes of counting performance under the Subgoals are comprised of numbers of home purchase mortgages in metropolitan areas, rather than numbers of dwelling units. Paragraph (i) also provides that, for purposes of addressing missing data or information for each Subgoal, the procedures in § 81.15(d) shall be implemented using numbers of home purchase mortgages in metropolitan areas and not single-family, owner-occupied dwelling units. Finally, the new paragraph provides that where a single home purchase mortgage finances the purchase of two or more owner-occupied units, the mortgage shall count once toward each Subgoal that applies to the GSE's mortgage purchase.

5. Low- and Moderate-Income Housing Goal, § 81.12

This section discusses the Department's consideration of the statutory factors in arriving at, and the comments received on, the new housing goal level for the Low- and Moderate-Income Housing Goal, which targets mortgages on housing for families with incomes at or below the area median income. After consideration of these factors, this final rule establishes the goal for the percentage of dwelling units to be financed by each GSE's mortgage purchases at 52 percent for 2005, 53 percent for 2006, 55 percent for 2007, and 56 percent for 2008.

Additional information analyzing each of the statutory factors is provided in Appendix A, “Departmental Considerations to Establish the Low- and Moderate-Income Housing Goal,” and Appendix D, “Estimating the Size of the Conventional Conforming Market for each Housing Goal.”

a. Market Estimate for the Low- and Moderate-Income Housing Goal

The Department estimates that dwelling units serving low- and moderate-income families will account for 51-56 percent of total units financed in the overall conventional conforming mortgage market during the period 2005 through 2008. HUD has developed this range, rather than a specific point estimate, to account for the projected effects of different economic and affordability conditions that can reasonably be anticipated. HUD estimates that the low-and-moderate-income share of the market averaged 57 percent between 1999 and 2002.

b. Past Performance of the GSEs Under the Low- and Moderate-Income Housing Goal

A number of changes in goal-counting procedures were adopted as part of HUD's Housing Goals final rule published on October 31, 2000 (65 FR 65044) (Housing Goals 2000 final rule). Thus, it is necessary to provide information using several different measures in order to track performance on the Low- and Moderate-Income Housing Goal over the 1996-2003 period. Table 4 shows performance under these measures.^[14]

Specifically, the following changes were made in counting procedures for measuring performance on the Low- and Moderate-Income Housing Goal for 2001-2003. HUD:

(1) Established “bonus points” (awarding double credit) for purchases of low- and moderate-income mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on two-to-four unit owner-occupied properties;

(2) Established a “temporary adjustment factor” (1.35 units credit, as revised by Congress for 2001-2003 from HUD's 1.2 unit credits in the Housing Goals 2000 final rule) that applied to Freddie Mac's purchases (but not Fannie Mae's purchases) of low- and moderate-income mortgages on large (more than 50-unit) multifamily properties; and

(3) Revised procedures that HUD had instituted regarding the treatment of missing data on unit affordability, the use of imputed or proxy rents for determining goal credit for multifamily mortgages, and the eligibility for goals credit for certain qualifying government-backed loans.

Based on the counting rules in effect at that time for 1996-2000, as shown under “official performance” for 1996-2000 in Table 4, Low- and Moderate-Income Housing Goal performance for Fannie Mae was consistently in the 44-46 percent range over the 1996-1999 period, before jumping to a peak of 49.5 percent in 2000. Freddie Mac's performance started at a lower level, but then increased in several steps, from 41-43 percent in 1996-1998 to 46.1 percent in 1999, and a record level of 49.9 percent in 2000. That was the only year prior to 2001 in which Freddie Mac's performance exceeded Fannie Mae's performance on this goal.

Based on the then current counting rules, including the bonus points and TAF, as shown under “official performance” in Table 4, Low- and Moderate-Income Housing Goal performance was 51.5 percent for Fannie Mae in 2001, 51.8 percent in 2002, and 52.3 percent in 2003. For Freddie Mac, performance was 53.2 percent in 2001, 50.5 percent in 2002, and 51.2 percent in 2003.

Immediately beneath the official Low- and Moderate-Income Housing Goal performance percentages in Table 4 are figures showing the GSEs' low- and moderate-income purchase percentages on a consistent basis for the entire 1996-2003 period. The assumptions used were the counting rules established in HUD's Housing Goals 2000 final rule except that bonus points and the Freddie Mac TAF (which were terminated at the end of 2003) are not applied. These figures are termed the “2001-2003 baseline assumptions.” For 1996-2000 these figures differ from the official performance figures because they incorporate the revised counting procedures described under point (c), above, which were not reflected in the official performance figures at that time. For 2001-2003 both sets of figures incorporate the revised counting procedures, but the baseline does not incorporate the bonus points and the Freddie Mac TAF.

In terms of the 2001-2003 baseline measure, both Fannie Mae's and Freddie Mac's low- and moderate-income performance reached its maximum in 2000 (Fannie Mae at 51.3 percent and Freddie Mac at 50.6 percent). Baseline performance fell somewhat for both GSEs in 2001, 2002, and 2003. Fannie Mae's baseline performance last year exceeded the level attained in 1999, but Freddie Mac's performance fell to the lowest level since 1998.

Overall, both GSEs' performance exceeded HUD's Low- and Moderate-Income Housing Goals by significant margins in 1996-1999, and by wide margins in 2000. New, higher goals were established for 2001-2003, and despite somewhat lower performance than the level attained in 2000, both GSEs' official performance exceeded the new goal levels in each year 2001-2003, with the inclusion of the bonus points and the TAF.

The decline in baseline performance in 2001-2003 can be attributed in large measure to the mortgage refinance wave that occurred in those years. Fannie Mae's overall volume of mortgage purchases (in terms of numbers of housing units) rose from 2.2 million in 2000 to 4.7 million in 2001, 6.4 million in 2002, and then to 10.1 million in 2003. Similarly, Freddie Mac's volume rose from 1.6 million in 2000 to 3.3 million in 2001, 4.3 million in 2002, and then to 5.8 million in 2003. For each GSE the increase in volume each year can be largely attributed to increases in purchase volumes for refinance mortgages relative to home purchase mortgages. For each GSE, the fraction of mortgages that qualified as Low- and Moderate-Income was less for refinance mortgages than for home purchase mortgages.

For 2005-2008, HUD is expanding the affordability estimation of units with missing affordability information. In addition to multifamily units, the GSEs will also be able to use estimates of affordability for single-family rental units with missing rents and owner-occupied units with missing borrower incomes for determining goal credit. HUD is also increasing the amount of the maximum allowed for affordability estimation for multifamily units.

Beneath the 2001-2003 baseline figures in Table 4 is another row of figures designated “With 2005 Assumptions.” These figures show the effects of applying 2000 Census data and the new specification of MSAs released by OMB in 2003 to the measurement of Low- and Moderate-Income purchase percentages with the same counting rules that were used for the 2001-2003 baseline in Table 4. The effect is to reduce the Goal-qualifying percentage by an average of 0.6 percentage points for Fannie Mae and 0.7 percentage points for Freddie Mac, over the 1999-2002 period.

However, for 2003, the effects are just the opposite—these assumptions increased Fannie Mae's performance by 0.8 percentage point (from 48.7 percent to 49.5 percent) and Freddie Mac's performance by 0.3 percentage point (from 45.0 percent to 45.3 percent). The difference in the direction of this impact between 1999-2002 and 2003 may be due to the need to apply estimation techniques in 1999-2002 but not in 2003. For 1999-2002 HUD had to estimate the effect based on data geocoded according to 1990 census tract definitions, while for 2003 the data were geocoded to 2000 census tracts. Further insight will be provided by analysis of data for 2004 and further years.

c. Low- and Moderate-Income Home Purchase Subgoal

The Department has determined to establish a Subgoal of 45 percent for each GSE's purchases of home purchase mortgages on single-family owner-occupied properties in metropolitan areas which are for low- and moderate-income families in 2005, with this Subgoal rising to 46 percent in 2006 and 47 percent in both 2007 and 2008.

The purpose of this Subgoal is to encourage the GSEs to increase their acquisitions of home purchase loans for low- and moderate-income families, many of whom are expected to enter the homeownership market over the next few years. Table 5 provides basic information on both the GSEs' low-mod performance and the primary market's low-mod performance for the years 1999 to 2003. Since the same format will be followed for the other housing subgoals, several points are made about the information in the Table 5, prior to discussing the low-mod subgoal.

Average Performance Data. In addition to individual year data, various averages of annual performance are provided at the bottom of Table 5 (1999-2003, 2001-2003, and 2002-2003); these averages provide a useful context for examining the feasibility of the subgoals and the degree to which they call for performance that is above past market levels. This table provides a picture of how much the low-mod subgoal targets move the GSEs above past market levels and how much of a stretch each subgoal will be for each GSE (as compared with that GSE's past performance). As will become clear below, Fannie Mae and Freddie Mac have shown different past performances, which means that the subgoal targets will appear to have different impacts on these two institutions.

Definitions of Primary Market. HUD's basic market definition is the conventional conforming market without B&C loans; in other words, the A-minus loans in the subprime market are included in the market definition but the more risky B&C portion is not included (see Appendix D of the final rule for further discussion of this). In its report for Freddie Mac, ICF indicated that small loans (those less than $15,000) should be excluded from any analysis that dealt with loans that might be available for purchase by the GSEs. Therefore, data are provided in Table 5 for (a) the market without B&C loans and (b) the market without both B&C and small loans less than $15,000. As shown in Table 5, dropping small loans reduces the low-mod share of the conventional conforming market by about one-half percentage point.

Projected 2000-Based Data. Table 5 is based on projected data that incorporates both 2000 Census geography and the new OMB definitions. Thus, the goals-qualifying percentages in this table differ from those reported earlier in this Preamble, the latter being historical, 1990-Census-based percentages. HUD had to reapportion the data for the years prior to 2003. For 2003, both HMDA and GSE data were defined in terms of 2000 Census geography, so no reapportionment was necessary; for this reason, the 2003 data are probably the most accurate. With these basics, the results for the low-mod subgoal can now be briefly summarized as follows:

Low-Mod Subgoals Compared With Market. The 45-percent subgoal for the first year (2005) is approximately two percentage points above 1999-2003 and 2001-2003 average market performance, one percentage point above 2002-2003 average market performance, and 0.6 percent (market without B&C loans) to 0.2 percent (market without both B&C and small loans) below peak market performance. The 46-percent subgoal for 2006 would add one percentage point to these comparisons, while the 47-percent subgoal for 2007 and 2008 would add two percentage points. For example, the 47-percent subgoal is approximately three percentage points above 2002-2003 average market performance, and 1.4 percent (market without B&C loans) to 1.8 percent (market without both B&C and small loans) above peak market performance.

Low-Mod Subgoals Compared With Past Freddie Mac Performance. To reach the 45-percent 2005 subgoal, Freddie Mac would have to improve its performance by 3.0 percentage points over its 2001-2003 average low-mod performance of 42.0 percent, by 1.8 percentage points over its 2002-2003 average low-mod performance of 43.2 percent, and by 0.8 percent over its previous peak performance of 44.2 percent in 2003. To reach the 47-percent subgoal, Freddie Mac would have to improve its performance by 3.8 percentage points over its 2002-2003 average low-mod performance, and by 2.8 percent over its previous peak performance.

Low-Mod Subgoals Compared With Past Fannie Mae Performance. To reach the 45-percent 2005 subgoal, Fannie Mae would have to improve its performance by 0.7 percentage points over its 2001-2003 average low-mod performance of 44.3 percent; Fannie Mae would meet the 45-percent subgoal based on its 2002-2003 average low-mod performance of 45.6 percent and its previous peak low-mod performance of 47.5 percent in 2003. To reach the 47-percent subgoal, Fannie Mae would have to improve its performance by 2.7 percent over its 2001-2003 average performance and by 1.4 percentage points over its 2002-2003 average performance; Fannie Mae would meet the 47-percent subgoal based on its previous peak performance of 47.5 percent in 2003.

The low-mod subgoal targets will be more challenging for Freddie Mac than Fannie Mae. The type of improvement needed to meet the new low-mod subgoal targets was demonstrated by Fannie Mae during 2001-2003, as Fannie Mae increased its low-mod purchases from 40.1 percent of its single-family-owner business in 2000 to 43.6 percent in 2002 to 47.5 percent in 2003, as shown in Table 5. The approach taken is for the GSEs to obtain their leadership position by staged increases in the subgoals; this will enable the GSEs to take new initiatives in a correspondingly staged manner to achieve the new subgoals each year. Thus, the increases in the housing subgoals are sequenced so that the GSEs can gain experience as they improve and move toward the new higher subgoal targets.

Section 4.b. above of this preamble, and Section I.3 of Appendix A to this rule, discuss the reasons why the Department is establishing the Subgoal for low- and moderate-income loans, as follows: (1) The GSEs have the resources and the ability to lead the market in providing mortgage funding for low- and moderate-income families; (2) except for Fannie Mae's recent performance, the GSEs have historically (over periods such as 1993-2003, 1996-2003, and 1999-2003) not led the market, even though they have had the ability to do so; (3) troublesome disparities in our housing and mortgage markets indicate a continuing need for increased GSE activity; and (4) there are ample opportunities for the GSEs to improve their low- and moderate-income performance in the home purchase market.

Although single-family owner-occupied mortgages comprise their principal line of business, Freddie Mac has always lagged behind the primary market in financing mortgages for low- and moderate-income families. Over the past three years Fannie Mae has closed its historical gap with the market and now leads the primary market in funding mortgages for low- and moderate-income families. Because home purchase loans account for a major share of the GSEs' purchases, the establishment of this Subgoal will aid their performance under the overall Low- and Moderate-Income Housing Goal.

For the foregoing reasons, the Department believes that the GSEs, and particularly Freddie Mac, can do more to raise the share of their home loan purchases serving low- and moderate-income families. This can be accomplished by building on efforts that the enterprises have already started, including their new affordable lending products, their many partnership efforts, their outreach to inner city neighborhoods, their incorporation of greater flexibility into their underwriting guidelines, and their purchases of seasoned CRA loans. A wide variety of quantitative and qualitative indicators indicate that the GSEs have the resources and financial strength to improve their affordable lending performance enough to lead the market serving low- and moderate-income families.

d. Summary of Comments

The majority of comments that addressed the housing goals focused on the highest goal in year 2008 for the Low- and Moderate-Income Housing Goal. While some commenters, such as affordable housing policy advocacy groups and housing and consumer coalitions, expressed support for more aggressive goals, stating that the goals should be set to challenge the GSEs to do more, most commenters expressed concerns about possible adverse affects on middle-income borrowers, including the potential for higher costs and for unrealistic goals to lead to credit allocation to the lower end of the housing market, thereby hindering the GSEs' ability to serve all homebuyers. Other concerns included issues related to HUD's market share methodology analysis and the effects of single-family refinance loans in high refinance years on the GSEs' ability to meet the higher goals. Many commenters recommended that HUD exempt refinances from the goals performance calculation. As described earlier in this rule, HUD is seeking public comments on how to address the effects of refinance loans when this annual volume is high. In addition, some expressed the belief that overly aggressive goals could weaken the FHA insurance program and could encourage over-investment in rental housing at a time when multifamily vacancy rates are high. HUD has addressed these concerns in earlier sections of this final rule preamble. Others felt that higher goal levels will encourage more investor-owned rental units that harm communities. Both Fannie Mae and Freddie Mac objected to the higher goal level for the Low- and Moderate-Income Goal. Each disputed HUD's market share analysis, citing the uncertainty of data, for example the size of the multifamily market, and the uncertainty about future economic conditions. Freddie Mac stated that HUD overestimated the low/mod market share by 4 percent. Both GSEs also stated that it was inappropriate to base the goals at the high end of market share ranges. Freddie Mac stated that this approach ignores the year-to-year variability of the market. Appendix D to this rule responds to these market issues raised by the GSEs.

With regard to the Low- and Moderate-Income Home Purchase Subgoal, most commenters did not address the subgoal levels proposed by HUD, and none specifically addressed the proposal levels for the Low- and Moderate-Income Subgoal. For those that did mention the subgoals, the comments were mixed with about half supportive of the subgoal proposals in general and half believing the subgoal levels were too high. Both GSEs commented on HUD's proposed subgoals. Fannie Mae stated that the levels were higher than any values observed in HMDA from 1999-2002, and that the concept was duplicative of the overall goal structure. Freddie Mac stated that HUD should withdraw the home purchase subgoals or HUD should re-estimate the market using reasonable assumptions and set both the goal and subgoal levels no higher than the midpoint of the resulting ranges.

e. HUD's Determination

The Low- and Moderate-Income Housing Goal established in this final rule is reasonable and appropriate having considered the factors set forth in FHEFSSA. For 2001-2003, HUD set the level of the housing goal conservatively, relative to the Department's market share estimates, in order to accommodate a variety of economic scenarios. Moreover, current examination of the gaps in the mortgage markets, along with the estimated size of the market available to the GSEs, demonstrate that the number of mortgages secured by housing for low- and moderate-income families is more than sufficient for the GSEs to achieve the new goal.

Therefore, having considered all the statutory factors including housing needs, projected economic and demographic conditions for 2005 to 2008, the GSEs' past performance, the size of the market serving low- and moderate-income families, and the GSEs' ability to lead the market while maintaining a sound financial condition, HUD has determined that the annual goal for mortgage purchases qualifying under the Low- and Moderate-Income Housing Goal will be 52 percent for 2005, 53 percent in 2006, 55 percent in 2007, and 56 percent in 2008. This reflects a reduction in the upper end of the market share range from 57 percent to 56 percent since HUD's publication of its proposed rule, resulting from changes in estimating market share as described at the end of section 3 (a), above, and in section F of Appendix D to this rule.

Further, the Department is establishing a Subgoal for each GSE's purchases of home purchase mortgages on single-family owner-occupied properties in metropolitan areas which are for low- and moderate-income families of 45 percent in 2005, with this Subgoal rising to 46 percent in 2006, and 47 percent in both 2007 and 2008. The reasons for increasing the Low- and Moderate-Income Housing Goal are discussed in sections a and b, above, and the reasons for establishing a Home Purchase Subgoal at the stated levels are set forth in section c.

While the GSEs have lagged the primary market in financing owner and rental housing for low- and moderate-income families, they appear to have ample room to improve their performance in that market. A wide variety of quantitative and qualitative indicators demonstrate that the GSEs have the expertise, resources and financial strength to improve their low- and moderate-income lending performance, including lending for low- and moderate-income home purchases, and achieve the levels of the goals being established.

6. Central Cities, Rural Areas, and Other Underserved Areas Housing Goal, § 81.13

This section discusses the Department's consideration of the statutory factors in arriving at, and the comments received on, the new housing goal levels for the Central Cities, Rural Areas, and Other Underserved Areas Goal, which focuses on areas currently underserved by the mortgage finance system. After consideration of the factors and the comments received, this final rule establishes the goal for the percentage of dwelling units to be financed by each GSE's mortgage purchases at 37 percent in 2005, 38 percent in 2006 and 2007, and 39 percent in 2008.

The 1995 final rule provided that mortgage purchases count toward the Underserved Areas Housing Goal if such purchases finance properties that are located in underserved census tracts. At 24 CFR 81.2 of HUD's current regulations, HUD defines “underserved areas” for metropolitan areas (in central cities and other underserved areas) as census tracts where either: (1) The tract median income is at or below 90 percent of the area median income (AMI); or (2) the minority population is at least 30 percent and the tract median income is at or below 120 percent of AMI. The AMI ratio is calculated by dividing the tract median income by the MSA median income. The minority percentage of a tract's population is calculated by dividing the tract's minority population by its total population. For properties in non-metropolitan (rural) areas, mortgage purchases have counted toward the Underserved Areas Housing Goal where such purchases finance properties that are located in underserved counties. As discussed above under the heading “Definitions” in this final rule, HUD is changing this specification from the county level to the census tract level. Mortgages will count toward the Underserved Areas Housing Goal where such purchases finance properties that are located in census tracts were either (1) the median income in the tract does not exceed 95 percent of the greater of the median incomes for the non-metropolitan portions of the state or the non-metropolitan portions of the nation as a whole, or (2) minorities comprise at least 30 percent of the residents of the tract and the median income in the tract does not exceed 120 percent of the greater of the median incomes for the non-metropolitan portions of the state or the non-metropolitan portions of the nation as a whole.

The level for the Underserved Areas Housing Goal is based on 2000 Census data on area median incomes and minority percentages for census tracts, MSAs, and the non-metropolitan portions of states and of the entire nation. HUD's analysis, which is set forth below and described in greater detail in Appendix B to this rule, is based on 2000 census data. The effect of using 2000 census data rather than 1990 data to determine whether areas are underserved increases the percentage of the GSEs' mortgage purchases in underserved areas by an estimated average of 5 percentage points for Fannie Mae and 4 percentage points for Freddie Mac, based on the geographic locations of properties financed by the GSEs' mortgage purchases in 1999 through 2003. This change reflects geographical shifts in population concentrations by income and minority status from 1990 to 2000.

After analyzing the statutory factors, HUD is: (a) establishing a Goal of 37 percent for the percentage of the total number of dwelling units financed by each GSE's mortgage purchases for properties located in underserved areas for 2005, 38 percent for 2006 and 2007, and 39 percent for 2008; (b) establishing census tracts as the spatial basis for establishing whether properties in non-metropolitan (rural) areas count toward the Underserved Areas Housing Goal, in place of counties as in the definition stated above, for the reasons described below; and (c) also establishing a Subgoal of 32 percent of the total number of dwelling units financed by each GSE's purchases of home purchase mortgages in metropolitan areas for properties located in underserved areas of metropolitan areas for 2005, rising to 33 percent for 2006 and 2007, and 34 percent for 2008.

A short discussion of the statutory factors reviewed follows. Additional information analyzing each of the statutory factors is provided in Appendix B to this rule, “Departmental Considerations to Establish the Underserved Areas Housing Goal,” and Appendix D to this rule, “Estimating the Size of the Conventional Conforming Market for each Housing Goal.”

a. Market Estimate for the Underserved Areas Housing Goal

The Department estimates that dwelling units in underserved areas will account for 35-39 percent of total units financed in the overall conventional conforming mortgage market during the period 2005 through 2008. HUD has developed this range, rather than a specific point estimate, to accommodate the projected effects of different economic and affordability conditions that can reasonably be anticipated. HUD estimates that the underserved areas market averaged 39 percent between 1999 and 2002.

b. Past Performance of the GSEs Under the Underserved Areas Housing Goal

As discussed above, a number of changes in goal-counting procedures were adopted as part of HUD's Housing Goals 2000 final rule. Thus it is necessary to provide information using several different measures in order to track changes in the GSEs' performance on the Underserved Areas Housing Goal over the 1996-2003 period. These are shown in Table 6.^[15] The same changes in counting rules described for the Low- and Moderate-Income Housing Goal are applicable to the Underserved Areas Housing Goal.

Based on the counting rules in effect at that time, as shown under “official performance” for 1996-2000 in Table 6, Underserved Areas Housing Goal performance for Fannie Mae generally fluctuated between 27 and 29 percent over the 1996-1999 period, before rising to a peak of 31.0 percent in 2000. Freddie Mac's performance started at a lower level, but then increased in several steps, from 25-26 percent in 1996-1998, to 27.5 percent in 1999, and a record level of 29.2 percent in 2000. Freddie Mac's performance in 1999 was the only year prior to 2001 in which it exceeded Fannie Mae's performance on this Goal.

Based on counting rules in effect for 2001-2003, including the bonus points and the TAF, as shown under “official performance” in Table 6, Underserved Areas Housing Goal performance for Fannie Mae was 32.6 percent in 2001, 32.8 percent in 2002, and 32.1 percent in 2003. Performance for Freddie Mac was 31.7 percent in 2001, slightly less than 31.0 percent in 2002, and 32.7 percent in 2003.

Immediately beneath the official Underserved Areas Housing Goal performance percentages in Table 6 are figures showing the GSEs' purchase percentages under this Goal on a consistent basis for the entire 1996-2003 period. The assumptions used were the counting rules established in HUD's Housing Goals 2000 final rule, except that bonus points and the Freddie Mac TAF (which terminated at the end of 2003) are not applied. These figures are termed the “2001-2003 baseline” assumptions. For 1996-2000 these figures differ from the official performance figures because they incorporate the revised counting procedures, which were not reflected in the official performance figures at that time. For 2001-2003 both sets of figures incorporate the revised counting procedures, but the baseline does not incorporate the bonus points and Freddie Mac TAF.

In terms of the 2001-2003 baseline measure, both Fannie Mae and Freddie Mac's Underserved Areas Housing Goal performance reached its maximum in 2000 (Fannie Mae at 31.0 percent and Freddie Mac at 29.2 percent) before declining somewhat over the 2001-2003 period. Both GSEs' baseline performance in 2001-2003 exceeded the level attained in 1999.

Overall, both GSEs' official performance exceeded their Underserved Areas Housing Goal by significant margins in 1996-1999, and by wide margins in 2000. New, higher Goals were established for 2001-2003, and despite somewhat lower performance than the level attained in 2000 (largely due to the 2001-2003 refinance wave), both GSEs' performance exceeded the new Goal levels in 2001 and 2003; Fannie Mae also exceeded its goal in 2002, while Freddie Mac fell slightly short.

Appendix B to this rule includes a comprehensive analysis of the GSEs' performance in funding mortgages for single-family-owner properties in underserved areas. (The data reported there are based on 2000 Census geography, which produces underserved area figures slightly over five percentage points higher than 1990-based geography.) Both GSEs have lagged the market in funding properties located in underserved neighborhoods. Between 1999 and 2003, 28.3 percent of Freddie Mac's purchases of home loans financed properties in underserved neighborhoods, as did 30.0 percent of Fannie Mae's purchases—compared with 31.4 percent of home purchase loans originated in the conventional conforming market (excluding B&C loans). Thus, Freddie Mac performed at 90 percent of the market level, while Fannie Mae performed at 96 percent of the market level. In 2003, underserved areas accounted for 29.0 percent of Freddie Mac's purchases, 32.0 percent of Fannie Mae's purchases, and 32.5 percent of market originations.

In evaluating the GSEs' past performance, it should be noted that while borrowers in underserved metropolitan areas tend to have much lower incomes than borrowers in other areas, this does not mean that GSE mortgage purchases in underserved areas must necessarily be mortgages on housing for lower income families. Between 1999 and 2001, housing for above median-income households accounted for nearly 60 percent of the single-family owner-occupied mortgages that the GSEs purchased in underserved areas.

Beneath the 2001-2003 baseline figures in Table 6 are two additional rows of figures designated “2005 Assumptions.” These figures show the effects of applying 2000 census data and the new specification of MSAs released by OMB in 2003 to the identification of underserved areas for purposes of measuring historical GSE goal performance. The second of the two rows also incorporates the effects of the Department's proposed change from counties to census tracts as the basis for identifying underserved areas outside of metropolitan areas beginning in 2005.

HUD's determination of underserved areas for purposes of computing the GSEs' performance on the Underserved Areas Housing Goal has, through 2003, been based on area median incomes and area minority percentages from the 1990 Census. HUD applied the existing numerical thresholds for minority percentages and median incomes to 2000 Census data and ascertained that the proportion of underserved census tracts and the proportion of housing units in underserved census tracts in metropolitan areas both have increased significantly from 1990 levels: from 47.6 percent to 51.3 percent of census tracts underserved and from 44.3 percent to 48.7 percent of population in underserved census tracts (including the effects of the 2003 re-specification of Metropolitan Statistical Areas).

Comparable shifts at the county level in non-metropolitan areas were found to be of much smaller magnitude. Further, HUD estimated the spatial distribution of GSE mortgage purchases across metropolitan census tracts and non-metropolitan counties for recent years. The findings were that for 2000, 2001, 2002, and 2003, Fannie Mae's performance figures are an estimated 7.2 percentage points, 6.0 percentage points, 5.5 percentage points, and 5.1 percentage points higher in terms of 2000 Census geography than with 1990 Census geography. The corresponding figures for Freddie Mac are 5.6 percentage points, 5.1 percentage points, 5.1 percentage points, and 3.9 percentage points larger, respectively.

With a further shift to tract-based definitions, the figures for Fannie Mae are reduced by 0.7 percentage point in 2000, 2001, and 2002, and for Freddie Mac by 0.7, 0.8, and 0.7 percentage point, respectively. The differences between county-based performance and tract-based performance were much smaller in 2003, with the latter falling below the former by only 0.2 percentage point for Fannie Mae and exceeding the former by 0.1 percentage point for Freddie Mac last year. As previously noted in the discussion of the Low- and Moderate-Income Housing Goals, the smaller differences between these two approaches in 2003 than in 2000-2002 may be due to the need to apply estimation techniques in 2000-2002 but not in 2003.

c. Underserved Areas Home Purchase Subgoal

The Department believes the GSEs can play a leadership role in underserved markets. To facilitate this leadership, the Department is establishing a Subgoal of 32 percent for each GSE's acquisitions of home purchase mortgages on properties located in the underserved census tracts of metropolitan areas for 2005, rising to 33 percent in 2006 and 2007, and 34 percent in 2008. The purpose of this Subgoal is to encourage the GSEs to improve their purchases of mortgages for homeownership in underserved areas, thus providing additional credit and capital for neighborhoods that historically have not been adequately served. As discussed in Appendix A to this rule, the GSEs have the ability to lead the primary market for single-family-owner loans, which is their “bread-and-butter” business. Both GSEs have been dominant players in the home purchase market for years, funding 61 percent of the single-family-owner mortgages financed between 1999 and 2002. Through their many new product offerings and their various partnership initiatives, the GSEs have shown that they have the capacity to operate in underserved neighborhoods. Even though they have the ability to lead the market, they have not done so, as both GSEs have lagged behind the primary market in serving underserved areas. As shown in Table 7, underserved areas (based on 2000 Census geography) accounted for 29.4 percent of Freddie Mac's purchases of home purchase mortgages in 2003, 32.0 percent of Fannie Mae” purchases, and 32.5 percent of market originations.^[16] The following points can be made about the data presented in Table 7 regarding the underserved areas subgoal:

Underserved Areas Subgoals Compared With Market. The 32-percent subgoal for the first year (2005) is approximately one percentage point above 1999-2003 and 2001-2003 average market performance (based on the market defined without B&C and small loans) and approximately at the 2002-2003 average market performance and the previous peak market performance. The 33-percent subgoal for 2006 and 2007 would add one percentage point to these comparisons, while the 34-percent subgoal for 2008 would add two percentage points. For example, the 34-percent subgoal is approximately three percentage points above both 1999-2003 and 2001-2003 average market performance, 1.8 percent (market without B&C loans) to 2.4 percent (market without both B&C and small loans) above 2002-2003 average market performance, and 1.5 percent (market without B&C loans) to 1.8 percent (market without both B&C and small loans) the market's previous peak performance in 2003.

Underserved Areas Subgoals Compared With Past Freddie Mac Performance. To reach the 32-percent 2005 subgoal, Freddie Mac would have to improve its performance by 2.7 percentage points over its 2001-2003 average underserved areas performance of 29.3 percent, by 1.6 percentage points over its 2002-2003 average underserved areas performance of 30.4 percent, and by 0.3 percent over its previous peak performance of 31.7 percent in 2002. To reach the 34-percent subgoal, Freddie Mac would have to improve its performance by 3.6 percentage points over its 2002-2003 average underserved areas performance, and by 2.3 percent over its previous peak performance. As noted in Table 7, Freddie Mac's performance jumped from 27.3 percent in 2001 to 31.7 percent in 2002, only to fall back to 29.0 percent in 2003. Thus, the 32-percent subgoal for 2005 is three percentage points above Freddie Mac's most recent experience (29.0 percent). However, as noted above, Freddie Mac's 31.7-percent performance in 2002 is only 0.3 percentage points below the 32-percent subgoal for 2005.

Underserved Areas Subgoals Compared With Past Fannie Mae Performance. To reach the 32-percent 2005 subgoal, Fannie Mae would have to improve its performance by 0.6 percentage points over its 2001-2003 average underserved areas performance of 31.4 percent; Fannie Mae would meet the 32-percent subgoal based on its 2002-2003 average underserved areas performance of 32.2 percent and its previous peak underserved areas performance of 32.3 percent in 2002. To reach the 34-percent subgoal, Fannie Mae would have to improve its performance by 2.6 percent over its 2001-2003 average performance, by 1.8 percentage points over its 2002-2003 average performance, and by 1.7 percent over its previous peak performance of 32.3 percent in 2003.

As with the other two home purchase subgoals, the underserved areas subgoal targets will be more challenging for Freddie Mac than Fannie Mae, particularly given Freddie Mac's low performance (29.0 percent) during the most recent year (2003). Again, the type of improvement needed to meet the new underserved areas subgoal targets was demonstrated by Fannie Mae during 2001-2003, as Fannie Mae increased its underserved areas purchases from 29.0 percent of its single-family-owner business in 2000 to approximately 32 percent in both 2002 and 2003. As noted above for the low-mod subgoals, staged increases in the underserved areas subgoal enable the GSEs to obtain their leadership position by gaining experience as they improve and move toward the new higher subgoal targets.

The type of improvement needed to meet this new underserved area subgoal was demonstrated by Fannie Mae during 2001 and 2002. During 2001, underserved area loans declined as a percentage of primary market originations (from 31.7 to 30.7 percent), but they increased as a percentage of Fannie Mae's purchases (from 29.0 to 29.8 percent); and during 2002, they increased further as a percentage of Fannie Mae's purchases (from 29.8 to 32.3 percent), placing Fannie Mae at the market level.

Section 4.b. above of this preamble and Section I.4 of Appendix B to this rule discuss the reasons why the Department is establishing a Subgoal for home purchase mortgages in underserved areas, namely: (1) the GSEs have the resources and the ability to lead the market in providing funding in underserved neighborhoods; (2) the GSEs lag the underserved areas market, even though they have the ability to lead; (3) troublesome disparities in our housing and mortgage markets indicate a continuing need for increased GSE activity; and (4) there are ample opportunities for the GSEs to improve their underserved area performance in the home purchase market.

Although single-family owner-occupied mortgages are the GSEs' principal line of business, the GSEs have lagged behind the primary market in financing properties in underserved areas. For the foregoing reasons, HUD believes that the GSEs can do more to raise the share of their home loan purchases in underserved areas. This can be accomplished by building on efforts that the GSEs have already started, including their new affordable lending products, their many partnership efforts, their outreach to inner city neighborhoods, their incorporation of greater flexibility into their underwriting guidelines, and their purchases of seasoned CRA loans.

A wide variety of quantitative and qualitative indicators demonstrate that the GSEs have the resources and financial strength to improve their affordable lending performance enough to lead the market in underserved areas.

d. Summary of Comments

The Department received no comments that specifically addressed the level of the Underserved Areas Goal. The majority of commenters that offered opinions on the level of the housing goals focused on the high year (2008) of the Low- and Moderate-Income Goal. Where commenters did mention the Underserved Area Goal, their remarks were in the context of better targeting through changes in the definition of underserved areas. HUD also received no comments specific to the Underserved Area Home Purchase Subgoal. Both Fannie Mae and Freddie Mac commented on the level of the Underserved Area Goal. Fannie Mae stated that its replication of HUD's market sizing assumptions did not justify an Underserved Area Goal of 38 or 40 percent. For example, Fannie Mae noted that in reaching a goal level of 40 percent, HUD relied on the most unlikely owner-occupied underserved share of 30 percent, a level reached only once in the past 11 years. With respect to the Underserved Area Subgoal, Fannie Mae stated generally that subgoals risk unintended consequences and that HUD has proposed subgoals in excess of the opportunity and business mix seen in the market. Freddie Mac commented in general that all the goals and subgoals were set beyond what the primary market is likely to originate. With respect to the underserved areas market share, Freddie Mac estimates that the core ranges are 3-4 percentage points below the upper limits of the Department's projected ranges.

e. HUD's Determination

The Underserved Areas Housing Goal established in this final rule is reasonable and appropriate having considered the factors set forth in FHEFSSA. For 2001-2003, HUD set the level of the housing goal conservatively, relative to the Department's market share estimates, in order to accommodate a variety of economic scenarios. Moreover, current examination of the gaps in the mortgage markets, along with the estimated size of the market available to the GSEs, demonstrate that the number of mortgages secured by housing in underserved areas is more than sufficient for the GSEs to achieve the new goal.

Therefore, having considered all the statutory factors including housing needs, projected economic and demographic conditions for 2005 to 2008, the GSEs' past performance, the size of the market serving low- and moderate-income families, and the GSEs' ability to lead the market while maintaining a sound financial condition, HUD has determined that the annual goal for mortgage purchases qualifying under the Underserved Areas Housing Goal will be 37 percent for 2005, 38 percent for 2006 and 2007, and 39 percent for 2008.

Further, the Department is establishing a Subgoal of 32 percent for each GSE's acquisitions of home purchase mortgages on properties located in the underserved census tracts of metropolitan areas for 2005, rising to 33 percent in 2006 and 2007, and 34 percent in 2008. This reflects a reduction in the upper end of the market share range from 35 percent to 34 percent since HUD's publication of its proposed rule, resulting from changes in estimating market share as described at the end of Section 3.a. above, and in Section G of Appendix D to this rule.

The reasons for increasing the Underserved Areas Housing Goal are discussed in Sections a. and b. above, and for establishing a Home Purchase Subgoal at the stated levels in section c. While the GSEs have lagged the primary market in funding loans in underserved areas, they appear to have ample room to improve their performance in that market. A wide variety of quantitative and qualitative indicators demonstrate that the GSEs have the expertise, resources, and financial strength to improve their low- and moderate-income lending performance, including lending for home purchases in underserved areas, and achieve the levels of the goals being established.

7. Special Affordable Housing Goal, § 81.14

This section discusses the Department's consideration of the statutory factors in arriving at, and the comments received on, the new housing goal level for the Special Affordable Housing Goal, which targets mortgages on housing for very low-income families and low-income families in low-income areas. After consideration of these statutory factors and the comments received, this final rule establishes the goal for the percentage of dwelling units to be financed by each GSE's mortgage purchases at 22 percent in 2005, 23 percent in 2006, 25 percent in 2007, and 27 percent in 2008.

After analyzing the statutory factors, HUD has determined to establish: (a) a Goal of 22 percent for the percentage of the total number of dwelling units financed by each GSE's mortgage purchases that are for special affordable housing, affordable to very low-income families and families living in low-income areas for 2005, rising to 23 percent in 2006, 25 percent in 2007, and 27 percent in 2008; (b) a Subgoal of 17 percent of the total number of each GSE's purchases of home purchase mortgages in metropolitan areas that are for housing affordable to very low-income families and low-income families in low-income areas for 2005 and 2006, rising to 18 percent in 2007 and 2008; and (c) a Subgoal of 1 percent of each GSE's combined annual average mortgage purchases in 2000, 2001, and 2002, for each GSE's special affordable mortgage purchases that are for multifamily housing in 2005-2008.

A short discussion of the statutory factors for establishing the Special Affordable Housing Goal follows. Additional information analyzing each of the statutory factors is provided in Appendix C, “Departmental Considerations to Establish the Special Affordable Housing Goal,” and Appendix D, “Estimating the Size of the Conventional Conforming Market for each Housing Goal.”

a. Market Estimate for the Special Affordable Housing Goal

The Department estimates that dwelling units serving very low-income families and low-income families living in low-income areas will account for 23-27 percent of total units financed in the overall conventional conforming mortgage market during the period 2005 through 2008. HUD has developed this range, rather than a point estimate, to account for the projected effects of different economic conditions that can reasonably be anticipated. HUD also estimates that the special affordable market averaged 28 percent between 1999 and 2002.

b. Past Performance of the GSEs under the Special Affordable Housing Goal

As discussed above, a number of changes in goal-counting procedures were adopted as part of HUD's Housing Goals 2000 final rule. Thus, it is necessary to provide information using several different measures in order to track changes in performance on the Special Affordable Housing Goal over the 1996-2003 period. These are shown in Table 8.^[17]

Based on the counting rules in effect at that time, as shown under “official performance” for 1996-2000 in Table 8, Special Affordable Housing Goal performance for Fannie Mae generally fluctuated in the range between 14 and 17 percent over the 1996-1999 period, before rising to a peak of 19.2 percent in 2000. Freddie Mac's performance started at a lower level, but then increased in several steps, from 14-16 percent in 1996-1998 to 17.2 percent in 1999, and to a record level of 20.7 percent in 2000. That was the only year prior to 2001 in which Freddie Mac's performance exceeded Fannie Mae's performance on the Special Affordable 9Housing Goal.

Based on counting rules in effect for 2001-2003, as shown under “official performance” in Table 8, Special Affordable Housing Goal performance for Fannie Mae was 21.6 percent in 2001, 21.4 percent in 2002, and 21.2 percent in 2003. Official performance for Freddie Mac was 22.6 percent in 2001, 20.4 percent in 2002, and 21.4 percent in 2003.

Immediately beneath the official Special Affordable Housing Goal performance percentages in Table 8 are figures showing the GSEs' special affordable purchase percentages on a consistent basis for the entire 1996-2003 period. The assumptions used were the counting rules established in HUD's Housing Goals 2000 final rule, except that bonus points and the Freddie Mac TAF (which were terminated at the end of 2003) are not applied. These are termed the “2001-2003 baseline” assumptions. In terms of this measure, both Fannie Mae and Freddie Mac's special affordable performance reached its maximum in 2000 (Fannie Mae at 21.4, percent and Freddie Mac at 21.0 percent) before declining somewhat in 2001, and then declining further in 2002 and 2003. Both GSEs' baseline performance in 2003 exceeded the level attained in 1999.

Overall, both GSEs' performance exceeded HUD's Special Affordable Housing Goals by significant margins in 1996-1999, and by wide margins in 2000. New, higher Goals were established for 2001-2003, and despite somewhat lower performance than the level attained in 2000 (largely due to the 2001-2003 refinance wave, as discussed under the Low- and Moderate-Income Housing Goal), both GSEs' performance exceeded the new Goal levels in 2001-2003.

The Special Affordable Housing Goal is designed, in part, to ensure that the GSEs maintain a consistent focus on serving the low- and very low-income portion of the housing market where housing needs are greatest. Appendices A and C to this rule use HMDA data and GSE loan-level data for home purchase mortgages on single-family owner-occupied properties in metropolitan areas to compare the GSEs' performance in special affordable lending to the performance of depositories and other lenders in the conventional conforming market. There are two main findings with respect to the special affordable category.

First, Freddie Mac and Fannie Mae have historically lagged depositories and the overall market in providing mortgage funds for special affordable borrowers over periods, such as 1993-2003, 1996-2003, and 1999-2003. Between 1993 and 2003, 12.2 percent of Freddie Mac's mortgage purchases were for special affordable borrowers, 13.3 percent of Fannie Mae's purchases, 15.4 percent of loans originated by depositories, and 15.5 percent of loans originated in the conventional conforming market (without estimated B&C loans). During the period between 1999 and 2003, the GSEs' performance was approximately 90 percent of the market'special affordable loans accounted for 15.1 percent of Fannie Mae's purchases, 14.5 percent of Freddie Mac's purchases, and 16.2 percent of loans originated in the conforming market. (See Table 9, which is based on 2000 Census geography.)

Second, while both GSEs have improved their performance over the past few years, Fannie Mae has made more progress than Freddie Mac in erasing its gap with the market. During 2003, the special affordable share of Fannie Mae's purchases was 17.7 percent, which was above the market share of 16.8 percent. In 2003, the special affordable share of Freddie Mac's purchases was 16.2 percent.

Section G in Appendix A to this rule discusses the role of the GSEs both in the overall special affordable market and in the different segments (single-family owner, single-family rental, and multifamily rental) of the special affordable market. The GSEs' special affordable purchases accounted for 41 percent of all special affordable owner and rental units that were financed in the conventional conforming market between 1999 and 2002. The GSEs' 41-percent share of the special affordable market was below their 55-percent share of the overall market. Even in the owner market, where the GSEs account for 61 percent of the market, their share of the special affordable market was only 52 percent. As noted above, Fannie Mae led the primary market in funding special affordable home loans during 2003. On the other hand, Freddie Mac continued to lag that market in 2003. The data indicate that there is room for Freddie Mac to improve its performance in purchasing affordable home loans at the lower-income end of the market.

The rental market (including both 1-to 4-family rental properties and multifamily rental properties) is especially important in the establishment of the Special Affordable Housing Goal for Fannie Mae and Freddie Mac because of the relatively high percentage of rental units meeting the Special Affordable Housing Goal. For example, between 1999 and 2002, 51 percent of units financed by Fannie Mae's rental mortgage purchases met the Special Affordable Housing Goal, representing 46 percent of units counted toward the Special Affordable Housing Goal, during a period when rental units represented only 18 percent of its total purchase volume. For Freddie Mac, 50 percent of units financed by rental mortgage purchases met the Special Affordable Housing Goal, representing 41 percent of units counted toward the Special Affordable Housing Goal, during a period when rental units represented only 16 percent of its total purchase volume.

c. Special Affordable Home Purchase Subgoal

The Department believes the GSEs can play a leadership role in the special affordable market generally, and the home purchase special affordable market in particular. Thus, the Department is establishing a Subgoal of 17 percent for each GSE's purchases of home purchase mortgages for special affordable housing located in metropolitan areas for 2005 and 2006, rising to 18 percent in 2007 and 2008.

The purpose of this Subgoal is to encourage the GSEs to improve their purchases of home purchase mortgages on special affordable housing, thus expanding homeownership opportunities for very-low-income borrowers and low-income borrowers in low-income areas, including minority first-time homebuyers who are expected to enter the housing market over the next few years. Table 9 provides information needed to compare the special affordable subgoal targets with past market and GSE performance.

Special Affordable Subgoals Compared With Market. The 17-percent subgoal for the first year (2005) is approximately one percentage point above the 1999-2003, 2001-2003, and 2002-2003 average market performance. The 17-percent subgoal is at the previous peak market performance (the 1999, 2000, and 2003 markets without B&C loans were about 17 percent) or slightly below the previous peak market performance (based on 2003 market without both B&C and small loans). The 18-percent subgoal for 2007 and 2008 would add one percentage point to these figures. Thus, the 18-percent subgoal is approximately two percentage points above the 1999-2003, 2001-2003, and 2002-2003 average market performance of approximately 16 percent. The 18-percent subgoal is one percentage point above the previous peak market performance (the 1999, 2000, and 2003 markets without B&C loans were about 17 percent) or 1.5 percentage points above the previous peak market performance based on the 2003 market without both B&C and small loans.

Special Affordable Subgoals Compared With Past Freddie Mac Performance. To reach the 17-percent 2005 subgoal, Freddie Mac would have to improve its performance by 1.9 percentage points over its 2001-2003 average special affordable performance of 15.1 percent, by 1.3 percentage points over its 2002-2003 average special affordable performance of 15.7 percent, and by 0.8 percent over its previous peak performance of 16.2 percent in 2003. To reach the 18-percent subgoal, Freddie Mac would have to improve its performance by 2.9 percentage points over its 2001-2003 average special affordable performance, 2.3 percent over its 2002-2003 average performance, and by about 1.8 percent over its previous peak performance.

Special Affordable Subgoals Compared With Past Fannie Mae Performance. To reach the 17-percent 2005 subgoal, Fannie Mae would have to improve its performance by 0.9 percentage points over its 2001-2003 average special affordable performance of 16.1 percent; Fannie Mae would essentially meet the 17-percent subgoal based on its 2002-2003 average special affordable performance of 16.8 percent and would surpass the 17-percent subgoal based on its peak special affordable performance of 17.7 percent in 2003. To reach the 18-percent subgoal, Fannie Mae would have to improve its performance by 1.9 percent over its 2001-2003 average performance and by 1.2 percentage points over its 2002-2003 average performance; Fannie Mae would meet the 18-percent subgoal based on its peak performance of 17.7 percent in 2003.

As with the low-mod and underserved areas subgoals, the special affordable subgoal targets will be more challenging for Freddie Mac than Fannie Mae. But, as with other goals, the type of improvement needed to meet the new special affordable subgoal targets was demonstrated by Fannie Mae during 2001-2003, as Fannie Mae increased its special affordable purchases from 13.4 percent of its single-family-owner business in 2000, to 15.8 percent in 2002, to 17.7 percent in 2003, as shown in Table 9. This subgoal is designed to encourage Fannie Mae and Freddie Mac to lead the special affordable market. As noted earlier, the approach taken is for the GSEs to obtain their leadership position by staged increases in the subgoals to enable the GSEs to gain experience as they improve and move toward the new higher subgoal targets.

The section above on considerations in establishing the Low- and Moderate-Income Home Purchase Subgoal and Section D of Appendix C to this rule further discuss reasons why the Department set the Subgoal for special affordable loans.

Both Fannie Mae and Freddie Mac questioned HUD's authority under FHEFSSA to establish any subgoals within the Special Affordable Housing Goal. The GSEs noted that both sections establishing the Low- and Moderate-Income and the Underserved Areas Housing Goals include language that HUD “may establish separate specific subgoals within the goal under this section and such subgoals shall not be enforceable * * * .” No such language appears in the section establishing the Special Affordable Housing Goal. The GSEs asserted that this omission is an indication that Congress intended to prohibit HUD from establishing any subgoals within the Special Affordable Housing Goal.

HUD has also considered the GSEs' claim that HUD lacks the statutory authority to impose any subgoals within the Special Affordable Housing Goal. These same arguments were presented by the GSEs during HUD's 1995 rulemaking establishing the housing goals. (See Housing Goals 1995 proposed rule published on February 16, 1995 at 60 FR 9154, and the final rule published on December 1, 1995 at 60 FR 1846.)

At that time, HUD stated that the absence of a similar subgoal provision under the Special Affordable Housing Goal section “is not an indication that subgoals or subcategories within the overall goal are prohibited; rather, such omission indicates that to the extent that subgoals or subcategories are promulgated for the Special Affordable Housing Goal, no bar exists to enforcing them.” (60 FR 61860.) The 1995 Housing Goals final rule established an enforceable subgoal for multifamily mortgages within the Special Affordable Housing Goal; this subgoal has been in place each year since then. This final rule does not change this longstanding agency interpretation.

d. Special Affordable Housing Goal: Multifamily Subgoals

Based on the GSEs' past performance on the Special Affordable Multifamily Subgoals, and on the outlook for the multifamily mortgage market, HUD proposed that these Subgoals be retained for the 2005-2008 period.

Unlike the overall Goals, which are expressed in terms of minimum Goal-qualifying percentages of total units financed, these Subgoals for 2001-2003 and in prior years have been expressed in terms of minimum dollar volumes of Goal-qualifying multifamily mortgage purchases. Specifically, each GSE's special affordable multifamily Subgoal is currently equal to 1.0 percent of its average total (single-family plus multifamily) mortgage volume over the 1997-1999 period. Under the proposal, the GSEs' purchases of mortgages financing dwelling units in multifamily housing for calendar years 2005-2008 will be 1.0 percent of the GSEs' average annual dollar volume of mortgage purchases in the calendar years 2000, 2001, and 2002. The proposal would increase the subgoal levels by roughly 90 percent compared to their current levels. Specifically, Fannie Mae's total eligible multifamily mortgage purchase volume increased from $4.6 billion in 1993 to $12.5 billion in 1998, and then jumped sharply to $18.7 billion in 2001, $18.3 billion in 2002, and $33.3 billion in 2003. As shown in Table 8, special affordable multifamily mortgage purchases followed a similar path, rising from $1.7 billion in 1993 to $3.5 billion in 1998 and $4.1 billion in 1999, and also jumping sharply to $7.4 billion in 2001, $7.6 billion in 2002, and $12.2 billion in 2003. As a result of its strong performance, Fannie Mae's purchases have been at least twice its minimum subgoal in every year since 1997—247 percent of the Subgoal in that year, 274 percent in 1998, 315 percent in 1999, 294 percent in 2000, and, under the new Subgoal level, 258 percent in 2001, 266 percent in 2002, and 426 percent in 2003.

Freddie Mac's total eligible multifamily mortgage purchase volume increased even more sharply, from $0.2 billion in 1993 to $6.6 billion in 1998, and then jumped further to $11.8 billion in 2001, $18.3 billion in 2002, and $21.5 billion in 2003. As shown in Table 8, special affordable multifamily mortgage purchases followed a similar path, rising from $0.1 billion in 1993 to $2.7 billion in 1998, and also jumping sharply to $4.6 billion in 2001, $5.2 billion in 2002, and $8.8 billion in 2003. As a result of its strong performance, Freddie Mac's purchases have also been at least twice its minimum Subgoal in every year since 1998—272 percent of the Subgoal in that year, 228 percent in 1999, 242 percent in 2000, and, under the new Subgoal level, 220 percent in 2001, 247 percent in 2002, and 417 percent in 2003.

The Special Affordable Multifamily Subgoals set forth in this final rule are reasonable and appropriate based on the Department's analysis of this market. The Department's decision to retain these Subgoals is based on HUD's analysis, which indicates that multifamily housing still serves the housing needs of lower-income families and families in low-income areas to a greater extent than single-family housing. By retaining the Special Affordable Multifamily Subgoal, the Department ensures that the GSEs continue their activity in this market, and that they achieve at least a minimum level of special affordable multifamily mortgage purchases that are affordable to lower-income families.

e. Summary of Comments

Comments regarding the Special Affordable Goal were received from numerous public advocacy groups and one trade association; however, only one public advocacy group commented on the level of the goal. The commenting group recommended that the 2004 Special Affordable Goal be maintained for the years 2005-2008.

No comments specific to the Special Affordable Home Purchase Subgoal were received from the public. Fannie Mae provided an analysis as part of its comments that illustrated, for the years 1999 through 2002, that the market did not perform up to the level of HUD's proposed Special Affordable Home Purchase Subgoal.

Regarding the Multifamily Special Affordable Subgoal, neither GSE objected to HUD's proposed subgoal levels for 2005-2008. One trade organization suggested that the subgoal has outlived its original purpose and should be discontinued. This organization stated that the subgoal was established to induce the GSEs to purchase multifamily loans at a time when heavy credit losses had caused them to back away from this market, and that the situation had changed greatly since then. The organization stated that the overall goals now provided sufficient incentive for the GSEs to focus on multifamily mortgage purchases. One multifamily lender expressed concern that increasing the Multifamily Special Affordable Subgoal will push the GSEs to extend credit to unqualified borrowers with poor quality properties that should not be eligible for long-term, low-cost financing. However, other commenters, including multiple public advocacy groups and a local government official, recommended that HUD increase the level of this subgoal. Several commenters specifically recommended that HUD set this subgoal between 2.5 percent and 3 percent of the GSEs' purchases in preceding years. They noted that the GSEs have far exceeded the subgoal levels in recent years and said that a higher subgoal level is needed to promote additional multifamily lending.

f. HUD's Determination

HUD concludes that the Special Affordable Housing Goal established in this final rule is reasonable and appropriate having considered the factors set forth in FHEFSSA. Current examination of the gaps in the mortgage markets, along with the estimated size of the market available to the GSEs, demonstrates that the number of mortgages secured by special affordable housing is more than sufficient for the GSEs to achieve the new goal.

Therefore, having considered all the statutory factors including housing needs, projected economic and demographic conditions, the GSEs' past performance, the size of the market serving low- and moderate-income families, and the GSEs' ability to lead the market while maintaining a sound financial condition, HUD has determined that the Special Affordable Housing Goal will be 22 percent for 2005, 23 percent for 2006, 25 percent for 2007, and 27 percent for 2008. This reflects a reduction in the upper end of the market share range from 28 percent to 27 percent since HUD's publication of its proposed rule, resulting from changes in estimating market share as described at the end of section 3.a, above, and in Section H of Appendix D to this rule.

The reasons for increasing the Special Affordable Housing Goal are discussed above in this preamble. Since the GSEs have historically lagged the primary market in purchasing loans on owner and rental properties that qualify as special affordable, they have ample room to improve their performance in that market. The GSEs' mortgage purchases between 1999 and 2002 accounted for 55 percent of the total (single-family and multifamily) conforming mortgage market, but they accounted for only 41 percent of the special affordable market. A wide variety of quantitative and qualitative indicators demonstrate that the GSEs have the expertise, resources, and financial strength to improve their special affordable lending performance and to close their gap with the market.

Further, the Department is establishing a Subgoal of 17 percent for each GSE's acquisitions of home purchase mortgages for special affordable housing in 2005 and 2006, rising to 18 percent in 2007 and 2008. The special affordable home purchase subgoal will ensure that Freddie Mac improves its performance enough not only to close its current gap with the primary market but also to place itself in a leadership position. The subgoal will also encourage Fannie Mae to improve further its current market-leading performance. A wide variety of quantitative and qualitative indicators demonstrate that the GSEs have the expertise, resources, and financial strength to improve their special affordable lending performance, including lending for home purchases for special affordable housing, and to achieve the levels of the subgoals being established.

Finally, the Department is establishing each GSE's Special Affordable Multifamily Subgoal at 1.0 percent of its average annual dollar volume of total (single-family and multifamily) mortgage purchases over the 2000-2002 period. In dollar terms, the level of the subgoal is $5.49 billion per year in special affordable multifamily mortgage purchases for Fannie Mae and $3.92 billion per year in special affordable multifamily mortgage purchases for Freddie Mac. These Subgoals would be less than the actual special affordable multifamily mortgage purchase volume in 2001-2003 for both GSEs. Thus, the Department believes that they would be feasible for the 2005-2008 period.

HUD believes that the proposed increase in the dollar level of the Special Affordable Multifamily Subgoal balances the need to promote GSE activity in this segment with the need to provide some protection in the event of a decline in overall mortgage market activity. Because this goal is set as a dollar amount rather than as a share of business, overall declines in residential mortgage lending would make this goal harder to achieve. Setting the subgoal level based on the GSEs' record multifamily loan purchases during 2000-2002 sets an appropriately high level for the next several years, in the Department's view. In recent years Fannie Mae and Freddie Mac have each purchased multifamily mortgages in at least twice the subgoal amount. The increase in that subgoal dollar level should serve to provide a more meaningful floor to the level of multifamily lending during the 2005-2008 period.

8. Missing Data/No-Doc Loans

Overview. Accurate measurement of the GSEs' performance under the three Housing Goals depends on the completeness of data on borrower income (or, in the case of non-owner-occupied units, the rent) and property location. With respect to property location data, there was a less than one percent incidence of missing or incomplete geographical data between 2000 and 2002 for mortgages purchased by the GSEs. The incidence of missing borrower income data has been greater—on the order of several percent each year.

One reason for the increase in missing income data is the market's recent increased use of mortgages, commonly called low documentation (Low Doc) and no documentation (No Doc) loans. These loans do not require the borrower to provide income information. In some cases, the borrower provides information on assets but not income because of circumstances that make assets easier to document. In other instances, mortgages are originated entirely on the basis of a credit report, property appraisal, and cash for the downpayment. These mortgages typically require relatively large downpayments and may also require a higher interest rate than fully documented mortgages.

The Housing Goals 2000 Final Rule provided that the GSEs may exclude from the denominator owner-occupied units which lack mortgagor income data and which are located in low- or moderate-income census tracts, i.e., tracts whose median income is no greater than the median income of the metropolitan area, or for properties located outside of metropolitan areas, the larger of the median incomes of the county or the statewide non-metropolitan area (see 24 CFR 81.15(d)).^[18]

In view of the increasing use of loans made without obtaining income information from the borrower, there is a question whether HUD's existing counting rules for missing-data situations are adequately reliable and create no more than a negligible statistical bias in the GSEs' Housing Goals performance figures relative to the values that they would have if complete income data could be obtained, and whether a more precise method for imputing incomes could be employed. For this reason, HUD requested comments from the public about the desirability and feasibility of implementing a standard econometrically based method for imputing the income distribution of mortgages purchased by each GSE that lack income data, based on known characteristics of the loan and the census tract.

Summary of Comments. Fannie Mae supported expanding affordability estimation to single-family rental and owner-occupied goal performance calculations and favored a more complex econometrically based affordability estimation methodology. For owner-occupied units Fannie Mae suggested a method based on the probability of mortgages/units qualifying for a goal based on census tract location. Fannie Mae stated that the multifamily affordability estimation methodology could also be applied to single-family rental units. Fannie Mae commented that if HUD were to adopt an econometrically based methodology, no limit should be placed on its implementation. With the current methodology, Fannie Mae requested that the limit for rental units be increased to 10 percent of total rental unit acquisitions.

Freddie Mac commented that HUD should adopt a simpler approach to missing data. For example, HUD should allow the GSEs to remove units with missing incomes from the calculation of the housing goals. Freddie Mac reasoned that the market numbers used in establishing the Housing Goals omit missing data and that omitting missing data from a GSE's performance would be consistent. Also, Freddie Mac stated that it historically has had a lower missing data rate than the market and that it has sufficient business related incentives to reduce missing data. Freddie Mac commented that any limits on adjustments for missing data should be related to overall missing data rates in the market, estimation parameters should be available at the beginning of the performance year, and estimation procedures should be simple and straightforward to implement.

Several other organizations endorsed a standardized procedure for estimating affordability for those units missing rent or income data, including an econometrically based methodology. Two commenters stated that HUD should require only actual data for determining whether a unit is affordable or not. In addition, some commenters strongly recommended that HUD disallow goals credit for all no-documentation subprime loans because such loans are likely to be predatory.

HUD's Determination. Having considered the comments received, HUD has determined that permitting some level of estimation for affordability data is reasonable and consistent with statutory intent that the GSEs serve the affordable housing needs of families even if actual data are not available. With regard to some commenters' objections that HUD should not permit the use of estimated data for—or even allow goals credit for—any loans that were underwritten for approval without borrower income data due to the potential for these loans to have predatory features, the Department does not find that these loans are inherently predatory in nature. Also, both GSEs have publicly announced that they will not finance any loans with predatory features, and the Department expects that they will continue to vigorously enforce these policies. Accordingly, this final rule implements several changes to the treatment of missing data. The first change amends § 81.15(d) of the General Requirements to provide an alternative treatment for single-family owner-occupied units where the mortgagor's income is missing. As provided in § 81.15(d), the GSEs may continue to exclude such units from the denominator as well as the numerator when they are located in census tracts with median income less than or equal to area median income according to the most recent census, up to a ceiling of one percent of total eligible units. Purchases in excess of the ceiling will be included in the denominator and excluded from the numerator if they are missing data.

However, in lieu of using this procedure, HUD is making available to the GSEs in § 81.15(d) an alternative method for missing income treatment that provides the GSEs with the ability to apply a HUD-approved affordability estimation methodology to all single-family owner-occupied units with missing borrower income data up to a specified maximum. This alternative provision specifies an approach that recognizes the distribution of borrower incomes within census tracts in determining how to treat loans with missing income data. Goal-qualifying units, by census tract, are estimated by multiplying the number of single-family owner-occupied units with missing borrower income information in properties securing mortgages purchased by the GSE, by the percentage of all single-family owner-occupied units from originations that would count toward achievement of the goal, as determined by HUD based on the most recent HMDA data available, for each census tract where the GSE acquired mortgage units. In establishing the maximum number of units where borrower income may be estimated under this alternative provision, HUD will apply two factors. The first of these is the rate of missing borrower income data for each census tract. This is calculated using HMDA data for the most recent years for which comparable data are available. The second factor is the number of single-family owner-occupied units purchased by a GSE during the performance year, by census tract. The maximum is calculated by multiplying the HMDA percentage of missing income data by the number of units that a GSE purchased in each tract. This number is summed up for all tracts to obtain the overall nationwide maximum for that GSE. HUD will provide each GSE with a dataset containing applicable tract-based HMDA missing income rates prior to the start of each year. The GSEs may choose which provision of § 81.15(d) they will use in any year. However, they may not combine the options available under this provision. If the maximum on missing single-family owner-occupied unit incomes is exceeded, the estimated goal-qualifying units will be adjusted by the ratio of the maximum amount divided by the total number of units with missing income information. Under each provision of § 81.15(d), units in excess of the specified maximum as well as units where affordability information is not available will remain in the denominator when calculating goal performance.

HUD is also in this final rule revising § 81.15(e)(6) to change the current maximum on the use of HUD-approved multifamily rent estimation data from 5 percent to 10 percent. In analyzing the GSEs' multifamily purchases for the past several years, HUD has determined that this change is statistically insignificant and will serve to promote further the financing of rental units that would otherwise be eligible for credit under the Housing Goals. In this final rule, HUD is also specifying a methodology that may be used to estimate affordability data for multifamily properties with missing rent data. This methodology is the same methodology that has been used in past years to estimate affordability data for multifamily properties with missing rent data.

With regard to single-family one-to-four unit rental properties financed with loans that are missing affordability data, the Department finds that a lack of data should not act as a disincentive for the GSEs to serve markets that historically are important sources of affordable housing. Under HUD's 2000 Rule, § 81.15(e)(6)(ii) permits the GSEs to exclude these units from both the numerator and the denominator when neither income nor rental data are available. While this provision does not penalize the GSEs for financing these properties by requiring that they be counted in the denominator towards goal calculation, it also does not allow them to obtain Housing Goals credit for financing mortgages that tend disproportionately to serve affordable housing. In this final rule, HUD is retaining the exclusion provision at § 81.15(e)(6)(ii) but is also adding an alternative provision that will permit the use of the same estimation methodology now used for multifamily loans with missing rent data. However, HUD is imposing separate maximum rates for the new provision as follows: a 5 percent maximum on unseasoned single-family rental units originated in the current year and a 20 percent maximum for seasoned loan units, that is, for loans that were originated more than 365 days prior to the date of acquisition by the GSE. HUD recognizes the greater difficulty of obtaining rent information on units from mortgages originated a year or more prior to acquisition by the GSE. Therefore, HUD is allowing the higher maximum on affordability estimation for these units. As with the estimating provisions permitted under § 81.15(d), the GSEs may use only one of the provisions permitted under § 81.15(e)(6)(ii) in any year.

In addition to the changes described herein, HUD is adding a provision to §§ 81.15(d)(2)(i), 81.15(e)(6)(i) and (ii) that permits the use of such other data source or methodology as may be approved by HUD. HUD is also clarifying that owner occupied units that exceed the maximum established under § 81.15(d)(2) for using any estimation methodology will remain in the denominator of the respective goal calculation.

9. Double Counting of Seasoned Mortgages

In addition to the preceding changes being made at this final rule stage, HUD is making a technical change to § 81.16(c)(6) for purposes of clarity. Paragraph (c)(6) addresses the treatment of seasoned mortgages. The paragraph, as currently codified, is a long one-sentence paragraph. HUD believes that dividing this paragraph into two subparagraphs would improve comprehensibility and clarity. This change is intended to clarify the restriction on double counting of seasoned mortgages in § 81.16(c)(6), i.e., the restriction that prohibits the counting of a GSE's purchase of a seasoned mortgage toward a goal where such mortgage has already been counted by the GSE toward the goal. This change makes clear that the restriction applies to all seasoned mortgages, regardless of whether any other counting rules under § 81.16(c) also apply. Section 81.16(c)(6) in this final rule reflects this technical change.

10. Bulk Purchases/Counting of Seasoned Loans

Overview. In its May 3, 2004, proposed rule, HUD sought comment on whether its current definition of a “mortgage purchase” should be revised to ensure that transactions, especially large transactions, are appropriately counted under the law and in accordance with the purposes of FHEFSSA and the GSEs' charter acts. HUD also sought comment on whether it should amend its counting rules at 24 CFR 81.15 and 81.16 to ensure that the GSEs' large-scale transactions further the requirements and purposes of the Housing Goals.

For example, HUD asked if commenters believe the current counting rules are specific enough to determine which seasoned mortgage transactions, including large-scale transactions, are substantially equivalent to mortgage purchases. HUD sought these comments primarily in response to certain large-scale transactions of seasoned loans undertaken by both GSEs in late 2003 for the purpose of meeting the 2003 Housing Goals. HUD questioned whether such transactions furthered the purposes of FHEFSSA, especially since the transactions, including a transaction between Freddie Mac and Washington Mutual Bank (WaMu), contained an option for dissolution in the following year. HUD sought public comment on its counting rules and definitions to ascertain the effect of the GSEs' bulk purchases, including those with special terms or conditions, on the market and on affordable housing.

Summary of Comments. HUD received several suggestions for revising its current definitions and counting rules. A trade association commented that HUD should specify the definition of mortgage purchase so as not to count transactions that are goals-oriented in form but not in substance. Some organizations commented that seasoned loans should be excluded from counting towards the goals altogether because they do not directly fund new housing supply. Likewise, some commenters believed that these transactions are contrary to the Charter requirement that the GSEs provide assistance to the secondary market on an on-going basis.

One policy group asked that HUD exclude loans with recourse clauses because these purchases do not alleviate risk from the market. Other commenters took the opportunity to request that the definitions and counting rules more closely match CRA loan definitions. These commenters did not suggest specific regulatory language for the definitions.

HUD also received comments that supported counting bulk purchases that occur late in the year towards the goals. One trade association described the efficiencies gained from large-scale transactions. For example, the market for multifamily units is large and fragmented, and seasoned portfolio transactions are an efficient means for the GSEs to acquire smaller loans in the under 50-unit segment of the market. Some commenters cautioned that changing the definition of mortgage purchase or the counting rules to clarify the treatment of large-scale seasoned mortgage transactions could have negative unintended consequences.

The GSEs responded to this issue with detailed comments. Fannie Mae stated that every mortgage purchase, whether executed through flow, large or seasoned transactions, contributes to its housing mission, and therefore, HUD should not change the qualification of mortgage purchases either for the size of the transaction or for the amount of seasoning involved. Fannie Mae also stated that large-scale mortgage purchases lower transactions costs for both the buyers and sellers of mortgages. Some lenders offer to sell the GSEs mortgages on a flow basis, but others prefer to bundle mortgages together and sell to the GSEs from their portfolios. Bulk transactions also serve the business needs of lenders who do not have a direct relationship with Fannie Mae. Fannie Mae said that two-thirds of its bulk purchases between 2001 and 2003 were not for seasoned loans. Fannie Mae characterized the purchase of seasoned loans as an important component of the liquidity of current mortgages. Knowing that there is a ready market allows financial institutions to hold some of their assets in the form of mortgages, and affords them the opportunity to sell these mortgages later to manage liquidity, improve profitability, strengthen their capital position, and manage certain risks.

In addition to the market benefits of seasoned mortgages, Fannie Mae also discussed the practical relationship of seasoned loan treatment and goals performance. The GSEs need bulk purchases of seasoned loans to meet the goals in years when the mix of business in the primary market deviates from the business mix anticipated at the time the goals were set. Fannie Mae pointed out that HUD cited late-year purchases of seasoned loans in the proposed rule as a useful method to meet the goals when market conditions change unexpectedly. Fannie Mae also discussed the attributes of dissolvable securities, stating that lenders sometimes request the option to dissolve securities swapped with the GSEs. Fannie Mae said that dissolution options are common terms in the marketplace because dissolution options grant lenders greater control over their balance sheets, capital position, and other financial concerns. Fannie Mae indicated that lenders request these options because they obtain more favorable rates and can make more loans.

Freddie Mac made many of the same points about bulk purchases of seasoned purchases as Fannie Mae and also discussed its recent bulk transaction with WaMu. For example, Freddie Mac commented that bulk purchases and dissolution options are common industry practices. Freddie Mac also stated that counting seasoned loans increased the value and liquidity of current loans. Knowledge that the GSEs stand ready to purchase mortgages under all market conditions gives other investors greater confidence because they have a viable exit strategy when providing funds to the real estate market.

Freddie Mac indicated that bulk purchases are an essential means of achieving the goals when market conditions take an unexpected turn, such as the conditions leading to its transaction with WaMu in 2003. Freddie Mac pointed out that, unlike FHA, which can manage its business to the cap on insurance commitments set annually by Congress, Freddie Mac instead must respond to a dynamic market in which the nature and magnitude of loan originations are volatile. In real time, it is extremely difficult to predict the volume and “mix” or proportion of goals-eligible mortgages those markets will produce. Market refinance forecasts for 2003 by Economy.com and Freddie Mac were off by over $2 trillion. Large transactions of mortgage purchases are essential because forecasts are not precise.

With respect to its transaction in 2003 with WaMu, Freddie Mac stated that it engaged in this transaction because HUD took a number of steps to strongly encourage the GSEs to participate in the small 5-50 multifamily mortgage market, including bonus points. The GSEs can only purchase on terms that sellers are willing to accept. Freddie Mac further stated that goals that force the GSEs to stretch their business mix in uncertain market conditions must eventually cause the GSEs to value some mortgages more than sellers do. Under these conditions, sellers will negotiate for more favorable terms. Freddie Mac stated that the seller “put” option in the WaMu transaction and a similar transaction with Citibank exemplify pro-seller terms and that these transactions advance the GSE's regulatory purposes as well as meet the letter of the law.

In response to concerns about the options included in the swap, Freddie Mac stated that “it is the GSE's affordable housing goal requirements, among other things, that give the sellers the negotiating power to obtain such options.” Both Fannie Mae and Freddie Mac concluded that HUD's definition of a mortgage purchase and the counting rules should not be changed.

HUD's Determination. HUD considered the comments received, with particular focus on the GSEs' comments regarding transactions that include dissolution options. HUD is concerned that transactions of this type, which both GSEs undertook in 2003 to achieve their affordable housing goals, are not fully consistent with the purposes of FHEFSSA, which are to award goals credit for mortgage purchases that increase market liquidity for affordable housing. When a seller can exercise its option to reverse or unwind a transaction and take back the mortgages within a specified time period, the transaction appears temporary in nature, and the liquidity that might result from the transaction also appears transitory.

The drafters of FHEFSSA intended that the GSEs provide liquidity for affordable housing where such liquidity would otherwise not exist or where it would be less reliable. HUD is aware that even short-term liquidity, as may occur with dissolution options, can be of value to mortgage sellers, especially for balance sheet management or other purposes, but sellers seeking such options are generally not constrained in locating short-term liquidity solutions, especially when these solutions are backed by seasoned mortgage loans.

Further, HUD believes that placing no constraints on goals eligibility for transactions with dissolution options would have the effect of encouraging transactions that are so short-term as to be dissolvable almost immediately after they have been counted towards the housing goals. Such an outcome is clearly at odds with FHEFSSA.

Therefore, HUD has determined to amend its counting rules to provide that for units acquired in transactions with seller dissolution options to count toward the housing goals, such options must provide for a lockout period that prohibits the exercise of the dissolution option for at least one year from the date on which the transaction was entered into and the transaction cannot be dissolved during the one-year period. The Secretary may grant an exception to the minimum lockout period, in response to a written request from a GSE, if the Secretary determines that the transaction furthers the GSE's statutory purposes and the purposes of FHEFSSA. Where a mortgage purchase involving a seller dissolution option has been counted toward the housing goals under a transaction subject to this provision, the transaction may not be dissolved (either by the exercise of the seller dissolution option, or by separate agreement entered into by the GSE and the seller) during the one-year minimum lockout period. If the seller of the mortgages and the GSE dissolve the transaction before that time, the transaction may no longer be counted toward the housing goals and the GSE's performance must be adjusted in accordance with this rule.

The Department defines seller dissolution option as an option for a seller of mortgages to the GSEs to dissolve or otherwise cancel a mortgage purchase agreement or loan sale. The Department, however, wishes to fully distinguish the arrangements established in these seller dissolution options from other types of agreements involving repurchases of securities or mortgages that involve the GSEs. For example, the GSE, as seller of a security, may agree to repurchase, or buy back, a previously sold mortgage-backed security on a negotiated basis from the holder of the security. HUD's regulation does not address that practice. Likewise, it does not address arrangements whereby a mortgage lender agrees to repurchase or replace a mortgage upon demand of the GSE if the mortgage defaults. The provision also does not apply to repurchase and resale agreements where the GSE is the purchaser of the security. Rather, the transactions addressed by HUD's regulation provide, as a term of the transaction, the mortgage lender/seller—and not the GSE—with the option of dissolving the transaction and having the mortgages returned to the mortgage lender/seller.

HUD believes the one-year lockout period will prevent potential misuse of these transactions but will still allow sellers of mortgages to manage their portfolios in the medium and long term. The limit on dissolution options applies to all transactions because it is the potential for misuse, not the size of the transaction that could conflict with FHEFSSA. HUD will continue to monitor the GSEs' use of dissolution options to ensure that the one-year minimum lockout requirement is accomplishing its intended purpose. If there is a question about whether a particular transaction complies with the one-year minimum lockout requirement, HUD expects that the GSE will seek clarification from HUD regarding the appropriate treatment of that transaction under the counting rules.

With regard to modifying its definition of a “mortgage purchase,” HUD has determined that defining mortgage purchases in terms of market effects would be cumbersome. The definition would have to be broad enough to encompass all of the statutory purposes, including market liquidity and market stability, and still narrow enough to exclude transactions that are legitimate in form but not in substance.

Similarly, while some commenters suggested that HUD exclude seasoned mortgages from its definition or that HUD impose a credit risk threshold for awarding goals credit, HUD believes that these measures could have unintended consequences that could potentially harm market liquidity for affordable housing. For example, HUD has encouraged the GSEs to buy seasoned portfolios of CRA loans as an important source of liquidity for these loans.

11. Responses to Other Issues Raised by Commenters Relating to the Housing Goals

a. Feasibility Determinations

Overview. Section 1336(b) of FHEFSSA, together with HUD's current regulations, provides a process for determining that one or more goal levels are infeasible. This process may be initiated either by HUD or by a GSE; nothing in FHEFSSA or in HUD's regulations limits a GSE's ability to request HUD to examine whether a particular goal may be infeasible. If HUD determines that a GSE has failed to meet a housing goal, or that there is a substantial probability that a GSE will fail to do so, HUD must notify the GSE and provide an opportunity for the GSE to respond. HUD must then determine whether or not the goal was feasible. If HUD determines that the goal was infeasible, then no further HUD action to enforce the goal is authorized.

HUD's proposed rule did not make any changes to the process for determining whether a goal was or was not feasible. However, HUD still received comments from both Fannie Mae and Freddie Mac regarding those provisions.

Summary of Comments. Fannie Mae commented that “uncertainty regarding HUD's potential feasibility determination would lead Fannie Mae and Freddie Mac to engage in whatever means necessary to meet the goals, potentially resulting in market distortions.” Fannie Mae recommended that the goals be set at levels that are more likely to be seen in the marketplace, rather than at the high end of market estimates.

Freddie Mac commented that an after-the-fact finding of “infeasibility” or an adjustment to the goals would not alleviate the burden imposed by unreasonable goals. Freddie Mac noted that it is very difficult to estimate the size and composition (or “goal mix”) of the mortgage market in advance. Freddie Mac also expressed concern that an after-the-fact feasibility determination would require HUD to second-guess innumerable business decisions made by the GSEs, with no certainty as to how HUD would make such determinations. Finally, Freddie Mac stated that its reputation would suffer great harm during the time HUD considered its feasibility determination, and that this harm could not be undone.

HUD's Determination. The final rule does not make any changes to the process for determining whether a goal is infeasible for a particular year. Although HUD has never had to make a determination that a goal is infeasible, HUD believes that the process that is currently in place provides an effective framework for making a timely determination of infeasibility. If in the future it is necessary to make a determination of whether a goal is or was infeasible, HUD will make every effort to expedite the process in an effort to minimize any potential costs and uncertainty associated with the process.

b. Specification of Underserved Areas

Summary of Comments. Several commenters suggested that HUD should redefine the Underserved Areas Goal. A consensus of these commenters stated that lowering the tract income criteria from 90 (120) percent to 80 (100) percent would make the Underserved Areas Goal consistent with CRA. Several of the commenters also stated that the current definition is too broad and that lowering the tract income criteria to 80 percent or 100 percent when the minority population is greater than 50 percent (as opposed to 30 percent currently) of the tract would focus the goal on truly underserved areas. One commenter suggested including a borrower income criteria, such as less than 80 percent of area median income, in the Underserved Areas Goal to further focus the goal on the underserved.

HUD's Determination. As discussed in Appendix B to this rule, HUD has determined not to go forward with redefining the Underserved Areas Goal at this time.

c. Reconciling the CRA and the Affordable Housing Goals

Summary of Comments. Several commenters from trade associations and policy organizations suggested that HUD could more sharply focus GSE activity on low- and moderate-income homebuyers by encouraging greater purchases of CRA loans. According to these commenters, this could be accomplished by establishing a new CRA goal or by establishing CRA subgoals under each of the current Housing Goals.

The CRA requires depository institutions to help serve the credit needs of their communities and authorizes federal regulators to examine the level of lending, investment, and service that these institutions provide. Commenters noted that under section 1335 of FHEFSSA, Fannie Mae and Freddie Mac are directed to “take affirmative steps to assist insured depository institutions to meet their obligations under the CRA which shall include developing appropriate and prudent underwriting standards, business practices, repurchase requirements, pricing, fees, and procedures.” These commenters noted, however, that under FHEFSSA, the definitions for key categories of borrowers served through affordable housing goals differ from those established for borrowers served under CRA.

Under FHEFSSA, the definition for “low income” is a borrower at or below 80 percent of area median income, while for CRA purposes, the definition of “low-income” is a borrower at or below 50 percent of area median income. Similarly, the affordable housing goal definition of a “moderate income” borrower is at or below 100 percent of area median income, while for CRA purposes, “moderate income” is defined as at or below 80 percent of median area income.

Commenters pointed out that these definitional discrepancies create a mismatch between the loans made by the primary market institutions and those purchased by the GSEs to meet affordable housing goals. The result is that the GSEs can meet their goals by purchasing loans to borrowers in higher income ranges than those mandated under CRA, which may result in less liquidity available to primary mortgage market lenders to make additional low and moderate income loans.

These commenters recommended that HUD find a way to resolve the apparent contradiction between the definitions. One commenter suggested that HUD has the authority to align the affordable housing goals with the CRA definitions without additional legislation. This commenter recommended that HUD require the GSEs to report low-income loans in two categories—“low income” and “very low income”—and conform the definitions of low-income and moderate income to the CRA definitions.

Other commenters however, indicated that legislative correction would be needed to accomplish such alignment. These commenters recommended that until that time, HUD should consult with federal bank and thrift regulators to determine the CRA-eligible market share and adjust the affordable housing goals for Fannie Mae and Freddie Mac accordingly.

Several commenters recommended that HUD should consider establishing specific “CRA loan sub-goals” under the existing goals for the GSEs. One commenter suggested that HUD could create a new goal that requires the GSEs to purchase stated amounts of CRA-eligible home purchase mortgages, with low and moderate income subgoals based on the CRA measures.

HUD's Determination. After close review of this issue, HUD has determined that full harmonization between affordable housing goals and CRA definitions will require legislative action. Income brackets for the goals under FHEFSSA and under CRA are statutorily defined, and CRA definitions allow for much greater discretion by examiners to determine CRA scoring. For example, under CRA, the distinction between home improvement loans and small business loans secured by housing may not match HUD's definitions of mortgage purchases. In contrast, HUD does not use a system of examiners to determine the goals eligibility of sellers dealing with the GSEs, and comparison areas are established through regulation.

In light of these legal constraints, HUD will not make any changes to the housing goals to address CRA concerns at this time.

d. Predatory Lending

Summary of Comments. Certain commenters urged the Department to adopt predatory lending safeguards in the final rule that would prohibit Housing Goals credit for purchases of loans that included mandatory arbitration clauses or loans with prepayment penalties beyond three years towards the goals. The GSEs did not specifically mention this issue in their comments to HUD. HUD's proposed rule did not suggest changes to its existing GSE regulations that address predatory lending practices.

HUD's Determination. The Department continues to vigorously oppose specific lending practices that are predatory or abusive in nature. As stated in the 2000 rulemaking, the GSEs should seek to ensure that they do not purchase loans that actually harm borrowers and support unfair lending practices. In that rulemaking, the Department determined that the GSEs should not receive the incentive of goals credit for purchasing high cost mortgages, including mortgages with unacceptable features.

The Department is authorized under 24 CFR 81.16 to determine whether to provide full, partial, or no credit toward achievement of any of the housing goals for any transaction. The Department's existing rules contain strong safeguards against abusive lending by excluding certain types of mortgages from counting towards the affordable housing goals. These include loans with excessive fees, and prepayment penalties in certain loans.

The Department is aware that certain practices that were not enumerated in the regulations adopted in 2000, such as loans with prepayment penalties after three years and loans with mandatory arbitration clauses, often lock borrowers into disadvantageous loan products. The Department will rely on existing regulatory authorities to monitor the GSEs' performance in this area. Should the Department later determine that there is a need to specifically enumerate additional prohibited predatory practices, it will address such practices at a future time.

e. Minority Subgoals/Goals

Summary of Comments. Among the many suggestions HUD received for subgoals and bonus points, several advocacy groups recommended that HUD directly target minority mortgage purchases such as those made to Native Americans. These groups note that homeownership rates are not equal across ethnic groups. Fewer Blacks and Hispanics own their own homes than the general population. Although the GSEs have made some progress in this area, the GSEs are still less likely to serve high minority areas than other lenders. In the view of these commenters, the absence of the GSEs has led to higher borrowing costs and harsher borrowing terms for minority borrowers because they are more likely to deal with nontraditional and predatory lenders.

HUD's Determination. Under FHEFSSA, HUD does not have statutory authority to establish goals beyond those enumerated in the statute. FHEFSSA directs HUD to establish a goal for underserved areas, and HUD's goal includes census tracts with high concentrations of minority households (and with median income below a certain level) as one category of underserved area. The statute does not empower HUD to establish a goal based on the characteristics of borrowers, other than by income of borrower.

Even without an explicit subgoal, HUD believes that the goals structure will address the concerns of minority borrowers. As discussed in the introduction, minorities and immigrants are a growing percentage of homebuyers and many more aspire to home ownership. Demographics dictate that these buyers will become increasing shares of the conventional conforming market. Requiring the GSEs to lead the market will encourage them to do even more to reach out to minorities.

f. Technical Change to § 81.16(c)(7)

In addition to the preceding changes being made at this final rule stage, HUD is making a technical change to § 81.6(c)(7) to correct a cross-reference. Paragraph (c)(7) addresses the treatment of refinanced mortgages. The paragraph includes a reference to § 81.14(f), which is not related to refinanced mortgages. Section 81.16(c)(7) in this final rule is revised to correct this cross-reference.

D. Subpart I—Other Provisions

1. Overview—Verification and Enforcement To Ensure GSE Data Integrity

HUD proposed to amend § 81.102 (Independent Verification Authority) of its regulations to incorporate certain data integrity procedures designed to ensure the accuracy, completeness, and timeliness of housing goal information submitted by the GSEs to the Department. These procedures included: (1) A requirement that the GSEs provide a certification with their Annual Housing Activity Reports (AHAR) and such other reports, data submissions, and information that the Department may request in writing be certified; (2) a procedure to adjust current year-end errors, omissions, and discrepancies in data submissions to HUD; and (3) a procedure for correcting prior year overstatements of performance due to reporting errors, omissions, or discrepancies in a GSE's AHAR. HUD also restated in the proposed amendment to § 81.102 the enforcement options and remedies under FHEFSSA and HUD's regulations that could result from a determination that a GSE's data submissions, information, or reports were not current, were incomplete, or otherwise contained an untrue statement of material fact.

In addition to comments provided by the GSEs, HUD received comments from groups that included mortgage lenders, non-profit and policy advocacy organizations, and trade associations. Most commenters supported the data verification provisions of the proposed rule. However, one mortgage lender stated that the proposed certification would impose a severe burden on the GSEs and lenders. Another suggested that the data integrity process should include some leeway for unintentional mistakes so that it does not become burdensome. A trade association stated that HUD should not enact regulations that would put additional data integrity burdens on lenders. Fannie Mae and Freddie Mac provided detailed comments on each proposal. These comments are discussed more fully in the following sections.

2. Independent Verification Authority—§ 81.102(a)

As it proposed, the Department is retaining and recodifying the provisions of the current § 81.102(a) that provide that HUD may independently verify the accuracy and completeness of data, information and reports submitted by a GSE in addition to the Department's existing authority to conduct on-site verifications and performance reviews. HUD is redesignating this section, as HUD proposed, as § 81.102(a).

3. Certification—§ 81.102(b)

To ensure the highest degree of corporate accountability, and to be consistent with the customary practice of regulators of financial institutions, the Department proposed that the GSEs be required to provide a certification with their AHAR reports and such other report(s), data submission(s), or information for which HUD requests certification in writing. HUD proposed a certification that consisted of the following four parts: (1) The GSE Certifying Official has reviewed the particular AHAR, other report(s), data submission(s) or information; (2) to the best of the GSE Certifying Official's knowledge and belief, the particular AHAR, other report(s), data submission(s), or information are current, complete, and do not contain any untrue statement of a material fact; (3) to the best of the GSE Certifying Official's knowledge and belief, the AHAR or other report(s), data submission(s), and information fairly present in all material respects the GSE's performance, as required to be reported; and (4) to the best of the Certifying Official's knowledge and belief, the GSE has identified in writing any areas in which the GSE's particular AHAR, other report(s), data submission(s), or information may differ from HUD's written articulations of its counting rules including, but not limited to, the regulations under 24 CFR part 81, and any other areas of ambiguity.

Summary of Comments. Fannie Mae and Freddie Mac commented on this proposal. Each expressed many similar objections to the certification language as proposed and offered many similar recommendations. For example, both GSEs stated that the certification language was overly broad and should be modified to the form authorized in FHEFSSA for submissions to OFHEO; namely, that the report is true and correct to the best of such officer's knowledge and belief. Each recommended that the words “fairly present” be deleted from the third proposed certification statement stating that these words are meaningful only in the context of Generally Accepted Accounting Practices (GAAP), which defines standards of determining “fairness” in financial reporting, but not performance reporting.

In addition, both GSEs questioned HUD's authority to impose a certification requirement, but stated that to the extent HUD does impose this requirement, it should be the certification used by OFHEO. They also stated that the phrases “errors, omissions, discrepancies, and ambiguities,” “written articulations of its counting rules,” and “any other areas of ambiguity” are vague and undefined, and that this vagueness makes it possible for HUD to arbitrarily implement the certification provision by interpreting it in a way that is not known by the GSEs. Freddie Mac also stated that HUD's informal written articulations are not enforceable and that it may not know about all of HUD's informal articulations. Both GSEs also stated that it is difficult to certify to the accuracy of information that must be included in the reports that they receive from third parties.

Freddie Mac suggested that the subject of the certification be limited to the year-end annual data tables and computerized loan-level data that it submits with its AHARs, and should not cover any narrative portions of the AHARs. Fannie Mae suggested that the certification should focus on the process it follows for generating its submissions and should cover only the final tables in the AHAR that it submits each year.

Both GSEs stated that no certification should be required for reports-in-progress, such as the housing goals progress reports each submits to HUD on a quarterly basis.

A policy advocacy group commented that the certification should be limited to reporting processes of the GSEs, not the accuracy of the underlying data obtained from individual lenders. A trade association commented that HUD should not put additional data integrity burdens on lenders.

HUD's Determination. HUD has considered the comments received and has determined to modify its proposal. HUD's reasons for requiring a certification were not disputed by commenters. However, HUD has revised the proposed rule language to address commenters' concerns regarding clarity. HUD has also included alternative language in the final rule that would specifically define terms as well as eliminate the language that the GSEs and others found to be ambiguous. As a result, the final rule includes a simplified certification that is much closer to the certification used by OFHEO. Section 81.102(b) has been amended to require the senior officer of each GSE who is responsible for submitting to HUD the fourth quarter Annual Mortgage Report and the AHAR under sections 309(m) and (n) of the Fannie Mae Charter Act or sections 307(e) and (f) of the Freddie Mac Act, as applicable, or for submitting to the Secretary such other report(s), data submission(s), or information for which certification is requested in writing by the Secretary to state that: “To the best of my knowledge and belief, the information provided herein is true, correct and complete.”

The Department has also included language to clarify that it may pursue enforcement action against a GSE that fails to provide the certification required under § 81.102(b). In addition, the Department may pursue enforcement action if a GSE submits the certification required under § 81.102(b), but the Secretary later determines that the data, information or report(s) are not true, correct and complete. For data, information and report(s) subject to § 81.102(c) or (d), the final rule makes clear that the Department will only pursue enforcement action against a GSE in connection with material errors, omissions or discrepancies, as those terms are defined therein.

The GSEs have asserted that HUD may not require certification of any information they submit because the Department has no express statutory authority to do so. The Department's authority to require certification of information submitted by the GSEs is authorized under HUD's general regulatory power over the GSEs under section 1321 of FHEFSSA as well as its authority to monitor and enforce the GSEs' compliance with the housing goals under section 1336. (See the preamble of HUD's proposed rule at 69 FR 24247-24248 for a full discussion of HUD's authority to require certification.)

In requiring this certification, HUD is fully aware that the GSEs collect millions of data elements from hundreds of sources and that the GSEs must depend upon these sources to provide accurate data. In requiring a certification, HUD intends that the GSEs will use and rely upon their internal controls and other due diligence processes and procedures for collecting, compiling, verifying the accuracy of, and reporting the data received from sellers. HUD will evaluate the sufficiency of this certification beginning with the 2005 fourth quarter Annual Mortgage Report and the AHAR to determine whether it is serving its function of providing adequate assurance as to the accuracy and completeness of information.

With respect to the scope of the certification, HUD believes it is appropriate and reasonable that the certification statement apply to the entire AHAR submission, including the narrative text, data tables, and computerized loan-level data. Section 309(n) of Fannie Mae's Charter Act and section 307(f) of the Freddie Mac Act specify the types of information each GSE is required to report, including narrative descriptions as well as data. HUD expects that all of the required information, not just the data and data tables, will be subjected to appropriate internal review processes by the GSEs. A certification regarding the entire report helps to ensure the GSEs' accountability for the information that they are required to report accurately under their charters.

Although Fannie Mae recommended that the certification should apply only to the tables in the AHARs and Freddie Mac recommended that the certification should apply only to the data tables in the AHAR and the loan-level data it submits with its AHAR, from time to time HUD requires one or both GSEs to submit other report(s), information, or data submission(s) that rise to a sufficient level of importance to HUD's oversight work that a certification statement is warranted. The final rule, therefore, retains this provision and further provides that the Secretary will issue a written notification to the GSE whenever such a certification is required. HUD expects that any additional certification requirements will be the exception rather than the rule to ensure that the routine and necessary flow of information is not impeded.

Both GSEs recommended that HUD not impose a certification on any progress reporting, such as the quarterly housing goals performance reports each submits to HUD. HUD did not propose that such reports be certified and reiterates that certification statements will not be required for the GSEs' first three quarterly housing goals reports and any other report(s), data submission(s) or information that represent incomplete “snapshots” or information that is being gathered but which is not in final form. Certification will be required for the fourth quarter report, i.e., the Annual Mortgage Report.

4. Adjustment To Correct Current Year-end Errors, Omissions or Discrepancies—§ 81.102(c)

HUD routinely conducts computerized consistency checks of loan-level data received from the GSEs as part of their AHAR reporting. This data are received on March 15th of each year for the previous year's performance. These reviews verify that the GSEs have applied HUD's counting rules and goals eligibility standards appropriately in determining their year-end performance. A key procedure involves applying HUD's counting rules to the GSEs' loan-level data for the purposes of replicating the performance figures computed by the GSEs in their AHARs. Also, in conjunction with other reports provided by the GSEs, including a report that reconciles all adjusted mortgage purchases (the denominator) with the GSE's total business volume as reported in the annual report to shareholders or other information filings, HUD's reviews also verify the completeness of the data. If HUD finds discrepancies between its results and those reported by the GSEs, HUD works with appropriate GSE staff to resolve the discrepancies after which HUD makes a final determination of year-end results and publishes these as HUD's official performance figures for the year.

HUD's proposed rule provided for a timeframe within which the GSEs may comment or otherwise respond to HUD's findings of errors, omissions, or discrepancies with additional information. If a GSE did not respond with information to correct or explain the error, omission, or discrepancy to HUD's satisfaction within five working days of HUD's initial notification, then HUD would notify the GSE in writing and seek clarification or additional information. At this point, the GSE would have 10 working days in which to respond and could request an extension of an additional 20 working days from HUD. If the GSE still did not respond in a manner that corrected the error, omission, or discrepancy, then HUD would determine the appropriate adjustment to the numerator and denominator of the applicable goal and/or subgoal. Currently, there are no required time limits within which the GSEs must respond to HUD's inquiries for additional information, and there is no procedure by which HUD can bring the process of reviewing a GSE's current year submission to closure absent voluntary assistance from the GSEs. The practical effect of not codifying a timetable for completion of this process is that HUD could be delayed in fulfilling its responsibilities to issue a timely, official report on the GSEs' performance for the year most recently ended and to produce the public use database.

Summary of Comments. In addition to the GSEs, many organizations, including policy advocates, trade associations, and one non-profit group, commented on the data verification provisions of HUD's proposed rule. Nearly all of these comments supported implementation of some type of data verification procedures. One trade group stated that data verification regulations should be enforced to get more accurate information. However, another trade group expressed concern that the data integrity process should include some leeway for unintentional mistakes to avoid becoming burdensome. Two advocacy organizations supported the proposed provisions regarding data verification but thought HUD should give the public the ability to comment on the GSEs' AHARs.

Both GSEs commented in detail on HUD's proposal. Both expressed concerns about the scope of this provision and questioned what procedures, especially adjustment notification and enforcement procedures, would be associated with its implementation. Freddie Mac augmented its comments with a legal opinion from outside counsel.

With respect to the words “errors, omissions and discrepancies,” the GSEs contended that these terms were vague and needed further definition. Freddie Mac stated that without such further definition, HUD could disallow counting of units based upon interpretations of its rules of which Freddie Mac was unaware, and thus violate the fair notice doctrine. Freddie Mac suggested that if HUD retained the use of these words in its regulation, it should explain how their meanings differ. Fannie Mae stated that potential adjustments should apply only to situations where the GSE failed to follow HUD's rules for data collection and reporting, and not where it failed to follow its own rules for procedures in data collection and reporting. Fannie Mae also contended that adjustments should be made only where the error, omission or discrepancy was in a data field that affected scoring and where it also had a material effect on compliance with a housing goal. Freddie Mac stated that adjustments should be made only for material errors or omissions. Fannie Mae stated that a GSE should be subject to additional enforcement action only when an error, omission or discrepancy is due to intentional or bad faith action.

Both GSEs stated that HUD's regulations should provide that HUD will issue a written determination to a GSE when it determines that an adjustment is necessary, that HUD should specify which official within HUD is authorized to issue orders under proposed § 81.102(c) and (d), and that the rule should provide for more lenient time frames for responding to HUD's inquiries. In addition, Freddie Mac commented that the regulations should state that an order requiring an adjustment constitutes “final agency action” for purposes of judicial review under the Administrative Procedure Act and that judicial review is immediately available.

Fannie Mae also commented on the title of HUD's provision stating that a provision to correct “current year end errors” is confusing because HUD cannot correct errors for a current year when it does not receive the data about any current year until the next year.

HUD's Determination. HUD has considered the comments and determined that a provision specifying what procedures HUD will use in developing its official performance numbers for the immediately preceding year is necessary. HUD notes that many of the concerns expressed by commenters, especially the GSEs, involve the lack of definition of the terms “errors, omissions or discrepancies” and a lack of clarity regarding how the regulation will be enforced. Accordingly, in the final rule, HUD has added a paragraph that defines an “error” as a technical mistake, such as a mistake in coding or calculating data. Mistakes of this type may also include, but not be limited to, systems errors, such as those related to geocoding or misapplication of HUD's most current data regarding median income or underserved areas. An “omission” is defined as a GSE's failure to count units in the denominator. A “discrepancy” is defined as any difference between HUD's analysis of data and the analysis contained in a GSE's submission of data, including a discrepancy in goal and/or Special Affordable subgoal performance.

The Department also clarifies in § 81.102(c)(5) of this final rule that an error, omission or discrepancy is “material” if it results in an overstatement of credit for a housing goal or Special Affordable subgoal and, without such overstatement, the GSE would have failed to meet such housing goal or Special Affordable subgoal for the immediately preceding year. Finally, the rule defines the term “year-end data” to mean data that HUD receives from the GSEs related to housing goals performance in the immediately preceding year and covering data reported in the fourth quarter Annual Mortgage Report and the GSE's AHAR.

With respect to procedures for notifying a GSE of any suspected error, omission or discrepancy, HUD is responding to the concerns raised by the commenters by amending the proposed rule to: (1) Provide that, with regard to each initial notification by HUD to a GSE, HUD may, in its own discretion, or upon a request by a GSE, extend the initial five working day response period for up to 20 additional working days; (2) establish that any person with delegated authority from the Secretary, or the Director of HUD's Financial Institution Regulation Division, or his or her designee, is responsible for issuing initial notifications regarding errors, omissions, or discrepancies, making determinations on the adequacy of responses received, approving any extensions of time permitted under this provision, and generally managing the data verification process; (3) establish that the Secretary or his designee will inform a GSE in writing of HUD's determination of official performance figures, including any adjustments, five working days prior to HUD's release of its official performance figures to the public; (4) provide that during the five working days prior to such public release, a GSE may request reconsideration in writing of HUD's final determination of its performance in which case the Secretary will decide whether to grant the request for reconsideration, and if the request is granted, make a final determination on the request for reconsideration within 10 working days of the Secretary's granting of the GSE's request for reconsideration; and (5) provide that, with the exception of the written determination of HUD's official performance figures, all other notifications under this provision may be by electronic mail.

HUD has also clarified through its definitions of errors, omissions and discrepancies, that an “adjustment” will be made in situations where a GSE failed to follow correct procedures in data compilation and reporting and/or where it failed to comply with HUD's regulation for determining eligible units. As has been the case in the past, HUD expects that any adjustments that it may make to the numerator or denominator, that result in a difference between the GSE's performance as stated in the GSE's AHAR for the immediately preceding year and HUD's official performance figures, will be well understood by the GSE because adjustments of this type occur routinely during HUD's verification work.

HUD is also clarifying that it intends to treat a GSE's material errors, omissions or discrepancies in, or failure to certify, data submissions under § 81.102(c) as a failure to submit information that the GSE is required to submit under its charter. Accordingly, the Department may pursue the additional enforcement remedies authorized under § 81.102(e).

With respect to events that could trigger enforcement under this provision, HUD does not intend that routine technical errors or omissions would warrant such enforcement. In order to trigger the enforcement provision, errors, omissions or discrepancies discovered during review of the immediately preceding year's performance must be material, as HUD has defined that term. The error, omission or discrepancy also must be one that indicates to HUD a serious problem in the GSE's internal procedures. Examples of errors, omissions, or discrepancies that could rise to this level under these criteria include a GSE counting units that are not eligible under HUD's rules for goals credit or a GSE underreporting units in the denominator. With respect to Freddie Mac's suggestion that HUD's regulations should state that this determination is “final agency action” for purposes of the Administrative Procedure Act and is immediately subject to judicial review, FHEFSSA already provides that the GSEs may obtain judicial review in connection with proceedings to enforce the housing goals, and that those proceedings shall be governed by the Administrative Procedure Act. Therefore, the Department declines to adopt Freddie Mac's suggestion.

To more clearly define the scope of this provision, HUD has renamed this provision in the final rule as Verification Procedure and Adjustment to Correct Errors, Omissions, or Discrepancies in AHAR Data for the Immediately Preceding Year.

5. Procedures for Prior Year Reporting Errors—§ 81.102(d)

The annual data verification review for the immediately preceding year described in § 81.102(c) was designed to ensure that reported goals performance was correctly calculated in accordance with HUD's regulations. Although these reviews can test for the reasonableness of some reported data, the reviews cannot generally determine the accuracy of the underlying loan-level data. To monitor data accuracy, HUD has implemented a second type of procedure, called performance reviews. Performance reviews are especially necessary because housing goals are calculated from information (e.g., number of dwelling units) that is not reported in the GSEs' financial statements and is, therefore, not subject to all GSE procedures designed to ensure the accuracy and completeness of reported financial information. HUD's performance reviews ensure that rigorous audit procedures, either similar or identical to those used to monitor the integrity of financial data, are also used in monitoring the accuracy, completeness, and timeliness of the data each GSE submits to HUD. Performance reviews include, but are not limited to, evaluating the GSEs' internal controls over the collection, management and reporting of loan-level mortgage data used in calculating housing goals performance. Performance reviews may also focus on the GSEs' quality control standards and procedures for information received from loan sellers and securities issuers and dealers and may include additional procedures to test random samples of data for accuracy and completeness. To supplement HUD's on-site performance review work, the Department has implemented specialized reporting by which each GSE informs HUD on a scheduled basis of key issues and findings relevant to goals reporting. For example, the GSEs report to HUD quarterly on the results of their own internal reviews and self-assessments related to housing goals. These reports cover all actions taken by the GSE to remove any findings related to weaknesses in controls or procedures, including those findings identified by HUD.

Because of the complexity of each GSE's business, as well as the complexity of many of the transactions that the GSEs undertake to meet their housing goals, there is a possibility that HUD may discover, during a performance review, that a serious overstatement of credit towards one or more housing goals occurred in the reported prior year under review. Currently, HUD has no procedure for ensuring that any such overstatement is corrected or otherwise adjusted in some manner unless the overstatement is discovered in the review of the immediately past year's data during the replication review described in § 81.102(c). To remedy this, HUD proposed a procedure that would adjust a GSE's current year performance by deducting from the numerator of the relevant housing goal or subgoal the number of overstated units from a prior year. A prior year was any one of the two years preceding the current reporting year.

Summary of Comments. Many organizations commented on HUD's data integrity provisions in general and nearly all of these organizations expressed support for data verification. The GSEs commented more specifically on HUD's proposals for adjustments to make up prior years' overstatements. The GSEs asserted that the Department does not have authority to either deduct credit from a current year's purchase that is entitled to credit under HUD's regulations or add to a current year's housing goal to compensate for the GSE's failure to meet its goals in a prior year. They also had other objections, including the objection that the only remedy provided in FHEFSSA for any failure to meet housing goals is the imposition of a housing plan, which may address only a probable failure to meet housing goals in the current year or actual failure to meet goals in a current year in the next calendar year. The GSEs stated that the Congress intended the statute to provide no remedy for their failure to meet a prior year's housing goal and, therefore, that the Department has no authority to fashion such a remedy. Based on this line of reasoning, they concluded that HUD may not take any action against a GSE when it discovers that it failed to meet a housing goal in a prior year, even though HUD could have taken action if the failure had been discovered within one year after the year in which it occurred.

Both GSEs also objected to the policy basis for HUD's proposal. For instance, Fannie Mae wanted the time period within which HUD might impose a prior-year correction shortened from up to 24 months to three months after HUD's receipt of AHAR loan-level data, which HUD receives on March 15th of each year. Fannie Mae cited OFHEO's ability to render a decision on its final capital classification within 90 days of the reporting quarter as evidence that complex determinations can be made within short time frames. Freddie Mac saw no reason why the necessary evaluations could not be accomplished within six months after the close of the immediately preceding year. Freddie Mac stated that HUD's policy justification does not support the proposal and that HUD did not point to any instance where the increasing complexity of transactions has led to overstatements in performance. Freddie Mac also commented that the Department already has the option of publicizing the discovery of any prior year mistakes—by press release, news conference or its Web site information—and of making Congress aware of these mistakes.

Freddie Mac requested that HUD withdraw the proposal entirely. If HUD opted to proceed to implement the proposal, then Freddie Mac suggested that HUD amend the provision to: (1) Limit application of the rule to large prior year overstatements that affect a material number of units under a goal (e.g., five percent); (2) provide that HUD will apply the rule only when a GSE acted in bad faith; (3) provide that HUD will not apply the rule cumulatively; that is, that HUD will not accumulate several years' over-counts and then deduct a cumulative total from the current year; and (4) clarify in the final rule which official within HUD will make decisions under this provision and provide that the basis for decisions be explained.

HUD's Determination. HUD has carefully considered both GSEs' comments, including their legal and policy arguments. The Department agrees that the only remedy Congress set out in FHEFSSA for failing to meet a housing goal is a housing plan under section 1336, and as the statute is written, the housing plan addresses only a current year's failure, either in that year or in the next calendar year. Therefore, the statute does not specifically address a GSE's failure to meet a housing goal in a prior year, i.e., a failure occurring in any one of the two years immediately preceding the latest year for which data on housing goals performance was reported to HUD. However, the Department does not agree that Congressional silence on this precise issue means either that Congress intended the GSEs to be allowed to fail to meet their housing goals as long as the Department does not discover that failure within a specific time or that the Department may not fashion a remedy to address this issue. This conclusion runs counter to Congress's purposes in enacting FHEFSSA, which directs HUD to establish and monitor the GSEs' compliance with the Housing Goals.

Section 1336 of FHEFSSA provides that the Secretary shall “monitor and enforce” the GSEs” compliance with the housing goals set by the Department. According to FHEFSSA's legislative history, in enacting FHEFSSA Congress intended “to establish a comprehensive framework of goals, data collection, reporting requirements and enforcement provisions.” S. Rep. No. 102-282, at 34 (1992)(emphasis added).

When discussing the GSEs' duties to meet housing goals set for low- and moderate-income housing and housing in central cities and rural areas, Congress stated:

The GSEs need to provide more leadership in all of these areas, and they have indicated a desire to do so. But direct and potentially forceful federal oversight is the only way to ensure that it will happen. Id. at 11.

Under the GSEs' suggested construction of FHEFSSA, HUD's ability to enforce the housing goals is totally dependent upon only one factor, namely how quickly HUD discovers that a GSE has failed to meet a goal. In order to determine whether a GSE has failed a goal, HUD must receive, verify and analyze massive amounts of data, as described above. Under the GSEs' suggested construction of FHEFSSA, only if HUD discovers that a GSE has failed to meet a housing goal or subgoal in the nine month period that runs from March 15th, when the GSEs submit current year-end data, to the end of that year—may HUD enforce the housing goals for that year. Such a construction is not only unreasonable on its face but it is contrary to the plain intent of Congress as expressed in the FHEFSSA and its legislative history. FHEFSSA and its legislative history indicate that Congress established a comprehensive regulatory scheme under which HUD would establish and enforce the Housing Goals through strict and pervasive regulation.

Furthermore, there is absolutely no statement in FHEFSSA or its legislative history to suggest that Congress intended that HUD must ignore or forgive a GSE's failure to meet its housing goals in any year for any reason, including the passage of a certain amount of time before HUD discovers this failure. The fact that FHEFSSA is silent on the issue of how to address a GSE's failure to meet a prior year's housing goal means that there is a gap in FHEFSSA's enforcement scheme regarding this precise issue. Under Chevron v. NRDC, 467 U.S. 837 (1984), the Department has discretion to fashion an appropriate remedy to fill this gap, and it has done so in § 81.102(d). Moreover, the Department has the discretion to fashion a remedy to correct prior year overstatements without which a GSE would have failed to meet a housing goal or Special Affordable subgoal under its general regulatory powers under section 1321 of FHEFSSA.

However, in light of the objections raised to the proposed regulation in the comments discussed above, HUD has revised § 81.102(d) to remove provisions that either provide for deduction of Housing Goals credit in a current year from purchases that qualify for credit, or that add requirements to a current year's Housing Goals due to errors, omissions or discrepancies in a prior year's data submissions. The final rule provides instead that the Secretary may require the GSEs to make up any overstatements of goal performance without which a GSE would have failed to meet a prior year's Housing Goal, no later than the year following the year in which HUD first notifies the GSE of this failure. (The rule now defines the term “prior year” to mean any one of the two years immediately preceding the latest year for which data on housing goals performance was reported to HUD.)

In order to remedy this failure, the Secretary may require the GSEs to purchase additional mortgages that finance a number of units that either (a) equal the number of units overstated in the prior year's goal performance, or for the Special Affordable subgoals the number or dollar amount, as applicable, of mortgage purchases that the Secretary has determined were overstated, or (b) that equal the percentage of the overstatement in the prior year's goal performance as applied to the most current year-end performance, whichever is less. Units purchased to remedy an overstatement must be eligible to qualify under the same goal or goals for which the overstatement occurred in the prior year. For example, a GSE may have overstated a prior year's performance by 5,000 units or .22 percent under the Low- and Moderate-Income Goal. To make up this overstatement, a GSE may purchase an additional 5,000 units that are eligible under the Low- and Moderate-Income Goal in the year immediately following the year in which HUD notifies the GSE of the overstatement or it may multiply the current year's total eligible purchases under the Low- and Moderate-Income Goal by the overstated percentage from a prior year (e.g., .22 percent) to determine the number of units that must be purchased, provided this number is less than 5,000 units. The same requirement also applies to the Special Affordable Home Purchase Subgoal. When an overstatement occurs under this Subgoal, the Secretary may require the GSE to make up the number of mortgages that were overstated using the lesser of the two procedures previously described. For overstatements under the Special Affordable Multifamily Subgoal, the GSE may be required to make up the dollar amount of overstatement by purchasing qualifying multifamily mortgages in an amount equal to the overstatement. The GSEs will not be required to make up any errors, omissions or discrepancies in prior years that were not material. As previously noted, the final rule provides that an error, omission or discrepancy is material if it results in an overstatement of credit for a housing goal or Special Affordable subgoal and, without such overstatement, the GSE would have failed to meet such housing goal or Special Affordable subgoal for the prior year.

Also, corrections for overstatement of goals performance under this provision will not be counted or reported under the GSEs' Annual Housing Activity Report, including calculation of housing goal performance in any year, but rather will be managed separately from the housing goals as directed by HUD.

If the GSE does not purchase a sufficient number of eligible units or mortgages, as described previously, then HUD may issue a notice that the GSE failed a housing goal or subgoal in a previous year, or seek additional enforcement remedies under § 81.102(e) or any other civil or administrative remedies that are available under applicable law. The Department is treating a GSE's material errors, omissions or discrepancies in, or failure to certify, a prior year's data submission as a failure to submit information that the GSE is required to submit under its charter.

Both GSEs also questioned the need for an adjustment period that could extend for up to 24 months from the close of a calendar year's performance, believing instead that any such review could be accomplished within six months of the close of the previous year, which is a time frame similar to that used by OFHEO to assess the adequacy of a GSE's capital. As HUD has stated previously, reviews conducted immediately upon receipt of a GSE's prior year loan-level data and pursuant to § 81.102(c) cannot generally gauge the accuracy of the data and cannot always determine whether the transaction itself complies with HUD's regulations for counting units towards goals performance. Assessments of this type require the application of procedures, either in whole or in part, that are characteristic of audit engagements. For example, it is customary for audits of a previous year's financial statements to require up to one year or more for completion due to the number of procedures involved and the volume of information to be reviewed, especially for exceedingly large and complex organizations. Similarly, the relatively new field of performance data auditing, including reviews based on some or all of these procedures, also requires a substantial commitment of time and resources if meaningful results are to be obtained. For these reasons, performance reviews are not analogous to OFHEO's evaluations of capital adequacy. HUD believes that its original proposal of allowing for up to 24 months after the close of the year under review is the appropriate time frame for completion of the performance review.

The GSEs also expressed some concerns about the potential for HUD to make determinations of error after the fact and without any prior notice to a GSE that a type of transaction and/or housing unit would not be eligible for goals credit. HUD believes it is useful to more fully describe the types of errors likely to trigger a finding that units were overstated. In the context of performance reviews, the words “errors, omissions or discrepancies” connote serious mistakes, such as those associated with violations of HUD's counting rules and other goals eligibility criteria as set forth in its regulations. HUD is aware that in collecting and reporting millions of data elements, some level of factual error is probably unavoidable. However, with regard to data accuracy in performance reviews, HUD is concerned with errors of a substantial nature, such as those that suggest a larger internal control problem, an example of which could be a pattern of incorrect rental data acquired from or generated by the same source. HUD is also concerned with types of transactions that are either expressly prohibited from goals eligibility, such as high cost mortgages, or for which HUD approval may have been required but not obtained prior to a GSE counting the units, such as the use of an affordability estimation methodology. Other similar types of problems may also trigger a HUD determination of error. In the event HUD supplements its regulations with letters to one or both GSEs regarding appropriate counting treatment, the GSE will be responsible for complying with only the specific directives it has received from HUD. In the final rule, HUD has reiterated that this procedure will apply only in those instances where an overstatement was material in nature; that is, the overstated units enabled the GSE to meet a housing goal that it otherwise would not have met. In the event that HUD undertakes a performance review that covers a two-year period and determines that material misstatements of housing goals or Special Affordable subgoals performance occurred in both years, then HUD will apply the same procedures as described previously for making up the overstatements. Upon a written request from a GSE, the Secretary may, in his discretion, determine to grant an extension of additional time to correct or compensate for the overstatement. For example, if overstatements were discovered for years 2005 and 2006 and the GSE is notified of the overstatements for both years in 2007, then the GSE could be required to make up the overstatements for both years in 2008. Similarly, if the overstatement was discovered for one year, 2005, and the GSE was notified of the overstatement in 2006, then the GSE could be required to make up the overstated units or mortgages in 2007. In both examples, upon receipt of a GSE's written request for an extension of time, the Secretary may grant an extension for completing make up of the overstated units or mortgages.

With regard to HUD's reasons for implementing a procedure that provides a mechanism by which overstated units of a material nature from a prior year can be made up in a subsequent year following the year a GSE is first notified of the overstated units, for reasons stated above, it is the Department's view that it has authority to do so, and that the procedure is needed at this time. The procedure is the only tool by which HUD can meet its statutory responsibility to assure the integrity of all of the housing goal data reported to the public, including the data reported in the GSE public use database and its duty to enforce the housing goals.

6. Additional Enforcement Option § 81.102(e)

The Department proposed a new § 81.102(e) that would provide HUD with additional enforcement options in the event it determines that a GSE has submitted data, information, or report(s) that are not current, are incomplete, or otherwise contain an untrue statement of material fact. Section 81.102(e) identified the data, information, or report(s) that would be subject to HUD's additional enforcement authority as those required under section 307(e) and (f) of the Freddie Mac Act, section 309(m) or (n) of the Fannie Mae Charter Act, or under 24 CFR part 81, subpart E.

The Department indicated in proposed § 81.102(e) that it could make a determination—either under its independent verification authority in § 81.102(a) or by “other means”—that such data, information or report(s) are not current, are incomplete, or otherwise contain an untrue statement of material fact. This reference to “other means” was intended to encompass the Secretary's authority under the three other provisions in § 81.102 that were also being proposed to ensure the accuracy, truthfulness and completeness of GSE submissions to HUD: (1) The proposed GSE certification in § 81.102(b); (2) the proposed procedure established in § 81.102(c) to verify the accuracy and completeness of the GSE's current year-end data; and (3) the proposed procedure established in § 81.102(d) to ensure the accuracy and completeness of the GSE's prior years' data.

The Department further provided in § 81.102(e) that the Secretary could regard a GSE's submission of data, information or report(s) that he or she determines under § 81.102(a), or by “other means” (i.e., pursuant to paragraphs (b), (c) or (d) of § 81.102), are not current, are incomplete, or that otherwise contain an untrue statement of material fact to be equivalent to the GSE's failure to submit such data. As a result of such failure of submission, proposed § 81.102(e) provided that the Department could initiate against the GSE, in accordance with the procedures in 24 CFR part 81, subpart G, an order to cease and desist, an action to seek civil money penalties, or any other remedies or penalties that may be available to the Secretary as a result of the GSE's failure to provide data submissions, information, and/or report(s) in accordance with § 81.102.

Summary of Comments. Several organizations commented, generally, on HUD's proposed requirements in § 81.102 for ensuring the accuracy and integrity of GSE data and other submissions, and almost all expressed support for HUD's proposals relating to data verification. The GSEs commented more specifically on HUD's proposal in § 81.102(e) relating to additional enforcement options.

Fannie Mae asserted that HUD's proposed additional enforcement options were overly broad, and exceeded the Department's authority under FHEFSSA to issue cease and desist orders, impose civil money penalties, and to punish GSE noncompliance by requiring the adoption of a housing plan. Fannie Mae stated that, if HUD decided to retain § 81.102(e), this provision should be redrafted more narrowly.

Freddie Mac, through a legal opinion prepared by outside counsel, asserted that sections 1341 and 1345 of FHEFSSA provide a two-step process before a GSE's failure to submit a housing plan, or its failure to comply with a feasible housing plan, could result in the Department's initiating an action for a cease and desist order or civil money penalties. Freddie Mac asserted that HUD's proposal expanded its enforcement authority beyond the FHEFSSA statutory limits by eliminating this two-step process. Freddie Mac also contended that HUD's enforcement powers under sections 1341 and 1345 of FHEFSSA extend only to instances of intentional non-compliance by the GSE, and that § 81.102 should be narrowed to reflect this limitation.

HUD's Determination. The Department has considered the GSEs' and other comments on § 81.102(e) and is making several changes in this final rule in response to these comments. In addition, the Department is making a number of conforming changes to § 81.102(e) to reflect changes that it has also decided to adopt in connection with the other provisions in § 81.102 (primarily in paragraphs (b), (c) and (d)), and is also making minor editorial corrections.

Specifically, the Department is providing in this final rule that:

The Department may pursue additional enforcement remedies under paragraph (e) under either of the following circumstances: (1) When a GSE fails to submit the certification required by § 81.102(b) in connection with data, information or report(s) required by section 309(m) or (n) of the Fannie Mae Charter Act, section 307(e) or (f) of the Freddie Mac Act, or under 24 CFR part 81, subpart E; or (2) when a GSE submits the certification required by § 81.102(b) in connection with such data, information or report(s), but the Secretary later determines that the data, information or report(s) are not “true, correct and complete” as provided in the certification. The final rule provides that, under either of the above two circumstances, the Secretary may regard a GSE's actions as tantamount to a failure to submit the data, information or report(s) which, in turn, authorizes the Secretary to take the additional enforcement remedies described in § 81.102(e).

The final rule also clarifies that for data, information or report(s) that are subject to § 81.102(c) or (d), the Secretary may only pursue additional enforcement remedies in connection with material errors, omissions or discrepancies. Moreover, if the data, information or report(s) are subject to § 81.102(d), the rule provides that the Secretary may only pursue additional enforcement remedies if the GSE has failed to purchase a sufficient amount or type of mortgages as required by the Secretary under § 81.102(d)(4).

It is the Department's view that § 81.102(e) is needed so that it can take appropriate action to ensure the accuracy and completeness of the GSEs' submissions to HUD, and also to implement the certification that is now established at § 81.102(b) of this final rule, while providing the Secretary with sufficient flexibility to exercise his or her discretion to determine whether enforcement action is appropriate in each instance.

The final rule clarifies that the proposed rule's reference in paragraph (e)(1) to “other means” by which the Secretary may determine that a GSE's data submission(s), information or report(s) fail to meet the prescribed regulatory standards is meant to refer to the Secretary's determinations under paragraphs (b), (c) or (d) of § 81.102 (i.e., the GSE certification in § 81.102(b), the procedure established in § 81.102(c) to verify the accuracy and completeness of the GSE's data for the immediately preceding year, and the procedure established in § 81.102(d) to ensure the accuracy and completeness of the GSE's prior years' data). In the final rule, the Department has deleted the reference to “other means” and has included a specific reference to paragraphs (b), (c) or (d) of § 81.102.

The final rule establishes a bifurcated approach with respect to the types of additional enforcement remedies that the Department may pursue under paragraph (e). This bifurcated approach recognizes that the Department's ability to pursue a cease and desist order, or to levy civil money penalties, applies specifically to data required by section 309(m) or (n) of the Fannie Mae Charter Act or section 307(e) or (f) of the Freddie Mac Act. The rule nevertheless provides that the Department may pursue other types of remedies against a GSE in connection with data that the GSE is required to submit under 24 CFR part 81, subpart E, but that the GSE is not required to submit under section 309(m) or (n) of the Fannie Mae Charter Act or section 307(e) or (f) of the Freddie Mac Charter Act.

The final rule provides that, in connection with either of the two remedial approaches now described in § 81.102(e)(2), the Secretary may pursue any civil or administrative remedies or penalties against the GSE that may be available to the Secretary by virtue of either of the circumstances described in 81.102(e)(1). If the Department elects to pursue a cease-and-desist order or civil money penalties against a GSE under § 81.102(e)(2)(i)(A) or (B), it will comply with the procedures applicable to such actions under 24 CFR part 81, subpart G. Alternatively, if the Department elects to pursue other civil or administrative remedies against a GSE under either §§ 81.102(e)(2)(i)(C) or 81.102(e)(2)(ii), it will pursue such remedies in accordance with applicable law.

Finally, the Department is replacing in paragraph (e) each reference to “HUD” with a reference to “the Secretary.” This replacement is designed to ensure that any additional enforcement action that may be pursued under § 81.102(e) will be considered at the highest levels within the Department.

III. Findings and Certifications

Executive Order 12866

The Office of Management and Budget (OMB) reviewed this final rule under Executive Order 12866, Regulatory Planning and Review, which the President issued on September 30, 1993. This rule was determined to be economically significant under E.O. 12866. Any changes made to this rule subsequent to its submission to OMB are identified in the docket file, which is available for public inspection between 8 a.m. and 5 p.m. weekdays in the Office of the Rules Docket Clerk, Office of General Counsel, Room 10276, Department of Housing and Urban Development, 451 Seventh Street, SW., Washington, DC. The Economic Analysis prepared for this rule is also available for public inspection in the Office of the Rules Docket Clerk and on HUD's Web site at http://www.hud.gov.

Congressional Review of Regulations

This rule is a “major rule” as defined in Chapter 8 of 5 U.S.C. This rule will be submitted for Congressional review in accordance with this chapter.

Paperwork Reduction Act

HUD's collection of information on the GSEs' activities has been reviewed and authorized by OMB under the Paperwork Reduction Act of 1995 (44 U.S.C. 3501-3520), as implemented by OMB in regulations at 5 CFR part 1320. The OMB control number is 2502-0514.

Environmental Impact

This final rule does not direct, provide for assistance or loan and mortgage insurance for, or otherwise govern or regulate real property acquisition, disposition, leasing, rehabilitation, alteration, demolition, or new construction; or establish, revise, or provide for standards for construction or construction materials, manufactured housing, or occupancy. Accordingly, under 24 CFR 50.19(c)(1), this rule is categorically excluded from environmental review under the National Environmental Policy Act of 1969 (42 U.S.C. 4321).

Regulatory Flexibility Act

The Secretary, in accordance with the Regulatory Flexibility Act (5 U.S.C. 605(b)), has reviewed this rule before publication and by approving it certifies that this rule would not have a significant economic impact on a substantial number of small entities. This rule is applicable only to the GSEs, which are not small entities for purposes of the Regulatory Flexibility Act. Therefore, the rule does not have a significant economic impact on a substantial number of small entities within the meaning of the Regulatory Flexibility Act.

Executive Order 13132, Federalism

Executive Order 13132 (“Federalism”) prohibits, to the extent practicable and permitted by law, an agency from promulgating a regulation that has federalism implications and either imposes substantial direct compliance costs on state and local governments and is not required by statute, or preempts state law, unless the relevant requirements of section 6 of the executive order are met. This rule does not have federalism implications and does not impose substantial direct compliance costs on state and local governments or preempt state law within the meaning of the executive order.

Unfunded Mandates Reform Act

Title II of the Unfunded Mandates Reform Act of 1995 (12 U.S.C. 1531-1538) (UMRA) establishes requirements for federal agencies to assess the effects of their regulatory actions on state, local, and tribal governments, and the private sector. This rule would not impose any federal mandates on any state, local, or tribal government, or on the private sector, within the meaning of UMRA.

List of Subjects in 24 CFR Part 81

• Accounting
• Federal Reserve System
• Mortgages
• Reporting and recordkeeping requirements
• Securities

For the reasons discussed in the preamble, HUD is amending 24 CFR part 81 as follows:

PART 81—THE SECRETARY OF HUD'S REGULATION OF THE FEDERAL NATIONAL MORTGAGE ASSOCIATION (FANNIE MAE) AND THE FEDERAL HOME LOAN MORTGAGE CORPORATION (FREDDIE MAC)

1. The authority citation for 24 CFR part 81 continues to read as follows:

Authority: 12 U.S.C. 1451 et seq., 1716-1723h, and 4501-4641; 28 U.S.C. 2461 note; 42 U.S.C. 3535(d) and 3601-3619.

2. In § 81.2(b), revise the definitions of “Metropolitan area” and “Minority,” and paragraph (2) of the definition of “Underserved area,” and add a new definition of the term “Home Purchase Mortgage,” in alphabetical order, to read as follows:

§ 81.2

Definitions.

* * * * *

(b) * * *

Home Purchase Mortgage means a residential mortgage for the purchase of an owner-occupied single-family property.

* * * * *

Metropolitan area means a metropolitan statistical area (“MSA”), or a portion of such an area for which median family income estimates are published annually by HUD.

Minority means any individual who is included within any one or more of the following racial and ethnic categories:

(1) American Indian or Alaskan Native—a person having origins in any of the original peoples of North and South America (including Central America), and who maintains tribal affiliation or community attachment;

(2) Asian—a person having origins in any of the original peoples of the Far East, Southeast Asia, or the Indian subcontinent, including, for example, Cambodia, China, India, Japan, Korea, Malaysia, Pakistan, the Philippine Islands, Thailand, and Vietnam;

(3) Black or African American—a person having origins in any of the black racial groups of Africa;

(4) Hispanic or Latino—a person of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin, regardless of race; and

(5) Native Hawaiian or Other Pacific Islander—a person having origins in any of the original peoples of Hawaii, Guam, Samoa, or other Pacific Islands.

* * * * *

Underserved area means * * *

(2) For purposes of the definition of “Rural area,” a whole census tract, a Federal or State American Indian reservation or tribal or individual trust land, or the balance of a census tract excluding the area within any Federal or State American Indian reservation or tribal or individual trust land, having:

(i) A median income at or below 120 percent of the greater of the State non-metropolitan median income or the nationwide non-metropolitan median income and a minority population of 30 percent or greater; or

(ii) A median income at or below 95 percent of the greater of the State non-metropolitan median income or nationwide non-metropolitan median income.

* * * * *

3. In § 81.12, revise the last sentence of paragraph (b) and revise paragraph (c), to read as follows:

§ 81.12

Low- and Moderate-Income Housing Goal.

* * * * *

(b) Factors. * * * A statement documenting HUD's considerations and findings with respect to these factors, entitled “Departmental Considerations to Establish the Low- and Moderate-Income Housing Goal,” was published in the Federal Register on November 2, 2004.

(c) Goals. The annual goals for each GSE's purchases of mortgages on housing for low- and moderate-income families are:

(1) For the year 2005, 52 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Low- and Moderate-Income Housing Home Purchase Subgoal, 45 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Low- and Moderate-Income Housing Goal in the year 2005 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(2) For the year 2006, 53 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Low- and Moderate-Income Housing Home Purchase Subgoal, 46 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Low- and Moderate-Income Housing Goal in the year 2006 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(3) For the year 2007, 55 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Low- and Moderate-Income Housing Home Purchase Subgoal, 47 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Low- and Moderate-Income Housing Goal in the year 2007 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(4) For the year 2008, 56 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Low- and Moderate-Income Housing Home Purchase Subgoal, 47 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Low- and Moderate-Income Housing Goal in the year 2008 unless otherwise adjusted by HUD in accordance with FHEFSSA; and

(5) For the year 2009 and thereafter HUD shall establish annual goals. Pending establishment of goals for the year 2009 and thereafter, the annual goal for each of those years shall be 56 percent of the total number of dwelling units financed by that GSE's mortgage purchases in each of those years. In addition, as a Low and Moderate Income Housing Home Purchase Subgoal, 47 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Low- and Moderate-Income Housing Goal in each of those years unless otherwise adjusted by HUD in accordance with FHEFSSA.

4. In § 81.13, revise the last sentence of paragraph (b) and revise paragraph (c), to read as follows:

§ 81.13

Central Cities, Rural Areas, and Other Underserved Areas Housing Goal.

* * * * *

(b) Factors. * * * A statement documenting HUD's considerations and findings with respect to these factors, entitled “Departmental Considerations to Establish the Central Cities, Rural Areas, and Other Underserved Areas Housing Goal,” was published in the Federal Register on November 2, 2004.

(c) Goals. The annual goals for each GSE's purchases of mortgages on housing located in central cities, rural areas, and other underserved areas are:

(1) For the year 2005, 37 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Central Cities, Rural Areas, and Other Underserved Areas Home Purchase Subgoal, 32 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Central Cities, Rural Areas, and Other Underserved Areas Housing Goal in the year 2005 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(2) For the year 2006, 38 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Central Cities, Rural Areas, and Other Underserved Areas Home Purchase Subgoal, 33 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Central Cities, Rural Areas, and Other Underserved Areas Housing Goal in the year 2006 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(3) For the year 2007, 38 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Central Cities, Rural Areas, and Other Underserved Areas Home Purchase Subgoal, 33 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Central Cities, Rural Areas, and Other Underserved Areas Housing Goal in the year 2007 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(4) For the year 2008, 39 percent of the total number of dwelling units financed by that GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Central Cities, Rural Areas, and Other Underserved Areas Home Purchase Subgoal, 34 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Central Cities, Rural Areas, and Other Underserved Areas Housing Goal in the year 2008 unless otherwise adjusted by HUD in accordance with FHEFSSA; and

(5) For the year 2009 and thereafter HUD shall establish annual goals. Pending establishment of goals for the year 2009 and thereafter, the annual goal for each of those years shall be 39 percent of the total number of dwelling units financed by that GSE's mortgage purchases in each of those years. In addition, as a Central Cities, Rural Areas, and Other Underserved Areas Home Purchase Subgoal, 34 percent of the total number of home purchase mortgages in metropolitan areas financed by that GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Central Cities, Rural Areas, and Other Underserved Areas Housing Goal in each of those years unless otherwise adjusted by HUD in accordance with FHEFSSA.

* * * * *

5. In § 81.14, revise the last sentence of paragraph (b) and revise paragraph (c), to read as follows:

§ 81.14

Special Affordable Housing Goal.

* * * * *

(b) * * * A statement documenting HUD's considerations and findings with respect to these factors, entitled “Departmental Considerations to Establish the Special Affordable Housing Goal,” was published in the Federal Register on November 2, 2004.

(c) Goals. The annual goals for each GSE's purchases of mortgages on rental and owner-occupied housing meeting the then-existing, unaddressed needs of and affordable to low-income families in low-income areas and very low-income families are:

(1) For the year 2005, 22 percent of the total number of dwelling units financed by each GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. The goal for the year 2005 shall include mortgage purchases financing dwelling units in multifamily housing totaling not less than 1.0 percent of the average annual dollar volume of combined (single-family and multifamily) mortgages purchased by the respective GSE in 2000, 2001, and 2002, unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Special Affordable Housing Home Purchase Subgoal, 17 percent of the total number of home purchase mortgages in metropolitan areas financed by each GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Special Affordable Housing Goal in the year 2005 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(2) For the year 2006, 23 percent of the total number of dwelling units financed by each GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. The goal for the year 2006 shall include mortgage purchases financing dwelling units in multifamily housing totaling not less than 1.0 percent of the average annual dollar volume of combined (single-family and multifamily) mortgages purchased by the respective GSE in 2000, 2001, and 2002, unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Special Affordable Housing Home Purchase Subgoal, 17 percent of the total number of home purchase mortgages in metropolitan areas financed by each GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Special Affordable Housing Goal in the year 2006 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(3) For the year 2007, 25 percent of the total number of dwelling units financed by each GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. The goal for the year 2007 shall include mortgage purchases financing dwelling units in multifamily housing totaling not less than 1.0 percent of the average annual dollar volume of combined (single-family and multifamily) mortgages purchased by the respective GSE in 2000, 2001, and 2002, unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Special Affordable Housing Home Purchase Subgoal, 18 percent of the total number of home purchase mortgages in metropolitan areas financed by each GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Special Affordable Housing Goal in the year 2007 unless otherwise adjusted by HUD in accordance with FHEFSSA;

(4) For the year 2008, 27 percent of the total number of dwelling units financed by each GSE's mortgage purchases unless otherwise adjusted by HUD in accordance with FHEFSSA. The goal for the year 2008 shall include mortgage purchases financing dwelling units in multifamily housing totaling not less than 1.0 percent of the average annual dollar volume of combined (single-family and multifamily) mortgages purchased by the respective GSE in 2000, 2001, and 2002, unless otherwise adjusted by HUD in accordance with FHEFSSA. In addition, as a Special Affordable Housing Home Purchase Subgoal, 18 percent of the total number of home purchase mortgages in metropolitan areas financed by each GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Special Affordable Housing Goal in the year 2008 unless otherwise adjusted by HUD in accordance with FHEFSSA; and

(5) For the year 2009 and thereafter HUD shall establish annual goals. Pending establishment of goals for the year 2009 and thereafter, the annual goal for each of those years shall be 27 percent of the total number of dwelling units financed by each GSE's mortgage purchases in each of those years. The goal for each such year shall include mortgage purchases financing dwelling units in multifamily housing totaling not less than 1.0 percent of the annual average dollar volume of combined (single-family and multifamily) mortgages purchased by the respective GSE in the years 2000, 2001, and 2002. In addition, as a Special Affordable Housing Home Purchase Subgoal, 18 percent of the total number of home purchase mortgages in metropolitan areas financed by each GSE's mortgage purchases shall be home purchase mortgages in metropolitan areas which count toward the Special Affordable Housing Goal in each of those years unless otherwise adjusted by HUD in accordance with FHEFSSA.

* * * * *

6. In § 81.15, revise paragraphs (d), (e)(6)(i), and (e)(6)(ii) and add a new paragraph (i), to read as follows:

§ 81.15

General requirements.

* * * * *

(d) Counting owner-occupied units. (1) For purposes of counting owner-occupied units toward achievement of the Low- and Moderate-Income Housing Goal or the Special Affordable Housing Goal, mortgage purchases financing such units shall be evaluated based on the income of the mortgagors and the area median income at the time of origination of the mortgage. To determine whether mortgages may be counted under a particular family income level, i.e., especially low, very low, low or moderate income, the income of the mortgagors is compared to the median income for the area at the time of the mortgage application, using the appropriate percentage factor provided under § 81.17.

(2)(i) When the income of the mortgagor(s) is not available to determine whether an owner-occupied unit in a property securing a single-family mortgage originated after 1992 and purchased by a GSE counts toward achievement of the Low- and Moderate-Income Housing Goal or the Special Affordable Housing Goal, a GSE's performance with respect to such unit may be evaluated using estimated affordability information in accordance with one of the following methods:

(A) Excluding from the denominator and the numerator single-family owner-occupied units located in census tracts with median incomes less than, or equal to, area median income based on the most recent decennial census, up to a maximum of one percent of the total number of single-family owner-occupied dwelling units eligible to be counted toward the respective housing goal in the current year. Mortgage purchases with missing data in excess of the maximum will be included in the denominator and excluded from the numerator;

(B) For home purchase mortgages and for refinance mortgages separately, multiplying the number of owner-occupied units with missing borrower income information in properties securing mortgages purchased by the GSE in each census tract by the percentage of all single-family owner-occupied mortgage originations in the respective tracts that would count toward achievement of each goal, as determined by HUD based on the most recent HMDA data available; or

(C) Such other data source and methodology as may be approved by HUD.

(ii) In any calendar year, a GSE may use only one of the methods specified in paragraph (d)(2)(i) of this section to estimate affordability information for single-family owner-occupied units.

(iii) If a GSE chooses to use an estimation methodology under paragraph (d)(2)(i)(B) or (d)(2)(i)(C) of this section to determine affordability for owner-occupied units in properties securing single-family mortgage purchases eligible to be counted toward the respective housing goal, then that methodology may be used up to nationwide maximums for home purchase mortgages and for refinance mortgages that shall be calculated by multiplying, for each census tract, the percentage of all single-family owner-occupied mortgage originations with missing borrower incomes (as determined by HUD based on the most recent HMDA data available for home purchase and refinance mortgages, respectively) by the number of single-family owner-occupied units in properties securing mortgages purchased by the GSE for each census tract, summed up over all census tracts. If this nationwide maximum is exceeded, then the estimated number of goal-qualifying units will be adjusted by the ratio of the applicable nationwide maximum number of units for which income information may be estimated to the total number of single-family owner-occupied units with missing income information in properties securing mortgages purchased by the GSE. Owner-occupied units in excess of the nationwide maximum, and any units for which estimation information is not available, shall remain in the denominator of the respective goal calculation.

(e) * * *

(6) * * *

(i) Multifamily. (A) When a GSE lacks sufficient information to determine whether a rental unit in a property securing a multifamily mortgage purchased by a GSE counts toward achievement of the Low- and Moderate-Income Housing Goal or the Special Affordable Housing Goal because neither the income of prospective or actual tenants, nor the actual or average rental data, are available, a GSE's performance with respect to such unit may be evaluated using estimated affordability information in accordance with one of the following methods:

(1) Multiplying the number of rental units with missing affordability information in properties securing multifamily mortgages purchased by the GSE in each census tract by the percentage of all rental dwelling units in the respective tracts that would count toward achievement of each goal, as determined by HUD based on the most recent decennial census. For units with missing affordability information in tracts for which such methodology is not possible, such units will be excluded from the denominator as well as the numerator in calculating performance under the respective housing goal(s); or

(2) Such other data source and methodology as may be approved by HUD.

(B) In any calendar year, a GSE may use only one of the methods specified in paragraph (e)(6)(i)(A) of this section to estimate affordability information for multifamily rental units.

(C) If a GSE chooses to use an estimation methodology under paragraph (e)(6)(i)(A) of this section to determine affordability for rental units in properties securing multifamily mortgage purchases eligible to be counted toward the respective housing goal, then that methodology may be used up to a nationwide maximum of ten percent of the total number of rental units in properties securing multifamily mortgages purchased by the GSE in the current year. If this maximum is exceeded, the estimated number of goal-qualifying units will be adjusted by the ratio of the nationwide maximum number of units for which affordability information may be estimated to the total number of multifamily rental units with missing affordability information in properties securing mortgages purchased by the GSE. Multifamily rental units in excess of the maximum set forth in this paragraph (e)(6)(i)(C), and any units for which estimation information is not available, shall be removed from the denominator of the respective goal calculation.

(ii) Rental units in 1-4 unit single-family properties. (A) When a GSE lacks sufficient information to determine whether a rental unit in a property securing a single-family mortgage purchased by a GSE counts toward achievement of the Low- and Moderate-Income Housing Goal or the Special Affordable Housing Goal because neither the income of prospective or actual tenants, nor the actual or average rental data, are available, a GSE's performance with respect to such unit may be evaluated using estimated affordability information in accordance with one of the following methods:

(1) Excluding rental units in 1-to 4-unit properties with missing affordability information from the denominator as well as the numerator in calculating performance under those goals;

(2) Multiplying the number of rental units with missing affordability information in properties securing single family mortgages purchased by the GSE in each census tract by the percentage of all rental dwelling units in the respective tracts that would count toward achievement of each goal, as determined by HUD based on the most recent decennial census. For units with missing affordability information in tracts for which such methodology is not possible, such units will be excluded from the denominator as well as the numerator in calculating performance under the respective housing goal(s); or

(3) Such other data source and methodology as may be approved by HUD.

(B) In any calendar year, a GSE may use only one of the methods specified in paragraph (e)(6)(ii)(A) of this section to estimate affordability information for single-family rental units.

(C) If a GSE chooses to use an estimation methodology under paragraph (e)(6)(ii)(A)(2) or (e)(6)(ii)(A)(3) of this section to determine affordability for rental units in properties securing single-family mortgage purchases eligible to be counted toward the respective housing goal, then that methodology may be used up to nationwide maximums of five percent of the total number of rental units in properties securing non-seasoned single-family mortgage purchases by the GSE in the current year and 20 percent of the total number of rental units in properties securing seasoned single-family mortgage purchases by the GSE in the current year. If either or both of these maximums are exceeded, the estimated number of goal-qualifying units will be adjusted by the ratio of the applicable nationwide maximum number of units for which affordability information may be estimated to the total number of single-family rental units with missing affordability information in properties securing seasoned or unseasoned mortgages purchased by the GSE, as applicable. Single-family rental units in excess of the maximums set forth in this paragraph (e)(6)(ii)(C), and any units for which estimation information is not available, shall be removed from the denominator of the respective goal calculation.

* * * * *

(i) Counting mortgages toward the Home Purchase Subgoals. (1) General. The requirements of this section, except for paragraphs (b) and (e) of this section, shall apply to counting mortgages toward the Home Purchase Subgoals at §§ 81.12 through 81.14. However, performance under the subgoals shall be counted using a fraction that is converted into a percentage for each subgoal and the numerator of the fraction for each subgoal shall be the number of home purchase mortgages in metropolitan areas financed by each GSE's mortgage purchases in a particular year that count towards achievement of the applicable housing goal. The denominator of each fraction shall be the total number of home purchase mortgages in metropolitan areas financed by each GSE's mortgage purchases in a particular year. For purposes of each subgoal, the procedure for addressing missing data or information, as set forth in paragraph (d) of this section, shall be implemented using numbers of home purchase mortgages in metropolitan areas and not single-family owner-occupied dwelling units.

(2) Special counting rule for mortgages with more than one owner-occupied unit. For purposes of counting mortgages toward the Home Purchase Subgoals, where a single home purchase mortgage finances the purchase of two or more owner-occupied units in a metropolitan area, the mortgage shall count once toward each subgoal that applies to the GSE's mortgage purchase.

7. In § 81.16, revise paragraphs (c)(6) and (c)(7), remove and reserve paragraphs (c)(10) and (c)(11), and add a paragraph (c)(14), to read as follows:

§ 81.16

Special counting requirements.

* * * * *

(c) * * *

(6) Seasoned mortgages. A GSE's purchase of a seasoned mortgage shall be treated as a mortgage purchase for purposes of these goals and shall be included in the numerator, as appropriate, and the denominator in calculating the GSE's performance under the housing goals, except where:

(i) The GSE has already counted the mortgage under a housing goal applicable to 1993 or any subsequent year; or

(ii) HUD determines, based upon a written request by a GSE, that a seasoned mortgage or class of such mortgages should be excluded from the numerator and the denominator in order to further the purposes of the Special Affordable Housing Goal.

(7) Purchase of refinanced mortgages. Except as otherwise provided in this part, the purchase of a refinanced mortgage by a GSE is a mortgage purchase and shall count toward achievement of the housing goals to the extent the mortgage qualifies.

* * * * *

(14) Seller dissolution option. (i) Mortgages acquired through transactions involving seller dissolution options shall be treated as mortgage purchases, and receive credit toward the achievement of the housing goals, only when:

(A) The terms of the transaction provide for a lockout period that prohibits the exercise of the dissolution option for at least one year from the date on which the transaction was entered into by the GSE and the seller of the mortgages; and

(B) The transaction is not dissolved during the one-year minimum lockout period.

(ii) The Secretary may grant an exception to the one-year minimum lockout period described in paragraph (c)(14)(i)(A) and (B) of this section, in response to a written request from an enterprise, if the Secretary determines that the transaction furthers the purposes of FHEFSSA and the GSE's charter act;

(iii) For purposes of this paragraph (c)(14), “seller dissolution option” means an option for a seller of mortgages to the GSEs to dissolve or otherwise cancel a mortgage purchase agreement or loan sale.

* * * * *

8. Revise § 81.102 to read as follows:

§ 81.102

Verification and enforcement to ensure GSE data integrity.

(a) Independent verification authority. The Secretary may independently verify the accuracy and completeness of the data, information, and reports provided by each GSE, including conducting on-site verification, when such steps are reasonably related to determining whether a GSE is complying with 12 U.S.C. 4541-4589 and the GSE's Charter Act.

(b) Certification. (1) The senior officer of each GSE who is responsible for submitting to HUD the fourth quarter Annual Mortgage Report and the AHAR under sections 309(m) and (n) of the Fannie Mae Charter Act or sections 307(e) and (f) of the Freddie Mac Act, as applicable, or for submitting to the Secretary such other report(s), data, or information for which certification is requested in writing by the Secretary, shall certify such report(s), data or information.

(2) The certification shall state as follows: “To the best of my knowledge and belief, the information provided herein is true, correct and complete.”

(3) If the Secretary determines that a GSE has failed to provide the certification required by paragraphs (b)(1) and (b)(2) of this section, or that a GSE has provided the certification required by paragraph (b) in connection with data, information or report(s) that the Secretary later determines are not true, correct and complete, the Secretary may pursue the enforcement remedies under paragraph (e) of this section. For data, information or report(s) subject to paragraphs (c) or (d) of this section, the Secretary may pursue the enforcement remedies described in paragraph (e) only in connection with material errors, omissions or discrepancies as those terms are defined in § 81.102(c) or (d).

(c) Verification procedure and adjustment to correct errors, omissions or discrepancies in AHAR data for the immediately preceding year. (1) This paragraph (c) pertains to the GSEs' submission of year-end data. For purposes of this paragraph, “year-end data” means data that HUD receives from the GSEs related to housing goals performance in the immediately preceding year and covering data reported in the fourth quarter Annual Mortgage Report and the GSE's AHAR. An “error” means a technical mistake, such as a mistake in coding or calculating data. An “omission” means a GSE's failure to count units in the denominator. A “discrepancy” means any difference between HUD's analysis of data and the analysis contained in a GSE's submission of data, including a discrepancy in goal or Special Affordable subgoal performance.

(2) If HUD finds errors, omissions or discrepancies in a GSE's year-end data submissions relative to HUD's regulations, HUD will first notify the GSE by telephone or e-mail transmission of each such error, omission or discrepancy. The GSE must respond within five working days of each such notification. HUD may, in its discretion or upon a request by a GSE within the five working day period, extend the response period for up to an additional 20 working days. Information exchanges during the five working day period following initial notification, and any subsequent extensions of time that may be granted, may be by electronic mail. Any person with delegated authority from the Secretary, or the Director of HUD's Financial Institution Regulation Division, or his or her designee, shall be responsible for issuing initial notifications regarding errors, omissions, or discrepancies; making determinations on the adequacy of responses received; approving any extensions of time permitted under this provision; and managing the data verification process.

(3) If each error, omission or discrepancy is not resolved to HUD's satisfaction during the initial five working day period from notification, and any extension period, the Secretary will notify the GSE in writing and seek clarification or additional information to correct the error, omission or discrepancy. The GSE shall have 10 working days (or such longer period as the Secretary may establish, not to exceed 30 working days) from the date of the Secretary's written notice to respond in writing to the notice. If the GSE fails to submit a written response to the Secretary within this period, or if the Secretary determines that the GSE's written response fails to correct or otherwise resolve each error, omission or discrepancy in its reported year-end data to the Secretary's satisfaction, the Secretary will determine the appropriate adjustments to the numerator and the denominator of the applicable housing goal(s) and Special Affordable subgoal(s) due to the GSE's failure to provide the Secretary with accurate submissions of data.

(4) The Secretary, or his or her designee, shall inform a GSE in writing, at least five working days prior to HUD's release of its official performance figures to the public, of HUD's determination of official goals performance figures, including any adjustments. During the five working days prior to such public release, a GSE may request, in writing, a reconsideration of HUD's final determination of its performance and must provide the basis for requesting the reconsideration. If the request is granted, the Secretary will consider the GSE's request for reconsideration of its determination of goals performance and make a final determination regarding the GSE's performance, within 10 working days of the Secretary's granting of the GSE's written request for reconsideration.

(5) Should the Secretary determine that additional enforcement action against the GSE is warranted for material errors, omissions or discrepancies with regard to a housing goal or Special Affordable subgoal, it may pursue additional remedies under paragraph (e) of this section. An error, omission or discrepancy is material if it results in an overstatement of credit for a housing goal or Special Affordable subgoal, and, without such overstatement, the GSE would have failed to meet such housing goal or Special Affordable subgoal for the immediately preceding year.

(d) Adjustment to correct prior year reporting errors, omissions or discrepancies. (1) General. The Secretary may require a GSE to correct a material error, omission or discrepancy in a GSE's prior year's data reported in the fourth quarter Annual Mortgage Report and the GSE's AHAR under sections 309(m) and (n) of the Fannie Mae Charter Act or sections 307(e) and (f) of the Freddie Mac Act, as applicable. An error, omission or discrepancy is material if it results in an overstatement of credit for a housing goal or Special Affordable subgoal and, without such overstatement, the GSE would have failed to meet such housing goal or Special Affordable subgoal for the prior year. A “prior year” for purposes of this section is any one of the two years immediately preceding the latest year for which data on housing goals performance was reported to HUD.

(2) Procedural requirements. In the event the Secretary determines that a GSE's prior year's fourth quarter Annual Mortgage Report or AHAR contain a material error, omission or discrepancy, the Secretary will provide the GSE with an initial letter containing written findings and determinations within 24 months of the end of the relevant GSE reporting year. The GSE shall have an opportunity, not to exceed 30 days from the date of receipt of the Secretary's initial letter, to respond in writing with supporting documentation, to contest the Secretary's initial determination that there was a material error, omission or discrepancy in a prior year's data. The Secretary shall then issue a final determination letter within 60 days of the date of HUD's receipt of the GSE's written response or, if no response is received, within 90 days of the date of the GSE's receipt of the Secretary's initial letter. The Secretary may extend the period for issuing a final determination letter by an additional 30 days and may grant the GSE an opportunity, for a period not to exceed 10 working days from the date of the GSE's receipt of the determination letter to request that the determination be reconsidered.

(3) If the Secretary determines that a GSE's prior year's fourth quarter Annual Mortgage Report or AHAR contained a material error, omission or discrepancy, the Secretary may direct the GSE to correct the overstatement by purchasing mortgages to finance the number of units that HUD has determined were overstated in the prior year's goal performance (or, for the Special Affordable subgoal, the number or dollar amount, as applicable, of mortgage purchases that HUD has determined were overstated), or that equal the percentage of the overstatement in the prior year's goal or Special Affordable subgoal performance as applied to the most current year-end performance, whichever is less. Units or mortgages purchased to remedy an overstatement in the housing goals or the Special Affordable subgoal must be eligible to qualify under the same goal or Special Affordable subgoal that HUD has determined were overstated in the prior year.

(4) If a GSE does not purchase a sufficient amount or type of mortgages to meet the requirements set forth in paragraph (d)(3) of this section as directed by the Secretary by no later than the end of the calendar year immediately following the year in which the Secretary notifies the GSE of such overstatement (unless, upon written request from the GSE, the Secretary, in his or her discretion, determines that a grant of additional time is appropriate to correct or compensate for the overstatement) the Department may pursue any or all of the following remedies:

(i) Issue a notice that the GSE has failed a housing goal or Special Affordable subgoal in the prior year;

(ii) Seek additional enforcement remedies under paragraph (e) of this section;

(iii) Pursue any other civil or administrative remedies as are available to it.

(e) Additional enforcement options. (1) General. In the event the Secretary determines, either as a result of his or her independent verification authority described in paragraph (a) of this section, or by the authority set forth in paragraphs (b), (c) or (d) of this section, that any of the following circumstances has occurred with respect to data, information or report(s) required by sections 309(m) or (n) of the Fannie Mae Charter Act, sections 307(e) or (f) of the Freddie Mac Act, or subpart E of this part, the Secretary may regard this as a GSE's failure to submit such data, information or report(s) and, accordingly, the Secretary may take the additional enforcement actions authorized by paragraph (e)(2) of this section:

(i) A GSE fails to submit the certification required by paragraphs (b)(1) and (b)(2) of this section in connection with such data, information or report(s); or

(ii) A GSE submits the certification required by paragraph (b) of this section, but the Secretary later determines that the data, information or report(s) are not true, correct and complete. For data, information or report(s) subject to paragraphs (c) or (d) of this section, the Secretary may pursue the additional enforcement remedies under paragraph (e)(2) only in connection with material errors, omissions or discrepancies, as those terms are defined in § 81.102(c) or (d). In addition, the Secretary may only pursue such remedies in connection with material errors, omissions or discrepancies arising under paragraph (d) of this section if the GSE has failed to purchase a sufficient amount or type of mortgages, as provided in paragraphs (d)(3) and (d)(4) of this section.

(2) Remedies. (i) Submissions required under the GSE's charter acts. After the Secretary makes a determination under paragraph (e)(1) of this section that any of the circumstances described in paragraphs (e)(1)(i) or (ii) has occurred with respect to data, information, or report(s) required by sections 309(m) or (n) of the Fannie Mae Charter Act, or by sections 307(e) or (f) of the Freddie Mac Act, the Secretary may pursue any or all of the following remedies in accordance with paragraph (e)(3), or applicable law, as appropriate:

(A) A cease-and-desist order against the GSE for failing to submit the required data, information or report(s) in accordance with this section;

(B) Civil money penalties against the GSE for failing to submit the required data, information or report(s) in accordance with this section;

(C) Any other civil or administrative remedies or penalties against the GSE that may be available to the Secretary by virtue of the GSE's failing to submit or certify the required data, information or report(s) in accordance with this section.

(ii) Submissions required under subpart E of this part. After the Secretary makes a determination under paragraph (e)(1) of this section that any of the circumstances described in paragraphs (e)(1)(i) or (ii) has occurred with respect to data, information or report(s) required under subpart E of this part (but that are not required by sections 309(m) or (n) of the Fannie Mae Charter Act or by sections 307(e) or (f) of the Freddie Mac Act), the Secretary may pursue any civil or administrative remedies or penalties against the GSE that may be available to the Secretary. The Secretary shall pursue such remedies under applicable law.

(3) Procedures. The Secretary shall comply with the procedures set forth in subpart G of this part in connection with any enforcement action that he or she may initiate against a GSE under paragraph (e) of this section.

Dated: October 22, 2004.

John C. Weicher,

Assistant Secretary for Housing—Federal Housing Commissioner.

Note:

The Appendices will not appear in the Code of Federal Regulations.

Appendix A

Departmental Considerations To Establish The Low- and Moderate-Income Housing Goal

A. Introduction

Sections 1 and 2 provide a basic description of the rule process. Section 3 discusses HUD's conclusions based on consideration of the factors.

1. Establishment of Low- and Moderate-Income Housing Goal

In establishing the Low- and Moderate-Income Housing Goals for the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac), collectively referred to as the Government-Sponsored Enterprises (GSEs), section 1332 of the Federal Housing Enterprises Financial Safety and Soundness Act of 1992 (12 U.S.C. 4562) (FHEFSSA) requires the Secretary to consider:

1. National housing needs;

2. Economic, housing, and demographic conditions;

3. The performance and effort of the enterprises toward achieving the Low- and Moderate-Income Housing Goal in previous years;

4. The size of the conventional mortgage market serving low- and moderate-income families relative to the size of the overall conventional mortgage market;

5. The ability of the enterprises to lead the industry in making mortgage credit available for low- and moderate-income families; and

6. The need to maintain the sound financial condition of the enterprises.

The Secretary also considered these factors in establishing a low- and moderate-income subgoal for home purchase loans on single-family-owner properties in metropolitan areas.

2. Underlying Data

In considering the statutory factors in establishing these goals, HUD relied on data from the 2001 American Housing Survey, the 2000 Censuses of Population and Housing, the 2001 Residential Finance Survey (RFS), the 1995 Property Owners and Managers Survey (POMS), other government reports, reports submitted in accordance with the Home Mortgage Disclosure Act (HMDA), and the GSEs. In order to measure performance toward achieving the Low- and Moderate-Income Housing Goal in previous years, HUD analyzed the loan-level data on all mortgages purchased by the GSEs for 1993-2003 in accordance with the goal counting provisions established by the Department in the December 1995 and October 2000 rules (24 CFR part 81).

3. Conclusions Based on Consideration of the Factors

The discussion of the first two factors covers a range of topics on housing needs and economic and demographic trends that are important for understanding mortgage markets. Information is provided which describes the market environment in which the GSEs must operate (for example, trends in refinancing activity). In addition, the severe housing problems faced by lower-income families are discussed, as are the barriers that minorities face when attempting to become homeowners. This discussion serves to provide useful background information for the discussion of the Geographically Targeted and Special Affordable Housing Goals in Appendixes B and C, as well as for the Low- and Moderate-Income Housing Goal in this Appendix.

The third factor (past performance) and the fifth factor (ability of the GSEs to lead the industry) are also discussed in some detail in this Appendix. With respect to home purchase mortgages, the past performance of the GSEs and their ability to lead the industry are examined for all three housing goals; that analysis provides the basis for establishing the three subgoals for the GSEs' acquisitions of home loans on single-family-owner properties.

The fourth factor (size of the market) and the sixth factor (need to maintain the GSEs' sound financial condition) are mentioned only briefly in this Appendix. Detailed analyses of the fourth factor and the sixth factor are contained in Appendix D and in the economic analysis of this rule, respectively.

The factors are discussed in sections B through H of this appendix. Section I summarizes the findings and presents the Department's conclusions concerning the Low- and Moderate-Income Housing Goal. Section I also gives the rationale for a low- and moderate-income subgoal for home purchase loans.

The consideration of the factors in this Appendix has led the Secretary to the following conclusions:

• Changing population demographics will result in a need for primary and secondary mortgage markets to meet nontraditional credit needs, respond to diverse housing preferences, and overcome information and other barriers that many immigrants and minorities face. Growing housing demand from immigrants (both those who are already here and those projected to come) and non-traditional homebuyers will help to offset declines in the demand for housing caused by the aging of the population. Immigrants and other minorities—who accounted for more than a third of household growth since the 1990s—will be responsible for almost two-thirds of the growth in the number of new households over the next ten years. As these demographic factors play out, the overall effect on housing demand will likely be sustained growth and an increasingly diverse household population from which to draw new renters and homeowners.
• Despite the record national homeownership rate of 68.3 percent in 2003, much lower rates prevailed for minorities, especially for African-American households (48.4 percent) and Hispanics (47.4 percent), and these lower rates are only partly accounted for by differences in income, age, and other socioeconomic factors.
• In addition to low incomes, barriers to homeownership that disproportionately affect minorities and immigrants include lack of capital for down payments and closing costs, poor credit history, lack of access to mainstream lenders, little understanding of the homebuying process, and continued discrimination in housing markets and mortgage lending.
• A HUD-published study of discrimination in the rental and owner markets found that while differential treatment between minority and white home seekers had declined over the past ten years, it continued at an unacceptable level in the year 2000. In addition, disparities in mortgage lending continued across the nation in 2002, when the loan denial rate was 7.8 percent for white mortgage applicants, but 20.1 percent for African Americans and 15.5 percent for Hispanics.^[1]

• Americans with the lowest incomes face persistent housing problems. Recent HUD analysis reveals that in 2001, 5.1 million households had “worst case” housing needs, defined as housing costs greater than 50 percent of household income or severely inadequate housing among unassisted very-low-income renter households. Among these households, 90 percent had a severe rent burden, 6 percent lived in severely inadequate housing, and 4 percent suffered from both problems.
• Over the past ten years, there has been a “revolution in affordable lending” that has extended homeownership opportunities to historically underserved households. Fannie Mae and Freddie Mac have been a substantial part of this “revolution in affordable lending”. During the mid-to-late 1990s, they added flexibility to their underwriting guidelines, introduced new low-down-payment products, and worked to expand the use of automated underwriting in evaluating the creditworthiness of loan applicants. HMDA data suggest that the industry and GSE initiatives are increasing the flow of credit to underserved borrowers. Between 1993 and 2003, conventional loans to low-income and minority families increased at much faster rates than loans to upper-income and non-minority families.
• The Low- and Moderate-Income Goal was set at 50 percent beginning in 2001. Effective on January 1, 2001, several changes in counting requirements came into effect, including (1) “bonus points” (double credit) for purchases of mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on 2-4 unit owner-occupied properties; and (2) a “temporary adjustment factor” (1.35 units credit) for Freddie Mac's purchases of mortgages on large (>50 unit) multifamily properties. With these two counting rules, Fannie Mae's performance was 51.5 percent in 2001, 51.8 percent in 2002 and 52.3 percent in 2003, and Freddie Mac's performance was 53.2 percent in 2001, 50.5 percent in 2002, and 51.2 percent in 2003; thus, both GSEs surpassed this higher goal in all three years.
• The bonuses and temporary adjustment factor expired at the end of 2003. Without these rules, Fannie Mae's performance would have been 51.3 percent in 2000, 49.2 percent in 2001, 49.0 percent in 2002, and 48.7 percent in 2003. Freddie Mac's performance would have been 50.6 percent in 2000, 47.7 percent in 2001, 46.1 percent in 2002, and 45.0 percent in 2003. Thus, both Fannie Mae and Freddie Mac would have surpassed the 50 percent goal in 2000 and fallen short in 2001, 2002, and 2003.
• This Appendix includes a comprehensive analysis of each GSE's performance in funding home purchase mortgages for borrowers and neighborhoods covered by the three housing goals—special affordable and low- and moderate-income borrowers and underserved. The GSEs' performance in funding first-time home buyers is also examined.
• While Freddie Mac has improved its affordable lending performance in recent years, it has consistently lagged the conventional conforming market in funding affordable home purchase loans for special affordable and low-moderate-income borrowers and underserved neighborhoods targeted by the housing goals.^[2] In 2003, its performance on the underserved areas goal was particularly low relative to both the performances of Fannie Mae and the market; in that year, underserved area loans accounted for only 24.0 percent of Freddie Mac's purchases compared with 26.8 percent of Fannie Mae's purchases and 27.6 percent of market originations. (These underserved area data are based on 1990 Census geography.)

• In general, Fannie Mae's affordable lending performance has been better than Freddie Mac's. But like Freddie Mac, Fannie Mae's average performance during past periods (e.g., 1993-2003, 1996-2003, 1999-2003) has been below market levels. However, it is encouraging that Fannie Mae markedly improved its affordable lending performance relative to the market during 2001, 2002, and 2003, the first three years under the higher housing goal targets that HUD established in the GSE Final Rule dated October 2000. Over this three-year period, Fannie Mae led the primary market in funding special affordable and low-mod loans but lagged the market in funding underserved areas loans. In 2003, Fannie Mae's increased performance placed it significantly above the special affordable market (a 17.1 percent share for Fannie Mae compared with a 15.9 percent share for the market) and the low-mod market (a 47.0 percent share for Fannie Mae compared with a 44.6 percent share for the market). However, Fannie Mae continued to lag the underserved areas market in 2003 (a 26.8 percent share for Fannie Mae compared with a 27.6 percent share for the market). In this case, which is referred to in the text as the “purchase year” approach, Fannie Mae's performance is based on comparing its purchases of all loans (both seasoned loans and newly-originated mortgages) during a particular year with loans originated in the market in that year. When Fannie Mae's performance is measured on an “origination year” basis (that is, allocating Fannie Mae's purchases in a particular year to the year that the purchased loan was originated), Fannie Mae also led the 2003 market in funding special affordable and low- and moderate-income loans, and lagged the market in funding underserved area loans.
• Both Fannie Mae and Freddie Mac lag the conventional conforming market in funding first-time homebuyers, and by a rather wide margin. Between 1999 and 2001, first-time homebuyers accounted for 27 percent of each GSE's purchases of home loans, compared with 38 percent for home loans originated in the conventional conforming market.
• The GSEs have accounted for a significant share of the total (government as well as conventional) market for home purchase loans, but their market share for each of the affordable lending categories (e.g., low-income borrowers and census tracts) has been less than their share of the overall market.
• The GSEs also account for a very small share of the market for important groups such as minority first-time homebuyers. Considering the total mortgage market (both government and conventional loans), it is estimated that the GSEs purchased only 14 percent of loans originated between 1999 and 2001 for African-American and Hispanic first-time homebuyers, or one-third of their share (42 percent) of all home purchase loans originated during that period. Considering the conventional conforming market and the same time period, it is estimated that the GSEs purchased only 31 percent of loans originated for African-American and Hispanic first-time homebuyers, or approximately one-half of their share (57 percent) of all home purchase loans in that market. The GSEs' small share of the first-time homebuyer market could be due to the preponderance of high (over 20 percent) downpayment loans in their mortgage purchases.
• This Appendix discusses the dynamic nature of the single-family mortgage market and the numerous changes that this market has undergone over the past few years. Some important trends that will likely factor into the GSEs' performance in meeting the needs of underserved borrowers include the growth of the subprime market, the increasing use of automated underwriting systems, and the introduction of risk-based pricing into the market.
• The long run outlook for the multifamily rental market is sustained, moderate growth, based on favorable demographics. The minority population, especially Hispanics, provides a growing source of demand for affordable rental housing. “Lifestyle renters” (older, middle-income households) are also a fast-growing segment of the rental population. Provision of affordable housing, however, will continue to challenge suppliers of multifamily rental housing and policy makers at all levels of government. Low incomes combined with high housing costs define a difficult situation for millions of renter households. Housing cost reductions are constrained by high land prices and construction costs in many markets. Government action—through land use regulation, building codes, and occupancy standards—are major contributors to those high costs.
• The market for financing multifamily apartments has grown to record volumes. Fannie Mae and Freddie Mac have been among those boosting volumes and introducing new programs to serve the multifamily market. Fannie Mae's multifamily purchases jumped from about $10 billion in 1999 and 2000 to $18.7 billion in 2001, $18.3 billion in 2002, and $33.3 billion in 2003—the last three years were characterized by heavy refinancing activity.
• Freddie Mac has re-entered the multifamily market, after withdrawing for a time in the early 1990s. Concerns regarding Freddie Mac's multifamily capabilities no longer constrain its performance with regard to the housing goals. Freddie Mac's multifamily purchases increased from a relatively low $3 billion in 1997 to approximately $7 billion during the next three years (1998 to 2000), before rising further to $11.9 billion in 2001, $13.3 billion in 2002, and $21.6 billion in 2003.
• The overall presence of both GSEs in the rental mortgage market falls short of their involvement in the single-family owner market. Between 1999 and 2002, the GSEs' purchases totaled for 61 percent of the owner market, but only 37 percent of the single-family rental and multifamily rental market. Certainly there is room for expansion of the GSEs in supporting the nation's rental markets, and that expansion is needed if the GSEs are to make significant progress in closing the gaps between the affordability of their mortgage purchases and that of the overall conventional conforming market.
• Considering both owner and rental properties, the GSEs' presence in the goals-qualifying market has been significantly less than their presence in the overall conventional conforming mortgage market. Specifically, HUD estimates that the GSEs accounted for 55 percent of all owner and rental units financed in the primary market between 1999 and 2002, but only 48 percent of units qualifying for the low-mod goal, 48 percent of units qualifying for the underserved areas goal, and 41 percent of units qualifying for special affordable goal.

B. Factor 1: National Housing Needs

This section reviews the general housing needs of lower-income families that exist today and are expected to continue in the near future. Affordability problems that lower-income families face in both the rental and owner markets are examined. The section also describes racial disparities in homeownership and the causes of these disparities. It also notes some special problems, such as the need to rehabilitate our older urban housing stock, that are discussed throughout this appendix.

1. Homeownership Gaps

Despite recent record homeownership rates, many Americans, including disproportionate numbers of racial and ethnic minorities, are shut out of homeownership opportunities. Although the national homeownership rate for all Americans was 68.3 percent in 2003, the rate for minority households was lower—for example, just 48.4 percent of African-American households and 47.4 percent of Hispanic households owned a home.^[3] Differences in income and age between minorities and whites do not fully explain these gaps. The Joint Center for Housing Studies estimated that if minorities owned homes at the same rates as whites of similar age and income, a homeownership gap of 10 percentage points would still exist.^[4]

a. Importance of Homeownership

Homeownership is one of the most common forms of property ownership as well as savings.^[5] Historically, home equity has been the largest source of wealth for most Americans, and wealth gains in housing have been more widely distributed among the population than gains in the stock market.^[6] With stocks appreciating faster than home prices over the past decade, home equity as a share of all family assets fell from 38 percent in 1989 to 33 percent in 1998 and 32 percent in 2001.^[7] However, many of the gains in the stock market were erased after 1999 and housing was once again a more significant asset in the household balance sheet than stocks in 2001.^[8] Even with a bull market through most of the 1990s, 59 percent of all homeowners in 1998 held more than half of their net wealth in the form of home equity.^[9] From 2001 to 2003, homes prices appreciated an average of 23 percent which meant $30,900 in housing equity accumulation for a typical homeowner.^[10] Moreover, unlike stock wealth, aggregate home equity has steadily increased over the past 40 years with only occasional small dips.^[11]

Among low-income homeowners (household income less than $20,000), home equity accounted for about 72 percent of household wealth, and approximately 55 percent for homeowners with incomes between $20,000 and $50,000. Median net wealth for low-income homeowners under 65 was twelve times that of a similar renter.^[12] Thus a homeownership gap continues to translate directly into a wealth gap. For this reason, President Bush issued the “Homeownership Challenge” in June 2002 to increase minority homeownership by 5.5 million by the end of the decade. By December of 2003, the Census estimated that the number of minority homeowners had increased by 1.53 million. Meaning that in the fourth quarter of 2003, for the first time ever, the majority of minority households are homeowners.^[13]

High rates of homeownership support economic stability within housing and related industries, sectors that contributed nearly one-third of the total gain in real GDP since the beginning of the decade.^[14] In addition, more than half of the refinancing mortgages in the first two years of the decade were cash-out, defined as refinancing procedures by which the mortgage balance is increased by more than five percent in order to tap into home equity. Cash-outs injected more than $300 billion into the economy between 2000 and 2002 and were responsible for one-fifth of real GDP growth since during that period.^[15] In addition to economic benefits such as jobs and residential investment, studies show that the better living environment associated with owning a home has positive impacts on children, in terms of lower rates of teenage pregnancy and higher reading other test scores. The current literature substantiates that the benefits of homeownership extend beyond individual homeowners and their families to society at large. Homeownership promotes social and community stability by increasing the number of stakeholders and reducing disparities in the distributions of wealth and income. The empirical literature is generally supportive of a relationship between homeownership and greater investment in property.^[16] Homeownership is also associated with neighborhood stability (lower mobility), greater participation in voluntary and political activities,^[17] and links to entrepreneurship.^[18]

b. Barriers to Homeownership ^[19]

Insufficient income, high debt burdens, and limited savings are obstacles to homeownership for younger families. As home prices skyrocketed during the late 1970s and early 1980s, real incomes also stagnated, with earnings growth particularly slow for blue collar and less educated workers. Through most of the 1980s, the combination of slow income growth and increasing rents made saving for home purchase more difficult, and relatively high interest rates required large fractions of family income for home mortgage payments. Thus, during that period, fewer households had the financial resources to meet down payment requirements, closing costs, and monthly mortgage payments.

Economic expansion and lower mortgage rates substantially improved homeownership affordability during the 1990s. Many young, low-income, and minority families who were closed out of the housing market during the 1980s re-entered the housing market during the last decade. Even with an economic slowdown in 2000-2001 and climbing house appreciation in 2002-2003, after-tax mortgage payments fell in 2003 for buyers of median priced homes because of historically low interest rates.^[20] However, many households still lack the earning power to take advantage of today's home buying opportunities. Several trends have contributed to the reduction in the real earnings of young adults without college education over the last 15 years, including technological changes that favor white-collar employment, losses of unionized manufacturing jobs, and wage pressures exerted by globalization. Over 42 percent of the nation's population between the ages of 25 and 34 had no advanced education in 2000^[21] and were therefore at risk of being unable to afford homeownership. African Americans and Hispanics, who have lower average levels of educational attainment than whites, are especially disadvantaged by the erosion in wages among less educated workers.

Immigrants and other minorities, who accounted for nearly 40 percent of the growth in the homeownership rate over the past five years, will be responsible for two-thirds of the growth in new households over the next ten years. These groups have unique housing needs and face numerous hurdles in becoming homeowners. In addition to low income, barriers to homeownership that disproportionately affect minorities and immigrants include:

(1) Lack of capital for down payment and closing costs;

(2) Poor credit history;

(3) Lack of access to mainstream lenders;

(4) Complexity and fear of the home buying process; and,

(5) Continued discrimination in housing markets and mortgage lending.

(i) Lack of Cash for Down Payment. In the 2002 Fannie Mae National Housing Survey, 40 percent of Hispanics reported not having enough money for a down payment as an obstacle to buying a home versus 32 percent of all Americans.^[22] A study by Gyourko, Linneman, and Wachter found significant racial differences in homeownership rates in “wealth-constrained” households while finding no racial differences in homeownership rates among households with wealth sufficient to meet down payment and closing costs.^[23] Minorities and immigrants are much less likely to receive gifts and inheritances from their parents to assist them in becoming a homeowner.

(ii) Poor Credit History. Poor credit history also differentially affects minority households. In the same Fannie Mae survey, nearly a third of African-American respondents said their credit rating would be an obstacle to buying a home versus 23 percent of all Americans.^[24] Because African-American and Hispanic borrowers are more likely than others to have little traditional credit history or a poorer credit history, they face increased difficulties in being accepted for mortgage credit. This is because credit history scores (such as a FICO score) are a major component of the new automated mortgage scoring systems. These systems are more likely to refer minority borrowers for more intensive manual underwriting, rather than to automatically accept them for the less costly, expedited processing. In these situations, there is the additional concern that “referred” borrowers may not always receive a manual underwriting for the loan that they initially applied for, but rather be directed to a high-cost subprime loan product.

(iii) Lack of Access to Mainstream Lenders. Minorities face heightened barriers in accessing credit because of their often limited access to mainstream lenders. Access to lenders becomes difficult when mainstream financial institutions are not located in neighborhoods where minorities live. The growth in subprime lending over the last several years has benefited credit-impaired borrowers—those who may have blemishes in their credit record, insufficient credit history, or non-traditional credit sources. Subprime lenders have allowed these borrowers to access credit that they could not otherwise obtain in the prime credit market. However, studies by HUD, The Woodstock Institute and others have shown that subprime lending is disproportionately concentrated in low-income and minority neighborhoods.^[25] While these studies recognize that differences in credit behavior explain some of the disparities in subprime lending across neighborhoods, they argue that the absence of mainstream lenders has also contributed to the concentration of subprime lending in low-income and minority neighborhoods. More competition by prime lenders in inner city neighborhoods could lower the borrowing costs of families who currently have only the option of a high-cost subprime loan. This issue of the lack of mainstream lenders in inner city neighborhoods is discussed further in subsection 2, below, in connection with disparities between neighborhoods.

(iv) Complexity and Fear of Homebuying Process. An additional barrier to homeownership is fear and a lack of understanding about the buying process and the risks of ownership. Many Americans could become homeowners if provided with information to correct myths, misinformation, and concerns about the mortgage process. Some potential homeowners, particularly minorities, are unaware that they may already qualify for a mortgage they can afford. The 2002 Fannie Mae survey revealed that 30 percent of Americans believe erroneously that they need to pay 20 percent of the cost of a home up-front. In addition, Fannie Mae reported that half of Americans are only “somewhat” or “not at all” comfortable with mortgage terms.^[26] Freddie Mac reports that six of 10 Hispanics are uncomfortable with home buying terminology, and think they need “perfect credit” to buy; and less than four in 10 are aware that lenders are not required by law to give them the lowest interest rate possible.^[27] A study using focus groups with renters found that even among those whose financial status would make them capable of homeownership, many felt that the buying process was insurmountable because they feared rejection by the lender or being taken advantage of.^[28]

(v) Discrimination in the Housing and Mortgage Markets. Finally, differential treatment of minorities in the sales and rental markets and in the mortgage lending market has been well documented. The continued discrimination in these markets is discussed in the next section.

2. Disparities in Housing and Mortgage Markets

Sales and Rental Markets. In 2002, HUD released its third Housing Discrimination Study (HDS) in the sale and rental of housing. The study, entitled Discrimination in Metropolitan Housing Markets: National Results from Phase I of The Housing Discrimination Study was conducted by the Urban Institute.^[29] This results of this HDS were based on 4,600 paired tests of minority and non-minority home seekers conducted during 2000 in 23 metropolitan areas nationwide. The report showed large decreases between 1989 and 2000 in the level of discrimination experienced by Hispanics and African Americans seeking to buy a home. There has also been a modest decrease in discrimination toward African Americans seeking to rent a unit. This downward trend, however, has not been seen for Hispanic renters, who now are more likely to experience discrimination in their housing search than do African-American renters. But while generally down since 1989, the report found that housing discrimination still exists at unacceptable levels. The greatest share of discrimination for Hispanic and African-American home seekers can still be attributed to being told units are unavailable when they are available to non-Hispanic whites, and being shown and told about fewer units than comparable non-minority home seekers. Although discrimination is down on most areas for African-American and Hispanic homebuyers, there remain worrisome upward trends of discrimination in the areas of geographic steering for African Americans and, relative to non-Hispanic whites, the amount of help agents provide to Hispanics with obtaining financing. On the rental side, Hispanics were more likely in 2000 than in 1989 to be quoted a higher rent than their white counterpart for the same unit.

Another HUD-sponsored study asked respondents to a nationwide survey if they “thought” they had ever been discriminated against when trying to buy or rent a house or an apartment.^[30] While the responses were subjective, they are consistent with the findings of the HDS. African Americans and Hispanics were considerably more likely than whites to say they have suffered discrimination—24 percent of African Americans and 22 percent of Hispanics perceived discrimination, compared to only 13 percent of whites.

Mortgage Lending Market. Research based on Home Mortgage Disclosure Act (HMDA) data suggests pervasive and widespread disparities in mortgage lending across the Nation. For 2001, the mortgage denial rate for white mortgage applicants was 23 percent, while 36 percent of African-American and 35 percent of Hispanic applicants were denied.

Two recent HUD-sponsored studies of paired-testing at the mortgage pre-application stage also points to discrimination by mortgage lenders. Based on its review of pair tests conducted by the National Fair Housing Alliance, the Urban Institute concluded that differential treatment discrimination at the pre-application level occurred at significant levels in at least some cities.^[31] Minorities were less likely to receive information about loan products, received less time and information from loan officers, and were quoted higher interest rates in most of the cities where tests were conducted. A second HUD-sponsored study by the Urban Institute used the paired testing methodology in Los Angeles and Chicago and found similar results. African Americans and Hispanics faced a significant risk of unequal treatment when they visited mainstream mortgage lending institutions to make pre-application inquiries.^[32]

Several possible explanations for these lending disparities have been suggested. A study by the Boston Federal Reserve Bank found that racial disparities cannot be explained by reported differences in creditworthiness.^[33] In other words, minorities are more likely to be denied than whites with similar credit characteristics, which suggests lender discrimination. In addition, loan officers, who may believe that race is correlated with credit risk, may use race as a screening device to save time, rather than devote effort to distinguishing the creditworthiness of the individual applicant.^[34] This violates the Fair Housing Act.

Underwriting rigidities may fail to accommodate creditworthy low-income or minority applicants. For example, under traditional underwriting procedures, applicants who have conscientiously paid rent and utility bills on time but have never used consumer credit would be penalized for having no credit record. Applicants who have remained steadily employed, but have changed jobs frequently, would also be penalized. As discussed in Section C below, lenders, private mortgage insurers, and the GSEs have been adjusting their underwriting guidelines to take into account these special circumstances of lower-income families. Many of the changes recently undertaken by the industry focused on finding alternative underwriting guidelines to establish creditworthiness that do not disadvantage creditworthy minority or low-income applicants. However, because of the enhanced roles of credit scoring and automated underwriting in the mortgage origination process, it is unclear to what degree the reduced rigidity in industry standards will benefit borrowers who have been adversely impacted by the traditional guidelines as discussed in section C.7, some industry observers have expressed a concern that the greater flexibility in the industry's written underwriting guidelines may not be reflected in the numerical credit and mortgage scores which play a major role in the automated underwriting systems that the GSEs and others have developed.

Disparities Between Neighborhoods. Mortgage credit also appears to be less accessible in low-income and high-minority neighborhoods. As discussed in Appendix B, 2001 HMDA data show that mortgage denial rates are nearly twice as high in census tracts with low-income and/or high-minority composition, as in other tracts (16.8 percent versus 8.7 percent). Numerous studies have found that mortgage denial rates are higher in low-income census tracts, even accounting for other loan and borrower characteristics.^[35] These geographical disparities can be the result of cost factors, such as the difficulty of appraising houses in these areas because of the paucity of previous sales of comparable homes. Sales of comparable homes may also be difficult to find due to the diversity of central city neighborhoods. The small loans prevalent in low-income areas are less profitable to lenders because up-front fees to loan originators are frequently based on a percentage of the loan amount, although the costs incurred are relatively fixed. As noted above, racial disparities in mortgage access may be due to the fact that mainstream lenders are not doing business in certain inner city neighborhoods. There is evidence that mainstream lenders active in white and upper-income neighborhoods are much less active in low-income and minority neighborhoods—often leaving these neighborhoods to unregulated subprime lenders. Geographical disparities in mortgage lending are discussed further in Section C.8 below (which examines subprime lending) and in Appendix B (which examines the Geographically Targeted Goal).

3. Affordability Problems and Worst Case Housing Needs

The severe affordability problems faced by low-income homeowners and renters are documented in HUD's “Worst Case Housing Needs” reports. These reports, which are prepared biennially for Congress, are based on the American Housing Survey (AHS), conducted every two years by the Census Bureau for HUD. The latest detailed report analyzes data from the 1999 AHS. Although it focuses on the housing problems faced by very-low-income renters, it also presents basic data on families and households in owner-occupied housing.^[36]

The “Worst Case” report measures three types of problems faced by homeowners and renters:

1. Cost or rent burdens where housing costs or rent exceed 50 percent of income (a “severe burden”) or range from 31 percent to 50 percent of income (a “moderate burden”);

2. The presence of physical problems involving plumbing, heating, maintenance, hallway, or the electrical system, which may lead to a classification of a residence as “severely inadequate” or “moderately inadequate;” and,

3. Crowded housing, where there is more than one person per room in a residence.

The study reveals that in 2001, 5.1 million very low income renter households had “worst case” housing needs, defined as housing costs greater than 50 percent of household income or severely inadequate housing among unassisted very-low-income renter households.^[37] Among the 5.1 million worst case needs renters, 4.8 million (94 percent) had a severe rent burden and 10 percent of renters lived in housing that was severely inadequate.

a. Problems Faced by Owners

Of the 68.8 million owner households in 1999, 5.8 million (8 percent) confronted a severe cost burden and another 8.7 million (12.7 percent) faced a moderate cost burden. There were 870,000 households with severe physical problems, 2 million with moderate physical problems and 905,000 that were overcrowded. The report found that 25 percent of American homeowners faced at least one severe or moderate problem.

Not surprisingly, problems were most common among very low-income owners.^[38] Almost a third of these households (31 percent) faced a severe cost burden, and an additional 22 percent faced a moderate cost burden. And 8 percent of these families lived in severely or moderately inadequate housing, while 2 percent faced overcrowding. Only 42 percent of very-low-income owners reported no problems.

Over time the percentage of owners faced with severe or moderate physical problems has decreased, as has the portion living in overcrowded conditions. However, affordability problems have become more common—the shares facing severe (moderate) cost burdens were only 3 percent (5 percent) in 1978, but rose to 5 percent (11 percent) in 1989 and 8 percent (13 percent) in 1999. The increase in affordability problems apparently reflects a rise in mortgage debt in the late 1980s and early 1990s, from 21 percent of homeowners' equity in 1983 to 36 percent in 1995.^[39] The Joint Center for Housing Studies also attributes this to the growing gap between housing costs and the incomes of the nation's poorest households.^[40] As a result of the increased incidence of severe and moderate cost burdens, the share of owners reporting no problems fell from 84 percent in 1978 to 78 percent in 1989 and 75 percent in 1999.

Between 1999 and 2001, the number of low income owners with severe cost burdens (meaning those with incomes below 120 percent of AMI and spending more than half of their reported income on housing) shot up by one million. This increase proved to be the main cause of a highly significant nine percent jump in the overall number of low and moderate income owners and renters with critical housing needs. Part of this could be due to the heavy home equity borrowing that has characterized the housing market from the late 1990s to the present day, as well as the fact that increases in house prices have outpaced increases in household income. As a corollary, subprime lending, especially in minority communities, rose by about ten percentage points from the early 1990s to 2001.^[41]

b. Problems Faced by Renters

Problems of all three types listed above are more common among renters than among homeowners. In 1999 there were 6.3 million renter households (19 percent of all renters) who paid more than 50 percent of their income for rent.^[42] Another 7.1 million faced a moderate rent burden. Thus in total 40 percent of renters paid more than 30 percent of their income for rent.

Among very-low-income renters, 71 percent faced an affordability problem, including 40 percent who paid more than half of their income in rent. Almost one-third (31 percent) of renters with incomes between 51 percent and 80 percent of area median family income also paid more than 30 percent of their income for rent.

Affordability problems have increased over time among renters. The shares of renters with severe or moderate rent burdens rose from 32 percent in 1978 to 36 percent in 1989 and 40 percent in 1999.

The share of households living in inadequate housing in 1999 was higher for renters (11 percent) than for owners (4 percent), as was the share living in overcrowded housing (5 percent for renters, but only 1 percent for owners). Crowding and inadequate housing were more common among lower-income renters, but among even the lowest income group, affordability was the dominant problem. The prevalence of inadequate and crowded rental housing diminished over time until 1995, while affordability problems grew.

Other problems faced by renters discussed in the most recent detailed “Worst Case” report include a sharp decline (of 2.3 million, or 14 percent) between 1991 and 1999 in the number of rental units affordable to very-low-income families, and a worsening of the national shortage of units affordable and available to extremely-low-income families (those with incomes below 30 percent of area median income). In 2001, the shortage for extremely-low-income families was approximately 5 million units, not statistically different from the 1999 number. However, between 1999 and 2001, the number of units available to renters with incomes below 50 percent of AMI dropped from 78 units to 76 units per 100 renters, in part because more of the units affordable to this group of renters were occupied by higher-income renters. Shortages of units affordable and available to extremely-low-income households were most pressing in the West and Northeast, especially in metropolitan areas in those regions. In 2001, the West was the only region to experience a significant decline in number of units affordable to renters with incomes below 50 percent of AMI. This decline occurred even in the wake of an increase in affordable units in the West during the 1990s.

4. Rehabilitation and Other National Housing Needs

In addition to the broad housing needs discussed above, there are additional needs confronting specific sectors of the housing and mortgage markets. One example of these specific needs concerns the rehabilitation of the nation's older housing stock. A major problem facing lower-income households is that low-cost housing units continue to disappear from the existing housing stock. Older properties are in need of upgrading and rehabilitation. These aging properties are concentrated in central cities and older inner suburbs, and they include not only detached single-family homes, but also small multifamily properties that have begun to deteriorate. But obtaining the funds to fix up older properties can be difficult. The owners of small rental properties in need of rehabilitation may be unsophisticated in obtaining financing. The properties are often occupied, and this can complicate the rehabilitation process. Lenders may be reluctant to extend credit because of a sometimes-inaccurate perception of high credit risk involved in such loans. The GSEs and other market participants have recently begun to pay more attention to these needs for financing of affordable rental housing rehabilitation. However, extra effort is required, due to the complexities of rehabilitation financing, as there is still a need to do more.

The rehabilitation of our aging housing stock is but one example of the housing and mortgage issues that need to be addressed. Several other examples will be provided throughout the following sections on the economic, housing, and demographic conditions in the single-family and multifamily markets, as well as in Appendices B-D. The discussion will cover a wide range of topics, such as subprime lending, predatory lending, automated underwriting systems, manufactured housing, the special needs of the single-family rental market, and challenges associated with producing affordable multifamily housing—just to name a few.

C. Factor 2: Economic, Housing, and Demographic Conditions: Single-Family Mortgage Market

This section discusses economic, housing, and demographic conditions that affect the single-family mortgage market. After a review of housing trends and underlying demographic conditions that influence homeownership, the discussion focuses on specific issues related to the single-family owner mortgage market. This subsection includes descriptions of recent market interest rate trends, refinance and home purchase activity, homebuyer characteristics, and the state of affordable lending. Other special topics examined include the growth in subprime lending, the increased use of automated underwriting, and the remaining homeownership potential among existing renters. Section D follows with a discussion of the economic, housing, and demographic conditions affecting the mortgage market for multifamily rental properties.

1. Recent Trends in the Housing Market

While most other sectors of the economy were weak or declining during 2001 and 2002, the housing sector showed remarkable strength. Again in 2003, the housing market enjoyed an outstanding year. The numbers of single-family permits, starts, completions, new home sales, and existing home sales were record-breaking. Home ownership was also at an all-time high, and mortgage interest rates continued to stay under six percent on average. In addition, the prosperity of the market stimulated GDP, contributing 0.37 percent to its overall growth rate of 3.1 percent. Although the multifamily sector experienced high vacancies and low lease-up rates, the vitality of the single family market was strong enough to result in a spectacular peak in total permits and starts as well as builders' attitudes and housing affordability.^[43]

Single-Family Permits, Starts, and Completions. Builders took out 1,440,400 single-family permits in 2003, up 6 percent from 2002. The 2003 level was the highest number of single-family permits ever reported in the 44-year history of this series. Single-family starts totaled 1,498,500 housing units, up 10 percent from 2002, a new single-family record. Construction was completed on 1,386,200 single-family housing units, up 5 percent from 2002.

Sales of New and Existing Homes. After leveling out in 2000, housing sales have boomed in the past three years, reaching record highs in 2001, 2002, and again in 2003. New single family home sales, which increased an average 6.3 percent per year between 1992 and 2002, reached a record high of 1,085,000 units in 2003, an increase of 12 percent over 2002 sales. The market for new homes has been strong in the Mid Atlantic, Midwest and Great Plains.

The National Association of Realtors reported that 6.1 million existing homes were sold in 2003, overturning the old record set in 2002 by almost 9 percent, and setting an all-time high in the 35-year history of the series. Combined new and existing home sales set a national record of 6.2 million in 2002 and a record of almost 7.2 million in 2003.

One of the strongest sectors of the housing market in past years had been manufactured homes, but that sector has declined recently. Between 1991 and 1996, manufactured home shipments more than doubled, peaking in 1998 at 373,000. However, shipments fell more than 20 percent in both 2000 and 2001. In 2002, the industry shipped 169,000 new manufactured homes, down 12.4 percent from 2001. This was the lowest number of manufactured home shipments since 1963. In 2003, the number of new manufactured homes shipped plummeted to 131,000, down 22.5 percent from 2002. Repossession has been cited as a cause for the sales drop-off, as has the popularity of conventional stick-built housing.

Homeownership Rate. In 1980, 65.6 percent of Americans owned their own home, but due to the unsettled economic conditions of the 1980s, this share fell to 63.8 percent by 1989. But since 1994, gains in the homeownership rate have occurred in each year, with the rate reaching another record mark of 68.3 percent in 2003.

Gains in homeownership have been widespread over the last eight years.^[44] As a result, the homeownership rate rose from:

• 42.0 percent in 1993 to 48.8 percent in 2003 for African American households,
• 39.4 percent in 1993 to 46.7 percent in 2003 for Hispanic households,
• 73.7 percent in 1993 to 79.1 percent in 2003 for married couples with children,
• 65.1 percent in 1993 to 68.4 percent in 2003 for household heads aged 35-44, and
• 48.9 percent in 1993 to 52.3 percent in 2003 for central city residents.

However, as these figures demonstrate, sizable gaps in homeownership remain.

Economy/Housing Market Prospects. Job growth has been less robust in the recent recovery than some previous recoveries. However, the economy grew at a rate of 2.2 percent in 2002 and even faster in 2003.^[45] Although the Federal Reserve has recently begun raising short term interest rates, mortgage interest rates remain low, supporting housing affordability.

Fannie Mae expects existing home sales to reach 5.7 million in 2004 and 2005.^[46] Projected at 1.84 million in 2003, the National Association of Home Builders expects housing starts to decline to 1.77 million in 2004 and 1.71 million in 2005.^[47] The Mortgage Bankers Association forecasts that 2004 housing starts will total 1.73 million units and the 30-year fixed mortgage rate will average 6.1 percent.^[48] After more than doubling from a relative trough in 2000 to an estimated $2.6 trillion in 2002, Fannie Mae projected in December 2003 that mortgage originations will drop to $1.8 trillion in 2004 and $1.5 trillion in 2005.^[49]

2. Underlying Demographic Conditions

Between 2000 and 2025, the U.S. population is expected to grow by an average of 2.5 million per year.^[50] This will likely result in at least 1.1 million new households per year.^[51] Recently revised increases in population projections by the Census Bureau push population figures higher with the Joint Center estimating new household growth at 13.3 million from 2005 to 2015.^[52] This section discusses important demographic trends behind these overall household numbers that will likely affect housing demand in the future. These demographic forces include the baby-boom, baby-bust and echo baby-boom cycles; immigration trends; non-traditional and single households; “trade-up buyers;” and the growing income inequality between people with different levels of education. HUD's Office of Policy Development and Research funded a study, Issue Papers on Demographic Trends Important to Housing, which analyzes effects of demographic conditions on the housing market. The findings are presented throughout the sections that follow.^[53]

As explained below, the role of traditional first-time homebuyers, 25-to-34-year-old married couples, in the housing market will be smaller in the current decade due to the aging of the population. For the first time in history, the population will have roughly equal numbers of people in every age group. Between 2000 and 2025, the Census Bureau projects that the largest growth in households will occur among householders 65 and over.^[54] Thus, an increasing percentage of the population will be past their home buying peak in the next two decades. However, because homeownership rates do not peak until population groups reach 65 to 74 years of age, this age cohort will continue to provide housing demand. According to Riche, the increasing presence of older households should increase the proportion of the population that owns, rather than rents housing.^[55]

Growing housing demand from immigrants and non-traditional homebuyers will help to offset declines in the demand for housing caused by the aging of the population. Riche's study estimates that minorities will account for two-thirds of the growth in U.S. households over the next 25 years,^[56] and by 2025, non-family households will make up a third of all households. The “echo baby-boom” (that is, children of the baby-boomers) will also add to housing demand in the current and next decades. Finally, the growing income inequality between people with and without a post-secondary education will continue to affect the housing market.

The Baby-Boom Effect. The demand for housing during the 1980s and 1990s was driven, in large part, by the coming of home buying age of the baby-boom generation, those born between 1945 and 1964. Homeownership rates for the oldest of the baby-boom generation, those born in the 1940s, rival those of the generation born in the 1930s. Due to significant house price appreciation in the late-1970s and 1980s, older baby-boomers have seen significant gains in their home equity and subsequently have been able to afford larger, more expensive homes. Circumstances were not so favorable for the middle baby-boomers. Housing was not very affordable during the 1980s, their peak home buying age period. As a result, the homeownership rate, as well as wealth accumulation, for the group of people born in the 1950s lags that of the generations before them.^[57]

As the youngest of the baby-boomers (those born in the 1960s) reached their peak home buying years in the 1990s, housing became more affordable. While this cohort has achieved a homeownership rate equal to the middle baby-boomers, they live in larger, more expensive homes. As the baby-boom generation ages, demand for housing from this group is expected to wind down.^[58]

The baby-boom generation was followed by the baby-bust generation, from 1965 through 1977. Since this population cohort is smaller than that of the baby boom generation, it reduced housing demand in the preceding decade and is expected to do the same in the current decade, though, as discussed below, other factors kept the housing market very strong in the 1990s. However, the echo baby-boom generation (the children of the baby-boomers, who were born after 1977), while smaller than the baby-boom generation, will reach peak home buying age later in the first decade of the millennium.

Immigrant Homebuyers. Past, present, and future immigration will also contribute to gains in the homeownership rate. During the 1990s, 9.8 million legal immigrants entered the United States, as compared to 6.3 million entering in the 1980s and 4.2 million during the 1970s. Overall, the increase in the immigrant population directly accounted for 35 percent of the nation's rise in population in the 1990s.^[59] As a result, the foreign-born population of the United States more than tripled from 9.6 million in 1970 to 31.1 million in 2000. Immigrants who become citizens buy homes at rates nearly as high as their same-aged native-born counterparts and for those aged 25 to 34, the gap is virtually nonexistent.^[60] Moreover, U.S.-born children of immigrants often have higher homeownership rates than the same-age children of native-born parents.^[61] However, there are concerns about the assimilation into homeownership of recent Hispanic immigrants who are less educated than earlier cohorts of immigrants. Many immigrants also locate in high-priced housing markets, which makes it more difficult for them to achieve homeownership.

Although net foreign immigration is projected to decline in the current decade after 2002, high levels of immigration in the late 1980s and throughout the 1990s will have lasting positive effects on housing demand. New immigration in the current and next decades is projected to create 6.9 million net new households, but the majority of household growth in the period (16.9 million) will come from people already resident in the U.S. including the foreign-born population.^[62] While immigrants tend to rent their first homes upon arriving in the United States, homeownership rates are substantial for those that have lived here for at least 6 years. In 1996, the homeownership rate for recent immigrants was 14.7 percent while it was 66.9 percent for foreign-born naturalized citizens after six years.^[63] Higher-than-average foreign-born fertility rates and high rates of homeownership for immigrants living in the country for several years and among the children of immigrants suggest that past immigration will continue to create housing demand.

Past and future immigration will lead to increasing racial and ethnic diversity, especially among the young adult population. As immigrant minorities account for a growing share of first-time homebuyers in many markets, HUD and others will have to intensify their focus on removing discrimination from the housing and mortgage finance systems. The need to meet nontraditional credit needs, respond to diverse housing preferences, and overcome the information barriers that many immigrants face will take on added importance. In order to address these needs, the mortgage industry must offer innovative products and improve outreach efforts to attract minority homebuyers.

Nontraditional and Single Homebuyers. While overall growth in new households has slowed down, nontraditional households have become more important in the homebuyer market. As the population ages both relatively and absolutely, the nation's households will become smaller and more diverse. Riche notes that in 2000, traditional family households represented fewer than one in four households and were surpassed by both single-person households and married couples without children. With later marriages and more divorces, single-parent and single-person households have increased rapidly. In fact, single-parent households grew from 4 percent of family households in 1950 to 12 percent in 2000. Single-person households are now the nation's second most numerous household type, accounting for over 25 percent of all households. In the future, longer life expectancies and the continuing preference for one or two children will make households without children even more numerous. Projected to compose 80 percent of all households by 2025, nontraditional family households will play an increasingly important role in the home buying market.^[64]

Trade-up Buyers. Due to weak house price appreciation, traditional “trade-up buyers” stayed out of the market during the early 1990s. Their absence may explain, in part, the large representation of nontraditional homebuyers during that period. However, since 1995 home prices have increased more than 30 percent.^[65] The greater equity resulting from recent increases in home prices should lead to a larger role for “trade-up buyers” in the housing market during the next 10 to 15 years. In addition, the growing number of higher-income, mid-life households will increase households' potential to “trade up” to more expensive housing.^[66]

Growing Income Inequality. The Census Bureau reported that the top 5 percent of American households received 22.4 percent of aggregate household income in 2001, up from 21.4 percent in 1998 and up sharply from 16.1 percent in 1977. The share accruing to the lowest 80 percent of households fell from 56.5 percent in 1977 to 50.8 percent in 1998 and again to 49.8 percent in 2001. The share of aggregate income accruing to households between the 80th and 95th percentiles of the income distribution was virtually unchanged from 1977 to 2001.^[67]

The increase in income inequality over past decades has been especially significant between those with and those without post-secondary education. The Census Bureau reports that by 1999, the annual earnings of workers with a bachelor's degree were 1.8 times the annual earnings of workers with a high school education.^[68] The inflation-adjusted median earnings of high school graduates were at the same level in 2001 as in 1991 while the earnings of bachelor degree-holders rose nearly 9 percent over the same period.^[69]

So, while homeownership is highly affordable, those without post-secondary education often lack the financial resources to take advantage of the opportunity. As discussed earlier, the days of the well-paying unionized factory job have passed. They have given way to technological change that favors white-collar jobs requiring college degrees, and wages in the manufacturing jobs that remain are experiencing downward pressures from economic globalization. The effect of this is that workers without the benefit of a post-secondary education find their demand for housing constrained. This is especially problematic for recent immigrants who are more likely to have limited educational attainment and English language proficiency.

Summary. Over the next two-and-a-half decades, the number of U.S. households is projected to increase by nearly 27 million. Of these new households, non-Hispanic white and traditional households will contribute only one-third and one-tenth of the growth, respectively. As the baby-boomers aged out of their peak home buying stage and the baby-bust generation aged into their peak home buying stage in the late 1980s, demand for housing was dampened by demographic factors during the 1990s. (Of course, other factors such as low interest rates propelled the housing market to record levels during this period.) As the echo baby-boomers begin to enter their peak home buying age, housing demand should pick up again through the remainder of the current decade and into the next. As these demographic factors play out, the overall effect on housing demand will likely be sustained growth and an increasingly diverse household population from which to draw new homeowners. There are continuing concerns about the increasing income inequality of our population and those recent immigrants and other persons who have limited education.

3. Basic Trends in the Single-Family Mortgage Market

Mortgage lending in the nation is growing at unprecedented levels. Residential mortgage originations soared to $2.5 trillion in 2002, a 22 percent increase over the previous record of $2.06 trillion set in 2001.^[70] Originations then jumped to $3.8 trillion in 2003, with refinances accounting for 66 percent (or $2.5 trillion) of this total.

This boom in lending over the past three years can be attributed to low mortgage interest rates and a record number of refinances. Approximately 40 percent of mortgage debt outstanding, or $2.5 trillion, was refinanced during the 2001-02 refinance boom. Freddie Mac calculates total home equity cashed out in 2002 at 105.4 billion and estimates that number will increase to 138.8 billion in 2003.^[71] This section focuses on recent interest rate trends, the refinance market, the home purchase market, and first-time homebuyers. The section concludes by examining the GSEs' acquisitions as a share of the primary single-family mortgage market, and provides mortgage market prospects.

a. Mortgage Characteristics

Interest Rate Trends and Volatility. Historically low mortgage interest rates in the late 1990s and 2001-2003 helped maintain consumer confidence in the housing sector as the economy emerged from its first recession in almost a decade. After high and fluctuating mortgage rates in the 1980s and early 1990s, recent years have seen a period of lower and more stable rates. The 1980s began with interest rates on mortgages for new homes above 12 percent but quickly rose to more than 15 percent.^[72] By 1987-88, rates dipped into single digits but were rising again by 1989-90. Rates declined in the early 1990s, reaching a low of 6.8 percent in late 1993. An upturn in rates in 1994 and 1995 peaked at 8.3 percent in early 1995. By 1998, 30-year fixed conventional mortgages averaged 6.95 percent, the lowest level since 1968 but saw a rise in 1999 to 7.44 percent. Mortgage rates then continued to rise in 2000, averaging 8.05 percent for the year, before falling to a low of 6.62 percent in October 2001 and averaging 6.97 percent for 2001 as a whole.^[73] Rates averaged 5.83 percent during 2003 ^[74] , reaching a low of 5.23 in June. Rates in 2004 have averaged 5.83 through August, reaching a low of 5.45 in March.^[75]

Other Loan Terms. When mortgage rates are low, most homebuyers prefer to lock in a fixed-rate mortgage (FRM). Adjustable-rate mortgages (ARMs) are more attractive when rates are high, because they carry lower rates than FRMs and because buyers may hope to refinance to an FRM when mortgage rates decline. The Federal Housing Finance Board (FHFB) reports that the ARM share of the market fell from 20 percent in 1993 to a record low of 12 percent in 1998, before rising back to 21 percent in 1999. The ARM share continued to rise to 24 percent in 2000, but then fell dramatically to a low of 12 percent in 2001 as mortgage rates decreased. However, in 2002 and 2003, there was a rebound in the ARM share of the market. Though it still is nowhere near the size it was in the mid to late 1990s, the past two years have seen the share climb to 17 and 19 percent, respectively.^[76]

In 2003, the term-to-maturity was 30 years for 80 percent of conventional home purchase mortgages, continuing to decline after steadily climbing to a high of 90 percent in 2000. The other major term of maturity in 2003 was 15 years (16 percent).^[77]

Low- and no-point mortgages continue to be a popular option for mortgage purchases. FHFB reports that average initial fees and charges (“points”) have decreased from 2.5 percent of loan balance in the mid-1980s to 2 percent in the late-1980s, 1.5 percent in the early 1990s, and less than 1 percent in 1995-97. The downward trend continued throughout the late 1990s with the average initial fees and charges reaching a low of one-half percent in 2001, staying there in 2002, and dipping even further down in 2003. Coupled with declining interest rates, these lower transactions costs have increased the propensity of homeowners to refinance their mortgages.^[78]

Another major change in the conventional home mortgage market has been the proliferation and then diminution of high loan-to-value ratio (LTV) mortgages. According to data from the Federal Housing Finance Board, loans with LTVs greater than 90 percent (that is, down payments of less than 10 percent) made up less than 10 percent of the market in 1989-91, but 25 percent of the market in 1994-97, gradually decreasing to an average of 20 percent of the market in 2003. Loans with LTVs less than or equal to 80 percent fell from three-quarters of the market in 1989-91 to an average of 56 percent of the market in 1994-97, but then rose to an average of 63 percent of mortgages originated in 1998-2001, and rose again to an average of 70 percent of mortgages originated in 2002-2003.^[79] As a result, the average LTV rose from 75 percent in 1989-91 to nearly 80 percent in 1994-97, and then declined to 76.2 percent in 2001, 75.1 percent in 2002, and 73.5 percent in 2003.^[80]

b. Refinance Mortgages

Over the past ten years, refinance booms occurred three times, during 1992-93, 1998, and 2001-03. Refinancing has fueled the growth in total mortgage originations, which were $638 billion in 1995 (a period of low refinance activity), but topped $2.5 trillion in 2002 (a period of heavy refinance activity). The refinance share of total mortgage originations rose to 50 percent in 1998, then decreased to 19 percent in 2000 before jumping to 57 percent in 2001, and 59 percent in 2002. During the 2001-02 refinance boom, approximately 40 percent of the $2.5 trillion in mortgage debt outstanding was refinanced. In 2003, the refinance share of total mortgage originations hit 66 percent, though late 2003 saw a steep drop-off from a 68 percent share in the third quarter to a 49 percent share in the fourth.^[81]

In 1989-90 interest rates exceeded 10 percent, and refinancings accounted for less than 25 percent of total mortgage originations.^[82] The subsequent sharp decline in mortgage rates drove the refinance share over 50 percent in 1992 and 1993 and propelled total single-family originations to more than $1 trillion in 1993—twice the level attained just three years earlier.

The refinance wave subsided after 1993, because most homeowners who found it beneficial to refinance had already done so and because mortgage rates rose once again.^[83] Total single-family mortgage originations bottomed out at $638 billion in 1995, when the refinance share was only 21 percent. Total originations, driven by the volume of refinancings, amounted to $1.507 trillion in 1998, nearly 50 percent higher than the previous record level of $1.02 trillion attained in 1993.

The refinance wave from late 1997 through early 1999 reflected other factors besides interest rates, including greater borrower awareness of the benefits of refinancing, a highly competitive mortgage market, and the enhanced ability of the mortgage industry, utilizing automated underwriting and mortgage origination systems to handle an unprecedented volume of originations. The refinance share decreased to 19 percent in 2000 before jumping to a record 57 percent in 2001.

Historically low interest rates and declining mortgage transaction costs have driven the latest refinancing boom. Given these conditions, the after-tax cost saving on a new, lower-rate loan is much greater than the transaction costs of refinancing. In addition, the appreciation of housing prices has also contributed to the increase in refinancing. Over the past five years, the value of housing rose by approximately $5 trillion, and the rise in value has enabled lenders to service refinancing homeowners because of greater confidence in the creditworthiness of borrowers.^[84]

Over the past few years, homeowners have become more willing to draw on the rising equity in their homes. According to Fannie Mae's 2002 National Housing Survey, homeowners that refinanced during 2001 withdrew about $110 billion in accumulated home equity wealth.^[85] Freddie Mac estimates that more than one-half of all refinance mortgages in the past two years involved cash-out refinancing.^[86]

The refinancing boom contributed to an estimated one-fifth of the national economy's real GDP growth since late 2000.^[87] During 2001 and 2002, roughly $270 billion was raised in cash-out refinancing. Approximately one-half of cash from cash-out refinancing has enabled consumers to finance more spending for expenses such as home improvements, medical payments, education, and vehicles during a weakened economy. Roughly one-third of the cash from cash-out refinancing has allowed consumers to repay other debt.^[88] The remaining cash from cash-out refinancing has enabled consumers to invest in other assets. Refinancing households save approximately $10 billion in their annual interest payments on their mortgage and consumer installment liabilities.

The refinancing boom will have lingering effects. Mortgage borrowers that were able to secure low long-term interest rates through fixed rate mortgages will have more of their budgets to spend on other items. Meanwhile, cash-out borrowers, who are just receiving their money, will spend this year. It must be noted there is some concern regarding the potential for increased credit risk stemming from mortgage debt from cash out borrowers. According to a 2002 Regional Finance Review article, the mortgage liabilities of households have been growing at a rate more than double the growth in household incomes. However, this potential credit risk is moderated by the strong growth in housing values. The ratio of mortgage debt to housing values, the aggregate loan-to-value ratio, has remained fairly stable for a decade.^[89]

c. Home Purchase Mortgages

The volume of home purchase mortgages was $505 billion in 1995, rose to $848 billion in 1999, and remained in the $829-$873 billion range between 1999-2001 before jumping to $1.02 trillion in 2002 and $1.30 trillion in 2003. The Mortgage Bankers Association (MBA) forecasts that the home purchase volume will be $1.52 trillion in 2004 as the home purchase share rises to 57 percent of all originations.^[90] The home purchase share of total mortgage originations was 79 percent in 1995, declined to 50 percent in 1998, rose to 81 in 2000, and sharply fell to 43 percent in 2001, 41 in 2002, and 34 percent in 2003, as refinance mortgage volume grew. This section discusses the important issue of housing affordability and then examines the value of homeownership as an investment.

The National Association of Realtors (NAR) has developed a housing affordability index, calculated as the ratio of median household income to the income needed to qualify for a median price home (the latter income is called the “qualifying income”). In 1993, NAR's affordability index was 133, which meant that the median family income of $37,000 was 33 percent higher than that income needed to qualify for the median priced home. Housing affordability remained at about 130 for 1994-97, with home price increases and somewhat higher mortgage rates being offset by gains in median family income.^[91] Falling interest rates and higher income led to an increase in affordability to 143 in 1998, reflecting the most affordable housing in 25 years. Affordability remained high in 1999, despite the increase in mortgage rates. NAR's affordability index declined from 140 in 1999 to 129 in 2000 as mortgage rates increased. The index turned upward to 136 in 2001 as mortgage rates fell and maintained this average in 2002, before rising further to 140 in 2003.^[92]

Although the share of home purchase loans for lower-income households and/or households living in lower-income communities increased over the past decade, affordability still remains a challenge for many. The median sales price of existing single-family homes in the United States continues to rise, reaching $158,100 in 2002 and $170,000 in 2003. The production of affordable housing and low interest rates could offset the negative impact of rising house prices, which undermine housing affordability for many Americans, particularly in several high-cost markets on the east and west coasts.

As discussed earlier, barriers are preventing many potential homeowners from becoming homeowners, thus reducing the possible amount of home purchase loans. While the strong housing sector has provided financial security for many Americans, a 2002 Fannie Mae survey found that “information barriers still keep many financially qualified families-particularly minority Americans from becoming homeowners or obtaining the lowest-cost financing available to them.” ^[93]

These homeownership barriers pose a serious problem for many Americans who view homeownership as a smart, safe, long-term investment, rating homeownership as a better investment than the stock market. Home equity is the single most important asset for approximately two-thirds of American households that are homeowners. Considering that half of all homeowners held at least 50 percent of their net wealth in home equity in 1998, increasing housing affordability is important for many Americans.^[94]

First-time Homebuyers. First-time homebuyers are a driving force in the nation's mortgage market. The recent low interest rates have made it an opportune time for first-time homebuyers, which are typically people in the 25-34 year-old age group that purchase modestly priced houses. As the post-World War II baby boom generation ages, the percentage of Americans in this age group decreased from 28.3 percent in 1980 to 25.4 percent in 1992.^[95] Even though this cohort is smaller, first-time homebuyers increased their share of home sales. According to Chicago Title data for major metropolitan areas, the first-time buyer share of the homebuyer market increased from roughly 40 percent in the beginning of the 1990s to 45-47 percent during the-mid and late 1990s.^[96] Since the late 1990s, industry survey data suggest that the first-time homebuyer percentage has decreased slightly. In the first quarter of 2003, the share of all home purchases by first-time homebuyers was 40 percent compared to 42 percent in 2001.^[97]

In the 1990s, lenders developed special programs targeted to first-time homebuyers and revised their underwriting standards to enhance homeownership opportunities for low-income families with special circumstances. The disproportionate growth in the number of first-time homebuyers and minority homebuyers largely drove the rising trend in total home purchases. Analysis of the American Housing Survey (AHS) indicates there were 1.3 million new first-time homebuyers during 1991, in comparison with over two million in each year between 1996 and 2001. In addition, first-time homebuyers comprised approximately 60 percent of all minority home purchases during the 1990s, compared with about 35 percent of all home purchases by non-Hispanic white families.

In comparison to repeat homebuyers, first-time homebuyers are more likely to be younger, have lower incomes, and purchase less expensive houses. According to the AHS, more than one-half or first-time homebuyers were below the age of 35, compared with less than one-quarter of repeat buyers in the 1990s. Thirty-nine percent of first-time buyers had incomes below 80 percent of the median compared to 30 percent of repeat buyers. Fifty-four percent of first-time buyers purchased homes priced below $100,000, compared to 37 percent of repeat buyers. Minorities comprise a higher proportion of first-time buyers (32 percent) compared to repeat buyers (14 percent). Compared to repeat buyers, first-time homebuyers are more likely to purchase a home in the central city and more likely to be a female-headed household.^[98]

The National Association of Realtors reports that the average first-time homebuyer in the first quarter of 2003 was 32 years old with a household income of $54,800, compared to an average age of 46 years and average household income of $74,600 for repeat buyers. The average first-time homebuyers made a downpayment of 6 percent on a home that cost $136,000 while the average repeat buyer made a downpayment of 23 percent on a home costing $189,000. In the NAR survey, 37 percent of first-time homebuyers were single compared to 28 percent of repeat buyers.^[99]

Many African Americans and Hispanics are likely to purchase homes in the coming years, contributing to the number of first-time home-buyers fueling growth in the housing sector. The number of homeowners will rise by an average of 1.1 million annually over the next two decades. The sizeable rise in the foreign-born population since the 1970's coupled with the increase in Latin American and Asian immigration will also contribute much to this growth.^[100]

d. GSEs' Acquisitions as a Share of the Primary Single-Family Mortgage Market

Purchases by the GSEs of single-family mortgages amounted to $519 billion during the heavy refinancing year of 1993, stood at $215 billion in 1995, and were at $618 billion during the heavy refinancing year of 1998. Purchases then fell to $395 billion in 2000 before reaching record levels during the heavy refinancing years of 2001 ($961 billion) and 2002 ($1,090 billion). Purchases by Fannie Mae decreased from $316 billion in 1999 to $227 billion in 2000, before rising to $568 billion in 2001, $800 billion in 2002, and $1.3 trillion in 2003. Freddie Mac's single-family mortgage purchases followed a similar trend, falling from $233 billion in 1999 to $168 billion in 2000, and then rising to $393 billion in 2001 and $475 billion in 2002.^[101]

The Office of Federal Housing Enterprise Oversight (OFHEO) estimates that the GSEs' share of total originations in the conventional single-family mortgage market, measured in dollars, declined from 37 percent in 1996 to 32 percent in 1997—well below the peak of 51 percent attained in 1993. OFHEO attributes the 1997 downturn in the GSEs' role to increased holdings of mortgages in portfolio by depository institutions and to increased competition with Fannie Mae and Freddie Mac by private label issuers. However, OFHEO estimates that the GSEs' share of the conventional market rebounded sharply in 1998-99, to 43-42 percent. The GSEs' share then decreased to approximately 30 percent of the single-family conventional mortgages originated in 2000, and then increased sharply to 40 percent in 2001. Total GSE purchases, including loans originated in prior years, amounted to 46 percent of conventional originations in 2001^[102] and approximately 38 percent of family home mortgage originations in 2002.^[103]

e. Mortgage Market Prospects

The Mortgage Bankers Association (MBA) reports that mortgage originations in 2001 were $2.0 trillion, which is almost twice the volume of originations in 2000. Mortgage originations then increased to record levels of $2.5 trillion in 2002 and $3.8 trillion in 2003, with refinancings representing 66 percent of originations and the purchase volume amounting to $1.3 trillion. Estimates indicate that ARMs accounted for 19 percent of total mortgage originations in 2003.^[104] In its September 17, 2004 forecast, MBA predicts that single-family mortgage originations will amount to $2.7 trillion in 2004 and $1.8 trillion in 2005, with refinancings representing 43 percent and 25 percent of originations respectively.

4. Affordable Lending in the Mortgage Market: New Products and Outreach

Extending homeownership opportunities to historically underserved households has been a growing concern for conventional lenders, private mortgage insurers and the GSEs. The industry has responded in what some have called a “revolution in affordable lending”. The industry has offered more customized mortgage products, more flexible underwriting, and expanded outreach so that the benefits of the mortgage market can be extended to those who have not been adequately served through traditional products, underwriting, and marketing.

Fannie Mae and Freddie Mac have been a part of this “revolution in affordable lending”. During the mid-to-late 1990s, they added flexibility to their purchase guidelines, they introduced new low-down-payment products, and they worked to expand the use of credit scores and automated underwriting in evaluating the creditworthiness of loan applicants. These major trends reflect changes in the GSEs' underwriting that have impacted affordable lending. Through these trends, Fannie Mae and Freddie Mac have attempted to increase their capacity to serve low- and moderate-income homebuyers.

This section summarizes recent initiatives undertaken by the GSEs and others in the industry to expand affordable housing. The end of this section will present evidence that these new industry initiatives are working, as increased mortgage credit has been flowing to low-income and minority families. The following section will continue the affordable lending theme by examining the performance of different market sectors (e.g., depositories, GSEs, etc.) in funding loans for low-income and minority families. That section will also discuss the important role that FHA plays in making affordable housing available to historically underserved groups as well as the continuing concern that participants in the conventional market could be doing even more to help underserved families.

a. Lowering Down Payments and Up-Front Costs

Numerous studies have concluded that saving enough cash for a down payment and for up-front closing costs is the greatest barrier that low-income and minority families face when considering homeownership.^[105] To assist in overcoming this barrier, the industry (including lenders, private mortgage insurers and the GSEs) began offering in 1994 mortgage products that required down payments of only 3 percent, plus points and closing costs. Other industry efforts to reduce borrowers' up-front costs included zero-point-interest-rate mortgages and monthly insurance premiums with no up front component. These new plans eliminated large up-front points and premiums normally required at closing.

During 1998, Fannie Mae introduced its “Flexible 97” and Freddie Mac introduced its “Alt 97” low down payment lending programs. Under these programs, borrowers were required to put down only 3 percent of the purchase price. The down payment, as well as closing costs, could be obtained from a variety of sources, including gifts, grants or loans from a family member, the government, a non-profit agency and loans secured by life insurance policies, retirement accounts or other assets. Fannie Mae continues to offer the “Flexible” line of products, and Freddie Mac continues to list “Alt 97.”

In 2000, Fannie Mae launched the “MyCommunityMortgage” suite of products, which provides high loan-to-value product options for low- and moderate-income borrowers. In 2003, Fannie Mae purchased or securitized more than $2.27 billion of MyCommunityMortgage products, which helped provide affordable housing solutions for 20,400 households. In addition, Fannie Mae enhanced the MyCommunityMortgage to help lenders further expand affordable financing to underserved families. Examples of these enhancements included adding MyCommunityMortgage to Desktop Underwriter in order to provide lenders easier access to customized CRA-targeted loan products, adding new credit and income flexibilities for borrowers purchasing single family homes, Community HomeChoice which offers more flexible requirements for persons with disabilities, Community 2-4 FamilyTM to help make the purchase of 2-4 unit homes more affordable for first time homebuyers, and Community RenovationTM 1-4 Family Pilot to help borrowers with home improvement and housing preservation costs.^[106] Additionally, in 2003, Fannie Mae enhanced Community 2-4 Family and Community Renovation 1-4 Family pilots. This product provides lower down payments and flexible parameters for owner-occupants of 1-4 unit properties.^[107]

Fannie Mae also expanded its “Flexible” product line with the “Flexible 100” product, which eliminates the requirement for a down payment by providing 100 percent loan-to-value financing. The borrower is required to make either a minimum of 3% (of the lesser of the sales price or appraised value) from approved flexible sources or making a minimum contribution of $500 from their own funds. The 3% may come from a variety on sources such as gifts, grants, or unsecured loans from relatives, employers, public agencies, or nonprofits. In 2003, Fannie Mae purchased $13.7 billion in Flexible loans that benefited 100,866 households.^[108]

Fannie Mae has also developed products specifically geared toward populations with unique needs such as seniors, Native Americans and families living near public transit routes. Examples of these targeted products include the Home Equity Conversion Mortgage (HECM) which allows seniors to convert the equity in their homes to receive cash. In 2003, Fannie Mae purchased 27,644 HECM's for a total value of $1.87 billion. PaymentPower^TM allows borrowers with strong credit to skip their regularly scheduled monthly payment up to two times during a twelve-month period and up to ten times during the life of the loan. This pilot was launched in July 2002 and by year-end 2003, Fannie Mae purchased 963 PaymentPower^TM mortgages totaling $126 million. Navajo Community Guaranty Initiative allows Navajo families to contribute a minimum of $500 or 1% of the purchase price, whichever is lower. This initiative, announced in 2003, will provide $3 million in home financing to help 60 families currently living on a reservation. The Smart Commute^TM Initiative, which targets borrowers purchasing homes near a public transit route, recognizes that homebuyers will save commuting expenses and therefore have more disposable income to pay housing expenses. In 2003 Fannie Mae purchased approximately $5 million in Smart Commute^TMInitiative loans.^[109]

In 2000, Freddie Mac introduced its “Freddie Mac 100” product, which is designed to assist borrowers who have good credit but lack the ability to provide a large down payment. “Freddie Mac 100” allows a 100 percent loan-to-value ratio with the condition that the borrower has the funds for closing costs. In 2003, a refinance option was added to Freddie Mac 100 and the cost of the loan was reduced through lower mortgage insurance coverage and a lower fee for the product. These changes have made the Freddie Mac 100 available to borrowers who may not have been able to take advantage of the refinance boom as a result of low or no equity in their homes.^[110]

Another Freddie Mac product, Affordable Gold^® 97 permits borrowers to make 3% down payments from personal cash and to use other sources to cover their closing costs, and offers flexible ratio and reserves guidelines. In 2003 this product was enhanced with a refinance option allowing more borrowers to take advantage of the low rates in the market. The Affordable Gold^® 100 provides 100 percent financing to low- and moderate-income borrowers for the purchase price of a home in California. Affordable Gold^® 100 combines mortgage insurance benefits provided by a state insurance fund, the secondary mortgage market, and a team of the nation's leading mortgage lenders.^[111]

Additional Freddie Mac products include the Alt 97^[SM] for borrowers who have good credit but limited cash for a down payment. In 2003, this product was enhanced with a refinance option and reduced fees. The Two-Family 95 Percent LTV Program offers low down payment loans to purchasers of two-family properties when the borrowers occupy one of the units as their primary residence.^[112] Other initiatives include policies aimed at improving the homeownership rate among immigrant families and the Section 8 Rental to Homeownership program, which allows people currently receiving Section 8 rental subsidies to use them toward mortgage payments.^[113] Freddie Mac purchases loans in which the borrower's down payment consists of funds that have been matched through an Individual Development Account homebuyer savings program. And in 2003, Freddie Mac provided increased liquidity for affordable housing through a series of targeted investments in Mortgage Revenue Bonds containing state and local housing finance agency mortgages.^[114]

b. Partnerships—Fannie Mae

In addition to developing new affordable products, lenders and the GSEs have been entering into partnerships with local governments and nonprofit organizations to increase mortgage access to underserved borrowers. Fannie Mae operates 55 partnership offices throughout the country, including the West Virginia Partnership Office, which opened in 2003. These offices coordinate Fannie Mae's programs with local governments, lenders, public officials, housing organizations, community nonprofits, real estate professionals, and other local stakeholders.^[115]

Fannie Mae continues to reach out to national groups and work with local affiliates to expand homeownership. Fannie Mae has established multi-year partnerships to increase affordable housing opportunities with organizations such as: The Enterprise Foundation, The Neighborhood Reinvestment Corporation, ACORN Housing Corporation, The National Council of La Raza, and many others engaged in promoting affordable housing. In 2003, Fannie Mae financed $1.3 billion of mortgages with these national partners and participating lenders, which resulted in 9,597 loans. For example, Fannie Mae maintains a partnership with the National Urban League (NUL) and the JP Morgan Chase Bank to increase NUL's homeownership counseling capacity by providing the necessary technology and tools to support the effort, and to purchase $50 million in mortgage products over five years that are specifically targeted to increase homeownership among minorities. In 2003, approximately $6 million in loans were originated through this initiative. Another example is Fannie Mae's partnership with the AFL-CIO Housing Investment Trust (HIT) and Countrywide Home Loans, which launched “HIT HOME” in 2001. HIT HOME is an affordable home mortgage initiative that targets 13 million union members in 35 cities throughout the nation to provide union members with a variety of affordable mortgage choices that enable them to qualify for competitively priced loans with new re-payment terms. In 2003, over $132 million worth of mortgages were originated through this partnership.^[116]

In order to meet the needs of underserved and low- and moderate-income populations, Fannie Mae has targeted specific populations for initiatives. These include the Section 8 Homeownership Initiative, which purchased 81 Section 8 loans and funded an additional 55 loans through a Community Development Financial Institution investment; the Native American Homeownership Initiative, which has committed to invest at least $350 million to support homeownership strategies for 4,600 Native American families and to work with 100 tribes; the Minority- and Women-Owned Lenders Initiative, to reach underserved communities and to develop innovative solutions for increasing business opportunities for these lenders; The Employer-Assisted Housing Initiative, designed to assist employers in developing a company benefit that helps employees meet their housing needs; and the Initiative to Reduce Barriers to Affordable Housing, which has established local partnerships in seven new states and localities in 2003. Additionally, Fannie Mae conducts various underwriting experiments aimed at eliminating obstacles faced by prospective homebuyers across the country. In 2003, Fannie Mae approved $222 million worth of Housing and Community Development place-based commitments for a total of 55 experiments.^[117]

Fannie Mae's American Dream Commitment is part of its National Minority Homeownership Initiative which has pledged to contribute at least $700 billion in private capital to serve 4.6 million families towards President George W. Bush's goal of expanding homeownership to 5.5 million new minority Americans by the end of the decade. Towards this goal, in 2003, Fannie Mae executed 17 new Housing and Community Development lender partnerships which seek to provide $394 billion in affordable housing lending to minority families.^[118]

Under the American Dream Commitment, Fannie Mae has committed to establishing 250 faith-based homeownership partnerships in communities across the country by the end of the current decade. The objective of this initiative is to build strong partnerships with national faith-based organizations in order to reach potential new homeowners, work with faith-based and nonprofit partners to help increase access to homeownership information and education, partner with lenders to increase access to mortgage financing, and provide faith-based organizations with the tools, training, and resources needed to advance their community development efforts. Fannie Mae's work under the Faith-Based Initiative in 2003 resulted in $125 million in mortgage financing to underserved families across the country.^[119] Additionally, Fannie Mae attended more than 12 faith-based symposiums providing training and technical assistance to over 2,000 symposium attendees.^[120]

c. Partnerships—Freddie Mac

Freddie Mac does not have a partnership office structure similar to Fannie Mae's, but it has undertaken a number of initiatives in specific metropolitan areas.^[121] Freddie Mac works with affordable housing lenders to design creative solutions to meet homeownership needs of specific populations in targeted areas; explore efficient use of public subsidies to make homeownership more affordable and develop homebuyer education/counseling and debt management assistance programs.^[122] In 2001, Freddie Mac joined the Congressional Black Caucus to launch a new initiative, “With Ownership Wealth,” designed to increase African-American homeownership with one million new families by 2005.^[123] Freddie Mac has partnered with the National Council of La Raza (NCLR), 20 community based NCLR affiliated housing counseling organizations, the National Association of Hispanic Real Estate Professionals (NAHREP), EMT Applications and participating Freddie Mac Seller/Servicers including Bank of America, U.S. Bank and Wells Fargo Home Mortgage on the “En Su Casa” initiative. This $200 million homeownership initiative combines technology tools with flexible mortgage products to meet the needs of Hispanic borrowers. Mortgage products include low down payments, flexible credit underwriting and debt-to-income ratios, and streamlined processing for resident alien borrowers.^[124]

In 2002, Freddie Mac joined with the City of Boston and the U.S. Conference of Mayors to make available the “Don't Borrow Trouble” predatory lending educational campaign to approximately 1,100 cities. As of the end of 2003, the campaign has been launched in more than 30 localities. Additionally, in late 2003, Freddie Mac sponsored a national Don't Borrow Trouble summit. Attorneys, community activists and local leaders from 23 cities convened to share campaign experiences and to learn about emerging predatory lending trends from some of the nation's leading community lending experts.^[125]

In addition, Freddie Mac joined with Rainbow/PUSH and the National Urban League to promote the CreditSmart® financial educational curriculum that helps consumers understand, obtain and maintain good credit, thereby preparing them for homeownership and other personal financial goals. Rainbow/PUSH has organized CreditSmart® classes with more than 80 churches across the nation, reaching more than 2,500 congregants. Bilingual curriculum was launched for this program in December 2002, and during 2003 CreditSmart® Español conducted a total of 23 Train-the-Trainer workshops for their partners and their local partners resulting in 326 trainers who are authorized to teach the CreditSmart® Español curriculum. Thus far 503 adults have been trained in the CreditSmart® Español financial literacy program.^[126] The CreditSmart®/Homeownership Development Initiative with the National Urban League has nine affiliates located in Birmingham, AL; Charlotte, NC; Louisville, KY; Greenville, SC; Oklahoma City, OK; Springfield, IL; and Washington, DC; with Orlando, FL and Knoxville, TN added in 2003. Since the initiative's launch in early 2002, 41 CreditSmart® financial literacy workshops have been presented to more than 600 minority participants. Those participants are proceeding to the next steps to achieving homeownership, and in 2003 313 loans have closed as a direct result.^[127]

In 2002 and 2003, Freddie Mac joined with the American Community Bankers, the Credit Union National Association, and the Independent Community Bankers of America in strategic alliances to better enable member banks and credit unions access to the secondary market.^[128]

In June 2002, President George W. Bush challenged the nation's housing industry to invest more than $1 trillion to make homeownership a reality for 5.5 million more minority households for the decade. Freddie Mac responded to the challenge with Catch the Dream which is a comprehensive set of 25 high impact initiatives aimed at accelerating the growth in minority homeownership. The initiatives range from homebuyer education and outreach, to new technologies with innovative mortgage products. Freddie Mac has committed to purchase $400 billion in mortgages made to minority families by the end of the decade.^[129] Catch the Dream represents a collaborative effort with lenders, nonprofit housing and community-based organizations, and other industry participants to expand homeownership opportunities for America's minorities.^[130] In 2003 initiatives were implemented in Birmingham, Charlotte, Atlanta, DeKalb County (GA), Lansing, and San Antonio. In 2003, single-family owner occupied mortgage purchases financed homes for almost 700,000 minority families, including mortgages for 133,000 African-American and 250,000 Hispanic families (this comprised 16% of Freddie Mac's single-family, owner-occupied mortgage purchases and 22.6% of their first-time homebuyer mortgage purchases).^[131]

The programs mentioned above are examples of the partnership efforts undertaken by the GSEs. There are more partnership programs than can be adequately described here. Fuller descriptions of these programs are provided in their Annual Housing Activity Reports.

d. Underwriting and GSE Purchase Guidelines

Lenders, mortgage insurers, and the GSEs have also been modifying their mortgage underwriting standards to address the needs of families who have historically found it difficult to qualify under traditional guidelines. In addition to the changes in underwriting standards, the use of automated underwriting has dramatically transformed the mortgage application process. This section focuses on changes to traditional underwriting standards and recent GSE initiatives for credit-impaired borrowers. Subsequent sections will provide more details on the impact of automated underwriting.

The GSEs modified their underwriting standards to address the needs of families who find qualifying under traditional guidelines difficult. The goal of these underwriting changes is not to loosen underwriting standards, but rather to identify creditworthiness by alternative means that more appropriately measures the unique circumstances of low-income, immigrant, and minority households. Examples of changes that the GSEs and others in the industry have made to their underwriting standards include the following:

• Using a stable income standard rather than a stable job standard (or a minimum period of employment). This particularly benefits low-skilled applicants who have successfully remained employed, even with frequent job changes.
• Using an applicant's history of rent and utility payments as a measure of creditworthiness. This measure benefits lower-income applicants who have not established a credit history.
• Allowing pooling of funds for qualification purposes. This change benefits applicants with extended family members. Freddie Mac, for example, allows income from relatives who live together to pool their funds to cover downpayment and closing costs and to combine their incomes for use in calculating the borrower's stable monthly income.

These underwriting changes have been accompanied by homeownership counseling to ensure homeowners are ready for the responsibilities of homeownership. In addition, the industry has engaged in intensive loss mitigation to control risks.

In 1999, HUD commissioned a study by the Urban Institute to examine the underwriting criteria that the GSEs use when purchasing mortgages from primary lenders.^[132] According to the study, while the GSEs had improved their ability to serve low- and moderate-income borrowers, it did not appear at that time that they had gone as far as some primary lenders to serve these borrowers. From the Urban Institute's discussion with lenders, it was found that primary lenders were originating mortgages to lower-income borrowers using underwriting guidelines that allow lower down payments, higher debt-to-income ratios and poorer credit histories than allowed by the GSEs' guidelines.

From this and other evidence, the Urban Institute concluded that the GSEs were lagging the market in servicing low- and moderate-income and minority borrowers. Furthermore, the Urban Institute found “that the GSEs' efforts to increase underwriting flexibility and outreach has been noticed and is applauded by lenders and community advocates. Despite the GSEs' efforts in recent years to review and revise their underwriting criteria, however, they could do more to serve low- and moderate-income borrowers and to minimize disproportionate effects on minorities.” ^[133] Since the Urban Institute study, Freddie Mac and Fannie Mae have been playing a larger role in financing low-income and minority borrowers. (See Section E.2.)

In addition to offering low-down-payment programs, the GSEs' recent efforts have also centered around their automated underwriting systems and their treatment of borrowers with blemished credit, the latter being perhaps the most controversial underwriting issue over the past few years. Freddie Mac has a variety of products and initiatives aimed at providing borrowers with impaired credit more mortgage choices. These products include: CreditWorks^SM which helps borrowers with excessive debt and impaired credit to become eligible for a prime market rate mortgage faster than would otherwise be possible, Affordable Merit Rate^SM Mortgage which permits borrowers to qualify at an initial interest rate that in many cases is lower than the usual subprime rate, and LeasePurchase Plus Initiative, which provides closing cost and down payment assistance in addition to extensive counseling for borrowers who have had credit issues in the past or who have never established a credit history. During 2003, Freddie Mac entered into several new markets under the LeasePurchase Plus Initiative and purchased more than $16 million in loans.^[134]

According to Freddie Mac, its automated underwriting system, “Loan Prospector” has reduced costs, made approving mortgages easier and faster, and increased the consistency of the application of objective underwriting criteria. In addition, Freddie Mac states that “Loan Prospector” extends the benefits of the mortgage finance system to borrowers with less traditional credit profiles and limited savings by more accurately measuring risk. Since its introduction in 1995, Freddie Mac reports that they have doubled their share of mortgage purchases with loan-to-value rations of 95 percent or above.^[135] In 2003, lenders and brokers used Loan Prospector to evaluate 9.5 million loan applications and Loan Prospector has evaluated more than 35 million mortgage applications since its introduction in 1995.^[136] Freddie Mac reports that its automated underwriting system, Loan Prospector, has resulted in higher approval rates for minority borrowers than under traditional manual underwriting because of improved predictive powers. As mentioned in Section C.7, the 2000 version of LP approved 87.1 percent of loans generated through affordable housing programs, compared to 51.6 percent approved by manual underwriting. The Freddie Mac study found automated mortgage scoring less discriminatory and more accurate in predicting risk. However, as noted below in the automated mortgage scoring section, there are concerns that the codification of certain underwriting guidelines could result in unintentional discrimination or disparate treatment across groups. In response to the potential disparate impact of automated underwriting, Freddie Mac have launched initiatives to make the mortgage process more transparent by disclosing both credit and non-credit factors that Loan Prospector consider when evaluating a loan application. In 2000, Freddie Mac has launched an initiative that published a list of all of the factors that Loan Prospector uses to analyze loans, and put the list on the Freddie Mac website.^[137]

In 2003, Fannie Mae released two versions of its automated underwriting service, “Desktop Underwriter” (DU), to expand its mortgage product offerings and to update underwriting guidelines. Desktop Underwriter® 5.3 outlined new eligibility requirements for mortgages secured by manufactured homes. It also expanded the InterestFirst^TM mortgage product line to offer borrowers greater purchasing power by allowing lower initial monthly payments than those available with traditional loan products. Desktop Underwriter® 5.3.1 enhanced the Flexible 100 mortgage to allow borrowers to contribute as little as $500 of their own funds to the transaction. The remainder of the funds can come from flexible sources of funds and interested party contributions subject to Fannie Mae's standard contribution limit.^[138] In addition, Fannie Mae added MyCommunityMortgage to Desktop Underwriter® in 2003, providing lenders easier access to customized CRA-targeted loan products.^[139] Automated mortgage scoring and the potential for disparate impacts on borrowers will be further discussed in a later section.

5. Affordable Single-family Lending: Data Trends

a. 1993-2003 Lending Trends

HMDA data suggest that the industry and GSE initiatives are increasing the flow of credit to underserved borrowers. Between 1993 and 2003, conventional loans to low-income and minority families increased at much faster rates than loans to higher income and non-minority families. As shown below, conventional home purchase originations to African Americans more than doubled between 1993 and 2003 and those to Hispanic borrowers more than tripled. Home loans to low-income borrowers and to low-income and high-minority census tracts also more than doubled during this period.

	1993-2003 Growth rate: all home loans (percent)	1993-2003 Growth rate: conventional home loans (percent)
African-American Borrowers	106	206
Hispanic Borrowers	235	357
White Borrowers	44	64
Low-Income Borrower (Less than 80% of AMI)	101	150
Upper-Income Borrower (More than 120% of AMI)	88	108
Low-Income Census Tract (only 1993-2002)	99	143
Upper-Income Census Tract (only 1993-2002)	64	78
High-Minority Tract (only 1993-2002) (50% or more minority)	113	167
Predominantly-White Tract (only 1993-2002) (Less than 10% minority)	53	64

GSE purchases showed similar trends, as indicated by the following 1993-to-2003 percentage point increases for metropolitan areas: African-American borrowers (199 percent), Hispanic borrowers (259 percent), and low-income borrowers (212 percent). While their annual purchases of all home loans increased by 60 percent between 1993 and 2003, their purchases of mortgages that qualify for the three housing goals increased as follows: special affordable by 287 percent; low- and moderate-income by 156 percent; and underserved areas by 121 percent.

While low interest rates and economic expansion certainly played an important role in the substantial increase in conventional affordable lending in recent years, most observers believe that the efforts of lenders, private mortgage insurers, and the GSEs were also important contributors. In addition, many observers believe that government initiatives such as the GSE housing goals and the Community Reinvestment Act have also played a role in the growth of affordable lending over the past 10 years.

b. Affordable Lending Shares by Major Market Sector

Section E below compares the GSEs' performance with the performance of primary lenders in the conventional conforming market. To provide a useful context for that analysis, this section examines the role of the conventional conforming market in funding low-income and minority families and their neighborhoods. Information on the mortgage market's funding of homes purchased by first-time homebuyers is also provided. In addition, this section compares the GSEs with other sectors of the mortgage market. The important role of FHA in the affordable lending market is highlighted and questions are raised about whether the conventional conforming market could be doing a better job helping low-income and minority borrowers obtain access to mortgage credit.

Table A.1 reports borrower characteristics and Table A.2 reports neighborhood characteristics for home purchase mortgages insured by FHA, purchased by the GSEs, originated by depository institutions (mainly banks and thrift), and originated in the conventional conforming market and in the total market for owner-occupied properties in metropolitan areas.^[140] In this case, the “total” market consists of both the conventional conforming market and the government (mainly FHA and VA loans) market; “jumbo” loans above the conventional conforming loan limit are excluded from this analysis.^[141]

HMDA is the source of the FHA, depository, and market data, while the GSEs provide their own data. Low-income, African-American, Hispanic, and minority borrowers are covered in Table A.1. Table A.2 provides information on four types of neighborhoods—low-income census tracts, tracts where minorities (or African Americans) account for more than 30 percent of the census tract population, and underserved areas as defined by HUD. The average data reported in Tables A.1 and A.2 for the years 1999 to 2003 offer a good summary of recent lending to low-income and minority borrowers and their communities.^[142] Individual year data are also provided.

The focus of different market sectors on affordable lending is summarized by the percentages reported in Tables A.1 and A.2. These percentages show each sector's “distribution of business,” defined as the share of loans originated (or, for the GSEs, purchased) that had a particular borrower or neighborhood characteristic. The interpretation of the “distribution of business” percentages can be illustrated using the FHA percentage for low-income borrowers: Between 1999 and 2003, 51.2 percent of all FHA-insured home purchase loans in metropolitan areas were originated for borrowers with an income less than 80 percent of the local area median income. These percentages are to be contrasted with “market share” percentages, which are presented below in Section E. A “market share” percentage is the share of loans with a particular borrower or neighborhood characteristic that was funded by a particular market sector (e.g., FHA-insured, GSEs, depositories). As will discussed below, FHA's “market share” for low-income borrowers during the 1999-to-2003 period was estimated to be 24 percent which is interpreted as follows: Of all home purchase loans originated for low-income borrowers in metropolitan areas between 1999 and 2003, 24 percent were FHA-insured loans. Thus, in this example, the “distribution of business” percentage measures the importance (or concentration) of low-income borrowers in FHA's overall business while the “market share” percentage measures the importance of FHA to the market's overall funding of loans for low-income borrowers. Both concepts are important for evaluating performance—for an industry sector such as FHA or the GSEs to have a significant impact on lending to a targeted group, that sector's business must be concentrated on the targeted group and that sector must be of some size. The discussion below will focus on the degree to which different mortgage sectors concentrate on targeted groups, while Section E will also provide estimates of market shares.

The main insights from the “distribution of business” percentages in Tables A.1 and A.2 pertain to four topics.

(i) FHA-Insured Loans. FHA has traditionally been the mechanism used by borrowers who face difficulty obtaining mortgage financing in the private conventional market. FHA has long been recognized as the major source of funding for first-time, low-income and minority homebuyers who are not often able to raise cash for large downpayments.^[143] Tables A.1 and A.2 show that FHA places much more emphasis on affordable lending than the other market sectors. Between 1999 and 2003, low-income borrowers accounted for 51.2 percent of FHA-insured loans, compared with 27.8 percent of the home loans purchased by the GSEs, 29.1 percent of home loans originated by depositories, and 29.2 percent of all originations in the conventional conforming market (see Table A.1). Likewise, 40.7 percent of FHA-insured loans were originated in underserved census tracts, while only 24.1 percent of the GSE-purchased loans, 25.9 percent of home loans originated by depositories, and 26.9 percent of conventional conforming loans were originated in these tracts (see Table A.2).^[144] As discussed in Section E, FHA's share of the minority lending market is particularly high. While FHA insured only 16 percent of all home purchase mortgages originated below the conforming loan limit in metropolitan areas between 1999 and 2003, it is estimated that FHA insured 29 percent of all home loans originated for African-American and Hispanic borrowers.

(ii) Conventional and GSE Minority Lending. The affordable lending shares for the conventional conforming sector are low for minority borrowers, particularly African-American and Hispanic borrowers. These borrowers accounted for only 15.2 percent of all conventional conforming loans originated between 1999 and 2003, compared with 34.4 percent of FHA-insured loans and 19.2 percent of all loans originated in the total (government and conventional conforming) market. Not surprisingly, the minority lending performance of conventional lenders has been subject to much criticism. Recent studies contend that primary lenders in the conventional market are not doing their fair share of minority lending which forces minorities, particularly African-American and Hispanic borrowers, to rely on more costly FHA and subprime loans.^[145] Thus, it appears that conventional lenders could be doing a better job helping minority borrowers obtain access to mortgage credit.

• The GSEs' funding of minority loans can be compared with mortgages originated for minority borrowers in the conventional conforming market, although the latter may be a poor benchmark, as discussed above. Between 1999 and 2003, home purchase loans to African-American and Hispanic borrowers accounted for 10.4 percent of Freddie Mac's purchases, 14.0 percent of Fannie Mae's purchases, and 15.2 percent of loans originated in the conventional conforming market (or 14.3 percent if B&C loans are excluded from the market definition). Thus, since 1999, the African-American and Hispanic share of the GSEs' purchases has been lower than the corresponding share for the conventional conforming market.^[146]

• As the above comparisons show, Fannie Mae has had a much better record than Freddie Mac in funding loans for minority families. And Fannie Mae significantly increased its purchases of loans for African-American and Hispanic borrowers during 2001, raising the share of its purchases to market levels—13.7 percent for both Fannie Mae and the conforming market (without B&C loans). In 2002, Fannie Mae surpassed the conventional conforming market in funding African-American and Hispanic borrowers (a 15.8 percent share for Fannie Mae and a 15.0 share for the market), but in 2003 fell slightly behind the market (a 16.6 percent share for Fannie Mae and a 16.9 percent share for the market). When all minority borrowers are considered, Fannie Mae has purchased mortgages for minority borrowers at a higher rate (years 2001, 2002 and 2003) than these loans were originated by primary lenders in the conventional conforming market (without B&C loans). Freddie Mac, on the other hand, lagged behind both the market and Fannie Mae in funding loans for minority borrowers during 2001-2003, as well as during the entire 1999-to-2003 period. The share of Freddie Mac's purchases for African-American and Hispanic borrowers declined from 10.9 percent in both 2000 and 2001 to 10.1 percent in 2002 before rising slightly to 10.7 percent in 2003.
• Considering the minority census tract data reported in Table A.2, Fannie Mae lagged behind the conforming market (without B&C loans) in high-minority neighborhoods and in high-African-American neighborhoods during the 1999-to-2003 period. However, Fannie Mae improved its mortgage purchases in African-American neighborhoods after 2001 and essentially matched the market in 2001-2003. And during 2001, 2002 and 2003, Fannie Mae also purchased loans in high-minority census tracts at a higher rate than loans were originated by conventional lenders in these tracts. While Freddie Mac has generally lagged the primary market in funding minority neighborhoods, note in Table A.2 that high African-American tracts increased from 3.9 percent of Freddie Mac's purchases in 2001 to 5.3 percent in 2002, placing Freddie Mac above the conventional conforming market level (4.6 percent) in 2002. However, in 2003, Freddie Mac fell behind the market.

(iii) Low-Income Lending by the GSEs. Information is also provided on the GSEs' purchases of home loans for low-income borrowers (A.1) and for families living in low-income neighborhoods (A.2). Historically, the GSEs have lagged behind the conventional conforming market in funding affordable loans for these groups. During the 1999-to-2003 period, low-income borrowers (census tracts) accounted for 27.4 (9.7) percent of Freddie Mac's purchases, 28.1 (10.1) percent of Fannie Mae's purchases, 29.1 (11.2) percent of loans originated by depositories, and 29.1 (11.1) percent of home loans originated by conventional conforming lenders (without B&C loans). By the end of this period, Fannie Mae had significantly improved its performance relative to the market. In 2003, low-income borrowers accounted for 31.0 percent of Fannie Mae's purchases, compared with 29.2 percent for the conforming market. It is also interesting that even though Freddie Mac lagged the market in funding home loans for low-income borrowers during 2002 (28.6 percent versus 29.1 percent), it surpassed the market in financing properties in low-income census tracts (11.3 percent versus 11.1 percent). During 2003, Freddie Mac's performance was again below the market in low-income census tracts (a 10.3 share for Freddie Mac and a 11.5 percent share for the market). A more complete analysis of the GSEs' recent improvements in purchasing home loans that qualify for the housing goals is provided below in Section E.

(iv) Depositories. Within the conventional conforming market, depository institutions (mainly banks and thrifts) are important providers of affordable lending for lower-income families and their neighborhoods.^[147] Between 1999 and 2003, underserved areas accounted for 26.9 percent of loans held in depository portfolios, which compares favorably with the underserved areas percentage (26.2 percent) for the overall conventional conforming market.^[148] Depository lenders have extensive knowledge of their communities and direct interactions with their borrowers, which may enable them to introduce flexibility into their underwriting standards without unduly increasing their credit risk. The Community Reinvestment Act provides an incentive for banks and thrifts to initiate affordable lending programs with underwriting flexibility and to reach out to lower income families and their communities.^[149] Many of the CRA loans are held in portfolio by lenders, rather than sold to Fannie Mae or Freddie Mac.

(v) First-time Homebuyers. As explained in Section E, market information on first-time homebuyers is not as readily available as the HMDA data reported in Tables A.1 and A.2 on the income and racial characteristics of borrowers and census tracts served by the mortgage market. However, the limited market data that are available from the American Housing Survey, combined with the first-time homebuyer data reported by FHA and the GSEs, indicate a rather large variation in the funding of first-time homebuyers across the different sectors of the mortgage market. Based on the American Housing Survey (AHS), it is estimated that first-time homebuyers accounted for 42.3 percent of all home purchase loans originated throughout the market between 1999 and 2001,^[150] and for 37.6 percent of home loans originated in the conventional conforming market. The AHS defines a first-time homebuyer as someone who has never owned a home. Using a more liberal definition of a first-time homebuyer (someone who has not owned a home in the past three years), FHA reports that first-time homebuyers accounted for 80.5 percent of all home loans that it insured between 1999 and 2001 and the GSEs report that first-time homebuyers accounted for 26.5 percent of the home loans purchased by each GSE during that same period. Given FHA's low downpayment requirements, it is not surprising that FHA focuses on first-time homebuyers. The GSEs, on the other hand, fall at the other end of the continuum, with their first-time homebuyer share (26.5 percent) falling far short of the first-time homebuyer share (37.6 percent) of the conventional conforming market. Section E will include a more detailed comparison of the GSEs and the conventional conforming market in serving first-time homebuyers. In addition, Section E will conduct a market share analysis that examines the funding of minority first-time homebuyers. Consistent with the earlier discussion, that analysis suggests that conventional lenders and the GSEs have played a relatively small role in the market for minority first-time homebuyers. One analysis reported in Section E estimates that mortgage purchases by the GSEs between 1999 and 2001 totaled 41.5 percent of all home loans originated, but they accounted for only 14.3 percent of home loans originated for first-time African-American and Hispanic homebuyers.

c. Community Reinvestment Act

The Community Reinvestment Act (CRA) requires depository institutions to help meet the credit needs of their communities.^[151] CRA loans are typically made to low-income borrowers earning less than 80 percent of area median income, and in moderate-income neighborhoods. CRA provides an incentive for lenders to initiate affordable lending programs with underwriting flexibility. CRA loans are usually smaller than typical conventional mortgages and also are more likely to have a higher LTV, higher debt-to-income ratios and no payment reserves, and may not be carrying private mortgage insurance (PMI). Generally, at the time CRA loans are originated, many do not meet the underwriting guidelines required in order for them to be purchased by one of the GSEs. Therefore, many of the CRA loans are held in portfolio by lenders, rather than sold to Fannie Mae or Freddie Mac. Evidence is growing that CRA-type lending to low-income families can be profitable, particularly when combined with intensive loss mitigation efforts to control credit risk. In a recent survey conducted by the Federal Reserve, lenders reported that most CRA loans are profitable although not as profitable as the lenders' standard products.^[152]

Some anticipate that the big growth market over the next decade for CRA-type lending will be urban areas. There has been some movement of population back to cities, consisting of aging Baby Boomers (so-called “empty nesters”), the children of Baby Boomers (the Echo Boomers aged 18-25), and immigrants, particularly Hispanics but also Asians.^[153] The current low homeownership in inner cities (compared with the suburbs) also suggests that urban areas may be a potential growth market for lenders. Lenders are beginning to recognize that urban borrowers are different from suburban borrowers. A new or recent immigrant may have no credit history or, more likely, a loan-worthy credit history that can't be substantiated by the usual methods.^[154] Products for duplexes and four-plexes are not the same as a mortgage for a subdivision house in the suburbs. Programs are being implemented to meet the unique needs of urban borrowers. One program emphasizing urban areas was initiated by the American Community Bankers (ACB). Under the ACB program, which made $16.2 billion in loans in 2002, lenders originated a variety of affordable products for first-time homebuyers and non-traditional borrowers that are then sold to Fannie Mae, Freddie Mac, Countrywide, or other investors that are partnering with the ACB. It is reported that some lenders are making these non-traditional loans for the first time.

For banks and thrifts, selling their CRA loans will free up capital to make new CRA loans. As a result, the CRA market segment provides an opportunity for Fannie Mae and Freddie Mac to expand their affordable lending programs. Section E.3c below presents data showing that purchasing targeted seasoned loans has been one strategy that Fannie Mae has chosen to improve its goals performance. Fannie Mae has been offering CRA programs since mid-1997, when it launched a pilot program, “Community Reinvestment Act Portfolio Initiative,” for purchasing seasoned CRA loans in bulk transactions, taking into account track record as opposed to relying just on underwriting guidelines. Fannie Mae also started another pilot program in 1998, involving purchases of CRA loans on a flow basis, as they are originated. As part of the American Dream Commitment, Fannie Mae has committed to investing $20 billion in CRA-targeted business, and funding $530 billion in CRA-eligible investments. One CRA-eligible product in 2003 included the MyCommunityMortgage ^TM suite, which provides flexible product options for low- to moderate-income families, including minorities, immigrants, first-time homebuyers, and underserved borrowers living in rural areas. MyCommunityMortgage is offered by over 300 lender partners nationwide, and marries targeted pricing with affordability features, such as 100 percent loan-to-value ratios with only $500 from the borrower's own funds.^[155] In 2003, Fannie Mae purchased or securitized more than $2.27 billion of MyCommunityMortgage products, which helped provide affordable housing solutions for 20,400 households.^[156]

In addition, Freddie Mac is also purchasing seasoned affordable mortgage portfolios originated by depositories to help meet their CRA objectives. In 2003, Freddie Mac developed credit enhancements that enable depositories to profitably sell their loans to Freddie Mac—these transactions facilitate targeted affordable lending activity by providing immediate liquidity. Freddie Mac also increased its ability to purchase smaller portfolios opening this option to many community banks that otherwise would not have an outlet for their portfolios.^[157] The billions of dollars worth of CRA loans that will be originated, as well as the CRA loans being held in bank and thrift portfolios, offer both GSEs an opportunity to improve their performance in the single-family area.

6. Potential Homebuyers

While the growth in affordable lending and homeownership has been strong in recent years, attaining this Nation's homeownership goals will not be possible without tapping into the vast pool of potential homebuyers. Due to record low interest rates, expanded homeownership outreach, and new flexible mortgage products, the homeownership rate reached an annual record of 67.9 percent in 2002, reaching 68.6 percent in the fourth quarter of 2003.^[158] This section discusses the potential for further increases beyond those resulting from current demographic trends.

The potential homeowner population over the next decade will be highly diverse, as growing housing demand from immigrants (both those who are already here and those projected to come) and non-traditional homebuyers will help to offset declines in the demand for housing caused by the aging of the population. As noted in the above discussion of CRA, many of these potential homeowners will be located in urban areas. As noted in the above discussion of underlying demographic conditions (section C.2.), immigrants and other minorities—who accounted for nearly 40 percent of the growth in the nation's homeownership rate over the past five years—will be responsible for almost two-thirds of the growth in the number of new households over the next ten years. This trend does not depend on the future inflow of new immigrants, as immigrants don't enter the housing market until they have been in this country for eleven years. As noted by Fannie Mae staff, “there are enough immigrants already in this country to keep housing strong for at least six and perhaps even 10 more years”.^[159] As these demographic factors play out, the overall effect on housing demand will likely be sustained growth and an increasingly diverse household population from which to draw new homeowners.

Surveys indicate that these demographic trends will be reinforced by the fact that most Americans desire, and plan, to become homeowners. According to the 2002 Fannie Mae Foundation annual National Housing Survey, Americans rate homeownership as the best investment they can make, far ahead of 401Ks, retirement accounts, and stocks. The percentage of Americans who said it was a good time to buy a home was at its highest level since 1994 at 75 percent, a jump of 21 percentage points since May 2001.^[160] In addition, the survey found that 27 percent of Americans report they are likely to buy in the next three years, and 23 percent of those have started to save or have saved enough money for a down payment.^[161]

Further increases in the homeownership rate depend on whether or not recent gains in the home owning share(s) of specific groups are maintained. Minorities accounted for 17 percent of owner households in 2001, but the Joint Center for Housing Studies reports that minorities were responsible for more than 40 percent (a total of 5.2 million) of the net growth in homeowners between 1993 and 2002.^[162] As reported by the Fannie Mae survey, 42 percent of African-American families reported that they were “very or fairly likely” to buy a home in the next three years, up from 38 percent in 1998 and 25 percent in 1997. Among Hispanics and Hispanic immigrants, the numbers reached 37 percent and 34 percent respectively. The 2002 survey also reports that more than half of Hispanic renters cite homeownership as being “one of their top priorities”. In addition, nearly a third (31 percent) of baby boomers said they are “very or fairly likely” to buy a home in the next three years.

In spite of these trends, potential minority and immigrant homebuyers see more obstacles to buying a home, compared with the general public. These barriers to homeownership are discussed in detail in section B.1.b above and include: lack of capital for down payment and closing costs; poor credit history; lack of access to mainstream lenders; complexity and fear of the homebuying process; and, continued discrimination in housing markets and mortgage lending. To address the needs of the new group of potential homeowners, the mortgage industry will have to address these needs on several fronts, such as expanding education and outreach efforts, introducing new products, and adjusting current underwriting standards to better reflect the special circumstances of these new households.

Thus, the new group of potential homeowners will have unique needs. To tap this potential homeowner population, the mortgage industry will have to address these needs on several fronts, such as expanding education and outreach efforts, introducing new products, and adjusting current underwriting standards to better reflect the special circumstances of these new households.

The Bush administration has outlined a plan to expand minority homeownership by 5.5 million families by the end of the decade. The Joint Center for Housing Studies has stated that if favorable economic and housing market trends continue, and if additional efforts to target mortgage lending to low-income and minority households are made, the overall homeownership rate could reach 70 percent by 2010.^[163]

7. Automated Underwriting Systems and Mortgage Scorecards

This, and the following two sections, discuss special topics that have impacted the primary and secondary mortgage markets in recent years. They are automated mortgage scoring, subprime loans, and risk-based pricing. The GSEs' use of automated underwriting and mortgage scoring systems was briefly discussed in the earlier section on underwriting standards. This section expands on issues related to automated underwriting, a process that has spread throughout the mortgage landscape over the past five years, due mainly to the efforts of Fannie Mae and Freddie Mac.

Automated mortgage scoring was developed as a high-tech tool with the purpose of identifying credit risks in a more efficient manner. Automated mortgage scoring has grown as competition and decreased profit margins have created demands to reduce loan origination costs. As a result, automated mortgage scoring has become the predominant (around 60 to 70 percent) mortgage underwriting method.^[164]

According to Freddie Mac economists, automated mortgage scoring has enabled lenders to expand homeownership opportunities, particularly for underserved populations.^[165] There is growing evidence that automated mortgage scoring is more accurate than manual underwriting in predicting borrower risks. Mortgage scorecards express the probability that an applicant will default as a function of several underwriting variables such as the level of down payment, monthly-payment-to-income ratios, cash reserves, and various indicators of an applicant's creditworthiness or credit history. Mortgage scorecards are statistically estimated regression-type equations, based on historical relationships between mortgage foreclosures (or defaults) and the underwriting variables. The level of down payment and credit history indicators, such as a FICO score, are typically the most important predictors of default in mortgage scoring systems.

For example, HUD has developed FHA TOTAL Scorecard to evaluate the credit risk of FHA loans submitted to an automated underwriting system. The Scorecard works with Fannie Mae's Desktop Underwriter® to provide a recommended level of underwriting and documentation for FHA loans and to determine a loan's eligibility for insurance with FHA. In 2003, Fannie Mae conducted a market test of the Scorecard with 18 FHA approved Desktop Underwriter® lenders. Over 3,000 loans were submitted to the Total Scorecard through Desktop Underwriter® during the market test period.^[166]

This increased accuracy in risk assessment of mortgage scorecards has allowed risk managers to set more lenient risk standards, and thus originate more loans to marginal applicants. Applicants who would otherwise be rejected by manual underwriting are being qualified for mortgages with automated mortgage scoring in part because the scorecard allows an applicant's weaker areas to be offset by stronger characteristics. Typically, applicants whose projected monthly debt payment (mortgage payment plus credit card payment plus automobile loan payment and so on) comprise a high percentage of their monthly income would be turned down by a traditional underwriting system that relied on fixed debt-to-income ratios (such as 36 percent). In a mortgage scoring system, these same applicants might be automatically accepted for a loan due to their stellar credit record or to their ability to raise more cash for a down payment. The entity funding or insuring the mortgage (i.e., a lender, private mortgage insurer, or a GSE) allows these positive characteristics to offset the negative characteristics because its confidence in the ability of the empirically-based mortgage scorecard to accurately identify those applicants who are more likely or less likely to eventually default on their loan. The mortgage score is in essence a recommendation to the lender to accept the application, or to refer it for further review through manual underwriting. Accepted loans benefit from reduced document requirements and expedited processing.

In 2003, Fannie Mae conducted a study of automated underwriting systems and concluded that the production cost per loan decreased significantly as lenders moved automated underwriting closer to the point of sale. Specifically, retail lenders using an integrated automated underwriting system at the point of sale reported originations savings of more than $1,000 over manual underwriting.^[167] Freddie Mac also reported that Loan Prospector reduces the average time lenders spend underwriting most loans and reduces origination costs by about an average of $650 or more per loan.^[168] In addition, Freddie Mac analyzed about 1,000 loans originated in 1993 and 1994. Of the loans, manual underwriters rated 52 percent accept, compared to a Loan Prospector accept rate of 87 percent.^[169] In total, Freddie Mac reports that innovations in the originations process, including automated underwriting, have reduced mortgage transaction costs by more than 70 percent between 1990 and 2003 from 1.87 points to 0.46 points—a decline of $1,410 per $100,000 borrowed.^[170]

As explained above, automated mortgage scoring allows tradeoffs between risk factors to be quantified more precisely, providing the industry more confidence in “pushing the envelope” of acceptable expected default rates. The GSEs’ willingness to offer low-down-payment programs was based on their belief that their scoring models could identify the more creditworthy of the cash-constrained applicants. The GSEs' new “timely reward” products for subprime borrowers (discussed later) are integrated with their mortgage scoring systems. Automated mortgage scoring presents the opportunity to remove discrimination from mortgage underwriting, to accept all applicants, and to bring fair, objective, statistically based competitive pricing, greatly reducing costs for all risk groups. Some institutions have sought to better model and automate marginal and higher-risk loans, which have tended to be more costly to underwrite and more difficult to automate.^[171]

Along with the promise of benefits, however, automated mortgage scoring has raised concerns. These concerns are related to the possibility of disparate impact and the proprietary nature of the mortgage score inputs. The first concern is that low-income and minority homebuyers will not score well enough to be accepted by the automated underwriting system, resulting in their getting fewer loans. African-American and Hispanic borrowers, for example, tend to have a poorer credit history record than other borrowers, which means they are more likely to be referred (rather than automatically accepted) by automated mortgage scoring systems that rely heavily on credit history measures such as a FICO score. There is also a significant statistical relationship between credit history scores and the minority composition of an area, after controlling for other locational characteristics.^[172]

The second concern relates to the “black box” nature of the scoring algorithm. The scoring algorithm is proprietary and therefore it is difficult for applicants to know the reasons for their scores. However, it should be noted that the GSEs have taken steps to make their automated underwriting systems more transparent. Both Fannie Mae and Freddie Mac have published the factors used to make loan purchase decisions in Desktop Underwriter and Loan Prospector, respectively. In response to criticisms aimed at using FICO scores in mortgage underwriting, Fannie Mae's new versions of Desktop Underwriter (DU) 5.3 and 5.3.1 [the newest versions are 5.3 and 5.3.1—they probably keep the following practices, but add no substantive underwriting practices, but rather lower downpayment options] replaces credit scores with specific credit characteristics and provides expanded approval product offerings for borrowers who have blemished credit. The specific credit characteristics include variables such as past delinquencies; credit records, foreclosures, and accounts in collection; credit card line and use; age of accounts; and number of credit inquiries.^[173]

With automated mortgage scoring replacing traditional manual underwriting comes the fear that the loss of individual attention poses a problem for people who have inaccuracies on their credit report or for members of cultural groups or recent immigrants who do not use traditional credit and do not have a credit score. Some subprime lenders and underwriters have claimed that their manual underwriting of high-risk borrowers cannot be automated with mortgage scoring. Although automated mortgage scoring has greatly reduced the cost of many lower-risk loans that are easier to rate, the cost of manually underwriting gray-area and higher-risk applicants still remains high.^[174] There is also the fear that applicants who are referred by the automated system will not be given the full manual underwriting for the product that they initially applied for—rather they might be pushed off to higher priced products such as a subprime or FHA loan. In this case, the applicant may have had special circumstances that would have been clarified by the traditional manual underwriting, thus enabling the applicant to receive a prime loan consistent with his or her creditworthiness.

Banking regulators and legal analysts acknowledge the value of automated mortgage scoring, although some skeptics have noted concerns regarding fair lending, potential fraud, privacy issues, and the ability of models to withstand changing economic conditions.^[175] With the rise of automated mortgage scoring, the great difference in Internet usage known as the “digital divide” could result in informational disadvantages for less educated and lower-income consumers. In addition to the digital divide, the lack of financial literacy in the United States may also result in a disparate impact on low-income and minority borrowers.^[176]

2002 Urban Institute Study. The Urban Institute submitted a report to HUD in 2002 on subprime markets, the role of GSEs, and risk-based pricing.^[177] The study took a preliminary look at the use of automated underwriting systems for a small sample of lenders. After conducting interviews with both subprime and prime lenders, the report noted that all of the lenders in the study had implemented some type of automated underwriting system. These lenders stated that automated underwriting raised their business volume and streamlined their approval process. In addition, the lenders reported they were able to direct more underwriting resources to borderline applications despite an increase in business volume.

Even with the use of automated mortgage scoring, the lenders in the study continued to conduct at least a cursory review to validate the application material. The majority of the lenders still used manual underwriting to originate loans not recommended for approval with automated mortgage scoring. The lenders reported they formulated their policies and procedures to make certain that borrowers receive the best mortgage, according to product eligibility. This study will be further referenced in a following section regarding subprime markets.

2001 Freddie Mac Study. According to a Freddie Mac study published by the Fisher Center for Real Estate and Urban Economics at University of California at Berkeley, underserved populations have benefited from automated mortgage scoring because of the increased ability to distinguish between a range of credit risks. In this paper, Freddie Mac economists compared the manual and automated mortgage scoring approval rates of a sample of minority loans originated in 1993-94 and purchased by Freddie Mac. While manual underwriters rated 51 percent of the minority loans in the sample as accept, automated mortgage scoring would have rated 79 percent of the loans as accept.^[178]

In comparison to manual underwriting, this study found automated mortgage scoring not only less discriminatory but also more accurate in predicting risk. Two versions of Freddie Mac's automated underwriting system, Loan Prospector (LP), were used to review three groups of mortgage loans purchased by Freddie Mac.^[179] The study found that LP was a highly accurate predictor of mortgage default. The resulting improved accuracy translates into benefits for borrowers, who would otherwise be rejected by manual underwriting to qualify for mortgages.

Analysis of the first group of loans showed that loans rated as “caution” were four times more likely to default than the average for all loans. Minority borrowers whose loans were rated as “caution” were five times more likely to default, and low-income borrowers whose loans were rated as “caution” were four times more likely to default than the average for all loans. The 2000 version of LP approved 87.1 percent of loans generated through affordable housing programs, compared to a 51.6 percent approval rate when the same loans were assessed using manual underwriting procedures. Further, the study found LP more accurate than manual underwriting at predicting default risk even with a higher approval rate. The study also demonstrated that Freddie Mac's year 2000 version of LP was more accurate in predicting risk than its 1995 version.

Concluding Observations. Automated underwriting has enabled lenders to reach new markets and expand homeownership opportunities, as illustrated by the 2001 Freddie Mac study. Increased accuracy with automated mortgage scoring has led to the development of new mortgage products that would have been previously considered too risky. For example, Freddie Mac uses Loan Prospector to approve Alt A loans, which tend to have nontraditional documentation; A-minus loans, which pose a higher risk of default; and other higher-risk mortgages, like 100 percent LTV loans. Both GSEs have and continue to add new products to develop their automated underwriting systems to reach more marginal borrowers.

Despite the gains in automated mortgage scoring and other innovations, minorities are still less likely to be approved for a loan. The difference in minority and non-minority accept rates may reflect greater social inequities in financial capacity and credit, which are integral variables in both manual and automated underwriting. In the future, the accuracy of automated mortgage scoring will hinge on updating the models and making them more predictive while reducing the disparate impact on low-income and minority borrowers.^[180] The fairness of automated scoring systems will also depend importantly on whether referred applicants receive a traditional manual underwriting for the loan that they initially applied for, rather than being immediately offered a higher priced loan that does not recognize their true creditworthiness.

In addition to using automated underwriting systems as a tool to help determine whether a mortgage application should be approved, the GSEs' automated underwriting systems are being further adapted to facilitate risk-based pricing. With risk-based pricing, mortgage lenders can offer each borrower an individual rate based on his or her risk. The division between the subprime and the prime mortgage market will begin to fade with the rise of risk-based pricing, which is discussed in the next section on the subprime market.

8. Subprime Lending

The subprime mortgage market provides mortgage financing to credit-impaired borrowers—those who may have blemishes in their credit record, insufficient credit history, or non-traditional credit sources. This section examines several topics related to subprime lending including (a) the growth and characteristics of subprime loans, (b) the neighborhood concentration of subprime lending, (c) predatory lending, and (d) purchases of subprime mortgages by the GSEs. Section C.9 follows with a discussion of risk-based pricing.

a. The Growth and Characteristics of Subprime Loans

The subprime market has grown rapidly over the past several years, increasing from an estimated $35 billion in 1994 to $160 billion in 1999 and $173.3 billion in 2001, before rising to $213 billion in 2002. The subprime share of total market originations rose from 4.6 percent in 1994 to a high of 15 percent in 1999, and then fell to 8.5 percent in both 2001 and 2002.^[181] Various factors have led to the rapid growth in the subprime market: Federal legislation preempting state restrictions on allowable rates and loan features, the tax reform act of 1986 which encouraged tax-exempt home equity financing of consumer debt, increased demand for and availability of consumer debt, a substantial increase in homeowner equity due to house price appreciation, and a ready supply of available funds through Wall Street securitization.^[182] It is important to note that subprime lending grew in the 1990s mostly without the assistance of Fannie Mae and Freddie Mac.

Generally, there are three different types of products available for subprime borrowers. These include: Home purchase and refinance mortgages designed for borrowers with poor credit histories; “Alt A” mortgages that are usually originated for borrowers who are unable to document all of the underwriting information but who may have solid credit records; and high loan-to-value mortgages originated to borrowers with fairly good credit. Fannie Mae and Freddie Mac are more likely to serve the first two types of subprime borrowers.^[183]

Borrowers use subprime loans for various purposes, which include debt consolidation, home improvements, and an alternative source of consumer credit. Between 1999 and 2001, about two-thirds of subprime loans were refinance loans. It has been estimated that 59 percent of refinance loans were “cash out” loans.^[184] According to a joint HUD-Treasury report, first liens accounted for more than three out of four loans in the subprime market.

The subprime market is divided into different risk categories, ranging from least risky to most risky: A-minus, B, C, and D. While there are no clear industry standards for defining the subprime risk categories, Inside Mortgage Finance defines them in terms of FICO scores—580-620 for A-minus, 560-580 for B, 540-560 for C, and less than 540 for D. The A-minus share of the subprime market rose from 61.6 percent in 2000 to 70.7 percent in 2001.^[185] For the first nine months of 2002, the A-minus share accounted for 74 percent of the market, while the B share accounted for 11 percent, the C share accounted for 7.2 percent, and the D share accounted for 7.9 percent of the market.^[186]

Delinquency rates by type of subprime loan are as follows: 3.36 percent for A-minus loans, 6.67 percent for B, 9.22 percent for C, and 21.03 percent for D, according to the Mortgage Information Corporation.^[187] Because of their higher risk of default, subprime loans typically carry much higher mortgage rates than prime mortgages. Recent quotes for a 30-year Fixed Rate Mortgage were 8.85 percent for A-minus (with an 85 percent LTV), 9.10 percent for B credit (with an 80 percent LTV), and 10.35 percent for C credit (with a 75 percent LTV).^[188] As the low loan-to-value (LTV) ratios indicate, one loss mitigation technique used by subprime lenders is a high down payment requirement. Some housing advocates have expressed concern that the perceptions about the risk of subprime loans may not always be accurate, for example, creditworthy borrowers in inner city neighborhoods may be forced to use subprime lenders because mainstream lenders are not doing business in their neighborhoods (see below).

Subprime borrowers are much more likely to be low income and be a minority than other borrowers. Between 1999 and 2001, 43.1 percent of subprime loans in the conventional conforming market went to low-income borrowers, compared with 29.5 percent of conventional conforming loans. During that same period, 19.9 percent of subprime loans were for African-American borrowers, compared with 6.5 percent of all conventional conforming loans. However, what distinguishes subprime loans from other loans is their concentration in African-American neighborhoods.

b. The Neighborhood Concentration of Subprime Lending

The growth in subprime lending over the last several years has benefited credit-impaired borrowers as well as those borrowers who choose to provide little documentation for underwriting. However, studies showing that subprime lending is disproportionately concentrated in low-income and minority neighborhoods have raised concerns about whether mainstream lenders are adequately serving these neighborhoods. A study of subprime lending in Chicago by The Woodstock Institute concluded that a dual, hyper-segmented mortgage market existed in Chicago, as mainstream lenders active in white and upper-income neighborhoods were much less active in low-income and minority neighborhoods-effectively leaving these neighborhoods to unregulated subprime lenders.^[189] As part of the HUD-Treasury Task Force on Predatory Lending, HUD's Office of Policy Development and Research released a national level study—titled Unequal Burden: Income and Racial Disparities in Subprime Lending in America—that showed families living in low-income and African-American neighborhoods in 1998 relied disproportionately on subprime refinance lending, even after controlling for neighborhood income. An update of that analysis for the year 2000 yields the following trends: ^[190]

• In 2000, 36 percent of refinance mortgages in low-income neighborhoods were subprime, compared with only 16 percent in upper-income neighborhoods.
• Subprime lending accounted for 50 percent of refinance loans in majority African American neighborhoods—compared with only 21 percent in predominantly white areas (less than 30 percent of population is African American).
• The most dramatic view of the disparity in subprime lending comes from comparing homeowners in upper-income African-American and white neighborhoods. Among homeowners living in the upper-income white neighborhoods, only 16 percent turned to subprime lenders in 2000. But 42 percent of homeowners living in upper-income African-American neighborhoods relied upon subprime refinancing which is substantially more than the rate (30 percent) for homeowners living in low-income white neighborhoods.
• Similar results are obtained when the analysis is conducted for borrowers instead of neighborhoods. Upper-income African-American borrowers are twice as likely as low-income white borrowers to have subprime loans. Over one-half (54 percent) of low-income African-American borrowers turn to subprime lenders, as does over one-third (35 percent) of upper-income African-American borrowers. By comparison, only 24 percent of low-income white borrowers and 12 percent of upper-income white borrowers, rely upon subprime lenders for their refinance loans.^[191]

It does not seem likely that these high market shares by subprime lenders in low-income and African-American neighborhoods can be justified by a heavier concentration of households with poor credit in these neighborhoods. Rather, it appears that subprime lenders may have attained such high market shares by serving areas where prime lenders do not have a significant presence. The above finding that upper-income black borrowers rely more heavily on the subprime market than low-income white borrowers suggests that a portion of subprime lending is occurring with borrowers whose credit would qualify them for lower cost conventional prime loans. A lack of competition from prime lenders in low-income and minority neighborhoods has increased the chances that borrowers in these communities are paying a high cost for credit. As explained next, there is also evidence that the higher interest rates charged by subprime lenders cannot be fully explained solely as a function of the additional risks they bear. Thus, a greater presence by mainstream lenders could possibly reduce the high up-front fees and interest rates being paid by residents of low-income and minority neighborhoods.

The Freddie Mac study presented evidence that subprime loans bear interest rates that are higher than necessary to offset the higher credit risks of these loans.^[192] The study compared (a) the interest rate on subprime loans rated A-minus by the lenders originating these loans with (b) the interest rates on prime loans purchased by Freddie Mac and rated A-minus by a Freddie Mac underwriting model. Despite the fact that both loan groups were rated A-minus, on average the subprime loans bore interest rates that were 215 basis points higher. Even assuming that the credit risk of the subprime loans was in fact higher than the prime loans, the study could not account for such a large discrepancy in interest rates. Assuming that default rates might be three to four times higher for the subprime loans would account for a 90 basis point interest rate differential. Assuming that servicing the subprime loans would be more costly would justify an additional 25 basis point differential. But even after allowing for these possible differences, the Freddie Mac researchers concluded that the subprime loans had an unexplained interest rate premium of 100 basis points on average.^[193]

Banking regulators have recognized the link between the growth in subprime lending and the absence of mainstream lenders and have urged banks and thrifts that lending in these neighborhoods not only demonstrates responsible corporate citizenship but also profitable lending. Ellen Seidman, former Director of the Office of Thrift Supervision, stated that, “Many of those served by the subprime market are creditworthy borrowers who are simply stuck with subprime loans or subprime lenders because they live in neighborhoods that have too few credit or banking opportunities.”

With respect to the question of whether borrowers in the subprime market are sufficiently creditworthy to qualify for more traditional loans, Freddie Mac has said that one of the promises of automated underwriting is that it might be better able to identify borrowers who are unnecessarily assigned to the high-cost subprime market. Freddie Mac has estimated that 10-30 percent of borrowers who obtain mortgages in the subprime market could qualify for a conventional prime loan through Loan Prospector, Freddie Mac's automated underwriting system.^[194] Fannie Mae has stated that half of all mortgage borrowers steered to the high-cost subprime market are in the A-minus category, and therefore are prime candidates for Fannie Mae.^[195]

c. Predatory Lending

Predatory lending has been a disturbing part of the growth in the subprime market. Although questions remain about its magnitude, predatory lending has turned homeownership into a nightmare for far too many households. The growing incidence of abusive practices has been stripping borrowers of their home equity, threatening families with foreclosure, and destabilizing neighborhoods. Also, in some cities, there are indications that unscrupulous realtors, mortgage brokers, appraisers, and lenders are duping some FHA borrowers into purchasing homes at an inflated price or with significant undisclosed repairs. The problems associated with home equity fraud and other mortgage abuses are not new ones, but the extent of this activity seems to be increasing. The expansion of predatory lending practices along with subprime lending is especially troubling since subprime lending is disproportionately concentrated in low- and very-low income neighborhoods, and in African-American neighborhoods.

The term “predatory lending” is a short hand term that is used to encompass a wide range of abuses. While there is broad public agreement that predatory lending should have no place in the mortgage market, there are differing views about the magnitude of the problem, or even how to define practices that make a loan predatory. The joint HUD-Treasury report, Curbing Predatory Home Mortgage Lending, concluded that a loan can be predatory when lenders or brokers: charge borrowers excessive, often hidden fees (called “packing fees”); successively refinance loans at no benefit to the borrower (called “loan flipping”); make loans without regard to a borrower's ability to repay; and, engage in high-pressure sales tactics or outright fraud and deception. These practices are often combined with loan terms that, alone or in combination, are abusive or make the borrower more vulnerable to abusive practices. Vulnerable populations, including the elderly and low-income individuals, and low-income or minority neighborhoods, appeared to be especially targeted by unscrupulous lenders.

One consequence of predatory lending is that borrowers are stripped of the equity in their homes, which places them at an increased risk of foreclosure. In fact, high foreclosure rates for subprime loans provide the most concrete evidence that many subprime borrowers are entering into mortgage loans that they simply cannot afford. The high rate of foreclosures in the subprime market has been documented by HUD and others in recent research studies.^[196] These studies have found that foreclosures by subprime lenders grew rapidly during the 1990s and now exceed the subprime lenders' share of originations. In addition, the studies indicate that foreclosures of subprime loans occur much more quickly than foreclosures on prime loans, and that they are concentrated in low-income and African-American neighborhoods. Of course, given the riskier nature of these loans, a higher foreclosure rate would be expected. With the information available it is not possible to evaluate whether the disparities in foreclosure rates are within the range of what would be expected for loans prudently originated within this risk class. But findings from these studies about the high rate of mortgage foreclosure associated with subprime lending reinforce the concern that predatory lending can potentially have devastating effects for individual families and their neighborhoods.

At this time, there are open questions about the effectiveness of the different approaches being proposed for eradicating predatory lending and the appropriate roles of different governmental agencies—more legislation versus increased enforcement of existing laws, long-run financial education versus mortgage counseling, Federal versus state and local actions. In its recent issuance of predatory lending standards for national banks, the Office of the Comptroller of the Currency (OCC) cited the efforts of Fannie Mae and Freddie Mac in reducing predatory lending.^[197] The OCC advised banks against abusive practices, such as rolling single-premium life insurance into a loan. The agency cited guidelines developed by Fannie Mae and Freddie Mac as a “useful reference” or starting point for national banks. Following publication of HUD's proposed 2000 Rule inviting comments on disallowing goals credit for high cost mortgage loans, Fannie Mae and Freddie Mac told lenders they would no longer purchase loans with certain abusive practices, such as excessive fees and failing to consider a borrower's ability to repay the debt.

It is important to re-emphasize that predatory lending generally occurs in neighborhoods where borrowers have limited access to mainstream lenders. While predatory lending can occur in the prime market, it is ordinarily deterred in that market by competition among lenders, greater homogeneity in loan terms and greater financial information among borrowers. Thus, one solution to address this problem would be to encourage more mainstream lenders to do business in our inner city neighborhoods.

Certain commenters urged the Department to adopt predatory lending safeguards in the final rule that would prohibit the GSEs from counting loans that included mandatory arbitration clauses or loans with prepayment penalties beyond three years towards the goals. In the 2000 rulemaking, the Department determined that the GSEs should not receive goals credit for purchasing high cost mortgages including mortgages with unacceptable features as explained in the preamble. The Department is aware that certain practices that were not enumerated in the regulations adopted in 2000, such as loans with prepayment penalties after three years and loans with mandatory arbitration clauses, often lock borrowers into disadvantageous loan products. The Department will rely on existing regulatory authorities to monitor the GSEs' performance in this area. Should the Department later determine that there is a need to specifically enumerate additional prohibited predatory practices, it will address such practices in a future rulemaking.

d. Purchases of Subprime Mortgages by the GSEs

Fannie Mae and Freddie Mac have shown increasing interest in the subprime market since the latter half of the 1990s. The GSEs entered this market by purchasing securities backed by non-conforming loans. Freddie Mac, in particular, increased its subprime business through structured transactions, with Freddie Mac guaranteeing the senior classes of senior/subordinated securities. The two GSEs also purchase subprime loans on a flow basis. Fannie Mae began purchasing subprime loans through its Timely Payment Reward Mortgage program in June 1999, and Freddie Mac rolled out a similar product, Affordable Merit Rate, in May 2000 (described below). In addition to purchasing subprime loans for borrowers with blemished credit, the GSEs also purchase another non-conforming loan called an Alternative-A or “Alt-A” mortgage. These mortgages are made to prime borrowers who do not want to provide full documentation for loans. The GSEs' interest in the subprime market has coincided with a maturation of their traditional market (the conforming conventional mortgage market), and their development of mortgage scoring systems, which they believe allows them to accurately model credit risk. Although the GSEs account for only a modest share of the subprime market today, some market analysts estimate that they could purchase as much as half of the overall subprime market in the next few years.^[198]

Precise information on the GSEs' purchases of subprime loans is not readily available. Data can be pieced together from various sources, but this can be a confusing exercise because of the different types of non-conforming loans (Alt-A and subprime) and the different channels through which the GSEs purchase these loans (through securitizations and through their “flow-based” product offerings). Freddie Mac, which has been the more aggressive GSE in the subprime market, purchased approximately $12 billion in subprime loans during 1999—$7 billion of A-minus and alternative-A loans through its standard flow programs and $5 billion through structured transactions.^[199] In 2000, Freddie Mac purchased $18.6 billion of subprime loans on a flow basis in addition to another $7.7 billion of subprime loans through structured transactions.^[200] Freddie Mac securitized $9 billion in subprime and Alt-A product in 2001 and $11.1 billion in 2002.

Fannie Mae's anti-predatory lending strategy includes eight major components. These components include: establishing business guidelines that ensure that liquidity is provided for only responsible lenders; expanding the application of conventional conforming mortgage practices to more borrowers; advancing the Mortgage Consumer Bill of Rights Agenda; offering a broad range of alternative responsible products; leveraging technology and the Internet to expand markets and reduce costs for consumers; working with partners to keep borrowers in their homes; supporting the home-buyer education industry to empower educators to reach more consumers; and supporting the Fannie Mae Foundation in consumer education and outreach.^[201]

In recent years, Freddie Mac has instituted measures designed to protect consumers from predatory lending. For example, Freddie Mac has announced that, effective August 1, 2004, they will no longer invest in subprime mortgages originated after that date that contain mandatory arbitration clauses. Since 2000, Freddie Mac has prohibited purchases of mortgages that impose a prepayment premium for a term of more than five years, and in March 2002, this prohibition was reduced to no more than three years. Freddie Mac does not purchase high-rate or high-fee loans that are covered by the Home Ownership and Equity Protection Act of 1994 (HOEPA); and they do not purchase mortgages containing a prepaid single-premium credit life, credit disability, credit unemployment or credit property insurance policy. Freddie Mac also requires all lenders servicing their loans to report monthly borrower mortgage payments to all four major credit repositories, and conducts onsite reviews of their customers and holds them accountable if their business practices do not meet Freddie Mac standards.^[202]

Fannie Mae initiated its Timely Payments product in September 1999, under which borrowers with slightly damaged credit can qualify for a mortgage with a higher interest rate than prime borrowers. Under this product, a borrower's interest rate will be reduced by 100 basis points if the borrower makes 24 consecutive monthly payments without a delinquency. Fannie Mae has revamped its automated underwriting system (Desktop Underwriter) so loans that were traditionally referred for manual underwriting are now given four risk classifications, three of which identify potential subprime (A-minus) loans.^[203] Fannie purchased about $600 million of subprime loans on a flow basis in 2000.^[204] Fannie Mae securitized around $0.6 billion in subprime mortgages in 2000, before increasing to $5.0 billion in 2001 and 7.3 billion in 2002.^[205] In terms of total subprime activity (both flow and securitization activities), Fannie Mae purchased $9.2 billion in 2001 and over $15 billion in 2002, the latter figure representing about 10 percent of the market, according to Fannie Mae staff.^[206]

A greater GSE role in the subprime lending market will most likely have a significant impact on the subprime market. Currently, the majority of subprime loans are not purchased by GSEs, and the numbers of lenders originating subprime loans typically do not issue a large amount of prime loans. Partly in response to higher affordable housing goals set by HUD in its new rule set in 2000, the GSEs are increasing their business in the subprime market. In the 2000 GSE Rule, HUD identified subprime borrowers as a market that can assist Fannie Mae and Freddie Mac in reaching their higher affordable housing goals while also helping establish more standardization in the subprime market. According to an Urban Institute study in 2002, many subprime lenders believe that successful companies serving high-risk borrowers need to have specialized expertise in outreach, servicing, and underwriting, which is lacking among most prime lenders.^[207] These lenders do not believe the more standardized approaches of prime lenders and the GSEs will work with subprime borrowers, who require the more customized and intensive origination and loan servicing processes currently offered by experienced subprime lenders.

As noted above, both Fannie Mae and Freddie Mac make the claim that the subprime market is inefficient, pointing to evidence indicating that subprime borrowers pay interest rates, points, and fees in excess of the increased costs associated with serving riskier borrowers in the subprime market.^[208] A recent Freddie Mac study found automated mortgage scoring less discriminatory and more accurate in predicting risk than manual systems such as those currently used by subprime lenders.^[209] According to Fannie Mae, although a high proportion of borrowers in the subprime market could qualify for less costly prime mortgages, it remains unclear why these borrowers end up in the subprime market.^[210] Fannie Mae and Freddie Mac believe they can bring more efficiency to the subprime market by creating standardized underwriting and pricing guidelines in the subprime market. An expanded GSE presence in the subprime market could be of significant benefit to lower-income and minority families if it attracted more mainstream lenders and competition to those inner-city neighborhoods that are currently served mainly by subprime lenders.

Several commenters indicated that to obtain the higher housing goals the GSEs would increase their purchasing of subprime loans. While some industry commenters welcome the entrance of the GSEs into the subprime market because their presence brings stability and standardizes business practices, they are concerned that unrealistically high goals could force the GSEs to jump into the market in a manner that negatively distorts underwriting and pricing. These commenters report that the GSEs can bring capital and standards but must gradually and carefully enter the subprime market in order to have a positive effect.

In the past, Fannie Mae and Freddie Mac have voluntarily decided not to purchases subprime loans with features such as single-premium life, HOEPA loans, and prepayment penalty terms that exceed three years. Freddie Mac indicated that the increased goals would limit its ability to influence subprime lending practices.

Several commenters suggest that if the GSEs are pushed to serve more of the subprime market, they will skim a significant portion of the lower-risk borrowers from that market. The resulting smaller subprime market would be comprised of the neediest borrowers. Concerned was raised by commenters that these higher risk borrowers would pay more based on three factors. First lower risk borrowers would not be present to subsidize them. Second, the market's high fixed costs would be distributed across fewer borrowers. Finally, a significantly smaller subprime market for private lenders would drive some lenders out of business translating into less competition.

9. Risk-Based Pricing

The expanded use of automated underwriting and the initial uses of risk-based pricing are changing the mortgage lending environment, often blurring the distinctions between the prime and subprime market. Prime lenders are now using automated underwriting systems that are being adapted to facilitate risk-based pricing. For some time, the majority of prime mortgage borrowers have received loan rates based on average cost pricing. Generally, borrowers receive roughly the same Annual Percentage Rate ^[211] (APR), regardless of the risk of loss to the lender. The risk of all borrowers is averaged together, and the price is determined by the average risk.

In contrast, risk-based pricing enables mortgage lenders to offer each borrower an individualized interest rate based on his or her risk. Or, more broadly, to offer interest rates based on whether or not the borrower falls into a certain category of risk, such as specific loan-to-value and FICO score combination or specified mortgage score range. Lenders could also set the interest rate based on various factors including the probability of prepayment and characteristics of the underlying collateral, as well as the default risk of the borrower. Borrowers that pose a lower risk of loss to the lender would then be charged a comparatively lower rate than those borrowers with greater risk. Rather than lower risk borrowers cross-subsidizing higher risk borrowers like in average cost pricing, lower risk borrowers pay a relatively lower rate.

In response to the expanded use of automated underwriting and pressures from the GSEs, other purchasers of loans, mortgage insurers, and rating firms, lenders are increasing their use of risk-based pricing.^[212] In today's markets, some form of differential pricing exists for the various subprime categories, for new products targeted at credit-impaired borrowers (such as Fannie Mae's Timely Payments product), and for private mortgage insurance across all credit ranges. For example, private mortgage insurers use FICO scores and “Accept” determinations from the GSEs” automated underwriting systems to make adjustments to insurance premiums.^[213] Rating agencies vary subordination requirements based on the credit qualify of the underlying collateral.

Many believe there is cross-subsidization within the crude risk categories used in today's market. For example, some of the better quality subprime borrowers in the A-minus category may be inappropriately assigned to the subprime market. The GSEs and others are attempting to learn more about the subprime market, and their initial efforts suggest that there will be an increase in the use of risk-based pricing within this market, although it is recognized that the expansion of risk-based pricing depends importantly on these parties gaining a better understanding of the subprime borrower and the ability of their mortgage scoring systems to predict risk within this market. It must be noted that the power of the underlying algorithm in automated underwriting systems determines the ability of these systems to accurately predict risk and set prices.

If prime lenders adopted risk-based pricing, many would be willing to lend to riskier subprime borrowers because their risk would now be offset with an increase in price. In theory, the mortgage market should expand because all mortgages will be approved at a price commensurate with risk, rather than setting a risk floor and approving no one beneath the floor. Risk-based pricing could also expand the prime lenders' market by enabling them to reach a new group of underserved customers.^[214] Taking advantage of GSEs' lower cost of capital, GSEs may be able to offer borrowers who could not afford a rate in the subprime market a rate they can afford resulting from risk-based pricing.

Risk-based pricing also poses challenges on the mortgage market because some of the more risky borrowers (who are currently cross-subsidized by less risky borrowers) may not be able to afford their higher, risk-based interest rate. Also, the adoption of an automated risk-based pricing system may have an uncertain effect on minority groups, who tend to have lower credit scores, as discussed earlier. On the other hand, if minorities are eligible for prime financing, the cost of financing minorities may fall as will the potential for subprime lenders to draw minorities to their higher-priced products.

As the GSEs become more comfortable with subprime lending, the line between what today is considered a subprime loan versus a prime loan will likely deteriorate, making expansion by the GSEs look more like an increase in the prime market. This melding of markets could occur even if many of the underlying characteristics of subprime borrowers and the market's evaluation of the risks posed by these borrowers remain unchanged. Increased involvement by the GSEs in the subprime market will result in more standardized underwriting guidelines and the increased participation of traditional lenders. In fact, there are indications that mainstream players are already increasing their activity in this market. According to staff from Moody's Investors Service, the growing role of large mortgage aggregators in the subprime market has been a key factor in the improving credit qualify on deals issued in 2002.^[215] According to a representative from Washington Mutual, subprime credit qualify has also improved as lenders carve out new loan categories that fall somewhere between the large Alt A market and traditional subprime business.^[216] As the subprime market becomes more standardized, market efficiencies will reduce borrowing costs. Lending to credit-impaired borrowers will, in turn, increasingly make good business sense for the mortgage market.

D. Factor 2: Economic, Housing, and Demographic Conditions: Multifamily Mortgage Market

1. Introduction

At the time of the previous GSE rulemaking in 2000, the multifamily rental housing market was coming off several years of generally positive performance. Vacancies were low in most markets and rent increases were matching or exceeding economy-wide inflation. A key to this strong performance was the volume of new multifamily construction, which was at a level consistent with demand growth. Job growth and income gains helped many renters pay the higher rents without undue burden. As always, conditions varied from region to region, and across market segments, but the overall tone of the apartment market was quite healthy.

Much has changed in the subsequent years. An economic slowdown reduced apartment demand, and with new multifamily construction about unchanged, vacancies rose and rents softened. Provision of decent housing affordable to households of moderate or low incomes is a challenge even in strong economic times, and with the unemployment rate rising above 6 percent before falling to about 5 and a half percent, affordability problems increased for many, despite the softness in rents.

Despite the recent weakness in the apartment property market, the market for financing of apartments has grown to record volumes. The favorable long-term prospects for apartment investments, combined with record low interest rates, has kept investor demand for apartments strong and supported property prices. Refinancings too have grown, and credit quality has remained very high. Fannie Mae and Freddie Mac have been among those boosting volumes and introducing new programs to serve the multifamily market.

This section will review these market developments, interpret the performance of Fannie and Freddie within that market context, and discuss future prospects for the multifamily rental market, its financing, and the GSE role. The intention here is only to update the discussion from 2000. For general background information on the multifamily mortgage market and the GSEs, see the 2000 Rule and the HUD-sponsored research report, Study of Multifamily Underwriting and the GSEs' Role in the Multifamily Market (Abt Associates, 2001).

2. The Multifamily Rental Housing Market: 2000-2003

The definition of “good” market conditions in multifamily rental housing depends on one's perspective. Investors and lenders like low vacancies, steady rent increases, and rising property values. Developers like strong demand for new construction and favorable terms on construction financing. Consumers, in contrast, prefer low rents and a wide selection of available apartments.

The mid- to late-1990s were among the most successful of recent history, in that apartment market conditions were generally good for all of these interest groups. Investment returns were favorable, construction volumes were steady at sustainable levels, and many consumers had income gains in excess of their rent increases.

Market conditions for multifamily rental housing began to weaken toward the end of 2000. Early warnings came from the publicly traded apartment companies, some of which reported easing in demand growth in the first months of 2001, coinciding with a slowdown in job growth to its lowest level since 1992.

By 2003, rental units were experiencing record high vacancy rates and newly completed apartments faced record low absorption or ease-up rates. The rental sector vacancy rate averaged 9.8 percent in 2003, up 0.8 percent from 2002, and the highest annual vacancy rate I the more than 40-year history of the measure.^[217]

Apartments—especially those serving the top end of the rental market—appear to have performed worse than other rental housing in the past four years, after several years of rent growth and occupancies surpassing the rental market averages. The multifamily (5+ units in structure) vacancy rate has increased more than the overall rental market vacancy rate in each of the years 2000, 2001, 2002, and 2003. For example, the Census Bureau's estimate of a 0.9 percentage point increase in vacancies for multi-family apartments in 2003 exceeds the overall rental vacancy rate of 0.6 percent.^[218] Similarly, while rent growth has decelerated slightly for all rental housing according to the CPI, industry surveys of apartment rents show year-over-year declines in rents in many local markets.^[219] In 2003, asking rents remained flat nationally, as multifamily completions declined 5 percent.^[220]

a. Apartment Demand and Supply

The primary reason for the softening in the multifamily rental market has been a reduction in the growth of consumer demand for apartment housing. The general slowdown in economic activity meant fewer apartment customers, with less money, than if the economy were vigorously expanding. Persistent low interest rates have also enticed renters into the home purchase market as evidenced by the U.S. homeownership rate, which grew to 68.4 percent in 2003, further contributing to a weakness in rental demand.

The reduced demand is most evident in the national statistics on employment. Job growth began decelerating in late 2000 and throughout 2001, turning negative late that year. The largest year-over-year job loss of the economic downturn occurred in February 2002, and year-over-year losses have continued through October 2003.^[221] Apartment demand seems particularly sensitive to labor market conditions, given the importance of rental housing to mobile individuals and families accepting new jobs or transfers. Reis, Inc., a real estate market research firm, estimates that the total number of occupied apartments (in properties with 40+ units) actually declined in both 2001 and 2002 in the large markets nationwide that are monitored by the company.^[222] Job numbers showed some rebound in the subsequent period.

Households, not jobs, fill apartments, and for this reason household formations are a preferable indicator of demand for apartments as well as other types of housing. The Census Bureau estimates that the total number of renter households nationwide has been essentially unchanged at approximately 34.8 million since 1996. Yet during the late 1990s apartment demand was expanding, and apartments were apparently picking up market share from other rental housing. The past two or three years may have seen a reversal of that trend in share.

Long-term demographic trends are expected to be favorable for rental housing demand.^[223] The maturing of the “Baby Boom Echo” generation will increase the number of persons under age 25 who will seek rental housing, immigration is expected to continue to fuel demand for rental housing, and minority populations, while increasing their homeownership rates, are growing and will contribute to higher absolute demand for rental housing. Thus demographic trends support an improvement in the long-run demand for rental demand, which is likely to include higher multifamily rental demand.

Supply growth has been maintained, even though the current reduced multifamily demand warrants less new construction. Total multifamily starts (2+ units) have been running approximately 325-to-350 thousand annually for the past six years, according to Census Bureau statistics, adding about 1 percent annually to the total multifamily stock. Most of these new units are built for rental use, with only about 20 percent in recent years reported as being built as for-sale condominium units.

The reduced short-term demand has shown through in absorption speeds for new apartments. The percentage of newly completed unfurnished apartments rented within three months of completion fell from 72 percent during 2000 to 63 percent during 2001 and to 59 percent during 2002, the lowest level in the 33-year history of the data series, according to the Census Bureau. This percentage rose slightly to 60 percent in 2003.^[224]

b. Performance by Market Segments

Some segments of the multifamily rental market have been more affected than others by the recent softening. As mentioned earlier, the top end of the apartment market seems especially hard hit, as measured by rising vacancies and reduced rent growth. This segment is particularly dependent on job growth and transfers for new customers, and is particularly vulnerable to losses of residents and prospective customers to home purchase. According to reports by apartment REITs and other investors, these top-end properties have not been getting the job-related in-movers, but have still been losing a lot of customers to home purchase. These properties generally have annual resident turnover rates of above 50 percent, and thus are particularly quickly influenced by changes in demand. Furthermore, this is the segment of the apartment market into which most of the new construction is built.

Performance has varied geographically as well. Some of the coastal markets, especially in Northern California, saw the double-digit rent increases of the late 1990s replaced by double-digit declines, before stabilizing more recently. “Supply constrained markets” had been preferred by apartment investors during the 1990s, but recent market performance has reminded investors and analysts that all markets have their day. For example, Houston posted the biggest year-over-year rent increase of any major apartment market in 2001, despite a long-run history of moderate rent growth and few barriers to new apartment construction. Rent changes in the 27 metro markets for which estimates are available from the CPI ranged from a low of −0.3 percent to a high of 6.7 percent in the first half of 2003 relative to a year earlier. And across the 75 metro areas for which rental vacancy rates (apartments plus other rentals combined) are available, rates for the year 2002 ranged from 2.4 percent to 15.4 percent, according to the Census Bureau. In a historical context, this variation is moderate, although up somewhat since the late 1990s.

Conditions in the “affordable” segment of the apartment market are harder to track than in the high-end segment because of lesser investor interest and analyst coverage. Data for the late 1990s analyzed by the National Housing Conference saw affordability problems continuing, although a study of apartment renters by the National Multi Housing Council saw some improvement in affordability during the strong economic growth of 1997-1999.^[225] Other work noted that rent to income ratios for the lowest income quintile of renters rose during the late 1990s even as these ratios were stable or declining for other renters.^[226] Harvard's State of the Nation's Housing report for 2002 highlighted the variability of the affordability problem from place to place.^[227]

Little research is available on affordability trends since 1999. However, tabulations from the 2001 American Housing Survey indicate that income growth between 1999 and 2001 in the lowest quintile of renter households continued to lag that of higher income renters, and fell short of the average rent increases during this period. Together, these statistics suggest that affordability has deteriorated early this decade among at least this group of very low-income renters. Other work using the AHS found that the number of low-to moderate-income working families with severe rental cost burdens increased 24 percent between 1999 and 2001.^[228]

The low-income housing tax credit (LIHTC) continues to finance much of the newly built multifamily rental housing that is affordable to households with moderate income. Restricted to households with incomes no greater than 60 percent of the local median, this program financed approximately 75,000 units in 2001, according to the National Council of State Housing Agencies, after running in the mid-to high-60 thousand range the previous three years. About 70 percent of these units are newly built, and the rest are renovations of existing units.

Expenditures for improvements to existing rental apartments have grown in recent years. In 2001 the total of $11.3 billion was nearly twice the figure of three years earlier, according to the Census Bureau, and more than a third as large as construction spending for newly built multifamily structures, including owner-occupied condos. Many of these improvements are to older properties in high-demand neighborhoods. Improvements to the physical structures have external benefits. But often the renovations are in connection with re-positionings that move the apartments into a higher rent range and bring changes in the demographic composition of the resident base.

In 2002, expenditures on total improvements to existing apartments declined to $9.8 billion, while new construction spending increased $2 billion. This shift further suggests a re-positioning to apartments with a higher rent range. Excluding units financed with tax credits or other subsidies, most of the multifamily rental construction in recent years has been targeted on the upper end of the market, often the only segment for which unsubsidized new construction is economically feasible. The median asking rent on new unfurnished apartments completed in 2001 was $877, up 11 percent over the previous two years. In 2002 median asking rent for these properties was $905. Of those units brought to market in 2002, 45 percent were at rents at or above $950.

3. Multifamily Financing Trends

In contrast to the softening observed in the demand/supply balance for multifamily, mortgage financing of these properties has been at a record pace in the past three years.

a. Lending Volume

Total multifamily mortgage debt outstanding increased 11 percent in 1999, 8.7 percent in 2000, 11.2 percent in 2001, 9.6 percent in 2002, and 11.2 percent in 2003 according to the Federal Reserve's flow of funds accounts. The dollar volume for 2003, $544.2 billion, is above those of any previous year. The pace seems to have slowed for 2004, with the first quarter indicating an annualized growth of 4.9 percent. Furthermore, a 2003 survey by the Mortgage Bankers Association of America show that of 48 member firms surveyed, representing all large mortgage banking firms an a cross section of smaller mortgage companies, multifamily origination volume increased 21.5 percent in 2003—from $41 billion in 2002 to $49.8 billion in 2003.

The apparent inconsistency between current market fundamentals and financing can be explained by low interest rates. The same financial forces that lowered the mortgage rates for home purchasers to record lows by 2002 also reduced the financing costs of multifamily properties. The ten year Treasury yield, a common benchmark for multifamily loan pricing, fell to a 45-year low of 3.3 percent in June 2003 from 6.3 percent as recently as the end of 1999.

Another feature boosting investor demand for apartment properties and the resulting demand for debt to finance those purchases has been the lack of attractive returns on many financial assets and other alternative investments. Despite the current weak performance of apartments, investors apparently are looking through to the long-run outlook for these assets, which is generally thought to be favorable, as indicated most recently by investor surveys fielded by the Urban Land Institute and by Lend Lease and PriceWaterhouseCoopers.^[229]

The net change in mortgage debt outstanding is defined as loan originations less repayments and charge offs. As discussed in Appendix D, net change is a lower bound on originations. By all accounts, originations—for which no single source of estimates is available—are much higher than net change in most years. High levels of refinancings of existing multifamily mortgages in recent years has been a factor in originations exceeding the net change in debt outstanding.

Most mortgage lending is in the “conventional” market. Multifamily loan programs of the Federal Housing Administration accounted about $7 billion in new insured mortgages in fiscal year 2003—up from $6 billion in fiscal year 2002 and $5 billion in fiscal 2001. Despite the recent increase in FHA originations, and the likely continued strong performance for FHA multifamily programs in the foreseeable future, ^[230] FHA remains but a small portion of the total multifamily mortgage market. Outstanding FHA-insured multifamily mortgage debt was $55 billion at the end of the first quarter of 2003—only about 11 percent of all multifamily mortgage debt outstanding.

Multifamily lending has been spurred by new apartment construction, property sales, and refinancings. New multifamily construction was valued at $34.1 billion in 2003, according to the Census Bureau, up 21 percent from 2000.^[231] The number of new multifamily units completed over this period actually declined 12 percent, and the increased expenditures reflect higher costs per unit. The increase in asking rents described earlier suggests higher property values and greater debt carrying capacity.

b. Property Sales and Refinancings

Sales of existing apartment properties tend to be procyclical. Increasing asset values bring buyers to the market and tempt sellers to realize their capital gains. In soft markets, in contrast, the bid-ask spread generally widens and the volume of sales declines, as sellers perceive current offers as beneath the property's long run value and buyers are reluctant to pay for past performance or the hope of future gains. Sales tend to increase mortgage debt, because the loan originated to finance the purchaser's acquisition is typically considerably larger than the mortgage retired by the seller.

No source of apartment property sales statistics matches the comprehensive national coverage of the single-family market provided by the National Association of Realtors' monthly estimates. But surveys by the National Multi Housing Council and other apartment industry reports indicate that transactions volume dipped during 2001 but since then have grown appreciably in both number of sales and aggregate dollar value.

Mortgage lending volumes have recently been boosted by shifts in property ownership. Publicly traded real estate investment trusts had been the big gainers during most of the 1990s, and by 1999 owned nearly 6 percent of all apartments nationwide and a considerably larger share of all big (100+ unit) properties. But beginning in 1999 capital market developments made private buyers more competitive. Since then the number of apartments owned by large REITs has declined about 5 percent, with diverse private interests apparently picking up market share.

Private investors are able to use more leverage—greater debt—in financing their transactions than the market permits the public REITs. As a result, the very low mortgage rates recently have given them an advantage in bidding on properties. In addition, equity funding costs of REITs rose as their stock prices flattened or moved down as part of the broader equity market correction.

Refinancings have, by all accounts, also been strong. Despite the lockout provisions and yield maintenance agreements that constrain early refinancings of many multifamily loans, lenders reported very strong refinancing activity in 2001 and continuing into 2002. Although refinancing volume data for the entire market are not available, the trends in refinance volume for FHA and the GSEs show very strong increases in refinance activity during 2002 and 2003. For example, FHA's Section 223(a)(7) program, which is limited to refinancing of FHA multifamily mortgages, experienced an increase in origination volume of 133 percent in Fiscal Year 2003 and 181 percent in Fiscal Year 2002. ($1.73 billion in FY 2003, $0.74 billion in FY 2002, and $0.26 billion in FY 2001). Similarly, the GSEs increased their combined volume of refinances by 83 percent from 1999-2000 to 2001-2002, from $17.6 billion to $32.1 billion. Refinancings, especially when motivated by a desire to lower interest expense rather than to extract equity, do not add as much to debt outstanding as do purchase loans, which often are much larger than the seller's existing mortgage that is repaid at the time of sale. Nonetheless, refinancings represent a significant part of all multifamily mortgage lending.

c. Sources of Financing and Credit Quality

The sources of funding of multifamily mortgages shifted somewhat in the past few years, judging from the Flow of Funds accounts. As shown in Table A.4, four categories of lenders have dominated multifamily mortgage lending since the mid-1990s. Of those four, commercial banks have played a lesser, although still substantial, role in recent years, providing 20 percent of the $86 billion in net additional funding of multifamily mortgages during 2000 and 2001. The portfolio holdings of the GSEs, by contrast, have been much more important than during the last half of the 1990s. Mortgage backed securities, both from the GSEs and especially from other issuers, accounted for proportionally less of the growth in 2000-01 than in 1995-99, but between them still accounted for nearly half of all the net credit extensions. Some slight broadening of the base of multifamily lending in the past two years, as these four lender groups accounted for only 85 percent of the net credit extended in 2000 and 2001, compared to all of it in the previous five-year period.

The market values of apartment properties have generally held up well, although the most recent indicators suggest some flattening. Properties in the portfolios of pension funds continued to appreciate into the second half of 2002, according to the National Council of Real Estate Investment Fiduciaries, although at a reduced annual rate of less than 2 percent. And the sales price per square foot of “Class A” properties monitored by Global Real Analytics rose until turning down in early 2002, posting a 1.6 percent year over year decline in the second quarter.

The continuing value of collateral has helped keep loan quality high on multifamily mortgages. Delinquency rates from all major reporters are at or near record lows, and well below the rates reported for single-family mortgages and commercial properties. At commercial banks, the FDIC reports a 0.38 non-current loan percentage in the second quarter of 2002. In life insurance company portfolios the only 0.05 percent of residential mortgages were overdue at the end of 2002, and as of the third quarter of 2002 the GSEs were both reporting similarly miniscule delinquency rates of below 0.1 percent; all of these rates are below those of a year earlier.

Multifamily lenders have remained cautious in their underwriting and, together with their regulators; have avoided repeating the mistakes of the 1980s. Many of the senior loan officers surveyed quarterly by the Federal Reserve have reported tightening their terms on commercial mortgages, and that shift likely has occurred in their multifamily lending as well. Perhaps the best indicator of discipline in multifamily lending is the fact that, despite the strong apartment demand during the last half of the 1990s, construction never rose above its long-run sustainable level, unlike the rampant overbuilding that plagued the industry in the mid- and late-1980s.

4. Recent GSE Involvement in Multifamily Finance

As the multifamily mortgage market has expanded since 1999, Fannie Mae and Freddie Mac have increased their lending, picked up market share, introduced new programs, and enhanced others.

Beginning with their whole loans, the GSEs added 34 percent to their combined holdings of multifamily loans in 2001, and another 26 percent in 2002 (see Table A.6 below). The growth in multifamily MBS volume was nearly as dramatic, increasing 26 percent in 2001 and another 14 percent in 2002. The gains resulted in the GSEs increasing their share (whole loans and securities combined) of all multifamily debt outstanding to 22.8 percent by the third quarter of 2003, up from 19 percent at year-end 2001, 15 percent at year-end 1999 and 11 percent at the end of 1995. By this combined measure of portfolio holdings and MBS outstanding, at year-end 2002 Fannie Mae had nearly twice ($65 billion versus $37 billion) the multifamily business of Freddie Mac, although Freddie was growing its multifamily business more rapidly (67 percent increase between 2000 and 2002, compared to 46 percent increase for Fannie Mae). In 2003, Freddie Mac's multifamily business activities totaled $21.587 billion ($14.894 billion of mortgage purchases and $6.693 billion in investment activities). These activities financed rental housing for 549,083 families. Nearly 92 percent of these units were affordable to low- and moderate-income renters. Since 1993, Freddie Mac has purchased $75.5 billion in multifamily mortgages, financing housing for more than 2.2 million families.^[232]

Measures that focus on new multifamily activity, specifically gross mortgage purchase volumes and new security issuance, vary across recent years and between the GSEs. For the GSEs combined, these measures of current business activity show sharp gains of over 70 percent in 2001, following small decreases in activity in 2000. In 2002, the GSEs combined posted small declines for both measures. Measures of multifamily gross mortgage purchases and new security issuance diverged for the two GSEs in 2002. Fannie Mae experienced declines in these balance sheet and new business indicators in 2002 while Freddie Mac experienced gains, particularly in new security issuance. As discussed earlier, the credit quality of GSE multifamily loans has remained very high even with the large gains in loan volume.

Despite the substantial pickup in GSE multifamily activity, the position of these companies in the multifamily mortgage market remains well below their dominance in single-family mortgage finance. At the end of 2002, the GSEs' market share of single family debt outstanding was 44 percent, twice the share of multifamily debt held or securitized by these two companies, according to Federal Reserve statistics. Furthermore, the multifamily share of all housing units financed by the GSEs combined has declined from its 1997 level (Table A.5), although the annual statistics are heavily influenced by the volume of refinancings in the single-family market, which spiked in 1998 and again in 2001 and 2002 in response to the big decline in mortgage rates in those years. Because of lock-out agreements and other loan covenants, multifamily loans are not as prone to rate-induced refinancings as are single-family mortgages.

a. Contrasting Business Models

While both Fannie Mae and Freddie Mac have significantly increased their multifamily activities in recent years, they have pursued distinct business models in achieving that growth. As shown in Table A.6, most of Fannie Mae's multifamily growth has come in MBS products, whereas Freddie Mac has relied more on loans purchased and held in its portfolio. At the end of 2002, Fannie Mae had almost four dollars of outstanding MBSs for every dollar of portfolio holdings. Freddie Mac, on the other hand, more than three times as much volume in portfolio as it had in MBS outstanding.

The differing emphasis on portfolio holdings and securities issuance is related to the GSEs' contrasting approaches to credit underwriting.^[233] Fannie Mae has long had risk-sharing arrangements with its multifamily loan originators, and currently has over 25 Delegated Underwriters and Servicers who are authorized to originate loans meeting Fannie Mae's requirements for sale to the GSE without prior approval of individual transactions. These “DUS” lenders retain part of the credit risk on the loans sold to Fannie.

Freddie Mac has taken a different approach to credit underwriting. In the wake of large credit losses on its multifamily business in the late 1980s and 1990, Freddie Mac essentially withdrew from the market. When it re-entered in late 1993, the company elected to retain all underwriting in-house and not delegate this function to the loan originators participating in Freddie Mac's Program Plus network. Because Freddie Mac assumes the entire credit risk on loans it purchases, some commercial banks and other financial institutions desiring to remove multifamily loans and all related liabilities from their books find Freddie Mac's program preferable.

b. Affordable Multifamily Lending

Because most of the GSEs' multifamily lending is on properties affordable to households with low-or moderate incomes, financing of affordable multifamily housing by the GSEs has increased almost as much as their total multifamily lending. Approximately 87 percent of Fannie Mae's multifamily lending volume in 2003 qualified as affordable to low-or moderate income households, according to Fannie Mae's annual Housing Activity Report, as did 92 percent of Freddie Mac's multifamily units financed. For the entire multifamily rental market, HUD estimates that 90 percent of all housing units qualify as affordable to families at or below 100 percent of the area median income, the standard upon which the low- and moderate-income housing goal is defined.

Owing to this high propensity to qualify as affordable lending, financing of multifamily rental housing is especially important for the GSEs attainment of their affordable housing goals. Less than 8 percent of the units financed by the GSEs in 2002 were multifamily rentals, as described above. Yet 15 percent of the units qualifying as low- and moderate-income purchases were multifamily, according to Table 1 of the GSEs' activity reports for 2002.

The GSEs increased the volume of their affordable multifamily lending dramatically in 2001, the first year of the new, higher affordable housing goals set for the GSEs. As measured by number of units financed, the total affordable lending (shown in the “low-mod total” rows of Table A.7) more than doubled from a year earlier, especially after application of the upward adjustment factor authorized for Freddie Mac in the 2000 Rule. In 2003 the GSEs maintained a high volume of affordable multifamily lending.^[234]

The figures in Table A.7 are exclusive of the “Temporary Adjustment Factor (TAF)” granted to Freddie Mac as part of the 2000 Rule. The TAF was a response to Freddie Mac's limited opportunities for refinancing business because of its minimal involvement in the multifamily market in the early and mid-1990s.^[235] The TAF, which expired at the end of 2003, provided a 20 percent upward adjustment to multifamily units in properties with 50 or more units, for purposes of the affordable housing goals.

Multifamily financing made major contributions not only to the GSEs' attainment of the overall goal for affordable lending in 2002, but also to the “underserved areas” goal and “special affordable” goal. As shown in Table A.7, the 2001 increases in lending in each of these categories were substantial at both Fannie Mae and Freddie Mac, again leveling off for both in 2002. The GSEs also met the special multifamily affordable subgoal set in the 2000 Rule in both 2001 and 2002.

c. Multifamily Initiatives of the GSEs

Fannie Mae and Freddie Mac have taken a number of steps since 2000 to expand their multifamily lending and to respond specifically to the goals established in the 2000 Rule. These initiatives are summarized in the annual activity reports filed by the GSEs.^[236]

One focus of the 2000 Rule was on lending to small (5-to-50 units) multifamily properties, which the Rule identified as an underserved market. HUD-sponsored research has found that the supply of mortgage credit to small properties was impeded by the substantial fixed costs of multifamily loan originations, by owners' insufficient documentation of property income and expense, and by the limited opportunities for fees for underwriting and servicing small loans.^[237] As a result, many multifamily lenders focus on larger properties, which were found to have more loan products available to them and to pay lower interest rates than did small properties.

In an attempt to promote the supply of credit to small properties, the 2000 Rule provided incentives for the GSEs to step up their involvement in this segment of the multifamily mortgage market. The incentives likely contributed to the huge increases in small property lending posted by both Fannie Mae and Freddie Mac in 2001 and continuing into 2002 (Table A.7). The combined total of these units financed in 2001 and 2002 was almost 8 times those financed in the previous two years. This lifted the percentage of all GSE multifamily lending that was on small properties to their highest levels ever.

During 2003, multifamily business activity at Fannie Mae topped $33 billion which financed over 809,703 multifamily units. Of this total, over 87% were affordable to families at or below the median income of their communities.^[238] Freddie Mac multifamily business activities totaled a record $21.587 billion which financed rental housing for 549,083 families. Nearly 92 percent of these apartment units were affordable to low- and moderate income renters.^[239]

Programs introduced or enhanced by the GSEs in the past two years have contributed to these striking numerical results. Delegated Underwriting and Servicing (DUS) is Fannie Mae's principle product line for purchasing individual multifamily loans. This product line is offered through 26 lenders with expertise in financing multifamily properties. In 2003, 91% of the DUS loan activity served affordable housing needs, 42% of DUS loans in underserved markets, and 52% addressed “special affordable” needs.^[240] Believing that small multifamily properties are a vital part of the country's affordable housing stock, Fannie Mae has focused efforts on providing financing for these projects through the development of the MFlex Loan Product, the 3MaxExpress Streamlined Mortgage Loan Product and the Affordable Alliances Loan Product. The MFlex Loan Product was established in 2000 to target lending partners that serve small property borrowers and increase Fannie Mae's participation in the 5-50 unit property market. By 2003, Fannie Mae had seven MFlex lending partners and had purchased $1.6 billion of these loans. Fannie Mae markets its specialized 3MaxExpress Streamlined Mortgage Loan Product line for loans worth less than or equal to $3 million. In 2003, Fannie Mae provided $1 billion in financing, which assisted over 34,000 families living in small multifamily properties. The Affordable Alliances Loan Product is responsible for debt investments in rental housing targeted to persons of low- and moderate-income and to rental markets that are underserved. During 2003, these financing initiatives provided affordable housing for 3,850 families.^[241] Fannie Mae additionally has federal Low-Income Housing Tax Credit (LIHTC) programs and special financing projects for special use properties such as Seniors Housing. In 2003, Fannie Mae committed over $1.6 billion in LIHTC equity properties to help make affordable rental housing possible for over 30,000 families.^[242]

During 2003, Freddie Mac used innovative financing structures combined with prudent, flexible multifamily lending practices, which enabled them to reach a record level of multifamily mortgage purchases.^[243] The GSEs face strong competition in this market from small banks and other depository institutions that prefer to hold these loans in their own portfolios.^[244]

In 2003, Freddie Mac continued to test initiatives through pilots, and implement enhancements to existing multifamily mortgage products which cover a broad array of eligible mortgage products. Freddie Mac's tax-exempt bond credit enhancements with synthetic fixed-rate financing continued to be popular. Freddie Mac's innovations to certain cash products including various combinations of fixed-rate, adjustable-rate and interest-only mortgages have been adopted by others in the industry. For example, the Fixed-to-Float execution provides borrowers with a reduced fixed interest rate and a one-year extension of the mortgage term at a floating rate. In 2003, borrowers used Fixed-to-Float option for $4.0 billion in mortgages.^[245]

In 2003, Freddie Mac purchased $6.6 billion in mortgages to finance more than 181,000 apartment units in 5-to 50-unit properties. Freddie Mac committed to invest $958 million to Low Income Housing Tax Credits (LIHTC). Altogether, the LIHTC investments made by Freddie Mac are approaching the $3.6 billion mark and have constructed or rehabbed more than 216,000 rental units for very-low and low income families in close to 3,000 projects. In 2003, Freddie purchased $412 million in newly issued multifamily mortgage revenue bonds. These bonds, issued by state, county or city government agencies, finance the acquisition and rehabilitation of nonprofit borrowers or property owners who agree to keep rents at affordable levels. These multifamily bond purchases will finance 6,100 estimated units of affordable housing with an estimate that 58 percent of those units will be affordable to very low income families. In 2003, Freddie issued a record $7.7 billion of securities backed by multifamily mortgages through negotiated transactions. More than 85 percent of these securities financed mortgages for affordable housing.^[246]

The 2000 Rule discussed other ways in which the GSEs might help promote financing of affordable multifamily housing. Two of those were lending for property rehabilitation and leadership in establishing standards for affordable multifamily lending. Many affordable properties are old and in need of capital improvements if they are to remain in the housing stock. Rehabilitation lending is a specialized field, and one in which the GSEs for a variety of reasons have not been major players. Less than 1 percent of all GSE multifamily lending in 2002 was for property rehabilitation. In 2002, Fannie Mae hosted its first ever Preservation Advisory Meeting with leaders in the housing and real estate finance industry to identify best practices and formulate real world solutions to this critical policy issue.^[247]

Setting standards for affordable multifamily lending was identified in the 2000 Rule as another area where the GSEs could provide greater leadership. It was also noted, based on HUD-sponsored research underway at that time,^[248] that market participants believe the GSEs to be conservative in their approaches to affordable property lending and underwriting. Actions described in the GSEs' annual activity reports for 2001, 2002 and 2003 indicate attempts by the GSEs to promote market standards that will reduce the transactions costs of multifamily lending while also providing programs that have the flexibility needed to deal with unique circumstances.

5. Future Prospects

The outlook for the multifamily rental housing market is marked by near-term risks and longer-run optimism, according to most observers. The prospects for the next few quarters are dominated by the macroeconomy. In particular, job growth, with its implications for formations of households, will be a key for the resumption of growth in apartment demand. Many forecasters would ascribe to the Federal Reserve's forecast of a slight increase in GDP growth to 4.3 percent in 2004 ^[249] , while also agreeing with the Fed's warning that “An unusual degree of uncertainty attends the economic outlook at present, in large measure, but not exclusively, because of potential geopolitical developments.” ^[250]

When consumer demand does pick up, recovery should be reasonably fast. While the recent production levels have outpaced demand, they have been near the middle of the long run historical range and very close to the average of the last half of the 1990s. Judging from the firm tone to rents and vacancies during that period, total multifamily completions production of 275,000 to 350,000 units is a sustainable level of annual production—that is, the level consistent with long run demographic trends and replacement of units lost from the stock.

Because new construction has remained moderate, there is no massive overhang of product that will need to be absorbed. With increased demand, vacancies should fall and rents firm reasonably promptly. A key assumption behind this forecast for vacancies and rents is that new apartment construction will not rise appreciably from its current level.

Recovery in the apartment market may also, perversely, be promoted by the recent unprecedented strength of the single-family market. Typically, economic recoveries bring strong growth in single-family housing demand, some of that coming from apartment renters seeking more space. With single-family activity already near record highs, boosted by historically low mortgage interests rates and despite the recently soft economy, it is uncertain how much higher single-family demand—and the accompanying losses of apartment customers to homeownership—can go.

A stronger economy will put the multifamily rental market back onto a long-run path that appears to promise sustained, moderate growth. As discussed in the 2000 Rule, the demographic outlook is favorable for apartment demand. Even if the homeownership rate increases further and the total number of renter households grows only slowly, as described in the discussion of the single-family housing market earlier in this Rule, apartment demand can be expected to increase more rapidly than that for other rental housing, owing to the likely changes in age composition and reductions in average household size. One estimate projects the annual growth in apartment households to be one percent.^[251]

a. The Outlook for Multifamily Housing Supply

Regarding supply, one of the secrets of the success of the multifamily sector during the 1990s was that production never rose above its long-run sustainable level. The discipline of developers, investors, and their lenders that brought that result needs to be continued if the apartment market is to maintain stability.

Multifamily housing may benefit in the future from more favorable public attitudes and local land use regulation. Higher density housing is a potentially powerful tool for preserving open space, reducing sprawl, and promoting transportation alternatives to the automobile. The recently heightened attention to these issues may increase the acceptance of multifamily rental construction to both potential customers and their prospective neighbors.

Provision of affordable housing will continue to challenge suppliers of multifamily rental housing and policy makers at all levels of governments. Low incomes combined with high housing costs define a difficult situation for millions of renter households. Housing cost reductions are constrained by high land prices and construction costs in many markets. Government action—through land use regulation, building codes, and occupancy standards—are major contributors to those high costs, as is widely recognized by market participants, including the leaders of the GSEs.^[252] Reflecting the preferences of the electorate, these regulated constraints are unlikely to change until voter attitudes change.

b. The Future Role of the GSEs

Regarding the mortgage financing of multifamily rental apartments, it is hard to anticipate events that might disrupt the flow or alter the sources of mortgage credit to apartments. In the past, certain events have triggered such changes—notably the savings and loan debacle of the 1980s and Freddie Mac's withdrawal from the market following large losses in the early 1990s—but these are, by definition, surprises. The current structure and performance of the multifamily mortgage market provide some comfort that the risks are slight. The lender base is not overly dependent on any one institution or lender type for either loan originations or funding. Lending discipline appears to have been maintained, given the low mortgage delinquency rates even during the weak economy of the past two years. The near term outlook of most market participants is for ample supply of mortgage financing at historically low interest rates.^[253] Yet complacency would be a mistake.

Responding to both market incentives and their public charters, Fannie Mae and Freddie Mac can be expected to build on their recent records of increased multifamily lending and continue to be leaders in financing volumes, in program innovations, and in standards setting. Certainly there is room for expansion of the GSEs' share of the multifamily mortgage market, which, as mentioned earlier, is by the measure of dollar volume outstanding currently only about half the market share enjoyed by the GSEs in single-family lending. And from the perspective of units financed, the statistics from Table A.5 combined with data from the 2001 American Housing Survey indicate that, while the GSEs financed 7.2 percent of all the nation's year-round housing units that year, the percentage of multifamily rental units (that is renter-occupied units and vacant rental units in structures with at least five units) was only 5.7 percent.

The sharp gains since 2000 in small property lending by Fannie Mae and Freddie Mac demonstrate that it is feasible for this important segment of the affordable housing market to be served by the GSEs. Building on the expertise and market contacts gained in the past three years, the GSEs should be able to make even greater in-roads in small property lending, although the challenges noted earlier will continue.

The GSEs' size and market position between loan originators and mortgage investors makes them the logical institutions to identify and promote needed innovations and to establish standards that will improve market efficiency. As their presence in the multifamily market continues to grow, the GSEs will have both the knowledge and the “clout” to push simultaneously for market standardization and for programmatic flexibility to meet special needs and circumstances, with the ultimate goal of increasing the availability and reducing the cost of financing for affordable and other multifamily rental properties.

E. Factor 3: Performance and Effort of the GSEs Toward Achieving the Low- and Moderate-Income Housing Goal in Previous Years

This section first discusses each GSE's performance under the Low- and Moderate-Income Housing Goal over the 1996-2003 period.^[254] The data presented are “official results—i.e., they are based on HUD's analysis of the loan-level data submitted to the Department by the GSEs and the counting provisions contained in HUD's regulations in 24 CFR part 81, subpart B. As explained below, in some cases these “official results” differ from goal performance reported by the GSEs in the Annual Housing Activities Reports (AHARs) that they submit to the Department.

The main finding of this section concerning the overall housing goals is that both Fannie Mae and Freddie Mac surpassed the Department's Low- and Moderate-Income Housing Goals for each of the eight years during this period. Specifically:

• The goal was set at 40 percent for 1996; Fannie Mae's performance was 45.6 percent and Freddie Mac's performance was 41.1 percent.
• The goal was set at 42 percent for 1997-2000. Fannie Mae's performance was 45.7 percent in 1997, 44.1 percent in 1998, 45.9 percent in 1999, and 49.5 percent in 2000; and Freddie Mac's performance was 42.6 percent in 1997, 42.9 percent in 1998, 46.1 percent in 1999, and 49.9 percent in 2000.
• In the October 2000 rule, the low- and moderate-income goal was set at 50 percent for 2001-03. As of January 1, 2001, several changes in counting provisions took effect for the low- and moderate-income goal, as follows: “bonus points” (double credit) for purchases of goal-qualifying mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on 2-4 unit owner-occupied properties; a “temporary adjustment factor” (1.20 units credit, subsequently increased by Congress to 1.35 units credit) for Freddie Mac's purchases of goal-qualifying mortgages on large (more than 50 units) multifamily properties; changes in the treatment of missing data; a procedure for the use of imputed or proxy rents for determining goal credit for multifamily mortgages; and eligibility of purchases of certain qualifying government-backed loans to receive goal credit. These changes are explained below. Fannie Mae's low-mod goal performance was 51.5 percent in 2001, 51.8 percent in 2002, and 52.3 percent in 2003; Freddie Mac's performance was 53.2 percent in 2001, 50.5 percent in 2002, and 51.2 percent in 2003, thus both GSEs surpassed this higher goal in all three years. This section discusses the October 2000 counting rule changes in detail below, and provides data on what goal performance would have been in 2001-03 without these changes.^[255]

After the discussion of the overall housing goals in Sections E.1 to E.5, Sections E.6 to E.12 examine the role of the GSEs in funding home purchase loans for lower-income borrowers and for first-time homebuyers. A summary of the main findings from that analysis is given in Section E.6. Section E.13 then summarizes some recent studies on the GSEs' market role and section E.14 discusses the GSEs' role in the financing of single-family rental properties.

1. Performance on the Low- and Moderate-Income Housing Goal in 1996-2003

HUD's December 1995 rule specified that in 1996 at least 40 percent of the number of units financed by each of the GSEs that were eligible to count toward the Low- and Moderate-Income Goal should qualify as low-or moderate-income, and at least 42 percent of such units should qualify in 1997-2000. HUD's October 2000 rule made various changes in the goal counting rules, as discussed below, and increased the Low- and Moderate-Income Goal to 50 percent for 2001-03.

Table A.8 shows low-mod goal performance over the 1996-2003 period, based on HUD's analysis. The table shows that Fannie Mae surpassed the goals by 5.6 percentage points and 3.7 percentage points in 1996 and 1997, respectively, while Freddie Mac surpassed the goals by narrower margins, 1.1 and 0.6 percentage points. During the heavy refinance year of 1998, Fannie Mae's performance fell by 1.6 percentage points, while Freddie Mac's performance rose slightly, by 0.3 percentage point. Freddie Mac showed a gain in performance to 46.1 percent in 1999, exceeding its previous high by 3.2 percentage points. Fannie Mae's performance in 1999 was 45.9 percent, which, for the first time, slightly lagged Freddie Mac's performance in that year.

Both GSEs exhibited sharp gains in goal performance in 2000—Fannie Mae's performance increased by 3.6 percentage points, to a record level of 49.5 percent, while Freddie Mac's performance increased even more, by 3.8 percent percentage points, which also led to a record level of 49.9 percent. Fannie Mae's performance was 51.5 percent in 2001, 51.8 percent in 2002, and 52.3 percent in 2003; Freddie Mac's performance was 53.2 percent in 2001, 50.5 percent in 2002, and 51.2 percent in 2003. However, as discussed below, using consistent accounting rules for 2000-03, each GSE's performance in 2001-03 was below its performance in 2000.

The official figures for low-mod goal performance presented above differ from the corresponding figures presented by Fannie Mae and Freddie Mac in their Annual Housing Activity Reports to HUD by 0.2-0.3 percentage point in both 1996 and 1997, reflecting minor differences in the application of counting rules. These differences also persisted for Freddie Mac for 1998-2000, but the goal percentages shown above for Fannie Mae for these three years are the same as the results reported by Fannie Mae to the Department. Fannie Mae reported its performance in 2001 as 51.6 percent and Freddie Mac reported its performance as 53.6 percent—both were slightly above the corresponding official figures of 51.5 percent and 53.4 percent, respectively. For 2002, Fannie Mae's reported performance was the same as reported by HUD (51.8 percent), while Freddie Mac's reported performance was 51.3 percent, slightly above HUD's official figure of 50.5 percent. For 2003, Fannie Mae's reported performance on this goal was 51.8 percent, somewhat below HUD's official figure of 52.3 percent, while Freddie Mac's reported performance (51.1 percent) was essentially the same as HUD's official figure of 51.2 percent.

Fannie Mae's performance on the Low- and Moderate-Income Goal was in the range between 44 percent and 46 percent between 1996 and 1999, but jumped sharply in just one year, from 45.9 percent in 1999 to 49.5 percent in 2000. Freddie Mac's performance was in the range between 41 percent and 43 percent between 1996 and 1998, and then rose to 46.1 percent in 1999 and 49.9 percent in 2000. As discussed above, official performance rose for both GSEs in 2001-02, but this was due to one-time changes in the counting rules—abstracting from counting rule changes, performance fell for both GSEs.

Fannie Mae's performance on the Low- and Moderate-Income Goal surpassed Freddie Mac's in every year through 1998. This pattern was reversed in 1999, as Freddie Mac surpassed Fannie Mae in goal performance for the first time, though by only 0.2 percentage point. This improved relative performance of Freddie Mac was due to its increased purchases of multifamily loans, as it re-entered that market, and to increases in the goal-qualifying shares of its single-family mortgage purchases. Freddie Mac's performance also slightly exceeded Fannie Mae's performance in 2000, 49.9 percent to 49.5 percent. Freddie Mac's official performance also exceeded Fannie Mae's official performance in 2001, but this reflected a difference in the counting rules applicable to the two GSEs that was enacted by Congress; if the same counting rules were applied to both GSEs (that is, Freddie Mac did not receive the 1.35 Temporary Adjustment Factor), Fannie Mae's performance would have exceeded Freddie Mac's performance, by 51.5 percent to 50.5 percent.

In 2002, Freddie Mac's performance on the low mod-goal (50.5 percent) fell short of Fannie Mae's performance (51.8 percent), even though Freddie Mac had the advantage of the Temporary Adjustment Factor. The gap would have been wider without this factor, and in fact Freddie Mac's performance would have been short of the goal, at 49.2 percent. This same pattern prevailed in 2003, when Freddie Mac's performance on this goal (51.2 percent) was significantly below Fannie Mae's performance (52.3 percent), even though Fannie Mae did not have the advantage of the Temporary Adjustment Factor. The gap in performance between the GSEs would have been much wider without this factor, as Freddie Mac's performance would again have fallen short of the goal, at 48.4 percent.

2. Changes in the Goal Counting Rules for 2001-03

A number of changes in the counting rules underlying the calculation of low- and moderate-income goal performance took effect beginning in 2001, as follows:

• Bonus points for multifamily and single-family rental properties. During the 2001-03 period the Department awarded “bonus points” (double credit in the numerator) for goal-qualifying units in small (5-50 unit) multifamily properties and, above a threshold, 2-4 unit owner-occupied properties whose loans were purchased by the GSEs. By letters dated December 24, 2003, the Department notified the GSEs that these bonus points would not be in effect after December 31, 2003.
• Freddie Mac's Temporary Adjustment Factor. As part of the Consolidated Appropriations Act of 2000, Congress required the Department to award 1.35 units of credit for each unit financed in “large” multifamily properties (i.e., those with 51 or more units) in the numerator in calculating performance on the housing goals for Freddie Mac for 2001-03.^[256] This “temporary adjustment factor” (TAF) did not apply to goal performance for Fannie Mae during this period. By letters dated December 24, 2003, the Department notified Freddie Mac that this factor would not be in effect after December 31, 2003.

• Missing data for single-family properties. In the past, if a GSE lacked data on rent for rental units or on borrower income for owner-occupied units in single-family properties whose mortgages it purchased, such units were included in the denominator, but not in the numerator, in calculating goal performance. Since some of these units likely would have qualified for one or more of the housing goals, this rule lowered goal performance. Under the new counting rules for the low- and moderate-income goal and the special affordable goal that took effect in 2001, the GSEs are allowed to exclude loans with missing borrower income from the denominator if the property is located in a below-median income census tract. This exclusion is subject to a ceiling of 1 percent of total owner-occupied units financed. The enterprises are also allowed to exclude single-family rental units with missing rental information from the denominator in calculating performance for these two goals; there is no ceiling or restriction to properties located in below-median income census tracts for this exclusion of single-family rental units. No single-family loans can be excluded from the denominator in calculating performance on the underserved areas goal—that is, if a GSE does not have sufficient information to determine whether or not a property is located in an underserved area, all units in such a property are included in the denominator, but not in the numerator, in calculating performance on this goal.
• Missing data and proxy rents for multifamily properties. In the past, if a GSE lacked data on rent for rental units in multifamily properties whose mortgages it purchased, such units were included in the denominator, but not in the numerator, in calculating goal performance. Since some of these units likely would have qualified for one or more of the housing goals, this rule lowered goal performance. Under the new counting rules that took effect in 2001, if rent is missing for multifamily units, a GSE may estimate “proxy rents,” and, up to a ceiling of 5 percent of total multifamily units financed, may apply these proxy rents in determining whether such units qualify for the low- and moderate income goal and special affordable goal. If such proxy rents cannot be estimated, these multifamily units are excluded from the denominator in calculating performance under these goals. No multifamily loans can be excluded from the denominator in calculating performance on the underserved areas goal—that is, if a GSE does not have sufficient information to determine whether or not a property is located in an underserved area, all units in such a property are included in the denominator, but not in the numerator, in calculating performance on this goal.
• Purchases of certain government-backed loans. Prior to 2001, purchases of government-backed loans were not taken into account in determining performance on the GSEs' low- and moderate-income and underserved area housing goals. That is, all such loans were excluded from both the numerator and the denominator in calculating goal performance on these two goals, and in accordance with Section 1333(b)(1)(A) of the Federal Housing Enterprises Financial Safety and Soundness Act of 1992, purchases of only certain government-backed loans were included in determining performance on the GSEs' special affordable goals. In October 2000 the Department took steps to encourage the enterprises to play more of a role in the secondary market for several types of government-backed loans where it appeared that greater GSE involvement could increase the liquidity of such mortgages. Home equity conversion mortgages (HECMs) were developed in the late-1980s by the Federal Housing Administration (FHA); these mortgages allow senior citizens to draw on the equity in their homes to obtain monthly payments to supplement their incomes. Thus purchases of FHA-insured HECMs now count toward the low- and moderate-income housing goals if the mortgagor's income is less than median income for the area. Similarly, purchases of mortgages on properties on tribal lands insured under FHA's Section 248 program or HUD's Section 184 program may qualify for the GSEs' housing goals. And purchases of mortgages under the Rural Housing Service's Single Family Housing Guaranteed Loan Program may also count toward all of the housing goals.^[257]

3. Effects of Changes in the Counting Rules on Goal Performance in 2001-03

Because of the changes in the low- and moderate-income goal counting rules that took effect in 2001, direct comparisons between official goal performance in 2000 and 2001-03 are somewhat of an “apples-to-oranges comparison.” For this reason, the Department has calculated what performance would have been in 2000 under the 2001-03 rules; this may be compared with official performance in 2001-03—an “apples-to-apples comparison.” HUD has also calculated what performance would have been in 2001-03 under the 1996-2000 rules; this may be compared with official performance in 2000—an “oranges-to-oranges comparison.” These comparisons are presented in Table A.9.

Specifically, Table A.9 shows performance under the low- and moderate-income goal in three ways. Baseline A represents performance under the counting rules in effect in 1996-2000. Baseline B incorporates the technical changes in counting rules—changes in the treatment of missing data (including use of proxy rents), and eligibility for the goals of certain government-backed loans. Baseline C incorporates in addition to the technical changes the bonus points and, for Freddie Mac, the temporary adjustment factor. Baseline B corresponds to the counting approach proposed in this rule to take effect in 2005. Boldface figures under Baseline A for 1999-2000 and under Baseline C for 2001-03 indicate official goal performance, based on the counting rules in effect in those years—e.g., for Fannie Mae, 45.9 percent in 1999, 49.5 percent in 2000, 51.5 percent in 2001, 51.8 percent in 2002, and 52.3 percent in 2003.

• Performance on the Low- and Moderate-Income Goal under 1996-2000 Counting Rules Plus Technical Changes. If the “Baseline B” counting approach had been in effect in 2000-03 and the GSEs' had purchased the same mortgages that they actually did purchase in those years, both Fannie Mae and Freddie Mac would have surpassed the low- and moderate-income goal in 2000 and fallen short in 2001, 2002, and 2003. Specifically, Fannie Mae's performance would have been 51.3 percent in 2000, 49.2 percent in 2001, 49.0 percent in 2002, and 48.7 percent in 2003. Freddie Mac's performance would have been 50.6 percent in 2000, 47.7 percent in 2001, 46.1 percent in 2002, and 45.0 percent in 2003.
• Performance on the Low- and Moderate-Income Goal under 2001-2003 Counting Rules. If the 2001-03 counting rules had also been in effect in 2000 and the GSEs' had purchased the same mortgages that they actually did purchase in those years (i.e., abstracting from any behavioral effects of “bonus points,” for example), both GSEs would have substantially surpassed the low- and moderate-income goal in all four years, but both GSEs' performance figures would have deteriorated somewhat from 2000 to 2001, and, for Freddie Mac, from 2001 to 2002 and 2003. Specifically, Fannie Mae's “Baseline C” performance would have been 52.5 percent in 2000, 51.5 percent in 2001, 51.8 percent in 2002, and 52.3 percent in 2003. Freddie Mac's performance would have been 55.1 percent in 2000, surpassing its official performance level of 53.2 percent in 2001, 50.5 percent in 2002, and 51.2 percent in 2003. Measured on this consistent basis, then, Fannie Mae's performance fell by 1.0 percentage point in 2001, and Freddie Mac's by 1.9 percentage points in 2001 and an additional 2.0 percentage points in 2002-03. These reductions were primarily due to 2001-03 being years of heavy refinance activity.

Details of Effects of Changes in Counting Rules on Goal Performance in 2001-03. As discussed above, counting rule changes that took effect in 2001 had significant positive impacts on the performance of both GSEs on the low- and moderate-income goal in that year—3.8 percentage points for Fannie Mae, and 6.0 percentage points for Freddie Mac. This section breaks down the effects of these changes on goal performance for both GSEs; results are shown in Table A.9.

• Freddie Mac. The largest impact of the counting rule changes on Freddie Mac's goal performance was due to the application of the temporary adjustment factor for purchases of mortgages on large multifamily properties, as enacted by Congress; this added 2.7 percentage points to goal performance in 2001, as shown in Table A.9. Bonus points for purchases of mortgages on small multifamily properties added 1.5 percentage points to performance, and bonus points for purchase of mortgages on owner-occupied 2-4 unit rental properties added 1.4 percentage points to performance. The remaining impact (0.5 percentage point) was due to technical changes in counting rules—primarily, the exclusion of single-family units with missing information from the denominator in calculating goal performance. Credit for purchases of qualifying government-backed loans played a minor role in determining Freddie Mac's goal performance. These same patterns also appeared in 2002. But in 2003, bonus points for purchases of low-mod mortgages on single-family rental properties had a larger impact on Freddie Mae's low-mod goal performance than Freddie Mac's temporary adjustment factor.
• Fannie Mae. The temporary adjustment factor applies to Freddie Mac's goal performance, but not to Fannie Mae's performance, thus counting rule changes had less impact on its performance than on Freddie Mac's performance in 2001. The largest impact of the counting rule changes on Fannie Mae's goal performance was due to the application of bonus points for purchases of mortgages on owner-occupied 2-4 unit rental properties, which added 1.6 percentage points to performance, and for purchases of mortgages on small multifamily properties, which added 0.7 percentage point to performance. The remaining impact (1.3 percentage points) was due to technical changes—primarily, the exclusion of single-family units with missing information from the denominator in calculating goal performance.^[258] Credit for purchases of qualifying government-backed loans and the use of proxy rent for multifamily properties played a minor role in determining Fannie Mae's goal performance. These same patterns also appeared in 2002 for Fannie Mae, but for 2003 bonus points for purchases of low-mod mortgages on small multifamily properties had more impact on performance than bonus points for single-family rental properties.

4. Bonus Points for the Low- and Moderate-Income Goal

As discussed above, the Department established “bonus points” to encourage the GSEs to step up their activity in 2001-03 in two segments of the mortgage market—the small (5-50 unit) multifamily mortgage market, and the market for mortgages on 2-4 unit properties where 1 unit is owner-occupied and 1-3 units are occupied by renters. Bonus points did not apply to purchases of mortgages for owner-occupied 1-unit properties, for investor-owned 1-4 unit properties, and for large (more than 50 units) multifamily properties, although as also discussed above, a “temporary adjustment factor” applied to Freddie Mac's purchases of qualifying mortgages on large multifamily properties.

Bonus points for small multifamily properties. Each unit financed in a small multifamily property that qualified for any of the housing goals was counted as two units in the numerator (and one unit in the denominator) in calculating goal performance for that goal. For example, if a GSE financed a mortgage on a 40-unit property in which 10 of the units qualified for the low- and moderate-income goal, 20 units would be entered in the numerator and 40 units in the denominator for this property in calculating goal performance.

Small multifamily bonus points thus encouraged the GSEs to play a larger role in this market, and also to purchase mortgages on such properties in which large shares of the units qualified for the housing goals. Some evidence may be gleaned from the data provided to HUD by the GSEs for 2001-03.

Fannie Mae financed 37,403 units in small multifamily properties in 2001 that were eligible for the low- and moderate-income goal, 58,277 such units in 2002, and 214,619 such units in 2003, as compared with only 7,196 such units financed in 2000. Small multifamily properties also accounted for a greater share of Fannie Mae's multifamily business in 2001-03—7.4 percent of total multifamily units financed in 2001, 13.2 percent in 2002, and 28.6 percent in 2003, up from 2.5 percent in 2000. However, HUD's 2000 rule reported information from the 1991 Residential Finance Survey that small multifamily properties accounted for 37 percent of all multifamily units, thus Fannie Mae was still less active in this market than in the market for large multifamily properties.^[259]

Within the small multifamily market, there was no evidence that Fannie Mae targeted affordable properties to a greater extent in 2001-03 than in 2000. That is, 87 percent of Fannie Mae's small multifamily units qualified for the low- and moderate-income goal in 2000; this fell to 75 percent in 2001, rose to 89 percent in 2002, and then declined to 82 percent in 2003.

Freddie Mac financed 50,299 units in small multifamily properties in 2001 that were eligible for the low- and moderate-income goal, 22,255 such units in 2002, and 177,561 such units in 2003, as compared with only such units financed in 2000. Small multifamily properties also accounted for a significantly greater share of Freddie Mac's multifamily business in 2001-2003—16.1 percent of total multifamily units financed in 2001, 7.5 percent in 2002, and 25.4 percent in 2003, up from 1.8 percent in 2000.

Within the small multifamily market, there was some evidence that Freddie Mac targeted affordable properties to a greater extent in 2001-2002 than in 2000. That is, 87 percent of Freddie Mac's small multifamily units qualified for the low- and moderate-income goal in 2000; this rose to 96 percent in 2001, but declined back to 87 percent in 2002 and 2003.

In summary, then, there is strong evidence that bonus points for small multifamily properties had an impact on Fannie Mae's role in this market in 2001-2003 and an even larger impact on Freddie Mac's role in this market. In addition, Fannie Mae has announced a program to increase its role in this market further in future years.^[260]

Bonus points for single-family rental properties. Above a threshold, each unit financed in a 2-4 unit property with at least one owner-occupied unit (referred to as “OO24s” below) that qualified for any of the housing goals was counted as two units in the numerator (and one unit in the denominator) in calculating goal performance for that goal in 2001-2003. The threshold was equal to 60 percent of the average number of such qualifying units over the previous five years. For example, Fannie Mae financed an average of 50,030 low- and moderate-income units in these types of properties between 1996 and 2000, and 101,423 such units in 2001. Thus Fannie Mae received 71,405 bonus points in this area in 2001—that is, 101,423 minus 60 percent of 50,030. So 172,828 units were entered in the numerator for these properties in calculating low- and moderate-income goal performance.

Single-family rental bonus points thus encouraged the GSEs to play a larger role in this market, and also to purchase mortgages on such properties in which large shares of the units qualified for the housing goals. As for small multifamily bonus points, again some evidence may be gleaned from the data provided to HUD by the GSEs for 2001-03.

Fannie Mae financed 175,103 units in OO24s in 2001 that were eligible for the low- and moderate-income goal, 229,632 such units in 2002, and 355,994 such units in 2003, well above the 77,930 units financed in 2000. However, with the refinance boom, Fannie Mae's total single-family business increased at approximately the same rate as its OO24 business in 2001-03, thus the share of its business accounted for by OO24s was the same in 2001-03 as in 2000—4 percent.

Within the OO24 market, there was no evidence that Fannie Mae targeted affordable properties to a greater extent in 2001-03 than in 2000. That is, approximately 55-60 percent of Fannie Mae's OO24 units qualified for the low- and moderate-income goal in each of these three years.

Freddie Mac financed 96,050 units in OO24s in 2001 that were eligible for the low- and moderate-income goal, 146,222 such units in 2002, and 154,535 such units in 2003, as compared with the 49,993 units financed in 2000. However, Freddie Mac's total single-family business increased at approximately the same rate as its OO24 business in 2001-02, thus the share of its business accounted for by OO24s was the same in 2002 as in 2000—4 percent. And its total single-family business increased at a faster rate than its OO24 business in 2003, thus the share of its business accounted for by OO24s declined to 3 percent last year.

As for Fannie Mae, within the OO24 market there was no evidence that Freddie Mac targeted affordable properties to a greater extent in 2001-03 than in 2000. That is, 68-69 percent of Fannie Mae's OO24 units qualified for the low- and moderate-income goal in each year from 2000 through 2002; this decreased to 64 percent in 2003.

5. Effects of 2000 Census on Scoring of Loans Toward the Low- and Moderate-Income Housing Goal

Background. Scoring of housing units under the Low- and Moderate-Income Housing Goal is based on data for mortgagors' incomes for owner-occupied units, rents for rental units, and area median incomes, as follows:

For single-family owner-occupied units:

The mortgagors' income at the time of mortgage origination.

The median income of an area specified as follows: (i) For properties located in Metropolitan Statistical Areas (MSAs), the area is the MSA; and (ii) for properties located outside of MSAs, the area is the county or the non-metropolitan portion of the State in which the property is located, whichever has the larger median income, as of the year of mortgage origination (which may be for the current year or a prior year).

For rental units in single-family properties with rent data are available (assuming no income data available for actual or prospective tenants):

The unit rent (or average rent for units of the same type) at the time of mortgage origination.

The area median income as specified for single-family owner-occupied units.

For rental units in multifamily properties where rent data are available:

The unit rent (or the average rent for units of the same type) at the time of mortgage acquisition by the GSE.

The area median income as specified for single-family owner-occupied units, but as of the year the GSE acquired the mortgage.

For rental units in multifamily properties where rent data are not available, the GSE may apply HUD-estimated rents which are based on the following area data:

The median rent in the census tract where the property is located, as of the most recent decennial census.

The area median income as specified for single-family owner-occupied units, but as of the most recent decennial census.

Thus, scoring loans under the Low- and Moderate-Income Goal requires a data series showing annual median incomes for MSAs, non-metropolitan counties, and the non-metropolitan portions of states; and decennial census data on median incomes for census tracts.^[261]

For scoring loans purchased by the GSEs year-by-year from 1993 through 2002, area median income estimates produced by HUD's Economic and Market Analysis Division were used. An example will illustrate the estimation procedure. To generate the area median income estimates that were used to score GSE loans in 2002, data from the 1990 census on 1989 area median incomes were adjusted to 2002 using Bureau of Labor Statistics survey data on rates of cha nge in average incomes for MSAs and counties between 1989 and 1999, data from the Census Bureau's Current Population Survey on rates of change in median family incomes for the nine Census Divisions between 1989 and 2000, and an assumed 4.0 percent per year inflation factor between 2000 and 2002.^{[262]  [263]}

2005 Procedure. Relative to the above procedure, scoring of loans purchased by the GSEs in and after 2005 will be affected by two factors. First, the Economic and Market Analysis Division has begun to incorporate data from the 2000 census into its procedure for estimating annual area median incomes and American Community Survey data are becoming available at increasingly finer levels of geographical detail for use in annual updating. Beginning in 2005 Bureau of Labor Statistics data on rates of inflation in average wages will not be used. For 2005, the procedure for estimating area median incomes will be to adjust 2000 census data on 1999 area median incomes to 2003 using data from the Census Bureau's American Community Survey (ACS) on rates of change in average incomes for States between 1999 and 2003, with a further adjustment to 2005 based on an appropriate annual inflation factor.^[264] Increasingly more detailed ACS data will be available and will be used in subsequent years, as ACS estimates for metropolitan and micropolitan areas and counties become available.

The second factor is the Office of Management and Budget's June, 2003, re-specification of MSA boundaries based on analysis of 2000 census data.^[265]

Analysis. For purposes of specifying the level of the Low- and Moderate-Income Housing Goal, HUD developed a methodology for scoring loans purchased by the GSEs in past years through 2002 as though the re-benchmarking of area median income estimates to the 2000 census and the 2003 re-designation of MSAs had been in effect and HUD had been using an ACS-based estimation procedure at the time the estimates for these years were prepared. For this purpose, HUD created a series of annual estimates of median incomes for MSAs, non-metropolitan counties, and the non-metropolitan portions of states. For 2000, the estimates were 1999 census medians trended by three-fourths of the 4.0 percent annual trending factor (to adjust the figures from mid-1999 to April 1, 2000). For 2001, the estimates were based on one-and-three-fourths years of trending, since no data would have been available to use for updating. The 2002 estimates would have used one year of data and 1.75 years of trending. The 2003 estimates would have used two years of data plus 1.75 years of trending. Area median incomes from 1989 to 1999 were estimated based on trend-lines between 1989 and 1999 census data. The 2003 OMB MSA designations were applied.

The resulting estimates of area median incomes for MSAs, non-metropolitan counties, and the non-metropolitan parts of States, were used to re-score loans purchased by the GSEs between 1999 and 2002, and were used further in estimating the share of loans originated in metropolitan areas that would be eligible to score toward the Low- and Moderate-Income Housing Goal, from HMDA data. The results of the retrospective GSE analysis are provided in Table A.10. The results of the GSE-HMDA comparative analysis are presented in the next section.

Table A.10 shows three sets of estimates for each GSE, based respectively on the counting rules in place in 2001-2002 (but disregarding the bonus points and Temporary Adjustment Factor), on the addition of 2000 census re-benchmarking, and finally on the addition of both 2000 census re-benchmarking and 2003 MSA specification. Re-benchmarking occurred to adjust for some differences between Census 1990 and Census 2000 tracts.

6. GSEs Compared With the Primary Conventional Conforming Mortgage Market

This section and the next five sections (Sections E.7 to E.12) provide a detailed analysis of the extent to which the GSEs' loan purchases mirror or depart from the patterns found in the primary mortgage market. As in Section C.5, the GSEs' affordable lending performance is also compared with the performance of depository lenders such as commercial banks and thrift institutions. Dimensions of lending considered include the three “goals-qualifying” categories—special affordable borrowers, less-than-median income borrowers, and underserved areas. The special affordable category consists mainly of very-low-income borrowers, or borrowers who have an annual income less than 60 percent of area median income. Because this category is more targeted than the broadly-defined less-than-median-income (or low-mod) category, the discussion below will often focus on the special affordable category as well as the underserved areas category which adds a neighborhood dimension (low-income and high-minority census tracts) to the analysis. This section will also compare the performance of Fannie Mae and Freddie Mac in funding first-time homebuyers with that of primary lenders in the conventional conforming market.

The remainder of this introductory section E.6 provides a list of the major and specific findings which are presented in detail in the following Sections E.7 through 12. Sections 7 and 8 define the primary mortgage market and discuss some technical issues related to the use of the GSE and HMDA data. Sections 8 and 9 compare the GSEs' performance with market performance for home purchase and first-time homebuyer loans, while Section 10 does the same for total single family loans (that is, refinance loans and home purchase loans). Section 11 examines GSE purchases in individual metropolitan areas. Following these analyses, Section 12 examines the overall market share of the GSEs in important submarkets such as first-time homebuyers.

a. Main Findings on GSEs' Performance in the Single-family Market

There are six main findings from this analysis concerning the GSEs' purchases of single-family-owner mortgages:

1. While Freddie Mac has improved its affordable lending performance in recent years, it has consistently lagged the conventional conforming market in funding affordable home purchase loans for special affordable and low-moderate-income borrowers and underserved neighborhoods targeted by the housing goals.^[266] In 2003, its performance on the underserved areas goal was particularly low relative to both the performances of Fannie Mae and the market; in that year, underserved area loans accounted for only 24.0 percent of Freddie Mac's purchases compared with 26.8 percent of Fannie Mae's purchases and 27.6 percent of market originations.

2. In general, Fannie Mae's affordable lending performance has been better than Freddie Mac's. But like Freddie Mac, Fannie Mae's average performance during past periods (e.g., 1993-2003, 1996-2003, 1999-2003) has been below market levels. However, it is encouraging that Fannie Mae markedly improved its affordable lending performance relative to the market during 2001, 2002, and 2003, the first three years under the higher housing goal targets that HUD established in the GSE Final Rule dated October 2000.

Over this three-year period, Fannie Mae led the primary market in funding special affordable and low-mod loans but lagged the market in funding underserved areas loans. In 2003, Fannie Mae's increased performance placed it significantly above the special affordable market (a 17.1 percent share for Fannie Mae compared with a 15.9 percent share for the market) and the low-mod market (a 47.0 percent share for Fannie Mae compared with a 44.6 percent share for the market). However, Fannie Mae continued to lag the underserved areas market in 2003 (a 26.8 percent share for Fannie Mae compared with a 27.6 percent share for the market). In this case, which is referred to in the text as the “purchase year” approach, Fannie Mae's performance is based on comparing its purchases of all loans (both seasoned loans and newly-originated mortgages) during a particular year with loans originated in the market in that year. When Fannie Mae's performance is measured on an “origination year” basis (that is, allocating Fannie Mae's purchases in a particular year to the year that the purchased loan was originated), Fannie Mae also led the 2003 market in funding special affordable and low- and moderate-income loans, and lagged the market in funding underserved area loans.

3. Both Fannie Mae and Freddie lag the conventional conforming market in funding first-time homebuyers, and by a rather wide margin. Between 1999 and 2001, first-time homebuyers accounted for 27 percent of each GSE's purchases of home loans, compared with 38 percent for home loans originated in the conventional conforming market.

4. The GSEs have accounted for a significant share of the total (government as well as conventional) market for home purchase loans, but their market share for each of the affordable lending categories (e.g., low-income borrowers and census tracts) has been less than their share of the overall market.

5. The GSEs also account for a very small share of the market for important groups such as minority first-time homebuyers. Considering the total mortgage market (both government and conventional loans), it is estimated that the GSEs purchased only 14 percent of loans originated between 1999 and 2001 for African-American and Hispanic first-time homebuyers, or less than half of their share (42 percent) of all home purchase loans originated during that period. Considering the conventional conforming market and the same time period, it is estimated that the GSEs purchased only 31 percent of loans originated for African-American and Hispanic first-time homebuyers, or about one-half of their share (57 percent) of all home purchase loans in that market.

6. The GSEs' small share of the first-time homebuyer market could be due to the preponderance of high (over 20 percent) downpayment loans in their mortgage purchases.

b. Specific Findings on GSE Performance in the Single-family Market

This section presents 17 specific findings from the analyses reported in Sections E.7 through 12; they are grouped under the following five topic-headings:

(b.1) Longer-term Performance of the GSEs;

(b.2) Performance of the GSEs During Recent Years;

(b.3) The GSEs' Funding First-time Homebuyer Loans;

(b.4) Performance of the GSEs Based on Total (Home Purchase and Refinance) Loans;

(b.5) GSE Market Shares; and,

(b.6) Additional Findings.

(b.1) Longer-Term Performance of the GSEs

The longer-run performance of the GSEs is examined between 1993 and 2003 (which covers the period since the housing goals were put into effect) and between 1996 and 2003 (which covers the period under the current definitions of the housing goals). Of the two borrower-income goals, the analysis below will typically focus on the special affordable category, which is a more targeted category than the rather broadly defined low- and moderate-income category.

(1) Since the early nineties, the mortgage industry has introduced new affordable lending programs and has allowed greater flexibility in underwriting lower-income loans. There is evidence that these programs are paying off in terms of more mortgages for low-income and minority borrowers. As noted earlier, Fannie Mae and Freddie Mac have played an active role in this upsurge of affordable lending, as indicated by the high growth rates of their goals-qualifying business.

• Between 1993 and 2003, the GSEs' purchases of home loans in metropolitan areas increased by 60 percent.^[267] Their purchases of home loans for the three housing goals increased at much higher rates—287 percent for special affordable loans, 156 percent for low- and moderate-income loans, and 121 percent for loans in underserved census tracts.

(2) Both Fannie Mae and Freddie Mac have improved their purchases of affordable loans since the housing goals were put in place, as indicated by the increasing share of their business going to the three goals-qualifying categories. (See Table A.15 in Section E.9.)

• Between 1992 and 2003, the special affordable share of Fannie Mae's business almost tripled, rising from 6.3 percent to 17.1 percent, while the underserved areas share increased more modestly, from 18.3 percent to 26.8 percent. The figures for Freddie Mac are similar. The special affordable share of Freddie Mac's business rose from 6.5 percent to 15.6 percent, while the underserved areas share also increased but more modestly, from 18.6 percent to 24.0 percent.

(3) While both GSEs improved their performance, they have lagged the primary market in providing affordable loans to low-income borrowers and underserved neighborhoods. Freddie Mac's average performance, in particular, fell far short of market performance during the 1990s. Fannie Mae's average performance was better than Freddie Mac's during the 1993-2003 period as well as during the 1996-2003 period, which covers the period under HUD's currently-defined housing goals.

• Between 1993 and 2003, 12.2 percent of Freddie Mac's mortgage purchases were for special affordable borrowers, compared with 13.3 percent of Fannie Mae's purchases, 15.4 percent of loans originated by depositories, and 15.5 percent of loans originated in the conventional conforming market (without estimated B&C loans).^[268]

• Considering the underserved areas category for the 1996-2003 period, 22.0 percent of Freddie Mac's purchases financed properties in underserved neighborhoods, compared with 24.0 percent of Fannie Mae's purchases, 25.1 percent of loans originated by depositories, and 25.7 percent of loans originated in the conventional conforming market.

(b.2) Performance of the GSEs During Recent Years

The recent performance of the GSEs is examined for the four-year period between 1999 and 2003 and then for 2001, 2002 and 2003, which were the first three years that the GSEs operated under the higher goal targets established by HUD in the 2000 Rule. As explained below, the most interesting recent trend concerned Fannie Mae, which improved its performance during 2001-2003, at a time when the conventional conforming market was showing little change in affordable lending.

(4) During the recent 1999-to-2003 period, both Fannie Mae and Freddie Mac fell significantly below the market in funding affordable loans.

• Between 1999 and 2003, special affordable loans accounted for 15.1 percent of Fannie Mae's purchases, 14.7 percent of Freddie Mac's purchases, and 16.2 percent of loans originated in the market; thus, the “Fannie-Mae-to-market” ratio was 0.93 and the “Freddie-Mac-to-market” ratio was also 0.91.
• During the same period, underserved area loans accounted for 24.7 percent of Fannie Mae's purchases, 23.1 percent of Freddie Mac's purchases, and 26.2 percent of loans originated in the market; the “Fannie-Mae-to-market” ratio was 0.94 and the “Freddie-Mac-to-market” ratio was only 0.88.^[269]

(5) After experiencing declines from 1997 to 1999, Fannie Mae's affordable lending performance improved between 2000 and 2003.

• After declining from 23.0 percent in 1997 to 20.4 percent in 1999, the share of Fannie Mae's purchases financing properties in underserved areas jumped by three percentage points to 23.4 percent in 2000, and then increased further to 26.7 percent in 2002 and 26.8 percent in 2003.
• After declining from 13.2 percent in 1998 to 12.5 percent in 1999, the share of Fannie Mae's purchases going to special affordable loans rebounded to 13.3 percent in 2000, 14.9 percent in 2001, 16.3 percent in 2002 and 17.1 percent in 2003.

(6) Freddie Mac's performance on the two borrower-income categories improved between 2000 and 2002, but not as much as Fannie Mae's performance. Freddie Mac's performance on the underserved areas category increased substantially between 2001 and 2002, but then declined between 2002 and 2003.

• The share of Freddie Mac's single-family-owner business going to special affordable home loans increased from 9.2 in 1997 to 14.7 percent in 2000 before falling to 14.4 percent in 2001 and rising to 15.8 percent in 2002 and 15.6 percent in 2003.
• Freddie Mac's purchases of underserved area loans increased at a modest rate from 19.7 percent in 1997 to 22.3 percent in 2001, before jumping to 25.8 percent in 2002 and then dropping to 24.0 percent in 2003.

(7) The long-standing pattern of Fannie Mae outperforming Freddie Mac was reversed during 1999 and 2000. But that pattern returned in 2001-2003 when Fannie Mae outperformed Freddie Mac on all three goals-qualifying categories.

• Fannie Mae and Freddie Mac had practically the same performance in 1992 on the three housing goal categories—special affordable loans accounted for 6.3 percent of Fannie Mae's purchases and 6.5 percent of Freddie Mac's purchases, for a “Fannie-Mae-to-Freddie-Mac” ratio of 0.97. The 1992 ratio for underserved areas was also 0.98 and that for low-mod, 1.02. Reflecting Fannie Mae's much better performance, the special affordable “Fannie-Mae-to-Freddie-Mac” ratio had risen to 1.27 by 1997, the underserved area ratio to 1.17, and the low-mod ratio to 1.10.
• However, in 1999, the “Fannie-Mae-to-Freddie-Mac” ratio for each of the three goals-qualifying categories fell to slightly below one. 1999 was the first year since 1992 that Freddie Mac had outperformed Fannie Mae in purchasing affordable home loans (although only by a very slight margin).
• In 2000, Freddie Mac's sharper increases in special affordable and low-mod purchases further reduced the “Fannie-Mae-to-Freddie-Mac” ratios for these two categories to 0.90 and 0.96, respectively. Fannie Mae's sharper increase in underserved areas funding resulted in the “Fannie-Mae-to-Freddie-Mac” ratio rising from slightly below one (0.98) in 1999 to 1.06 in 2000.
• Fannie Mae's stronger performance during 2001-2003 returned the “Fannie-Mae-to-Freddie-Mac” ratios for special affordable and low-mod loans to above one (1.10 and 1.09 respectively), indicating better performance for Fannie Mae in 2003. The “Fannie-Mae-to-Freddie-Mac” ratio for the underserved area category increased to 1.12 by 2003.

(8) While Freddie Mac has consistently improved its performance relative to the market, it continued to lag the market in funding affordable home loans during 2001-2003.

• Unlike Fannie Mae, Freddie Mac had not made any progress through 1997 in closing its gap with the market. The “Freddie Mac-to-market” ratio for the special affordable category actually declined from 0.63 in 1992 to 0.59 in 1996. But Freddie Mac's sharp improvement in special affordable purchases resulted in the “Freddie-Mac-to-market” ratio rising to 0.89 by 2000. After declining from 0.84 in 1992 to 0.79 in 1997, the “Freddie-Mac-to-market” ratio for underserved areas had risen only modestly to 0.84 by the year 2000. Thus, Freddie Mac's improvements prior to 2001 allowed it to close its gap with the market, mainly for the special affordable category where its gap had been the widest.
• During 2001, 2002 and 2003, Freddie Mac continued to close its gap with the market on the special affordable and low-mod categories. By 2003, these “Freddie-Mac-to-market” ratios were higher than in 2000, although they both continued to fall below one: at 0.98 for both categories. Between 2002 and 2003, Freddie Mac's market ratio for underserved areas fell from 0.98 to 0.87 (24.0 percent for Freddie Mac and 27.6 percent for Fannie Mae). Thus, during 2003, Freddie Mac lagged the market on all three goals-qualifying categories.

(9) Through 1998, Fannie Mae had significantly improved its performance relative to the market. But as a result of shifts in its purchases of affordable loans, Fannie Mae lagged the market even further in 2000 than it had in some earlier years. During 2001-2003, Fannie Mae again improved its performance relative to the market and, in 2003, Fannie Mae led the special affordable and low-mod markets but lagged the underserved areas market.

• The above analysis and the data reported under this specific finding (9) are based on the “purchase year” approach for measuring GSE activity. The purchase year approach assigns GSE purchases of both prior-year (seasoned) and newly-originated mortgages to the calendar year in which they were purchased by the GSE; this results in an inconsistency with the HMDA-reported market data, which covers only newly-originated mortgages. Sections E.9 and E.10 also report the results of an alternative “origination year” approach that assigns GSE purchases to their year of origination, placing them on a more consistent basis with the HMDA-reported market data. The findings from the origination-year approach are discussed under specific finding (10).
• Fannie Mae's decline in performance during 1999 resulted in the “Fannie-Mae-to-market” ratio falling sharply to 0.74 for special affordable, to 0.81 for underserved areas and to 0.89 for low-mod. In 2000, Fannie Mae improved and reversed its declining trend, as the “Fannie-Mae-to-market” ratios increased to 0.80 for special affordable purchases, to 0.89 for underserved area purchases, and to 0.93 for low-mod purchases.
• During 2001, Fannie Mae increased its special affordable percentage by 1.6 percentage points to 14.9 percent, which was only 0.7 percentage point below the market's performance of 15.6 percent. Fannie Mae increased its low-mod percentage from 40.8 percent to 42.9 percent at the same time that the low-mod share of the primary market was falling from 43.9 percent to 42.9 percent, placing Fannie Mae at the market's performance. Similarly, Fannie Mae increased its underserved area percentage from 23.4 percent in 2000 to 24.4 percent in 2001 while the underserved area share of the primary market was falling from 26.2 percent to 25.2 percent, placing Fannie Mae at 0.8 percentage point from the market's performance.
• During 2002, Fannie Mae continued to improve its performance on all three goals categories. Using the purchase-year approach to measure GSE performance, Fannie Mae slightly led the market on the special affordable category (16.3 percent for Fannie Mae and 16.1 percent for the market), led the market on the low-mod category (45.3 percent for Fannie Mae compared with 44.6 percent for the market), and led the market on the underserved area category (26.7 percent for Fannie Mae versus 26.3 percent for the market).
• During 2003, Fannie Mae's further improvement resulted in Fannie Mae leading the special affordable market (17.1 percent for Fannie Mae compared with 15.9 percent for the market) and continuing to lead the low-mod market (47.0 percent for Fannie Mae compared with 44.6 percent for the market). During 2003, Fannie Mae lagged behind the underserved areas market (26.8 percent for Fannie Mae compared with 27.6 percent for the market).

(10) This analysis addresses several technical issues involved in measuring GSE performance. The above analysis was based on the “purchase year” approach, as defined in (9) above. An alternative “origination year” approach has also been utilized, which assigns GSE purchases to their year of origination, placing them on a more consistent basis with the HMDA-reported market data. While the average results (e.g., 1999-2003 GSE performance) are similar under the two reporting approaches, GSE performance in any particular year can be affected, depending on the extent to which the GSE has purchased goals-qualifying seasoned loans in that particular year.

• The choice of which approach to follow particularly affected conclusions about Fannie Mae's performance relative to the market in 2002 (but not in 2001). Under the origination-year approach, Fannie Mae lagged the market on all three housing goal categories during 2001 and on the underserved area category during 2002. In 2002, Fannie Mae matched the market on the special affordable category and led the market on the low-mod category (45.5 percent for Fannie Mae compared with 44.6 percent of the market).
• During 2003, the origination year approach gives the similar results as the purchase year approach—Fannie Mae led the special affordable and low-mod markets and lagged the underserved areas market.

(b.3) The GSEs' Funding of First-time Homebuyer Loans

(11) The GSEs' funding of first-time homebuyers has been compared to that of primary lenders in the conventional conforming market. Both Fannie Mae and Freddie lag the market in funding first-time homebuyers, and by a rather wide margin.

• First-time homebuyers account for 27 percent of each GSE's purchases of home loans, compared with 38 percent for home loans originated in the conventional conforming market.

(b.4) Performance of the GSEs Based on Total (Home Purchase and Refinance) Loans

(12) The GSEs' acquisitions of total loans (including refinance loans as well as home purchase loans) were also examined. The main results indicate (a) Freddie Mac has improved its performance but has consistently lagged the market in funding loans (home purchase and refinance) that qualify for the housing goals; and (b) Fannie Mae has not only improved its performance but matched the low-mod market in 2001 and 2002 and led both the special affordable and low-mod markets in 2003. Fannie Mae, however, lagged the primary market in funding underserved areas during 2003. (See Table A.20 of Section E.10, which is based on the purchase-year approach for measuring GSE activity.)

• 1999-2003. During the recent 1999-to-2003 period, both Fannie Mae and Freddie Mac fell significantly below the market in funding affordable total (home purchase and refinance) loans. Between 1999 and 2003, special affordable loans accounted for 14.0 percent of Fannie Mae's purchases, 13.2 percent of Freddie Mac's purchases, and 15.6 percent of loans originated in the market; thus, the “Fannie-Mae-to-market” ratio was 0.93 and the “Freddie-Mac-to-market” ratio was 0.88 during this period.
• During the same period, underserved area loans accounted for 23.8 percent of Fannie Mae's purchases, 22.1 percent of Freddie Mac's purchases, and 25.2 percent of loans originated in the market; thus, the “Fannie-Mae-to-market” ratio was 0.94 and the “Freddie-Mac-to-market” ratio was 0.88.^[270]

• 2002 and 2003. During 2002, the first of these two years of heavy refinancing, Fannie Mae's performance was slightly above the market on the low-mod category and slightly below market performance on the special affordable and underserved areas categories; essentially, Fannie Mae matched the market on all three categories in 2002. In 2003, Fannie Mae led the market on the special affordable and low-mod categories and lagged the market on the underserved areas category. The 2003 “Fannie-Mae-to-market” ratios were 1.02 for special affordable loans, 1.03 for low-mod loans, and 0.97 for underserved area loans. In 2003, the “Freddie-Mac-to-market” ratios were much lower: 0.86 for special affordable loans, 0.90 for low-mod loans, and 0.82 for underserved area loans.

(b.5) GSE Market Shares

This analysis includes an expanded “market share” analysis that documents the GSEs” contribution to important segments of the home purchase and first-time homebuyer markets.

(13) The GSEs account for a significant share of the total (government as well as conventional conforming) market for home purchase loans. However, the GSEs' market share for each of the affordable lending categories is much less than their share of the overall market.

• The GSEs' purchases were estimated to be 46 percent of all home loans originated in metropolitan areas between 1999 and 2003 but only 30 percent of loans originated for African-American and Hispanic borrowers, 38 percent of loans originated for low-income borrowers, and 37 percent for properties in underserved areas. The GSEs' market share for the various affordable lending categories increased during 2001-2003, but the above-mentioned pattern remained.
• A study by staff from the Federal Reserve Board suggests that the GSEs have a much more limited role in the affordable lending market than is suggested by the data presented above.^[271] The Fed study, which combined market share, downpayment, and default data, concluded that the GSEs play a very minimal role in providing credit support and assuming credit risk for low-income and minority borrowers; for example, the study concluded that in 1995 the GSEs provided only four percent of the credit support going to African-Americans and Hispanic borrowers.

• Section V of this study begins to reconcile these different results by examining the role of the GSEs in the first-time homebuyer market and the downpayment characteristics of mortgages purchased by the GSEs.

(14) The market role of the GSEs appears to be particularly low in important market segments such as minority first-time homebuyers.

• Recent analysis has estimated that the GSEs' share of the market for first-time African-American and Hispanic homebuyers was only 14.3 percent between 1999 and 2001, or about one-third of their share (41.5 percent) of all home purchases during that period. This analysis includes the total market, including government and conventional loans.
• A similar market share analysis was conducted for the conventional conforming market. Between 1999 and 2001, the GSEs' purchases accounted for 56.6 percent of all home loans originated in the conventional conforming market of both metropolitan areas and non-metropolitan areas. Their purchases of first-time homebuyer loans, on the other hand, accounted for only 39.8 percent of all first-time homebuyer loans originated in that market.
• The GSEs have funded an even lower share of the minority first-time homebuyer market in the conventional conforming market. Between 1999 and 2001, the GSEs purchases of African-American and Hispanic first-time homebuyer loans represented 30.9 percent of the conventional conforming market for these loans. Thus, while the GSEs have accounted for 56.6 percent of all home loans in the conventional conforming market, they have accounted for only 30.9 percent of loans originated in that market for African-American and Hispanic first-time homebuyers.

(15) A noticeable pattern among the lower-income-borrower loans purchased by the GSEs is the predominance of loans with high downpayments. This pattern of purchasing mainly high downpayment loans is one factor explaining why the Fed study found such a small market role for the GSEs. It may be the explanation for the small role of Fannie Mae and Freddie Mac in the first-time homebuyer market. Further study of this issue is needed.

• During 2001 and 2002, approximately 50 percent of Fannie Mae's special affordable, low-mod, and underserved areas loans had downpayments of at least 20 percent, a percentage only slightly smaller that the corresponding percentage (53 percent) for all Fannie Mae's home loan purchases. Similar patterns of above-20-percent downpayments on goals-qualifying loans were evident in Freddie Mac's 2001, 2002, and 2003 purchases, as well as in prior years for both GSEs. During 2003, Fannie Mae's high downpayment share of their special affordable purchases dropped to 45 percent while the patterns for Fannie Mae's low-mod and underserved area purchases did not change, remaining about 50 percent.

(b.6) Additional Findings

This analysis examines two additional topics related to minority first-time homebuyers and the use of HMDA data for measuring the characteristics of loans originated in the conventional conforming market.

(16) The share of the GSEs' purchases for minority first-time homebuyers was much less than the share of newly-originated mortgages in the conventional conforming market for those homebuyers.

• Between 1999 and 2001, minority first-time homebuyers accounted for 6.6 percent of Fannie Mae's purchases of home loans, 5.8 percent of Freddie Mac's purchases, and 10.6 percent of home loans originated in the conventional conforming market. For this subgroup, Fannie Mae's performance is 62 percent of market performance, while Freddie Mac's performance is 55 percent of market performance.

(17) Some studies have concluded that HMDA data overstate the share of market loans going to low-income borrowers and underserved areas. This analysis does not support that conclusion.

• This compares the low-income and underserved areas characteristics of the GSEs' purchases of newly-originated (“current-year”) loans as reported both by the GSEs” own data and by HMDA data.^[272] For recent years, HMDA data on loans sold to the GSEs do not always have higher percentages of low-income and underserved areas loans than the GSEs' own data on their purchases of newly-originated mortgages. For example, from 1996-2003, both HMDA and Fannie Mae reported that special affordable loans accounted for about 13 percent of Fannie Mae's purchases of newly-originated loans. HMDA reported a 22.6 underserved areas percentage for Fannie Mae, which was rather similar to the underserved areas percentage (23.1 percent) reported by Fannie Mae itself. Given that similar patterns were observed for Freddie Mac's mortgage purchases, it appears that there is no upward bias in the HMDA-based market benchmarks used in this study.

7. Definition of Primary Market

Conventional Conforming Market. The market analysis section is based mainly on HMDA data for mortgages originated in the conventional conforming market of metropolitan areas during the years 1992 to 2003. Only conventional loans with a principal balance less than or equal to the conforming loan limit are included; the conforming loan limit was $322,700 in 2003—these are called “conventional conforming loans.” The GSEs' purchases of FHA-insured, VA-guaranteed, and Rural Housing Service loans are excluded from this analysis. The conventional conforming market is used as the benchmark against which to evaluate the GSEs because that is the market definition Congress requires that HUD consider when setting the affordable housing goals. However, as discussed in Section II, some have questioned whether lenders in the conventional market are doing an adequate job meeting the credit needs of minority borrowers, which suggests that this market provides a low benchmark.^[273]

Manufactured Housing Loans. Both GSEs have raised questions about whether loans on manufactured housing should be excluded when comparing the primary market with the GSEs. The GSEs purchase these loans, but they have not played a significant role in the manufactured housing loan market. As emphasized by HUD in its 2000 GSE Rule, manufactured housing is an important source of home financing for low-income families and for that reason, should be included in any analysis of affordable lending. However, for comparison purposes, data are also presented for the primary market defined without manufactured housing loans. Because this analysis focuses on metropolitan areas, it does not include the substantial number of manufactured housing loans originated in non-metropolitan areas.

Subprime Loans. Both GSEs also raised questions about whether subprime loans should be excluded when comparing the primary market with their performance. In its final 2000 GSE Rule, HUD argued that borrowers in the A-minus portion of the subprime market could benefit from the standardization and lower interest rates that typically accompany an active secondary market effort by the GSEs. A-minus loans are not nearly as risky as B&C loans and the GSEs have already started purchasing A-minus loans (and likely the lower “B” grade subprime loans as well). The GSEs themselves have mentioned that a large portion of borrowers in the subprime market could qualify as “A credit.” This analysis includes the A-minus portion of the subprime market, or conversely, excludes the B&C portion of that market.

Unfortunately, HMDA does not identify subprime loans, much less separate them into their A-minus and B&C components.^[274] Randall M. Scheessele at HUD has identified approximately 200 HMDA reporters that primarily originate subprime loans and account for about 60-70 percent of the subprime market.^[275] To adjust HMDA data for B&C loans, this analysis follows HUD's 2000 Rule which assumed that the B&C portion of the subprime market accounted for one-half of the loans originated by the subprime lenders included in Scheessele's list.^[276] As shown below, the effects of adjusting the various market percentages for B&C loans are minor mostly because the analysis in this section focuses on home purchase loans, which historically have accounted for less than one quarter of the mortgages originated by subprime lenders—the subprime market is mainly a refinance market.^[277]

Lender-Purchased Loans in HMDA. When analyzing HMDA data, Fannie Mae includes in its market totals those HMDA loans identified as having been purchased by the reporting lender, above and beyond loans that were originated by the reporting lender.^[278] Fannie Mae contends that there are a subset of loans originated by brokers and subsequently purchased by wholesale lenders that are neither reported by the brokers nor the wholesale lenders as originations but are reported by the wholesale lenders as purchased loans. According to Fannie Mae, these HMDA-reported purchased loans should be added to HMDA-reported originated loans to arrive at an estimate of total mortgage originations.

This rule's market definition includes only HMDA-reported originations; purchased loans are excluded from the market definition. While some purchased loans may not be reported as originations in HMDA (the Fannie Mae argument), there are several reasons for assuming that most HMDA-reported purchased loans are also reported in HMDA as market originations. First, Fed staff have told HUD that including purchased loans would result in double counting mortgage originations.^[279] Second, comparisons of HMDA-reported FHA data with data reported by FHA supports the Fed's conclusion. For instance, FHA's own data indicate that during 2001 FHA insured 752,319 home purchase loans in metropolitan areas; the sum of HMDA-reported purchased home loans and HMDA-reported originated home loans in metropolitan areas alone yields a much higher figure of 845,176 FHA-insured loans during 2001.^[280] While these calculations are for the FHA market (rather than the conventional market), they suggest that including HMDA-reported purchased loans in the market definition would overstate mortgage origination totals. Third, Abt Associates surveyed nine wholesale lenders and questioned them concerning their guidelines for reporting in HMDA loans purchased from brokers. Most of these lenders said brokered loans were reported as originations if they [the wholesale lender] make the credit decision; this policy is consistent with the Fed's guidelines for HMDA reporting. Abt Associates concluded that “brokered loans do seem more likely to be reported as originations * * *.” ^[281]

Finally, it should be noted that including purchased loans in the market definition does not significantly change the goals-qualifying shares of the market, mostly because borrower income data are missing for the majority of purchased loans. In addition, the low-income and underserved area shares for purchased and originated loans are rather similar. In 2001, the following differences in shares for the conventional conforming home purchase market were obtained for purchased and originated loans: Low-income (25.8 percent for purchased loans, 28.3 percent for market originations), low-mod income (41.3 percent, 43.2 percent), and underserved areas (24.2 percent, 25.8 percent). The comparisons were also similar for 2002.^[282]

8. Technical Issues: Using HMDA Data To Measure the Characteristics of GSE Purchases and Mortgage Market Originations ^[283]

This section discusses important technical issues concerning the use of HMDA data for measuring the GSEs' performance relative to the characteristics of mortgages originated in the primary market. The first issue concerns the reliability of HMDA data for measuring the borrower income and census tract characteristics of loans sold to the GSEs. Fannie Mae, in particular, has contended that HMDA data understates the percentages of its business that qualify for the three housing goals. In its comments on the proposed 2000 Rule, Fannie Mae questioned HUD's reliance on HMDA data for measuring its performance. As discussed below, HMDA data on loans sold to the GSEs do not include prior-year (seasoned) loans that are sold to the GSEs. Since about one-fourth of GSE purchases in any particular year involve loans originated in prior years, HMDA data will not provide an accurate measure of the goals-qualifying characteristics of the GSEs' total purchases when the characteristics of prior-year loans differ from those of newly-originated, current-year loans.

A related issue concerns the appropriate definition of the GSE data when making annual comparisons of GSE performance with the market. On the one hand, the GSE annual data can be expressed on a purchase-year basis, which means that all GSE purchases in a particular year would be assigned to that particular year. Alternatively, the GSE annual data can be expressed on an origination-year basis, which means that GSE purchases in a particular year would be assigned to the calendar year that the GSE-purchased mortgage was originated; for example, a GSE's purchase during 2001 of a loan originated in 1999 would be assigned to 1999, the year the loan was originated. These two approaches are discussed further below.

A final technical issue concerns the reliability of HMDA for measuring the percentage of goals-qualifying loans in the primary market. Both GSEs refer to findings from a study by Jim Berkovec and Peter Zorn concerning potential bias in HMDA data.^[284] Based on a comparison of the borrower and census tract characteristics between Freddie-Mac-purchased loans (from Freddie Mac's own data) and loans identified in 1993 HMDA data as sold to Freddie Mac, Berkovec and Zorn conclude that HMDA data overstate the percentage of conventional conforming loans originated for lower-income borrowers and for properties located in underserved census tracts. If HMDA data overstate the percentage of goals-qualifying loans, then HUD's market benchmarks (which are based on HMDA data) will also be overstated. The analysis below does not support the Berkovec and Zorn findings—it appears that HMDA data do not overstate the share of goals-qualifying loans in the market. The discussion below of the GSEs' purchases of prior-year and current-year loans also highlights the strategy of purchasing seasoned loans that qualify for the housing goals. The implications of this strategy for understanding recent shifts in the relative performance of Fannie Mae and Freddie Mac are discussed below in Section E.9.

a. GSEs' Purchases of “Prior-Year” and “Current-Year” Mortgages

There are two sources of loan-level information about the characteristics of mortgages purchased by the GSEs—the GSEs themselves and HMDA data. The GSEs provide detailed data on their mortgage purchases to HUD on an annual basis. As part of their annual HMDA reporting responsibilities, lenders are required to indicate whether their new mortgage originations or the loans that they purchase (from affiliates and other institutions) are sold to Fannie Mae, Freddie Mac or some other entity. There have been numerous studies by HUD staff and other researchers that use HMDA data to compare the borrower and neighborhood characteristics of loans sold to the GSEs with the characteristics of all loans originated in the market. One question is whether HMDA data, which is widely available to the public, provides an accurate measure of GSE performance, as compared with the GSEs' own data.^[285] Fannie Mae has argued that HMDA data understate its past performance, where performance is defined as the percentage of Fannie Mae's mortgage purchases accounted for by one of the goal-qualifying categories. As explained below, over the past six years, HMDA has provided rather reliable national-level information on the goals-qualifying percentages for the GSEs' purchases of “current-year” (i.e., newly-originated) loans, but not for their purchases of “prior-year” loans.^[286]

In any given calendar year, the GSEs can purchase mortgages originated in that calendar year or mortgages originated in a prior calendar year. In 2001 and 2002, for example, purchases of prior-year mortgages accounted for approximately 20 percent of the home loans purchased by each GSE.^[287] HMDA data provide information mainly on newly-originated mortgages that are sold to the GSEs—that is, HMDA data on loans sold to the GSEs will not include many of their purchases of prior-year loans. The implications of this for measuring GSE performance can be seen in Table A.11, which provides annual data on the borrower and census tract characteristics of GSE purchases, as measured by HMDA data and by the GSEs' own data. Table A.11 divides each of the GSEs' goals-qualifying percentages for a particular acquisition year into two components, the percentage for “prior-year” loans and the percentage for “current-year” loans.

Consider Fannie Mae's special affordable purchases in 2002. According to Fannie Mae's own data, 16.3 percent of its purchases during 2002 were special affordable loans. According to HMDA data, only 15.5 percent of loans sold to Fannie Mae fell into the special affordable category. In this case, HMDA data underestimate the special affordable share of Fannie Mae's purchases during 2002. What explains these different patterns in the GSE and HMDA data? The reason that HMDA data underestimate the special affordable percentage of Fannie Mae's 2002 purchases can be seen by disaggregating Fannie Mae's purchases during 2002 into their prior-year and current-year components. Table A.11 shows that the overall figure of 16.3 percent for special affordable purchases is a weighted average of 18.8 percent for Fannie Mae's purchases during 2002 of prior-year mortgages and 15.8 percent for its purchases of current-year purchases. The HMDA-reported figure of 15.5 percent is based mainly on newly-mortgaged (current-year) loans that lenders reported as being sold to Fannie Mae during 2002. The HMDA figure is similar in concept to the current-year percentage from the GSEs' own data. And the HMDA figure and the GSE current-year figure are practically the same in this case (15.5 versus 15.8 percent). Thus, the relatively large share of special affordable mortgages in Fannie Mae's purchases of prior-year mortgages explains why Fannie Mae's own data show an overall (both prior-year and current-year) percentage of special affordable loans that is higher than that reported for Fannie Mae in HMDA data.

b. Reliability of HMDA Data

With the above explanation of the basic differences between GSE-reported and HMDA-reported loan information, issues related to the reliability of HMDA data can now be discussed. Table A.12 presents the same information as Table A.11, except that the data are aggregated for the years 1993-5, 1996-2003, and 1999-2003. Comparing HMDA-reported data on GSE purchases with GSE-reported current-year data suggests that, on average, HMDA data have provided reasonable estimates of the goals-qualifying percentages for the GSEs' current-year purchases (with the exception of Freddie Mac's underserved area loans, as discussed below). For example, Fannie Mae reported that 13.7 percent of the current-year loans it purchased between 1996 and 2003 were for special affordable borrowers. In their HMDA submissions, lenders reported a nearly identical figure of 13.4 percent for the special affordable share of loans that they sold to Fannie Mae. The corresponding numbers for Freddie Mac were 12.8 percent reported by them and 12.1 percent reported by HMDA. During the same period, both Fannie Mae and HMDA reported that approximately 23 percent of current-year loans purchased by Fannie Mae financed properties in underserved areas. However, Freddie Mac reported that 21.3 percent of the current-year loans it purchased between 1996 and 2003 financed properties in underserved areas, a figure somewhat higher than the 19.6 percent that HMDA reported as underserved area loans sold to Freddie Mac during that period.^[288]

The facts that the Fannie Mae and HMDA figures for special affordable, low-mod and underserved area loans are similar, and that the Freddie Mac discrepancies are the result of Freddie Mac reporting higher percentages than HMDA, suggest that the Berkovec and Zorn conclusions about HMDA being upward biased are wrong.^[289] For the 1996-to-2003 period, the discrepancies reported in Table A.11 as well as Table A.12 are mostly consistent with HMDA being biased in a downward direction, not an upward direction as Berkovec and Zorn contend.^[290] In particular, the Freddie-Mac-reported underserved area percentage (as well as its special affordable percentage) being larger than the HMDA-reported underserved area percentage suggests a downward bias in HMDA. The more recent and complete (Fannie Mae data as well as Freddie Mac data) analysis does not support the Berkovec and Zorn finding that HMDA overstates the goals-qualifying percentages of the market.^[291]

c. Purchase-Year Versus Origination-Year Reporting of GSE Data

In comparing the GSEs' performance to the primary market, HUD has typically expressed the GSEs' annual performance on a purchase-year basis. That is, all mortgages (including both current-year mortgages and prior-year mortgages) purchased by a GSE in a particular year are assigned to the year of GSE purchase. The approach of including a GSE's purchases of both “current-year” and “prior-year” mortgages gives the GSE full credit for their purchase activity in the year that the purchase actually takes place; this approach is also consistent with the statutory requirement for measuring GSE performance under the housing goals. However, this approach results in an obvious “apples to oranges” problem with respect to the HMDA-based market data, which include only newly-originated mortgages (i.e., current-year mortgages). To place the GSE and market data on an “apples to apples” basis, HUD has also used an alternative approach that expresses the GSE annual data on an origination-year basis. In this case, all purchases by a GSE in any particular year would be fully reported but they would be allocated to the year that they were originated, rather than to the year they were purchased. Under this approach, a GSE's data for the year 2000 would not only include that GSE's purchases during 2000 of newly-originated mortgages but also any year-2000-originations purchased in later years (i.e., during 2001, 2002 and 2003 in this analysis). This approach places the GSE and the market data on a consistent, current-year basis. In the above example, the market data would present the income and underserved area characteristics of mortgages originated in 2000, and the GSE data would present the same characteristics of all year-2000-mortgages that the GSE has purchased to date (i.e., through year 2003).^[292]

Below, results will be presented for both the purchase-year and origination-year approaches. Following past HUD studies that have compared GSE performance with the primary market, most of the analysis in this section reports the GSE data on a purchase-year basis; however, the main results are repeated with the GSE data reported on an origination-year basis. This allows the reader to compare any differences in findings about how well the GSEs have been doing relative to the market.

9. Affordable Lending by the GSEs: Home Purchase Loans

This section compares the GSEs' affordable lending performance with the primary market for the years 1993-2003. The analysis in this section begins by presenting the GSE data on a purchase-year basis. As discussed above, the GSE data that are reported to HUD include their purchases of mortgages originated in prior years as well as their purchases of mortgages originated during the current year. The market data reported by HMDA include only mortgages originated in the current year. This means that the GSE-versus-market comparisons are defined somewhat inconsistently for any particular calendar year. Each year, the GSEs have newly-originated loans available for purchase, but they can also purchase loans from a large stock of seasoned (prior-year) loans currently being held in the portfolios of depository lenders. One method for making the purchase-year data more consistent is to aggregate the data over several years, instead of focusing on annual data. This provides a clearer picture of the types of loans that have been originated and are available for purchase by the GSEs. This approach is taken in Tables A.14 and A.15, which are discussed below. Another method for making the GSE and market data consistent is to express the GSE data on an origination-year basis; that approach is taken in Table A.16, which is discussed after presenting the annual results on a purchase-year basis.

a. Longer-Term Performance, 1993-2003 and 1996-2003

Table A.13 summarizes the funding of goals-qualifying mortgages by the GSEs, depositories and the conforming market for the ten-year period between 1993 and 2003. Data are also presented for two important sub-periods: 1993-95 (for showing how much the GSEs have improved their performance since the early-to-mid 1990s); and 1996-2003 (for analyzing their performance since the current definitions of the housing goals were put into effect). Given the importance of the GSEs for expanding homeownership, this section focuses on home purchase mortgages, and the next section will examine first-time homebuyer loans. Section IV below will briefly discuss the GSEs' overall performance, including refinance and home purchase loans. Several points stand out concerning the affordable lending performance of Freddie Mac and Fannie Mae over the two longer-term periods, 1993-2003 and 1996-2003.

Freddie Mac lagged both Fannie Mae and the primary market in funding affordable home loans in metropolitan areas between 1993 and 2003. During that period, 12.2 percent of Freddie Mac's mortgage purchases were for special affordable (mainly very-low-income) borrowers, compared with 13.3 percent of Fannie Mae's purchases, 15.4 percent of loans originated by depositories,^[293] and 15.5 percent of loans originated in the conforming market without B&C loans.^[294]

Although Freddie Mac consistently improved its performance during the 1990s, a similar pattern characterized the 1996-2003 period. During that period, 40.3 percent of Freddie Mac's purchases were for low- and moderate-income borrowers, compared with 42.2 percent of Fannie Mae's purchases, 43.1 percent of loans originated by depositories, and 43.6 percent of loans originated in the conventional conforming market. Over the same period, 22.0 percent of Freddie Mac's purchases financed properties in underserved neighborhoods, compared with 24.0 percent of Fannie Mae's purchases, 25.1 percent of depository originations, and 25.7 percent of loans originated in the primary market.

Fannie Mae's affordable lending performance was better than Freddie Mac's over the 1993 to 2003 period as well as during the 1996 to 2003 period. However, Fannie Mae lagged behind depositories and the overall market in funding affordable loans during both of these periods (see above paragraph). Between 1996 and 2003, the “Fannie-Mae-to-market” ratio was only 0.89 on the special affordable category, obtained by dividing Fannie Mae's performance of 14.1 percent by the market's performance of 15.9 percent. Fannie Mae's market ratio was 0.97 on the low-mod category and 0.93 on the underserved area category. The “Freddie-Mac-to-market” ratios for 1996-2003 were lower—0.83 for special affordable, 0.92 for low-mod, and 0.86 for underserved areas.

The above analysis has defined the market to exclude B&C loans, which HUD believes is the appropriate market definition. However, to gauge the sensitivity of the results to how the market is defined, Table A.14 shows the effects on the market percentages for different definitions of the conventional conforming market, such as excluding manufactured housing loans, small loans, and all subprime loans (i.e., the A-minus portion of the subprime market as well as the B&C portion). For example, the average special affordable (underserved area) market percentage for 1996-2003 would fall by about 1.6 (1.2) percentage points if both small loans (less than $15,000) and manufactured loans in metropolitan areas were also dropped from the market definition (see right-hand-side column in Table A.14). Except for Fannie Mae's relative performance on the low-mod category, the above findings with respect to the GSEs' longer-term performance are not much affected by the choice of market definition.

b. Recent Performance, 1999-2003

This and the next subsection focus on the average data for 1999-2003 in Table A.13 and the annual data reported in Table A.14. As explained below, the annual data are useful for showing shifts in the relative positions of Fannie Mae and Freddie Mac that began in 1999, and for highlighting the improvements made by Fannie Mae during 2001-2003 (which were the first three years under HUD's higher goal levels) and by Freddie Mac during 2002. Between 1993 and 1998, Freddie Mac's performance fell below Fannie Mae's, but a sharp improvement in Freddie Mac's performance during 1999 pushed it pass Fannie Mae on all three goals-qualifying categories. In 2000, Fannie Mae improved its underserved areas performance enough to surpass Freddie Mac on that category, while Freddie Mac continued to out-perform Fannie Mae on the borrower-income categories (special affordable and low-mod). By 2002, Fannie Mae had improved its performance enough to surpass Freddie Mac on all three goals-qualifying categories and to lead the special affordable and low-mod markets, while lagging the underserved areas market.

Consider first the average data for 1999-2003 reported in Table A.13. During this recent period, Freddie Mac's average performance was similar to Fannie Mae's performance for the special affordable category. Between 1999 and 2003, 14.7 percent of Freddie Mac's purchases and 15.1 percent Fannie Mae's mortgage purchases consisted of special affordable loans, compared with a market average of 16.2 percent. During this period, Freddie Mac purchased low-mod loans lower than the rate of Fannie Mae—42.6 percent for Freddie Mac, 43.6 percent for Fannie Mae, and 44.1 percent for the market. Freddie Mac (23.1 percent) also purchased underserved area loans at a lower rate than Fannie Mae (24.7 percent) and the primary market (26.2 percent). As these figures indicate, both Fannie Mae and Freddie Mac continued to lag the market during this recent four-year period. The GSEs' market ratios were 0.91-0.93 for special affordable loans and 0.97-0.99 for low-mod loans. Although less than one (where one indicates equal performance with the market), the “Fannie-Mae-to-market” ratio (0.94) for the underserved area category was much higher than the “Freddie-Mac-to-market” ratio (0.88).

Fannie Mae's performance in 1999 was significantly below its long run trend. Thus, averages for 2000-2003 are also presented in Table A.13, dropping 1999. These data show an increase in Fannie Mae's performance relative to the market. Between 2000 and 2003, special affordable (underserved area) loans accounted for 15.6 percent (25.5 percent) of Fannie Mae's purchases, compared with 16.0 percent (26.4 percent) for the market. During this 2000-2003 period, Fannie Mae slightly led the low-mod market (44.4 percent for Fannie Mae and 44.1 percent for the primary market).

Table A.14 shows the effects on the market percentages for 1999-2003 (as well as 2000-2003) of different definitions of the conventional conforming market. Excluding both small loans and manufactured housing loans (as well as B&C loans) in metropolitan areas would reduce the 1999-2003 market percentage for special affordable loans from 16.2 percent to 14.9 percent, which would place Fannie Mae slightly above the market and Freddie Mac close to the market. Similarly, excluding these loans would reduce the 1999-2003 market percentage for underserved areas from 26.2 percent to 25.2 percent, which would raise Fannie Mae's market ratio from 0.94 to 0.98 and Freddie Mac's, from 0.88 to 0.92. As shown in Table A.14, Fannie Mae is even closer to the market averages if the year 1999 is dropped—over the 2000-2003 period, Fannie Mae's performance on the underserved area category is practically at market levels under the above alternative definition of the market, and its performance on the special affordable and low-mod categories is above market levels.

Finally, Tables A.13 and A.14 report GSE and market data for the even more recent period, 2001-2003, which represents the first three years under the current housing goal targets (put in place by HUD in its Final Rule dated October 30, 2000). These data show that Freddie Mac's average performance during this period was below the market on each of the three housing goals (with market ratios of 0.96 for special affordable, 0.98 for low-mod, and 0.91 for underserved areas and that Fannie Mae's average performance was above the market on the special affordable and low-mod categories (with a market ratio of 1.02 on each category) but below the market on the underserved areas category (with a market ratio of 0.98).

c. GSEs' Performance—Annual Data

Freddie Mac's Annual Performance. As shown by the annual data reported in Table A.15, Freddie Mac significantly improved its purchases of goals-qualifying loans during the 1990s. Its purchases of loans for special affordable borrowers increased from 6.5 percent of its business in 1992 to 9.2 percent in 1997, and then jumped to 14.7 percent in 2000 before falling slightly to 14.4 percent in 2001 and rising again to almost 16 percent in 2002 and 2003. The underserved areas share of Freddie Mac's purchases increased at a more modest rate, rising from 18.6 percent in 1992 to 22.3 percent by 2001; it then jumped to 25.8 percent in 2002 but fell to 24.0 percent in 2003.

With its improved performance, Freddie Mac closed its gap with the market in funding goals-qualifying loans. In 2003, special affordable loans accounted for 15.6 percent of Freddie Mac's purchases and 15.9 percent of loans originated in the conventional conforming market, which produces a “Freddie-Mac-to-market” ratio of 0.98 (15.6 divided by 15.9). Table A.15 shows the trend in the “Freddie-Mac-to-market” ratio from 1992 to 2003 for each of the goals-qualifying categories. For the special affordable and low-mod categories, Freddie Mac's performance relative to the market remained flat (at approximately 0.60 and 0.80, respectively) through 1997; by 2003, the “Freddie-Mac-to-market” ratios had risen to 0.98 for both the special affordable and low-mod categories.

Surprisingly, Freddie Mac did not make much progress during the 1990s closing its gap with the market on the underserved areas category. The “Freddie-Mac-to-market” ratio for underserved areas was the same in 2000 (0.84) as it was in 1992 (0.84). While it rose to 0.88 in 2001, that was due more to a decline in the market level than to an improvement in Freddie Mac's performance. However, due to a substantial increase in Freddie Mac's underserved area percentage from 22.3 percent in 2001 to 25.8 percent in 2002, Freddie Mac's performance approached market performance (26.3 percent) during 2002.^[295] In the ten years under the housing goals, the year 2002 represented the first time that Freddie Mac's performance in purchasing home loans in underserved areas had ever been within two percentage points of the market's performance.^[296] But, as noted above, Freddie Mac's performance on the underserved areas goal fell to 24.0 percent in 2003, leaving it with a “Freddie Mac-to-Market” ratio of 0.87.

Fannie Mae's Annual Performance. With respect to purchasing affordable loans, Fannie Mae followed a different path than Freddie Mac. Fannie Mae improved its performance between 1992 and 1998 and made much more progress than Freddie Mac in closing its gap with the market. In fact, by 1998, Fannie Mae's performance was close to that of the primary market for some important components of affordable lending. In 1992, special affordable loans accounted for 6.3 percent of Fannie Mae's purchases and 10.4 percent of all loans originated in the conforming market, giving a “Fannie Mae-to-market” ratio of 0.61. By 1998, this ratio had risen to 0.86, as special affordable loans had increased to 13.2 percent of Fannie Mae's purchases and to 15.4 percent of market originations. A similar trend in market ratios can be observed for Fannie Mae on the underserved areas category. In 1992, underserved areas accounted for 18.3 percent of Fannie Mae's purchases and 22.2 percent of market originations, for a “Fannie Mae-to-market” ratio of 0.82. By 1998, underserved areas accounted for 22.7 percent of Fannie Mae's purchases and 24.2 percent of market originations, for a higher “Fannie Mae-to-market” ratio of 0.94.^[297]

The year 1999 saw a shift in the above patterns, with Fannie Mae declining in overall performance while the share of goals-qualifying loans in the market increased. Between 1998 and 1999, the special affordable share of Fannie Mae's business declined from 13.2 percent to 12.5 percent while this type of lending in the market increased from 15.4 percent to 17.0 percent. For this reason, the “Fannie-Mae-to-market” ratio for special affordable loans declined sharply from 0.86 in 1998 to 0.74 in 1999. The share of Fannie Mae's purchases in underserved areas also declined, from 22.7 percent in 1998 to 20.4 percent in 1999, which lowered the “Fannie-Mae-to-market” ratio from 0.94 to 0.81.

After declining in 1999, Fannie Mae's performance rebounded in 2000, particularly on the underserved areas category. Fannie Mae's underserved areas percentage jumped by three percentage points from 20.4 percent in 1999 to 23.4 percent in 2000. The 2000 figure was similar to its level in 1997 but below Fannie Mae's peak performances of 24-25 percent during 1994 and 1995. Between 1999 and 2000, the “Fannie-Mae-to-market” ratio for underserved areas increased from 0.81 to 0.89. Fannie Mae improved its performance on the special affordable goal at a more modest rate. Fannie Mae's special affordable percentage increased by 0.8 percentage points from 12.5 percent in 1999 to 13.3 percent in 2000. The 2000 figure was similar to its previous peak level (13.2 percent) in 1998. The “Fannie-Mae-to-market” ratio for special affordable loans increased from 0.74 in 1999 to 0.80 in 2000, with the latter figure remaining below Fannie Mae's peak market ratio (0.86) in 1998.

Fannie Mae continued its improvement in purchasing targeted home loans during 2001, at a time when the conventional conforming market was experiencing a decline in affordable lending; and again in 2002, at a time when the conventional conforming market was increasing enough to return approximately to its year-2000 level. Thus, during the 2000-to-2003 period, Fannie Mae significantly improved its targeted purchasing performance while the primary market originated targeted home loans at about the same rate in 2002 as it did in 2000. As a result, Fannie Mae's performance during 2001 approached the market on the special affordable and underserved area categories and matched the market on the low-mod category. In 2002, Fannie Mae outperformed the market on all three areas categories.

As shown in Table A.15, Fannie Mae increased its special affordable percentage by 1.6 percentage points, from 13.3 percent in 2000 to 14.9 percent in 2001, and then increased it further to 16.3 percent in 2002, the latter being slightly above the market's performance of 16.1 percent. The “Fannie-Mae-to-market” ratio for special affordable loans jumped from 0.80 in 2000 to 1.01 in 2002. In 2003, Fannie Mae's special affordable performance jumped to 17.1 percent while the market declined slightly to 15.9 percent, increasing Fannie Mae's market ratio to 1.08.

Between 2000 and 2001, Fannie Mae increased its low-mod percentage from 40.8 percent to 42.9 percent at the same time that the low-mod share of the primary market was falling from 43.9 percent to 42.9 percent, placing Fannie Mae at the market's performance in 2001. During 2002, the low-mod share of Fannie Mae's purchases of home loans increased further to 45.3 percent, placing Fannie Mae 0.7 percentage points above the market performance of 44.6 percent. Between 2002 and 2003 Fannie Mae's performance jumped to 47.0 percent, while the primary market remained at 44.6 percent, giving Fannie Mae a market ratio of 1.05 in 2003.

Fannie Mae increased its underserved area percentage from 23.4 percent in 2000 to 24.2 percent in 2001 while the underserved area share of the primary market was falling from 26.4 percent to 25.2 percent, placing Fannie Mae at less than one percentage point from the market's performance. The “Fannie-Mae-to-market” ratio for underserved area loans was 0.97 in 2001. During 2002, the underserved area share of Fannie Mae's purchases of home loans increased further to 26.7 percent, placing Fannie Mae slightly ahead of market performance (26.3 percent). However, between 2002 and 2003, Fannie Mae showed little improvement (rising to 26.8 percent) while the market increased to 27.6 percent, leaving Fannie Mae with a market ratio of 0.97.

As noted earlier, Tables A.13 and A.14 summarize Fannie Mae's average performance over the 2001-2003 period. During these first three years under the current housing goal targets, Fannie Mae led the special affordable market (average performance of 16.2 percent versus 15.9 percent for the market) and the low-mod market (average performance of 45.2 percent versus 44.1 percent for the market) but lagged the underserved areas market (average performance of 26.0 percent versus 26.4 percent for the market). Table A.14 also reports Fannie Mae's 2001-2003 performance under alternative definitions of the primary market. As shown there, the above findings of Fannie Mae's improvement relative to the market during 2001-2003 are further reinforced when lower market percentages are used. For example, Fannie Mae essentially matches the underserved areas market if manufactured housing loans in metropolitan areas (in addition to B&C loans) are excluded from the market definition (a Fannie Mae share of 26.0 percent and a market share of 26.1 percent).

Changes in the “Fannie-Mae-to-Freddie-Mac” Performance Ratio. The above discussion documents shifts in the relative performance of Fannie Mae and Freddie Mac over the past few years. To highlight these changing patterns, Table A.15 reports the ratio of Fannie Mae's performance to Freddie Mac's performance for each goals category for the years 1992 to 2003. As shown there, the “Fannie-Mae-to-Freddie-Mac” ratio for the special affordable category increased from approximately one in 1992 (indicating equal performance) to over 1.3 during the 1994-97 period, indicating that Fannie Mae clearly out-performed Freddie Mac during this period. Between 1997 and 2000, Freddie Mac substantially increased its special affordable share (from 9.2 percent to 14.7 percent), causing the “Fannie-Mae-to-Freddie-Mac” ratio to fall from 1.27 in 1997 to 0.90 in 2000 (indicating Freddie Mac surpassed Fannie Mae). But Fannie Mae's stronger performance during 2001-2003 returned the ratio to above one (1.03 in 2001 and 2002 and 1.10 in 2003), indicating better performance for Fannie Mae (e.g., 17.1 percent in 2002 versus 15.6 percent for Freddie Mac). The “Fannie-Mae-to-Freddie-Mac” performance ratio for low-mod loans followed a similar pattern, standing at 1.07 in 2003 (47.0 percent for Fannie Mae versus 43.8 percent for Freddie Mac).

Prior to 2000, the “Fannie-Mae-to-Freddie-Mac” ratio for underserved areas had also followed a pattern similar to that outlined above for special affordable loans, but at a lower overall level—rising from about one in 1992 (indicating equal performance) to approximately 1.2 during the 1994-97 period, before dropping to slightly below one (0.98) in 1999. However, Fannie Mae increased its underserved areas percentage from 20.4 percent in 1999 to 24.4 percent in 2001 while Freddie Mac only increased its percentage from 20.9 percent to 22.3 percent. This resulted in the “Fannie-Mae-to-Freddie-Mac” ratio rising from 0.98 in 1999 to 1.09 in 2001. But during 2002, Freddie Mac's underserved area percentage jumped by 3.5 percentage points to 25.8 percent, while Fannie Mae's increased at a more modest rate (by 2.3 percentage points) to 26.7 percent, with the result being that the “Fannie-Mae-to-Freddie-Mac” ratio for underserved area loans fell from 1.09 in 2001 to 1.03 in 2002. During 2003, Fannie Mae essentially maintained its performance (26.8 percent), while Freddie Mac reduced its performance by 1.8 percentage points to 24.0 percent. This increased the 2003 “Fannie Mae-to-Freddie Mac” ratio for underserved areas to 1.12.

To conclude, while Freddie Mac ended the 1990s on a more encouraging note than Fannie Mae, the past four years (2000, 2001, 2002 and 2003) have seen a substantial improvement in Fannie Mae's performance on all three goals-qualifying categories. Fannie Mae ended the 1990s with a decline in affordable lending performance at the same time that Freddie Mac was improving and the share of goals-qualifying loans was increasing in the market. Both GSEs' performance during 2000 was encouraging—Freddie Mac continued to improve, particularly with respect to the borrower-income categories, while Fannie Mae reversed its declining performance, particularly with respect to underserved areas. During 2000, Freddie Mac outperformed Fannie Mae on the special affordable and low-mod categories, while Fannie Mae purchased a higher percentage of loans in underserved areas. During 2001, Fannie Mae continued to improve its performance while Freddie Mac's performance remained about the same and the market's originations of affordable loans declined somewhat. The result was that during 2001 Fannie Mae outperformed Freddie Mac on all three goals-qualifying categories, and even matched the market on the low-mod category. During 2002, both Fannie Mae and Freddie Mac again improved their performance; Fannie Mae continued to outperform Freddie Mac and outperformed the market on all three goals-qualifying categories. While Freddie Mac lagged the market on all three goals-qualifying categories during 2002, it had significantly closed its gap by the end of 2002, particularly on the underserved area category. During 2003, Fannie Mae made significant improvement in the special affordable and low-mod categories, allowing it to lead the primary market. Freddie Mac, on the other hand, simply maintained its 2002 performance in these two categories, which meant it lagged further behind Fannie Mae. On the underserved area category, Fannie Mae maintained its 2002 performance during 2003 while Freddie Mac significantly reduced its performance, leaving both GSEs, but particularly Freddie Mac, behind the primary market on this category.

GSE Purchases of Seasoned Loans. When the GSE data are expressed on a purchase-year basis (as in the above analysis), one factor which affects each GSE's performance concerns their purchases of seasoned (prior-year) loans. As shown in Table A.11, Fannie Mae followed a strategy of purchasing targeted seasoned loans between 1996 and 1998, and again during 2000-2002—all years when Fannie Mae improved its overall affordable lending performance. For example, consider Fannie Mae's underserved area performance of 24.4 percent during 2001, which was helped by its purchases of seasoned mortgages on properties located in underserved neighborhoods. The underserved area percentage for Fannie Mae's purchases of newly-originated (current-year) mortgages was only 23.3 percent in 2001, or 1.9 percentage points below the market average of 25.2 percent. Fannie Mae obtained its higher overall percentage (24.4 percent) by purchasing seasoned loans with a particularly high concentration (28.3 percent) in underserved areas. Similarly, during 2001, the special affordable share of Fannie Mae's purchases of newly-originated mortgages was only 14.2 percent, or 1.4 percentage points below the market average of 15.6 percent. Again, Fannie Mae improved its overall performance by purchasing seasoned loans with a high percentage (18.1 percent) of special affordable loans, enabling Fannie Mae to reduce its gap with the market to 0.7 percentage points—14.9 percent versus 15.6 percent.

As shown in Table A.11, Freddie Mac also followed a strategy of purchasing seasoned special affordable loans mainly after 1999. Prior to 2000, Freddie Mac had not pursued such a strategy, or at least not to the same degree as Fannie Mae. During the 1997-99 period, Freddie Mac's purchases of prior-year mortgages and newly-originated mortgages had similar percentages of special affordable (and low-mod) borrowers. Over time, there have been small differentials between Freddie Mac's prior-year and newly-originated mortgages for the underserved areas category but they have been smaller than the differentials for Fannie Mae (see Table A.11).

d. GSEs' Annual Purchases of Home Loans—Origination-Year Basis

Table A.16 reports GSE purchase data for 1996 to 2003 on an origination-year basis. Recall that in this case, mortgages purchased by a GSE in any particular calendar year are allocated to the year that the mortgage was originated, rather than to the year that the mortgage was purchased (as in the above). This approach places the GSE and the market data on a consistent, current-year basis, as explained earlier.

In general, the comparisons of Freddie Mac's and the market's performance are similar to those discussed in Sections E.9.a-c above, except for some differences on the special affordable category. The “Freddie Mac to market” ratios in Table A.16 show that Freddie Mac has improved its performance but has also consistently lagged the primary market in funding mortgages covered by the housing goals.

The “Fannie Mae to market” ratios in Table A.16 show that Fannie Mae has improved its performance, has generally outperformed Freddie Mac, and led the market during 2003 on both the special affordable and low-mod goals. Under the origination-year approach, Fannie Mae lagged the market on all three housing goal categories during 2001 and on the underserved area category during 2002. Fannie Mae matched the market in funding special affordable loans during 2002 and led the market in funding low-mod loans. During 2003, Fannie Mae led the primary market on both the special affordable and low-mod categories but lagged the market on the underserved area category. For instance in 2003, low- and moderate-income loans accounted for 47.0 percent of Fannie Mae's purchases and 44.6 percent of the market originations, placing Fannie Mae 2.4 percentage points above the market. On the other hand, underserved areas accounted for 26.3 percent of Fannie Mae's purchases during 2003, which was 1.4 percentage points below market performance.

e. GSEs' Purchases of First-Time Homebuyer Mortgages—1999 to 2001

While not a specific housing goal category, mortgages for first-time homebuyers are an important component of the overall home loan market. Making financing available for first-time homebuyers is one approach for helping young families enter the homeownership market. Therefore, this section briefly compares the GSEs' funding of first-time homebuyer loans with that of primary lenders in the conventional conforming market.

During the past few years, the GSEs have increased their purchases of first-time homebuyer loans. For example, Fannie Mae's annual purchases of first-time homebuyer loans increased from approximately 287,000 in 1999 to 423,485 in 2003.^[298] However, since 1999, the first-time homebuyer share of the GSEs' purchases of home loans has remained relatively flat, varying within the 25-28 percent range.^[299]

Table A.17a compares the first-time homebuyer share of GSE purchases with corresponding share of home loans originated in the conventional conforming market. Readers are referred to recent work by Bunce and Gardner ^[300] for the derivation of the estimates of first-time homebuyer market shares reported in Table A.17a. Between 1999 and 2001, first-time homebuyers accounted for 26.5 percent of Fannie Mae's purchases of home loans, 26.5 percent of Freddie Mac's, and 37.6 percent of home loans originated in the conventional conforming market. Thus, both Fannie Mae and Freddie Mac fell substantially short of the primary market in financing first-time homebuyers during this time period. The GSEs' performance was only 70.5 percent of market performance (26.5 percent divided by 37.6 percent).

Table A.17a also reports first-time homebuyer shares for African Americans and Hispanics and for all minorities. Between 1999 and 2001, African-American and Hispanic first-time homebuyers accounted for 4.0 percent of Fannie Mae's purchases of home loans, 3.4 percent of Freddie Mac's purchases, and 6.9 percent of home loans originated in the conventional conforming market. for this subgroup, Fannie Mae's performance is 58 percent of market performance, while Freddie Mac's performance is 49 percent of market performance. The group of all minority first-time homebuyers accounted for 6.6 percent of Fannie Mae's purchases of home loans, 5.8 percent of Freddie Mac's purchases, and 10.6 percent of home loans originated in the conventional conforming market. In this case, Fannie Mae's performance is 62 percent of market performance, while Freddie Mac's performance is 55 percent of market performance.

Section E.12 below will continue this examination of first-time homebuyers by presenting market share analysis that estimates the GSEs' overall importance in the funding of first-time homebuyers.

f. Low- and Moderate-Income Subgoal for Home Purchase Loans

The Department is proposing to establishing a subgoal of 45 percent for each GSE's purchases of home purchase loans for low- and moderate-income families in the single-family-owner market of metropolitan areas for 2005, with the subgoal rising to 46 percent for 2006 and 47 percent for 2007 and 2008. If the GSEs meet this subgoal, they will be leading the primary market by approximately one percentage point in 2005 and by three percentage points in 2007-08, based on historical data (see below). This home purchase subgoal will encourage the GSEs to expand homeownership opportunities for lower-income homebuyers who are expected to enter the housing market over the next few years. As detailed in Section I, there are four specific reasons for establishing this subgoal: (1) The GSEs have the expertise, resources, and ability to lead the single-family-owner market, which is their “bread and butter” business; (2) except for the recent performance of Fannie Mae, the GSEs have historically lagged the primary market for low- and moderate-income loans, not led it; (3) the GSEs can improve their funding of first-time homebuyers and help reduce troublesome disparities in homeownership and access to mortgage credit; and (4) there are ample opportunities for the GSEs to expand their purchases in important and growing market segments such as the market for minority first-time homebuyers. Sections E.9 and G of this appendix provide additional information on opportunities for an enhanced GSE role in the home purchase market and on the ability of the GSEs to lead that market.

As shown in Tables A.13 and A.15, low- and moderate-income families accounted for an average of 44.1 percent of home purchase loans originated in the conventional conforming market of metropolitan areas between 1999 and 2003; the figure is 43.6 percent if the average is computed for the years between 1996 and 2003 or 44.1 percent if the average is computed for the more recent 2001-2003 period. Loans in the B&C portion of the subprime market are excluded from these market averages. To reach the 45-percent subgoal for 2005, Freddie Mac would have to improve its performance by one percentage point over its approximately 44 percent low-mod performance during 2002 and 2003, while Fannie Mae would have to maintain its performance of 45-47 percent over these two years. To reach the 47 percent subgoal in 2007-08, Freddie Mac would have to improve by three percentage points over its 2002-3 performance while Fannie Mae would have to maintain its 2003 performance of 47 percent.

As explained earlier, HUD will be re-benchmarking its median incomes for metropolitan areas and non-metropolitan counties based on 2000 Census median incomes, and will be incorporating the effects of the new OMB metropolitan area definitions. As shown in Table 17b, HUD projected the effects of these two changes on the low- and moderate-income shares of the single-family-owner market for the years 1999-2003. These estimates will be referred to as “projected data” while the 1990-based data reported in the various tables will be referred to as “historical data.” With the historical data, the average low-mod share of the conventional conforming market (without B&C loans) was 44.2 percent for home purchase loans (weighted average of 1999-2003 percentages in Table A.13); the corresponding average with the projected data was 43.5 percent, a differential of 0.7 percentage points. However, note that in 2003, the projected data for both GSEs and the market exhibit higher low-mod shares than the corresponding historical data. For 2003, the low-mod shares for the projected and historical data are as follows: Fannie Mae (47.5 percent for the projected data versus 47.0 percent for the historical data), Freddie Mac (44.2 percent versus 43.8 percent), and the market (45.6 percent versus 44.6 percent). Thus, based on 2003 experience, it appears that the low-mod share for single-family-owners in the conventional conforming market actually increase based on the re-benchmarking of area median incomes and the new OMB definitions of metropolitan areas. Thus, based on 2003 data, the 47-percent subgoal for 2007 is 2.4 percentage points above the 2003 market.

In terms of projected data, Fannie Mae could meet both the 2005 and 2007 subgoals by maintaining its projected 2003 low-mod performance of 47.5 percent. Freddie Mac's projected low-mod performance for 2003 was 44.2 percent, about 0.4 percentage points above its 2003 performance of 43.8 percent based on historical data. Thus, to reach the 45-percent subgoal for 2005, Freddie Mac would have to increase its 2003 projected performance by 0.8 percentage point, and to reach the 47-percent 2007 subgoal, Freddie Mac would have to increase its performance by 2.8 percentage points over its projected performance of 44.2 percent for 2003.

The subgoal applies only to the GSEs' purchases in metropolitan areas because the HMDA-based market benchmark is only available for metropolitan areas. HMDA data for non-metropolitan areas are not reliable enough to serve as a market benchmark. The Department is also setting home purchase subgoals for the other two goals-qualifying categories, as explained in Appendices B and C.

It should be noted that the findings in sub-sections 9.a-e above concerning the performance of the GSEs relative to the home purchase market do not change when projected, rather than historical data, are used.

10. GSEs Purchases of Total (Home Purchase and Refinance) Loans

Section E.9 examined the GSEs' acquisitions of home purchase loans, which is appropriate given the importance of the GSEs for expanding homeownership opportunities. To provide a complete picture of the GSEs' mortgage purchases in metropolitan areas, Tables A.18, A.19, A.20, and A.21 report the GSEs' purchases of all single-family-owner mortgages, including both home purchase loans and refinance loans.^[301]

Table A.18 provides a long-run perspective on the GSEs' overall performance. Between 1993 and 2003, as well as during the 1996-2003 period, the GSEs' performance was 81-91 percent of market performance for the special affordable category, 91-97 percent of market performance for the low-mod category, and 87-93 percent of market performance for the underserved areas category. For example, between 1996 and 2003, underserved areas accounted for 23.4 percent of Fannie Mae's purchases and 21.8 percent of Freddie Mac's purchases, compared with 25.2 percent for the conventional conforming market (without B&C loans). Similarly, for special affordable loans, both GSEs lagged the market during the 1996-2003 period—Fannie Mae and Freddie Mac averaged approximately 13.0 percent while the market was over two percentage points higher at 14.8 percent.

Similar to the patterns discussed for home purchase loans, Fannie Mae has tended to outperform Freddie Mac. This can be seen by examining the various “Fannie-Mae-to-Freddie-Mac” ratios in Table A.18, which are all equal to or greater than one. Over the recent 1999-2003 period, Fannie Mae and Freddie Mac continued to lag the overall market on all three goals-qualifying categories. Special affordable (underserved area) loans averaged 14.0 (23.8) percent of Fannie Mae's purchases, 13.2 (22.1) percent of Freddie Mac's purchases, and 15.0 (25.2) percent of market originations. For Fannie Mae, the market ratio was 0.93 for special affordable loans, 0.98 for low-mod loans, and 0.94 for underserved area loans. As with home purchase loans, dropping the year 1999 and characterizing recent performance by the 2000-2003 period improves the performance of both GSEs relative to the market, but particularly Fannie Mae. Over the 2000-2003 period, the “Fannie-Mae-to-market” ratio was 0.97 for special affordable loans, 1.00 for low-mod loans, and 0.96 for underserved area loans. Over the last three years (2001-2003), the “Fannie-Mae-to-market” ratios are even higher—1.00 for special affordable loans, 1.01 for low-mod loans, and 0.98 for underserved area loans. In other words, during the first three years under the current housing goal targets, Fannie Mae matched the special affordable market, led the low-mod market, and lagged the underserved areas market.

The above analysis has defined the market to exclude B&C loans. Table A.19 shows the effects on the market percentages of different definitions of the conventional conforming market. For example, the average 1999-2003 market share for special affordable (underserved areas) loans would fall to 14.4 (24.8) percent if small loans and manufactured housing loans in metropolitan areas were excluded from the market definition along with B&C loans. In this case, the market ratio for Fannie Mae (Freddie Mac) would be was 0.97 (0.92) for special affordable loans, 1.00 (0.95) for low-mod loans, and 0.96 (0.89) for underserved area loans.

Shifts in performance occurred during 2001-2003, the first three years under HUD's higher housing goal targets. Table A.20 shows that both GSEs improved their overall performance between 1999 and 2000, but they each fell back a little during the heavy refinancing year of 2001. But the primary market (without B&C loans) experienced a much larger decline in affordable lending during the refinancing wave than did either of the GSEs. Fannie Mae stood out in 2001 because of its particularly small decline in affordable lending. Between 2000 and 2001, Fannie Mae's special affordable lending fell by only 0.6 percentage points while Freddie Mac's fell by 2.8 percentage points and the market's fell by 3.6 percentage points. The corresponding percentage point declines for the underserved areas category were 1.0 for Fannie Mae, 1.9 for Freddie Mac, and 3.8 for the market. By the end of 2001, Fannie Mae led Freddie Mac in all three goals-qualifying categories, and had erased its gap with the low-mod market, but continued to lag the market on the special affordable and underserved areas categories.

During the refinancing wave of 2002, Fannie Mae improved slightly on the special affordable and low-mod categories and declined slightly on the underserved area category. Freddie Mac showed slight improvement on the special affordable and underserved area categories and remained about the same on the low-mod category. The result of these changes can be seen by considering the market ratios in Table A.20. In 2002, special affordable loans accounted for 14.3 percent of Fannie Mae's purchases and 14.4 percent of loans originated in the non-B&C portion of the conventional conforming market, yielding a “Fannie-Mae-to-market” ratio of 0.99. Since Fannie Mae's market ratio for the special affordable category stood at 0.80 in 2000, Fannie Mae substantially closed its gap with the market during 2001 and 2002. During this period, Fannie Mae also mostly eliminated its market gap for the other two goals-qualifying categories. In 2002, underserved area loans accounted for 24.0 percent of Fannie Mae's purchases and 24.2 percent of loans originated in the non-B&C portion of the conventional conforming market, yielding a “Fannie-Mae-to-market” ratio of 0.99, or approximately one. During 2002, low-mod loans accounted for 42.2 percent of Fannie Mae's purchases and 42.0 percent of loans originated in the market, yielding a “Fannie-Mae-to-market” ratio of 1.00 (also note that Fannie Mae slightly outperformed the low-mod market during 2001). Thus, during 2002, Fannie Mae essentially matched the market on each of the three goals-qualifying categories.

In 2003, Fannie Mae's continued to improve its performance on the special affordable and low-mod categories. In 2003, special affordable loans accounted for 14.3 percent of Fannie Mae's purchases and 14.0 percent of loans originated in the market, yielding a “Fannie-Mae-to-market” ratio of 1.02. During that year, low-mod loans accounted for 42.3 percent of Fannie Mae's purchases and 41.2 percent of total (home purchase and refinance) loans originated in the market, yielding a “Fannie-Mae-to-market” ratio of 1.03. On the underserved areas category, Fannie Mae continued to lag behind the market (a 23.7 percent share for Fannie Mae and a 24.5 percent share for the market).

Freddie Mac significantly lagged the single-family (home purchase and refinance loans combined) market during 2001-2003. In 2003, the “Freddie-Mac-to-market” ratios were 0.86 for special affordable loans, 0.98 for low-mod loans, and 0.82 for underserved area loans.

Subprime Loans. Table A.14 in Section E.9 showed that the goals-qualifying shares of the home purchase market did not change much when originations by subprime lenders are excluded from the analysis; the reason is that subprime lenders operate primarily in the refinance market. Therefore, in this section's analysis of the total market (including refinance loans), one would expect the treatment of subprime lenders to significantly affect the market estimates and, indeed, this is the case. For the year 2001, excluding subprime loans reduced the goal-qualifying shares of the total market as follows: special affordable, from 15.0 to 13.9 percent; low-mod, from 42.3 to 40.9 percent; and underserved areas, from 25.7 to 23.9 percent. (See Table A.19.) Similar declines take place in 2002 and 2003.

As explained earlier, the comparisons in this appendix have defined the market to exclude the B&C portion of the subprime market. Industry observers estimate that A-minus loans account for about two-thirds of all subprime loans while the more risky B&C loans account for the remaining one-third. As explained earlier, this analysis reduces the goal-qualifying percentages from the HMDA data by half the differentials between (a) the market (unadjusted) and (b) the market without the specialized subprime lenders identified by Scheessele. As shown in Table A.19, accounting for B&C loans in this manner reduces the year 2001 HMDA-reported goal-qualifying shares of the total (home purchase and refinance) conforming market as follows: special affordable, from 15.0 to 14.5 percent; low-mod, from 42.3 to 41.6 percent; and underserved areas, from 25.7 to 24.9 percent. Obviously, the GSEs' performance relative to the market will depend on which market definition is used (much as it did with the earlier examples of excluding manufactured housing loans in metropolitan areas from the market definition). For example, defining the conventional conforming market to exclude subprime loans, rather than only B&C loans, would increase Fannie Mae's 2001 special affordable (underserved area) market ratio from 0.96 to 1.00 (0.97 to 1.01). Similarly, it would increase Freddie Mac's special affordable (underserved area) market ratio from 0.92 to 0.96 (0.90 to 0.94). For the broader-defined low-mod category, redefining the 2001 market to exclude subprime loans, rather than only B&C loans, would increase Fannie Mae's (Freddie Mac's) market ratio from 1.00 to 1.02 (0.97 to 0.98).

Table A.21 reports GSE purchase data for total (home purchase and refinance) loans on an origination-year basis. The “Freddie Mac-to-market” ratios in Table A.21 show that Freddie Mac has lagged the primary market in funding mortgages covered by the housing goals. The “Fannie Mae-to-market” ratios in Table A.21 show that Fannie Mae has always lagged the primary market in funding home purchase and refinance mortgages for properties in underserved areas but, in 2002 and 2003, led the low-mod market, and in 2003 led the special affordable market.

11. GSE Mortgage Purchases in Individual Metropolitan Areas

While the above analyses, as well as earlier studies, concentrate on national-level data, it is also instructive to compare the GSEs' purchases of mortgages in individual metropolitan areas (MSAs). In this section, the GSEs' purchases of single-family owner-occupied home purchase loans are compared to the market in individual MSAs. There are three steps. First, goals-qualifying percentages for conventional conforming mortgage originations (without B&C loans) are computed for each year and for each MSA, based on HMDA data. Second, corresponding goals-qualifying percentages are computed for each GSE's purchases for each year and for each MSA. These two sets of percentages are the same as those used in the aggregate analysis discussed in the above sections. Third, the “GSE-to-market” ratio is then calculated by dividing each GSE percentage by the corresponding market percentage. For example, if it is calculated that one of the GSEs' purchases of low- and moderate-income loans in a particular MSA is 40 percent of their overall purchases in that MSA, while 44 percent of all home loans (excluding B&C loans) in that MSA qualify as low-mod, then the GSE-to-market ratio is 40/44 (or 0.91). The goals-qualifying ratios for Fannie Mae and Freddie Mac can be compared for each MSA in a similar manner.

Tables A.22, A.23, and A.24 summarize the performance of the GSEs within MSAs for 2001, 2002 and 2003 originations of home purchase loans. A GSE's performance is determined to be lagging the market if the ratio of the GSE housing goal loan purchases to their overall purchases is less than 99 percent of that same ratio for the market. (The analysis was conducted where the “lag” determination is made at 98 percent instead of 99 percent and the results showed little change.) In the example given in the above paragraph, that GSE would be considered lagging the market. Tables A.22 (2001), A.23 (2002) and A.24 (2003) report the number of MSAs in which each GSE under-performs the market with respect to each of the three housing goal categories. The following points can be made from this data:

Fannie Mae's improvement between 2001 and 2003 shows up in these tables. In 2001, Fannie Mae lagged the market in 264 (80 percent) of the 331 MSAs in the purchase of underserved area loans; this number decreased to 236 (71 percent) MSAs in 2002 and to 243 (73 percent) MSAs in 2003. Fannie Mae's improvement was even greater for special affordable and low-mod loans; in the latter case, Fannie Mae lagged the market in 51 (15 percent) MSAs in 2003, compared with 194 (59 percent) MSAs in 2001.

Freddie Mac's improvement between 2001 and 2003 was greater for underserved area loans. In 2001, Freddie Mac lagged the market in 261 (79 percent) of the 331 MSAs in the purchase of underserved area loans; this number decreased to 168 (51 percent) MSAs in 2002 before rising to 222 (67 percent) MSAs in 2003. Freddie Mac's made less improvement on the special affordable and low-mod categories; in the former case, Freddie Mac lagged the market in 234 (71 percent) MSAs in 2003, compared with 279 (84 percent) MSAs in 2001.

12. GSE Market Shares: Home Purchase and First-Time Homebuyer Loans

This section examines the role that the GSEs have played in the overall affordable lending market for home loans. There are two differences from the above analyses in Sections E.9 and E.10. The first difference is that this section focuses on “market share” percentages rather than “distribution of business” percentages. A “market share” percentage measures the share of loans with a particular borrower or neighborhood characteristic that is funded by a particular market sector (such as FHA or the GSEs). In other words, a “market share” percentage measures a sector's share of all home loans originated for a particular targeted group. The “market share” of a sector depends not only on the degree to which that sector concentrates its business on a targeted group (i.e., its “distribution of business” percentage) but also on the size, or overall mortgage volume, of the sector. If an industry sector has a large “market share” for a targeted group, then that sector is making an important contribution to meeting the credit needs of the group. Both “distribution of business” and “market share” data are important for evaluating the GSEs' performance. In fact, given the large size of the GSEs, one would expect that a “market share” analysis would highlight their importance to the affordable lending market.

The second difference is that this section also examines the role of the GSEs in the total market for home loans, as well as in the conventional conforming market. Such an approach provides a useful context for commenting on the contribution of Fannie Mae and Freddie Mac to overall affordable lending, particularly given evidence that conventional lenders have done a relatively poor job providing credit access to disadvantaged families, which renders the conventional market a poor benchmark for evaluating GSE performance. The analysis of first-time homebuyers conducts the market share analysis in terms of both the total market (Section E.12.b) and the conventional conforming market (Section E.12.c).

While the GSEs have accounted for a large share of the overall market for home purchase loans, they have accounted for a very small share of the market for important groups such as minority first-time homebuyers. But as this section documents, the GSEs have been increasing their share of the low-income and minority market, which provides an optimistic note on which to go forward.

Section E.12.a uses HMDA and GSE data to estimate the GSEs' share of home loans originated for low-income and minority borrowers and their neighborhoods. Sections E.12.b and E.12.c summarize recent research on the role of the GSEs in the first-time homebuyer market. Section E.12.d examines the downpayment characteristics of home loans purchased by the GSEs, a potentially important determinant of the GSEs' ability to reach first-time homebuyers.

a. GSEs' Share of Home Purchase Lending

Table A.25 reports market share estimates derived by combining HMDA market data with GSE and FHA loan-level data. To understand these estimates, consider the GSE market share percentage of 46 percent for “All Home Purchase Loans” at the bottom of the first column in the table. That market share percentage is interpreted as follows:

It is estimated that home loans acquired by Fannie Mae and Freddie Mac during the years, 1999 to 2003, totaled 46 percent of all home loans originated in metropolitan areas during that period.

It should be noted that “all home loans” refers to all government (FHA and VA) loans plus all conventional loans less than the conforming loan limit; in other words, only “jumbo loans” are excluded from this analysis.^[302] The analysis is restricted to metropolitan areas because HMDA data (the source of the market estimates) are reliable only for metropolitan areas. B&C originations are included in the market data, since the purpose here is to gauge the GSEs' role in the overall mortgage market. As discussed in Section E.9, excluding B&C loans, or even all subprime loans, would not materially affect this analysis of the home loan market since subprime loans are mainly for refinance purposes. The analysis below frequently combines purchases by Fannie Mae and Freddie Mac since previous sections have already compared their performance relative to each other.

The GSE market share percentage for “Low-Income Borrowers” at the top of the first column of Table A.25 has a similar interpretation:

It is estimated that home loans for low-income borrowers acquired by Fannie Mae and Freddie Mac between 1999 and 2003 totaled 38 percent of all home loans originated for low-income borrowers in metropolitan areas.

According to the data in Table A.25, the GSEs account for a major portion of the market for targeted groups. For example, purchases by Fannie Mae and Freddie Mac represented 38 percent of the low-income-borrower market and 36-38 percent of the markets in low-income, high-minority, and underserved census tracts. Thus, access to credit in these historically underserved markets depends importantly on the purchase activities of Fannie Mae and Freddie Mac. However, the data in Table A.25 show that the GSEs' role in low-income and minority markets is significantly less than their role in the overall home loan market. Fannie Mae and Freddie Mac accounted for 46 percent of all home loans but only 37 percent of the loans financing properties in underserved neighborhoods. Their market share was even lower for loans to African-American and Hispanic borrowers—30 percent, or 16 percentage points less than the GSEs' overall market share of 46 percent.

An encouraging finding is that the GSEs have increased their presence in the affordable lending market during 2002 and 2003, when they accounted for 40-44 percent of the loans financing properties in low-income, high-minority, and underserved neighborhoods and for 34 percent of loans for African-American and Hispanic borrowers. These market share figures for the GSEs are generally higher than their performance during the two earlier years, 2000 and 2001.

To provide additional perspective, Table A.25 also reports market share estimates for FHA.^[303] During the 1999-2003 period, FHA's overall market share was less than half of the GSEs' market share, as FHA insured only 16 percent of all home mortgages originated in metropolitan areas. However, FHA's shares of the underserved segments of the market were much higher than its overall market share. For instance, between 1999 and 2003, FHA insured 24 percent of all mortgages originated in low-income census tracts, even though it insured only 16 percent of all home loans. FHA's share of the market was particularly high for African-American and Hispanic borrowers, as FHA insured 29 percent of all home loans originated for these borrowers between 1999 and 2003—a figure only one percentage point higher than the GSEs’ share of 30 percent.^[304] Thus, during the 1999-2003 period, FHA's overall market share (16.0 percent) was about one-third of the GSEs' market share (45.6 percent), but its share of the market for loans to African-Americans and Hispanics was almost equal to the GSEs' share of that market.

The data for the two recent years (2002 and 2003) indicate a larger market role for Fannie Mae and Freddie Mac relative to FHA. While the GSEs continued to have a much larger share of the overall market than FHA (47-49 percent for the GSEs versus 11-14 percent for FHA), their share of home loans for African-Americans and Hispanics jumped to 34 percent during 2002 and 2003, which was higher than the percentage share for FHA (17-25 percent). The differentials in market share between FHA and the GSEs on the other affordable lending categories listed in Table A.25 were also higher in 2002 and 2003 than in earlier years.

b. The GSEs' Share of the Total First-Time Homebuyer Market

This section summarizes two recent analyses of mortgage lending to first-time homebuyers; these two studies examine the total mortgage market, including both government and conventional loans originated throughout the U.S. (i.e., in both metropolitan areas and non-metropolitan areas). Section E.12.c will summarize a third study of first-time homebuyers that focuses on the conventional conforming market. All three studies are market share studies that examine the GSEs' role in the first-time homebuyer market.

First, a study by Bunce concluded that the GSEs have played a particularly small role in funding minority first-time homebuyers.^[305] Because HMDA does not require lenders to report information on first-time homebuyers, Bunce used data from the American Housing Survey to estimate the number of first-time homebuyers in the market. Using American Housing Survey data on home purchases from 1997 to 1999, Bunce estimated that the GSEs' share of the market for first-time African-American and Hispanic homebuyers was only 10-11 percent, or less than one-third of their share (36 percent) of all home purchases during that period. FHA's share of this market was 36 percent, or twice its share (18 percent) of all home purchases.^[306] These data highlight the small role that the GSEs have played in the important market for minority first-time homebuyers.

Bunce, Neal and Vandenbroucke (BNV) recently updated through 2001 the study by Bunce. In addition, BNV developed an improved methodology that combined industry, HMDA and AHS data to estimate the number of first-time homebuyers (by race and ethnicity) in the mortgage market during the years 1996 to 2001.^[307] BNV's analysis includes the total mortgage market, that is, the government, conventional conforming, and jumbo sectors of the mortgage market.

Table A.26 presents the key market shares estimated by BNV for the GSEs and FHA. The first figure (40.7) in Table A.26 is interpreted as follows: purchases of home loans by Fannie Mae and Freddie Mac totaled 40.7 percent of all home loans financed between 1996 and 2001. Going down the first column shows that the GSEs' share of the first-time homebuyer market was 24.5 percent during the 1996-to-2001—a market share significantly lower than their overall market share of 40.7 percent.

FHA's greater focus on first-time homebuyers is also reflected in the market share data reported in Table A.26. While FHA insured only 16.6 percent of all home loans originated between 1996 and 2001, it insured 30.9 percent of all first-time-homebuyer loans during that period. The GSEs, on the other hand, accounted for a larger share (40.7 percent) of the overall home purchase market but a smaller share (24.5 percent) of the first-time homebuyer market.

Table A.26 also reports home purchase and first-time homebuyer information for minorities. During the more recent 1999-to-2001 period, the GSEs' loan purchases represented 41.5 percent of all home mortgages but only 24.3 percent of home loans for African-American and Hispanic families, and just 14.3 percent of home loans for African-American and Hispanic first-time homebuyers. During this period, the GSEs' role in the market for first-time African-American and Hispanic homebuyers was only one-third of their role in the overall home loan market (14.3 percent versus 41.5 percent).

FHA, on the other hand, accounted for a much larger share of the minority first-time homebuyer market than it did of the overall homebuyer market. Between 1999 and 2001, FHA insured 46.5 percent of all loans for African-American and Hispanic first-time homebuyers—a market share that was almost three times its overall market share of 16.4 percent.^[308] While FHA's market share was two-fifths of the GSEs' share of the overall home purchase market (16.4 percent versus 41.5 percent), FHA's market share was over three times the GSEs' share of the market for first-time African-American and Hispanic homebuyers (46.5 percent versus 14.3 percent). This finding that the GSEs have played a relatively minor role in the first-time minority market is similar to the conclusion reached by the Fed researchers (see below) and Bunce (2002) that the GSEs have provided little credit support to this underserved borrower group.

The results reported in Table A.26 for the year 2001 suggest some optimism concerning the GSEs' role in the first-time homebuyer market. As explained in earlier sections, both GSEs, but particularly Fannie Mae, improved their affordable lending performance during 2001, at a time when the overall market's performance was slightly declining. This improvement is reflected in the higher first-time market shares for the GSEs during the year 2001, compared with the two previous years, 1999 and 2000 (not reported). The GSEs' share of the market for first-time African-American and Hispanic homebuyers jumped from about 11-12 percent during 1999 and 2000 to 19.7 percent in 2001. Fannie Mae's share of this market almost doubled during this period, rising from 7.0 percent in 1999 to 12.6 percent in 2001. Thus, while the GSEs continue to play a relatively small role in the minority first-time homebuyer market, during 2001 they improved their performance in this area.^[309]

c. The GSEs' Share of the Conventional Conforming, First-time Homebuyer Market

Bunce and Gardner (2004) recently conducted an analysis of first-time homebuyers for the conventional conforming market. The Bunce and Gardner analysis used a similar methodology to the study by Bunce, Neal, and Vandenbroucke of first-time homebuyers in the total mortgage market. Bunce and Gardner restricted their analysis to the funding of first-time homebuyers in the conventional conforming market, which is the market where Fannie Mae and Freddie Mac operate. Their market share results are summarized in Table A.27.

Between 1999 and 2001, the GSEs' purchases accounted for 56.6 percent of all home loans originated in the conventional conforming market of both metropolitan areas and non-metropolitan areas. In other words, Fannie Mae and Freddie Mac funded almost three out of every five homebuyers entering the conventional conforming market between 1999 and 2001. Their purchases of first-time homebuyer loans, on the other hand, accounted for only 39.8 percent of all first-time homebuyer loans originated in that market. Thus, while the GSEs funded approximately two out of every five first-time homebuyers entering the conventional conforming market, their market share (39.8 percent) for first-time homebuyers was only 70 percent of their market share (56.6 percent) for all homebuyers.

As shown in Table A.27, the GSEs have funded an even lower share of the minority first-time homebuyer market. Between 1999 and 2001, the GSEs purchases of African-American and Hispanic first-time homebuyer loans represented 30.9 percent of the conventional conforming market for these loans. Thus, while the GSEs have accounted for 56.6 percent of all home loans in the conventional conforming market, they have accounted for only 30.9 percent of loans originated in that market for African-American and Hispanic first-time homebuyers.

The market share data in Table A.27 show some slight differences between the Freddie Mac and Fannie Mae in serving minority first-time homebuyers. During the 1999-to-2001 period, Freddie Mac's share (11.9 percent) of the African-American and Hispanic first-time homebuyer market was only one-half of its share (24.0 percent) of the home loan market. On the other hand, Fannie Mae's share (19.0 percent) of the African-American and Hispanic first-time homebuyer market was almost 60 percent of its share (32.5 percent) of the home loan market. Thus, while Fannie Mae performance in serving minority first-time homebuyers has been poor, it has been better than Freddie Mac's. This difference in performance between Fannie Mae and Freddie Mac was also seen in the portfolio percentages reported earlier in Table A.17a. Loans for African-American and Hispanic first-time homebuyers accounted for 6.9 percent of Fannie Mae's purchases of home loans between 1999 and 2001, a figure higher than Freddie Mac percentage of 5.3 percent. Loans for African-American and Hispanic first-time homebuyers accounted for 10.2 percent of all home loans originated in the conventional conforming market.

d. Downpayments on Loans Purchased by the GSEs

The level of downpayment can be an important obstacle to young families seeking their first homes. Examining the downpayment characteristics of the mortgages purchased by the GSEs might help explain why they have played a rather limited role in the first-time homebuyer market

Table A.28 reports the loan-to-value (LTV) distribution of home purchase mortgages acquired by the GSEs between 1997 and 2003. In Table A.29, LTV data are provided for the GSEs' purchases of home loans that qualify for the three housing goals—special affordable, low-mod, and underserved areas. Three points stand out.

First, the GSEs (and particularly Fannie Mae) have recently increased their purchases of home loans with low downpayments. After remaining about 4 percent of Fannie Mae's purchases between 1997 and 2000, over-95-percent-LTV loans (or less-than-five-percent downpayment loans) jumped to 7.1 percent during 2001, 7.7 percent in 2002 and 11.5 percent in 2003. It is interesting that this jump in less-than-five-percent downpayment loans occurred in the same years that Fannie Mae improved its purchases of loans for low-income homebuyers, as discussed in earlier sections. As a share of Freddie Mac's purchases, over-95-percent-LTV loans increased from 1.1 percent in 1997 to 5.9 percent in 2000, before falling to 4.3 percent in 2001, 4.8 percent in 2002 and 4.7 percent in 2003. If the low-downpayment definition is expanded to ten percent (i.e., over-90-percent-LTV loans), Freddie Mac had about the same percentage (25 percent) of low-downpayment loans during 2001 as Fannie Mae. In fact, under the more expansive definition, Freddie Mac had the same share of over-90-percent-LTV loans in 2001 as it did in 1997 (about 25 percent), while Fannie Mae exhibited only a modest increase in the share of its purchases with low downpayments (from 23.2 percent in 1997 to 25.4 percent in 2001). The share of over-90-percent-LTV loans in Freddie Mac's purchases fell sharply from 25.0 percent in 2001 to 21.9 percent in 2002 and 19.9 percent in 2003, while the share in Fannie Mae's purchases fell more modestly from 25.4 percent in 2001 to 24.2 percent in 2002 before rebounding to 25.3 percent in 2003.

Second, loans that qualify for the housing goals have lower downpayments than non-qualifying loans. In 2001 and 2002, over-95-percent-LTV loans accounted for about 15 percent of Fannie Mae's purchases of special affordable loans, 13 percent of low-mod loans, and 12 percent of underserved area loans, compared with about 7.5 percent of Fannie Mae's purchases of all home loans. (See Table A.29.) In 2003 these percentages increased to 23, 19 and 19 percent for special affordable, low-mod and underserved areas respectively. These low-downpayment shares for 2001, 2002 and 2003 were double those for 2000 when over-95-percent-LTV loans accounted for 8.4 percent of Fannie Mae's purchases of special affordable loans and about 7 percent of its purchases of low-mod and underserved area loans. Fannie Mae's low-downpayment shares during 2001 were higher than Freddie Mac's shares of 12.3 percent for special affordable loans and about 9 percent for low-mod and underserved area loans. Between 2001 and 2003, Freddie Mac's over-95-percent-LTV shares fell sharply to 3-4 percent for the three housing goal categories, while Fannie Mae's shares increased to the 13-23 percent range. Under the more expansive, over-90-percent-LTV definition, almost one-third of Fannie Mae's goals-qualifying purchases during 2001 would be considered low downpayment, as would a slightly smaller percentage of Freddie Mac's purchases. However, during 2003, Freddie Mac's over-90-percent-LTV shares for the goals-qualifying loans fell to 20-22 percent.

Third, a noticeable pattern among goals-qualifying loans purchased by the GSEs is the predominance of loans with high downpayments. For example, 54.3 percent of special affordable home loans purchased by Freddie Mac during 2003 had a downpayment of at least 20 percent, a percentage not much lower than the high-downpayment share (59.5 percent) of all Freddie Mac's home loan purchases. Similarly, 49.8 percent of the home loans purchased by Fannie Mae in underserved areas during 2003 had a twenty percent or higher downpayment, compared with 54.6 percent of all home loans purchased by Fannie Mae.

Thus, the data in Tables A.28 and A.29 show a preponderance of high downpayment loans, even among lower-income borrowers who qualify for the housing goals. The past focus of the GSEs on high-downpayment loans provides some insight into a study by staff at the Federal Reserve Board who found that the GSEs have offered little credit support to the lower end of the mortgage market.^[310] The fact that approximately half of the goals-qualifying loans purchased by the GSEs have a downpayment of over twenty percent is also consistent with findings reported earlier concerning the GSEs' minimal service to first-time homebuyers, who experience the most problems raising cash for a downpayment. On the other hand, the recent experience of Fannie Mae suggests that purchasing low-downpayment loans may be one technique for reaching out and funding low-income and minority families who are seeking to buy their first home.

13. Other Studies of the GSEs' Performance Relative to the Market

This section summarizes briefly the main findings from other studies of the GSEs' affordable housing performance. These include studies by the HUD and the GSEs as well as studies by academics and research organizations.

Freeman and Galster Study.^[311] A recent study by Lance Freeman and George Galster uses econometric analysis to test whether the Government-Sponsored Enterprises (GSEs) Fannie Mae and Freddie Mac purchases of home mortgages in neighborhoods traditionally underserved by financial institutions stimulate housing market activity in those neighborhoods. Specifically, this study analyzes data of single-family home sales volumes and prices of mortgages originated from 1993-1999 in Cleveland, OH.

The study concludes that aggressive secondary market purchasing behavior by non-GSE entities stimulated sales volumes and prices of homes in low-income and predominantly minority-occupied neighborhoods of Cleveland. The study results also showed a positive relationship between home transaction activity and the actions of the secondary mortgage market, and concludes that the secondary mortgage market (and the non-GSE sector in particular) purchases of mortgages had a positive effect on the number of sales transactions one year later. However, the study also concludes that although non-GSE purchases of non-home purchase mortgages appeared to boost prices one and two years later, no consistent impacts of purchasing rates on sales prices could be observed. In addition, there was no robust evidence that GSE purchasing rates were positively associated with single-family home transactions volumes or sales prices during any periods.

Urban Institute Rural Markets Study.^[312] A study by Jeanette Bradley, Noah Sawyer, and Kenneth Temkin uses both quantitative and qualitative data to explore the issue of GSE service to rural areas. The study first summarizes the existing research on rural lending and GSE service to rural areas. It then reviews the current underwriting guidelines of Fannie Mae, Freddie Mac, the USDA Rural Housing Service, and Farmer Mac, focusing on issues relevant to rural underwriting. The GSE public-use database is used to analyze GSE non-metro loan purchasing patterns from 1993-2000. Finally, the study presents the results of a series of discussions conducted with key national industry and lender experts and local experts in three rural sites in south-central Indiana, southwestern New Mexico and southern New Hampshire chosen for the diversity of their region, population, economic structures, and housing markets.

The authors of the study conclude that while Fannie Mae and Freddie Mac have increased their lending to rural areas since 1993, their non-metro loan purchases still lag behind their role in metro loan purchases, particularly in regard to the percentage of affordable loans. From the discussions with experts, the authors of the study make the following policy recommendations: Underserved populations and rural areas should be specifically targeted at the census-tract level; HUD should set manufactured housing goals; HUD should consider implementing a survey of small rural lenders or setting up an advisory group of small rural lenders in order to determine their suggestions for creating stronger relationships between the GSEs and rural lenders with the goal of increasing GSE non-metro purchase rates.

Urban Institute GSE Impacts Study.^[313] A report by Thomas Thibodeau, Brent Ambrose, and Kenneth Temkin analyzes the extent to which the GSEs' responses to the Federal Housing Enterprises Financial Safety and Soundness Act's (FHEFSSA) affordable housing goals have had their intended effect of making low- and moderate-income families better off. Specifically the report examines several methodologies determining that the conceptual model created by Van Order in 1996 ^[314] provided the most complete description of how the primary and secondary markets interact. This model was then applied in a narrow scope to capital market outcomes which included GSE market shares and effective borrowing costs, and housing market outcomes that include low- and moderate-income homeownership rates. Finally, metropolitan American Housing Survey (AHS) data for eight cities were used to conduct empirical analyses of the two categories of outcomes. These cities included areas surveyed in 1992, the year before HUD adopted the affordable housing goals, to provide the baseline for the analysis. Four metropolitan areas were surveyed in 1992 and again in 1996: Cleveland, Indianapolis, Memphis and Oklahoma City. Four cities were surveyed in 1992 and again in 1998: Birmingham, Norfolk, Providence and Salt Lake City.

The study's empirical analysis suggests that the GSE affordable goals have helped to make homeownership more attainable for target families. The assessment of the effects of the affordable goals on capital markets showed that the GSE share of the conventional conforming market has increased, especially for lower income borrowers and neighborhoods. The study also concludes that the affordable housing goals have an impact on the purchase decisions of Fannie Mae and Freddie Mac. The study also finds that interest rates are lower in markets in which Fannie Mae and Freddie Mac purchase a higher proportion of conventional loans. Finally, the study's analysis shows that overall lending volume in a metropolitan area increases when the GSEs purchase seasoned loans.

Specifically, that homeownership rates increased at a faster rate for low-income families when compared to all families, and that in a subset of MSAs, minority homeownership rates also grew faster when compared to overall homeownership changes in those MSAs.

Finally, the affordable housing goal effects were examined for 80 MSAs in relation to the homeownership rate changes between 1991 and 1997. The study found that the GSEs, by purchasing loans originated to low-income families, helped to reduce the disparity between homeownership rates for lower and higher income families, suggesting that the liquidity created when the GSEs purchase loans originated to low-income families is recycled into more lending targeted to lower income homebuyers.

The authors of the study qualify their results by stating that they are based on available data that does not provide the level of detail necessary to conduct a fully controlled national assessment.

Williams and Bond Study.^[315] Richard Williams and Carolyn Bond examine GSE leadership of the mortgage finance industry in making credit available for low- and moderate-income families. Specifically, it asks if the GSEs are doing relatively more of their business with underserved markets than other financial institutions, and whether the GSEs' leadership helps to narrow the gap in home mortgage lending that exists between served and underserved markets. The study uses HMDA data for metropolitan areas and the Public Use Data Base at HUD for compilations of GSE data sets for the entire nation (GSE PUDB File B) to conduct descriptive and multivariate analyses of nationwide lending between 1993 and 2000. Additionally, separate analyses are conducted that include and exclude loans from subprime and manufactured housing lenders.

The study concludes that the GSEs are not leading: They do not purchase relatively more underserved market loans than the primary market makes nor do they purchase as many of these loans as their secondary market competitors. Additionally, the study concludes that the disparities between the GSEs and the primary market are even greater once the growing role of subprime and manufactured housing is considered. The authors admit that there have been signs of progress, particularly in 1999 and 2000 when primary market lending to underserved markets increased and GSE purchases of underserved market loans increased even faster. Regardless, the study concludes that there continues to be significant racial, economic, and geographic disparities in the way that the benefits of GSE activities are distributed and that the benefits of GSE activities still go disproportionately to members of served rather than underserved markets.

14. The GSEs' Support of the Mortgage Market for Single-Family Rental Properties

The 1996 Property Owners and Managers Survey reported that 49 percent of rental units are found in the “mom and pop shops” of the rental market—“single-family” rental properties, containing 1-4 units. These small properties are largely individually-owned and managed, and in many cases the owner-managers live in one of the units in the property. They include many properties in older cities, in need of financing for rehabilitation. Single-family rental units play an especially important role in lower-income housing, over half of such units are affordable to very low-income families.

There is not, however, a strong secondary market for single-family rental mortgages. While single-family rental properties comprise a large segment of the rental stock for lower-income families, they make up a small portion of the GSEs' business. Between 1999 and 2002, single-family rental properties accounted for only 7.6 percent of total (both single-family and multifamily) units financed by the GSEs during this period. It follows that since single-family rentals make up such a small part of the GSEs business, they have not penetrated the single-family rental market to the same degree that they have penetrated the owner-occupant market. Table A.30 below shows that between 1999 and 2002, the GSEs financed 61 percent of owner-occupied dwelling units in the conventional conforming market, but only 40 percent of single-family rental units.

There are a number of factors that have limited the development of the secondary market for single-family rental property mortgages thus explaining the lack of penetration by the GSEs. Little is collectively known about these properties as a result of the wide spatial dispersion of properties and owners, as well as a wide diversity of characteristics across properties and individuality of owners. This makes it difficult for lenders to properly evaluate the probability of default and severity of loss for these properties.

Single-family rental properties could be important for the GSEs housing goals, especially for meeting the needs of lower-income families. Between 1999 and 2002, 87 percent of the GSEs' single-family rental units qualified for the Low- and Moderate-Income Goal, compared with 40 percent of one-family owner-occupied properties. (See Table A.30.) This heavy focus on lower-income families meant that single-family rental properties accounted for 14 percent of the units qualifying for the Low- and Moderate-Income Goal, even though they accounted for 7.6 percent of the total units (single-family and multifamily) financed by the GSEs.

Given the large size of this market, the high percentage of these units which qualify for the GSEs' housing goals, and the weakness of the secondary market for mortgages on these properties, an enhanced presence by Fannie Mae and Freddie Mac in the single-family rental mortgage market would seem warranted.^[316] Single-family rental housing is an important part of the housing stock because it is an important source of housing for lower-income households.

Despite the size and importance of single-family rental properties for low-income people, HUD received several comments advocating exclusion of single-family rentals from goals consideration. These commenters pointed out that single-family owner-occupiers often maintain their properties more effectively than single-family absentee landlords or their tenants. HUD was asked to exclude single-family investor owned properties to reduce these neighborhood effects.

Community associations raise an important issue for neighborhood development. However, they do not address the question of effective goals promotion for all segments of the housing market. They compare maintenance by owner-occupiers to maintenance by investors in the single-family market. This does not address the housing outcomes for tenants with access to single-family rental compared to tenants in multifamily rental. With nearly half of rental units in older cities composed of smaller single-family units, denial of goals eligibility for single-family investors would exclude a substantial proportion of housing units available to low income people.

Furthermore, single-family investors provide additional market benefits to the housing system. The whole structure of the GSEs provides liquidity to the housing market by allowing investors additional channels to fund mortgages. The question is not always between single-family investors and single-family owner-occupiers. Sometimes, the question is between a single-family investor and a property unable to be sold or even abandoned. Although the goals strongly support home ownership for low-income neighborhoods, investors in single-family properties also play an important role.

F. Factor 4: Size of the Conventional Conforming Mortgage Market Serving Low- and Moderate-Income Families Relative to the Overall Conventional Conforming Market

The Department estimates that dwelling units serving low- and moderate-income families will account for 51-56 percent of total units financed in the overall conventional conforming mortgage market during 2005-2008, the period for which the Low- and Moderate-Income Housing Goal will be effective. The market estimates exclude B&C loans and allow for much more adverse economic and market affordability conditions than have existed recently. The detailed analyses underlying these estimates are presented in Appendix D.

G. Factor 5: GSEs' Ability to Lead the Industry

FHEFSSA requires the Secretary, in determining the Low- and Moderate-Income Housing Goal, to consider the GSEs' ability to “lead the industry in making mortgage credit available for low- and moderate-income families.” Congress indicated that this goal should “steer the enterprises toward the development of an increased capacity and commitment to serve this segment of the housing market” and that it “fully expect[ed] [that] the enterprises will need to stretch their efforts to achieve [these goals].”^[317]

The Department and independent researchers have published numerous studies examining whether or not the GSEs have been leading the single-family market in terms of their affordable lending performance. This research, which is summarized in Section E, concludes that the GSEs have generally lagged behind primary lenders in funding first-time homebuyers, lower-income borrowers and underserved communities, although Fannie Mae's recent performance has placed it ahead of the special affordable and low-mod markets for single-family-owner loans. As required by FHEFSSA, the Department has produced estimates of the portion of the total (single-family and multifamily) mortgage market that qualifies for each of the three housing goals (see Appendix D). Congress intended that the Department use these market estimates as one factor in setting the percentage target for each of the housing goals. The Department's estimate for the size of the Low- and Moderate-Income market is 51-56 percent, which is higher than the GSEs' performance on that goal.

This section provides another perspective on the GSEs' performance by examining the share of the total conventional conforming mortgage market and the share of the goal-qualifying markets (low-mod, special affordable, and underserved areas) accounted for by the GSEs' purchases. This analysis, which is conducted by product type (single-family owner, single-family rental, and multifamily), shows the relative importance of the GSEs in each of the goal-qualifying markets.

1. GSEs' Role in Major Sectors of the Mortgage Market

Tables A.30 and A.31a compare GSE mortgage purchases with HUD's estimates of the numbers of units financed in the conventional conforming market. Table A.30 presents aggregate data for 1999-2002 while Table A.31a presents more summary market share data for individual years 2000, 2001 and 2002.^[318] (As explained below, Tables A.31b and A.31c repeat this information but for lower multifamily shares of the mortgage market.) HUD estimates that there were 47,551,039 owner and rental units financed by new conventional conforming mortgages between 1999 and 2002. Fannie Mae's and Freddie Mac's mortgage purchases financed 26,118,927 of these dwelling units, or 55 percent of all dwelling units financed. As shown in Table A.30, the GSEs have played a smaller role in the goals-qualifying markets than they have played in the overall market. Between 1999 and 2002, new mortgages were originated for 26,051,771 dwelling units that qualified for the Low- and Moderate-Income Goal; the GSEs low-mod purchases financed 12,608,215 dwelling units, or 48 percent of the low-mod market. Similarly, the GSEs' purchases accounted for 48 percent of the underserved areas market, but only 41 percent of the special affordable market. Obviously, the GSEs did not lead the industry during this period in financing units that qualify for the three housing goals. They need to improve their performance and it appears that there is ample room in the non-GSE portions of the goals-qualifying markets for them to do so. For instance, the GSEs were not involved in three-fifths of the special affordable market during the 1999-to-2002 period.

While the GSEs are free to meet the Department's goals in any manner that they deem appropriate, it is useful to consider their performance relative to the industry by property type. The GSEs accounted for 61 percent of the single-family owner market but only 35 percent of the multifamily market and 40 percent of the single-family rental market (or a combined 37 percent share of the rental market).

Single-Family Owner Market. As stated in the 2000 Rule, the single-family-owner market is the bread-and-butter of the GSEs' business, and based on the financial and other factors discussed below, the GSEs clearly have the ability to lead the primary market in providing credit for low- and moderate-income owners of single-family properties. However, the GSEs have historically lagged behind the market in funding single-family-owner loans that qualify for the housing goals and, as discussed in Section E, they have played a rather small role in funding minority first-time homebuyers. The market share data reported in Table A.30 for the single-family-owner market tell the same story. The GSEs' purchases of single-family-owner loans represented 61 percent of all single-family-owner loans originated between 1999 and 2002, compared with 57 percent of the low-mod loans that were originated, 55 percent of underserved area loans, and 52 percent of the special affordable loans.

The data in Table A.31a indicate the GSEs' growing market share during the heavy refinance years of 2001 and 2002. For example, the GSEs accounted for 74 percent of the overall single-family-owner market in 2002, and 67-69 percent of the markets covered by the three housing goal categories. While this improvement is an encouraging trend, there are ample opportunities for the GSEs to continue their improvement. Almost one-third of the goals-qualifying loans originated during 2002 remained available to the GSEs to purchase; there are clearly affordable loans being originated that the GSEs can purchase. Furthermore, the GSEs' purchases under the housing goals are not limited to new mortgages that are originated in the current calendar year. The GSEs can purchase loans from the substantial, existing stock of affordable loans held in lenders' portfolios, after these loans have seasoned and the GSEs have had the opportunity to observe their payment performance. In fact, based on Fannie Mae's recent experience, the purchase of seasoned loans appears to be one effective strategy for purchasing goals-qualifying loans.

The data in Table A.31a show a strong upward trend from 2000 and 2001 to 2002 in the GSE share of the single-family-owner market. Their share of 2000 financed units in the conventional conforming market totaled 48 percent. This increased to 55 percent in 2001 then to 74 percent in 2002. The large increase in 2002 can be attributed to the relatively low interest rates and heavy refinancing activity in 2003. During such a period, the share of fixed rate mortgage originations increases relative to adjustable rate mortgages. Due to the higher risk associated with fixed rate mortgages, less thrift institutions are willing to hold them, and, thus, more are sold to the GSEs. As a result, during low interest rate periods, the GSE share of mortgages increases.

Single-Family Rental Market. Single-family rental housing is a major source of low-income housing. As discussed in Appendix D, data on the size of the primary market for mortgages on these properties is limited, but available information indicate that the GSEs are much less active in this market than in the single-family owner market. HUD estimates that GSE purchases between 1999 and 2002 totaled only 40 percent of all newly-mortgaged single-family rental units that were affordable to low- and moderate-income families.

As explained in the 2000 Rule, many of these properties are “mom-and-pop” operations, which may not follow financing procedures consistent with the GSEs' guidelines. Much of the financing needed in this area is for rehabilitation loans on 2-4 unit properties in older areas, a market in which the GSEs' have not played a major role. However, this sector could certainly benefit from an enhanced role by the GSEs, and the data in Table A.30 indicate that there is room for such an enhanced role, as approximately two-thirds of this market remains for the GSEs to enter.

Once again, Table A.31a shows a large increase in the GSE share of newly-mortgaged units financed in 2002 compared to those financed in 2000 and 2001. As described above for the single-family owner market, this large increase is due to the large share of fixed-rate mortgages, compared to adjustable rate mortgages, originated during 2002.

Multifamily Market. Fannie Mae is the largest single source of multifamily finance in the United States, and Freddie Mac has made a solid reentry into this market over the last nine years. However, there are a number of measures by which the GSEs lag the multifamily market. For example, the share of GSE resources committed to the multifamily purchases falls short of the multifamily proportion prevailing in the overall mortgage market. HUD estimates that newly-mortgaged units in multifamily properties represented almost 15 percent of all (single-family and multifamily) dwelling units financed between 1999 and 2002.^[319] As shown in Table A.30, multifamily acquisitions represented 9.5 percent of dwelling units financed by the GSEs between 1999 and 2002.

The GSEs' role in the multifamily market is significantly smaller than in single-family. As shown in Table A.30, GSE purchases have accounted for 35 percent of newly financed multifamily units between 1999 and 2002—a market share much lower than their 61 percent share of the single-family-owner market. Stated in terms of portfolio shares, single-family-owner loans accounted for 83 percent of all dwelling units financed by the GSEs during this period, versus 75 percent of all units financed in the conventional conforming market.

While it is recognized that the GSEs have been increasing their multifamily purchases, a further enlargement of their role in the multifamily market seems feasible and appropriate, particularly in the affordable (lower rent) end of the market. As noted in Section D.3, market participants believe that the GSEs have been conservative in their approaches to affordable multifamily lending and underwriting.^[320] Certainly the GSEs face a number of challenges in better meeting the needs of the affordable multifamily market. For example, thrifts and other depository institutions may sometimes retain their best loans in portfolio, and the resulting information asymmetries may act as an impediment to expanded secondary market transaction volume.^[321] However, the GSEs have demonstrated that they have the depth of expertise and the financial resources to devise innovative solutions to problems in the multifamily market. The GSEs can build on their recent records of increased multifamily lending and innovative products to make further in-roads into the affordable market. As explained in Section D.3, the GSEs have the expertise and market presence to push simultaneously for market standardization and for programmatic flexibility to meet the special needs and circumstances of the lower-income portion of the multifamily market.

As discussed in Appendix D, the GSEs questioned HUD's historical estimates of the multifamily market as too high. Section C of Appendix D discusses these comments and responds. As indicated in Table A.30, multifamily loans accounted for 14.8 percent of all financed units in the market, excluding B&C loans. As reported in Appendix D, HUD also conducted sensitivity analyses that reduced its 1999-2002 multifamily shares for the market by approximately two percentage points. The results for these lower multifamily market shares are reported in Table A.31b (1999-2002 aggregate results) and Table A.31c (2000-2002 individual year results). In this case, 1999-2002 multifamily units decreased from 7,018,044 units to 5,991,036 units (reducing the multifamily share from 14.8 percent to 12.9 percent). With these reduced multifamily market numbers, the GSEs' share of the multifamily market increased from 35 percent to 41 percent. The GSEs also accounted for higher shares of the goals-qualifying multifamily market: 42 percent for low-mod units, 34 percent for underserved area units, and 37 percent for special affordable units. In this case, the GSEs' shares of the overall goals-qualifying markets increased as follows: low-mod—from 48 percent (see right column of Table A.30) to 50 percent (see right column of Table A.31b); underserved areas—from 48 percent to 49 percent; and special affordable—from 41 percent to 43 percent.

Conclusions. While HUD recognizes that some segments of the market may be more challenging for the GSEs than others, the data reported in Table A.30 and Tables A.31a-c show that the GSEs have ample opportunities to purchase goals-qualifying mortgages. Furthermore, if a GSE makes a business decision to not pursue certain types of goals-qualifying loans in one segment of the market, they are free to pursue goals-qualifying owner and rental property mortgages in other segments of the market. As market leaders, the GSEs should be looking for innovative ways to pursue this business. Furthermore, there is evidence that the GSEs can earn reasonable returns on their goals business. The Regulatory Analysis that accompanies this final rule provides evidence that the GSEs can earn financial returns on their purchases of goals-qualifying loans that are only slightly below their return on equity from their normal business.

2. Qualitative Dimensions of the GSEs' Ability to Lead the Industry

This section discusses several qualitative factors that are indicators of the GSEs' ability to lead the industry in affordable lending. It discusses the GSEs' role in the mortgage market; their ability, through their underwriting standards, new programs, and innovative products, to influence the types of loans made by private lenders; their development and utilization of state-of-the-art technology; the competence, expertise and training of their staffs; and their financial resources.

a. Role in the Mortgage Market

The GSEs have played a dominant role in the single-family mortgage market. As reported in Section C.3, mortgage purchases by the GSEs reached extraordinary levels in 2001-2003. Purchases by Fannie Mae stood at $568 billion in 2001 and $848 billion in 2002. Freddie Mac's single-family mortgage purchases were $393 billion in 2001 and $475 billion in 2002. The Office of Federal Housing Enterprise Oversight (OFHEO) estimates that the GSEs purchased 40 percent of newly-originated conventional mortgages in 2001. Total GSE purchases, including loans originated in prior years, amounted to 46 percent of conventional originations in 2001.

The dominant position of the GSEs in the mortgage market is reinforced by their relationships with other market institutions. Commercial banks, mutual savings banks, and savings and loans are their competitors as well as their customers—they compete to the extent they hold mortgages in portfolio, but at the same time they sell mortgages to the GSEs. They also buy mortgage-backed securities, as well as the debt securities used to finance the GSEs' portfolios. Mortgage bankers sell virtually all of their prime conventional conforming loans to the GSEs. Private mortgage insurers are closely linked to the GSEs, because mortgages purchased by the enterprises that have loan-to-value ratios in excess of 80 percent are normally required to be covered by private mortgage insurance, in accordance with the GSEs' charter acts.

b. Underwriting Standards for the Primary Mortgage Market

The GSEs' underwriting guidelines are followed by virtually all originators of prime mortgages, including lenders who do not sell many of their mortgages to Fannie Mae or Freddie Mac. The guidelines are also commonly followed in underwriting “jumbo” mortgages, which exceed the maximum principal amount which can be purchased by the GSEs (the conforming loan limit)—such mortgages eventually might be sold to the GSEs, as the principal balance is amortized or when the conforming loan limit is otherwise increased. Changes that the GSEs have made to their underwriting standards in order to address the unique needs of low-income families were discussed in Section C.4 of this Appendix. The GSEs' market influence is one reason these new, more flexible underwriting standards have spread throughout the market. Because the GSEs' guidelines set the credit standards against which the mortgage applications of lower-income families are judged, the enterprises have a profound influence on the rate at which mortgage funds flow to low- and moderate-income borrowers and underserved neighborhoods.

As discussed below, the GSEs' new automated underwriting systems are widely used to originate mortgages in today's market. As discussed in Sections C.7 and C.8, the GSEs have started adapting their underwriting systems for subprime loans and other loans that have not met their traditional underwriting standards.

c. State-of-the-Art Technology

Both GSEs are in the forefront of new developments in mortgage industry technology. Automated underwriting and online mortgage processing are a couple of the new technologies that have impacted the mortgage market, expanding homeownership opportunities. This section provides an overview of these new technologies and the extent of their use.

Each enterprise released an automated underwriting system in 1995—Freddie Mac's “Loan Prospector” (LP) and Fannie Mae's “Desktop Underwriter” (DU). During 2001 and 2002, roughly 60 percent of all newly-originated mortgages the GSEs purchased were processed through these systems. Lenders and brokers used LP to evaluate 7.3 million loan applications in 2001, 8.2 million in 2002,^[322] and 9.5 million in 2003. Similarly, DU was used to evaluate 8 million loans in 2001, over 10 million in 2002, and 14.8 million loans in 2003. The GSEs' systems have also been adapted for FHA and jumbo loans. Automated underwriting systems are being further adapted to facilitate risk-based pricing, which enables mortgage lenders to offer each borrower an individual rate based on his or her risk. As discussed earlier, concerns about the use of automated underwriting and risk-based pricing include the disparate impact on minorities and low-income borrowers and the “black box” nature of the score algorithm.

The GSEs are using their state-of-the-art technology in certain ways to help expand homeownership opportunities. For example, Fannie Mae has developed Fannie Mae Property GeoCoder a computerized mapping service offered to lenders, nonprofit organizations, and state and local governments to help them determine whether a property is located in an area that qualifies for Fannie Mae's community lending products designed to increase homeownership and revitalization in traditionally underserved areas. In addition, eFannieMae.com is Fannie Mae's business-to-business Web site where lenders can access product information and important technology tools, view upcoming events, and receive news about training opportunities. This site receives on average 80,000 visitors per week.^[323] Freddie Mac has introduced in recent years Internet-based debt auctions, debt repurchase operations, and debt exchanges. These mechanisms benefit investors by providing more uniform pricing, greater transparency and faster price discovery—all of which makes Freddie Mac debt more attractive to investors and reduces the cost of funding mortgages.^[324] In addition, Freddie Mac has provided automated tools for lenders to identify and work with borrowers most likely to encounter problems making their mortgage payments. EarlyIndicator has become the industry standard for default management technology. It can reduce the consequences of mortgage delinquency for borrowers, servicers and investors.^[325]

The GSEs are also expanding homeownership opportunities through the use of the Internet in processing mortgage originations. New online mortgage originations reached $267.6 billion in the first half of 2002, compared with $97 billion for the first six months of 2001. The 2002 six-month volume comprised 26.5 percent of the estimated $1.01 trillion in total mortgage originations for the same time period.^[326] Freddie Mac made Loan Prospector on the Internet service available to lenders for their retail operations. Freddie Mac also adopted the mortgage industry's XML (extensible markup language) data standard, which is integral to streamlining and simplifying Internet-based transactions. In addition, Congress enacted legislation that allows the use of electronic signature in contracts in 2001, making a completely electronic mortgage transaction possible. With the use of electronic signatures, electronic mortgages are expected to improve the mortgage process, further reducing origination and servicing costs. In October 2000, Freddie Mac purchased its first electronic mortgage under the new law.

The GSEs also offer a variety of other online tools and applications that have the potential to make the mortgage loan process more cost effective and efficient for lenders. Freddie Mac, for example, has launched dontborrowtrouble.com, which contains information on local anti-predatory lending campaigns, consumer tips on avoiding predatory lending, and information on how to start a local campaign and obtain additional resources.^[327] Fannie Mae offers “HomeBuyer Funds Finder,” a one-stop online resource designed for lenders and other housing professionals, enables users to access a database of local housing subsidy programs available for low- and moderate-income borrowers. In 2002, the HomeBuyer Funds Finder Web site received over 24,500 hits.^[328] “Home Counselor Online” provides homeownership counselors with the necessary tools to help consumers financially prepare to purchase a home. In 2003, 641 counselors representing over 2,000 organizations used Home Counselor Online.^[329] “True Cost Calculator 2.0” is designed to help homebuyers make informed home purchase decisions by helping them compare loan products and prices. Over 60 Fannie Mae partners officer the True Cost Calculator through their Web sites and a Spanish version is also available on Univision.com.^[330] A more complete list of Fannie Mae's online tool and applications can be found in its Annual Housing Activities Report. In 2002, Fannie Mae's total eBusiness volume was $1.1 trillion, up from $800 billion in 2000.^[331]

d. Staff Resources

Both Fannie Mae and Freddie Mac are well-known throughout the mortgage industry for the expertise of their staffs in carrying out their current programs, conducting basic and applied research regarding mortgage markets, developing innovative new programs, and undertaking sophisticated analyses that may lead to new programs in the future. The leaders of these corporations frequently testify before Congressional committees on a wide range of housing issues, and both GSEs have developed extensive working relationships with a broad spectrum of mortgage market participants, including various nonprofit groups, academics, and government housing authorities. Federal agencies and foreign governments and businesses seek them out for advice and consultation because of their expertise. The role that the GSEs have played in spreading the use of technology throughout the mortgage market reflects the enormous expertise of their staff.

e. Financial Strength

Fannie Mae. The benefits that accrue to the GSEs because of their GSE status, as well as their solid management, have made them two of the nation's most profitable businesses. Fannie Mae's net income was $3.9 billion in 1999, $4.4 billion in 2000, $5.9 billion in 2001, $4.6 billion in 2002,^[332] and $7.9 billion in 2003.^[333] Fannie Mae's return on equity averaged 24.0 percent over the 1995-99 period—far above the rates achieved by most financial corporations. Fannie Mae's return on equity was 26.0 percent in 2003, while this represented no change from 2002, it was an increase of 3 percent over 2001.^[334] In 2003, Fannie Mae's total stockholders' equity increased by 37% to $22.373 million, core business earnings grew by 14 percent ($7.3 billion), credit losses increased by $42 million to $111 million with the resulting credit loss ratio at .006% (represents credit losses divided by average single family mortgage credit book of business) and taxable equivalent revenues grew by 24 percent.^[335]

Fannie Mae's basic net earnings per common share increased from $3.75 in 1999 to $7.93 in 2003, dividends per common share have increased from $.96 in 1998 to $1.68 in 2003, a 27% increase over 2002, and operating earnings per diluted common share increased from 2002 to 2003 by 71% to $7.72.^[336]

Freddie Mac. Freddie Mac has shown similar trends. Freddie Mac's net income was $3.158 billion in 2001, $10.090 billion in 2002, and $4.891 billion in 2003, and total stockholder's equity increased by 10% over 2002 to $31.562 billion. Freddie Mac's return on equity averaged 23.4 percent over the 1995-1999 period, also well above the rates achieved by most financial corporations. Credit losses increased by $8 million to $82 million with the resulting credit loss ratio at 0.7 (represents annualized credit losses divided by average total mortgage portfolio). Basic earnings per common share (after cumulative effect of change in accounting principles, net of taxes) was $4.25 in 2001, $14.23 in 2002 and $6.80 in 2003. Dividends per common share have increased from 0.80 in 2001 to $1.04 in 2003, an 18% increase over 2002, and operating earnings per diluted common share (after cumulative effect of change in accounting principles, net of taxes) decreased from 2002 to 2003 by $7.39 to $6.79.^[337]

Other Indicators. Additional indicators of the strength of the GSEs are provided by various rankings of American corporations. Business Week has reported that among Standard & Poor's performance ranking of 500 companies in 2004, Fannie Mae was ranked 117, down from 91 in 2003 and Freddie Mac was listed as “INC” for 2004 and 16th for 2003. Additionally, Fannie Mae was ranked as 29th in overall market value, 17th in sales and 9th in profits, and Freddie Mac was ranked 59th in market value and “NR” in sales and profits.^[338] According to Fortune's annual listing of the 500 largest U.S. Corporations, Fannie Mae was ranked 20th in 2003, down from 16th in 2002, and Freddie Mac was “displaced” from the ranking in 2003, but ranked 32nd in 2002. Additionally, Fannie Mae ranked 11th for most profitable companies, 3rd for revenues per employee, and in the “Diversified Financials” category, they ranked 2nd out of 12 companies.^[339] And, according to Fortune's Global 500 listing of the world's largest corporations, Fannie Mae ranked 56th in 2003, (ranking 17th in highest profits) down from 45th in 2002, and Freddie Mac ranked 104th in 2003, down from 90th in 2002.^[340]

f. Conclusion About Leading the Industry

In light of these considerations, the Secretary has determined that the GSEs have the ability to lead the industry in making mortgage credit available for low- and moderate-income families.

H. Factor 6: The Need to Maintain the Sound Financial Condition of the GSEs

HUD has undertaken a separate, detailed economic analysis of this final rule, which includes consideration of (a) the financial returns that the GSEs earn on low- and moderate-income loans and (b) the financial safety and soundness implications of the housing goals. Based on this economic analysis and the Office of Federal Housing Enterprise Oversight review, HUD concludes that the goals raise minimal, if any, safety and soundness concerns.

I. Determination of the Low- and Moderate-Income Housing Goals

The annual goal for each GSE's purchases of mortgages financing housing for low- and moderate-income families is being established at 52 percent of eligible units financed in each of calendar years 2005, 53 percent in 2006, 55 percent in 2007, and 56 percent in 2008. This goal will remain in effect thereafter, unless changed by the Secretary prior to that time. In addition, a low- and moderate-income subgoal of 45 percent in 2005, 46 percent in 2006, and 47 percent in both 2007 and 2008 is being established for the GSEs' acquisitions of single-family-owner home purchase loans in metropolitan areas. This subgoal is designed to encourage the GSEs to lead the primary market in offering homeownership opportunities to low- and moderate-income families. The Secretary's consideration of the six statutory factors that led to the choice of these goals is summarized in this section.

1. Housing Needs and Demographic Conditions

Affordability Problems. Data from the 2000 Census and the American Housing Surveys demonstrate that there are substantial housing needs among low- and moderate-income families. Many of these households are burdened by high homeownership costs or rent payments and will likely continue to face serious housing problems. There is evidence of persistent housing problems for Americans with the lowest incomes. Since 1977, the percentage of U.S. households with worst case needs has hovered around five percent, with the worst year being 1983 (6.03 percent) and the best year being 1999 (4.72 percent). The proportion in 2001 was 4.77 percent, which is not significantly different from the 1999 figure. HUD's analysis of American Housing Survey data reveals that, in 2001, 5.1 million unassisted very-low income renter households had “worst-case” housing needs, defined as housing costs greater than 50 percent of household income or severely inadequate housing. Among these households, 90 percent had a severe rent burden, 6 percent lived in severely inadequate housing, and 4 percent suffered from both problems. Among the 34 million renters in all income categories, 6.3 million (19 percent) had a severe rent burden and over one million renters (3 percent) lived in housing that was severely inadequate.

Demographic Trends. Changing population demographics will result in a need for the primary and secondary mortgage markets to meet nontraditional credit needs, respond to diverse housing preferences and overcome information and other barriers that many immigrants and minorities face. It is projected that there will be 1.2 million new households each year over the next decade. The aging of the baby-boom generation and the entry of the baby-bust generation into prime home buying age will have a dampening effect on housing demand. However, the continued influx of immigrants will increase the demand for rental housing, while those who immigrated during the 1980s and 1990s will be in the market for owner-occupied housing. Immigrants and other minorities—who accounted for nearly 40 percent of the growth in the nation's homeownership rate over the past five years—will be responsible for almost two-thirds of the growth in the number of new households over the next ten years. Non-traditional households have become more important, as overall household formation rates have slowed. With later marriages, divorce, and non-traditional living arrangements, the fastest growing household groups have been single-parent and single-person households. As these demographic factors play out, the overall effect on housing demand will likely be sustained growth and an increasingly diverse household population from which to draw new renters and homeowners. According to the National Association of Homebuilders, annual housing demand will average from 1.84 to 2.19 million units over the next decade.^[341]

Growth in Single-Family Affordable Lending. Many younger, minority and lower-income families did not become homeowners during the 1980s due to the slow growth of earnings, high real interest rates, and continued house price increases. Over the past ten years, economic expansion, accompanied by low interest rates and increased outreach on the part of the mortgage industry, has improved affordability conditions for these families. As this appendix has explained, there has been a “revolution in affordable lending” that has extended homeownership opportunities to historically underserved households. The mortgage industry has offered more customized mortgage products, more flexible underwriting, and expanded outreach to low-income and minority borrowers. Fannie Mae and Freddie Mac have been a big part of this “revolution in affordable lending”. HMDA data suggest that the industry and GSE initiatives are increasing the flow of credit to underserved borrowers. Between 1993 and 2003, conventional loans to low-income and minority families increased at much faster rates than loans to upper-income and non-minority families. Thus, the 1990s and the early part of the current decade have seen the development of a strong affordable lending market.

Disparities in Housing and Mortgage Markets. Despite this strong growth in affordable lending, serious disparities in the nation's housing and mortgage markets remain. The homeownership rate for African-American and Hispanic households is about 25 percentage points below that of white households. In addition to low income, barriers to homeownership that disproportionately affect minorities and immigrants include: lack of capital for down payment and closing costs; poor credit history; lack of access to mainstream lenders; little understanding of the homebuying process; and continued discrimination in housing markets and mortgage lending. With respect to the latter, a recent HUD-sponsored study of discrimination in the rental and owner markets found that while differential treatment between minority and white home seekers had declined over the past ten years, it continued at an unacceptable level in the year 2000. In addition, disparities in mortgage lending continued across the nation in 2003, when the loan denial rate for African-American applicants was almost three times that for white applicants, even after controlling for income of the applicant. HUD studies also show that African-Americans and Hispanics are subject to discriminatory treatment during the pre-qualification process of applying for a mortgage.

Single-Family Mortgage Market. Heavy refinancing due to low interest rates increased single-family mortgage originations to record levels during 2001-2003. Demographic forces, industry outreach, and low interest rates also kept lending for home purchase at record levels as well. As noted above, the potential homeowner population over the next decade will be highly diverse, as growing demand from immigrants and minorities are expected to sustain the home purchase market, as our population ages. Single-family housing starts are expected to continue in the 1.65-1.70 million range over the next few years. Refinancing of existing mortgages, which accounted for about 60 percent of originations during 2001-2003 is expected to return to more normal levels. As this Appendix has explained, the GSEs will continue to play a dominant role in the single-family market and will both impact and be affected by major market developments such as the growth in subprime lending and the increasing use automated underwriting.

Multifamily Mortgage Market. The market for financing of multifamily apartments has grown to record volumes. The favorable long-term prospects for apartments, combined with record low interest rates, have kept investor demand for apartments strong and supported property prices. As explained above, Fannie Mae and Freddie Mac have been among those boosting volumes and introducing new programs to serve the multifamily market. The long run outlook for the multifamily rental market is sustained, moderate growth, based on favorable demographics. The minority population, especially Hispanics, provides a growing source of demand for affordable rental housing. “Lifestyle renters” (older, middle-income households) are also a fast growing segment of the rental population. However, provision of affordable housing will continue to challenge suppliers of multifamily rental housing and policy makers at all levels of government. Low incomes combined with high housing costs define a difficult situation for millions of renter households. Housing cost reductions are constrained by high land prices and construction costs in many markets. Government action—through land use regulation, building codes, and occupancy standards—are major contributors to those high costs. In addition to fewer regulatory barriers and costs, multifamily housing would benefit from more favorable public attitudes. Higher density housing is a potentially powerful tool for preserving open space, reducing sprawl, and promoting transportation alternatives to the automobile. The recently heightened attention to these issues may increase the acceptance of multifamily rental construction to both potential customers and their prospective neighbors.

2. Past Performance of the GSEs

This section reviews the low- and moderate-income performance of Fannie Mae and Freddie Mac. It first reviews the GSEs' performance on the Low- and Moderate-Income Goal, then reviews findings from Section E.2 regarding the GSEs' purchases of home loans for historically underserved families and their communities. Finally, it reviews findings from Section G concerning the GSEs' presence in owner and rental markets.

a. Housing Goals Performance

In the October 2000 rule, the low- and moderate-income goal was set at 50 percent for 2001-03. Effective on January 1, 2001, several changes in counting requirements came into effect for the low- and moderate-income goal, as follows: (a) “Bonus points” (double credit) for purchases of mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on 2-4 unit owner-occupied properties; (b) a “temporary adjustment factor” (1.35 units credit) for Freddie Mac's purchases of mortgages on large (more than 50 units) multifamily properties; (c) changes in the treatment of missing data; and (d) a procedure for the use of imputed or proxy rents for determining goal credit for multifamily mortgages. Fannie Mae's performance was 51.5 percent in 2001, 51.8 percent in 2002, and 52.3 percent in 2003; Freddie Mac's performance was 53.2 percent in 2001, 50.5 percent in 2002, and 51.2 percent in 2003—thus both GSEs surpassed this higher goal in all three years.

Counting requirements (a) and (b) expired at the end of 2003, while (c) and (d) will remain in effect after that. If this counting approach—without the bonus points and the “temporary adjustment factor”—had been in effect in 2000 and 2001, and the GSEs had purchased the same mortgages that they actually did purchase in both years, then Fannie Mae's performance would have been 51.3 percent in 2000, 49.2 percent in 2001, 49.0 percent in 2002, and 48.7 percent in 2003. Freddie Mac's performance would have been 50.6 percent in 2000, 47.7 percent in 2001, 46.1 percent in 2002, and 44.6 percent in 2003. Thus, both Fannie Mae and Freddie Mac would have surpassed the low- and moderate-income goal of 50 percent in 2000 and fallen short in 2001 through 2003. (See Figure A.1.)

b. Single-Family Affordable Lending Market

The GSEs have played a major role in the single-family mortgage market over the past ten years. Their purchases of single-family-owner mortgages accounted for 61 percent of all mortgages originated in the single-family conventional conforming market between 1999 and 2002. Their underwriting and purchase guidelines are market standards, used in all segments of the mortgage market. The GSEs have worked to improve their affordable lending record—they have introduced new low-downpayment products targeted at lower-income families; they have customized their underwriting standards to recognize the unique needs of immigrant and minority families; and, they have entered into numerous partnerships with lenders and non-profit groups to reach out to underserved populations. The enterprises' role in the mortgage market is also reflected in their use of cutting edge technology, such as the development of Loan Prospector and Desktop Underwriter, the automated underwriting systems developed by Freddie Mac and Fannie Mae, respectively. Both GSEs are also entering new and challenging fields of mortgage finance, such as purchasing subprime mortgages.

Despite these efforts and the overall gains in goal performance, the Department remains concerned about the GSEs' support of home lending for the lower-income end of the market and for first-time homebuyers. The shares of the GSEs' purchases are too low, particularly for underserved areas and groups such as minority first-time homebuyers.

This appendix included a comprehensive analysis of the GSEs' performance in funding home purchase mortgages for families and communities that historically have not been well served by the mortgage market. The following findings are offered with respect to the GSEs' acquisitions of home purchase loans that qualify for the three housing goals (special affordable and underserved areas as well as low- and moderate-income) and their acquisitions of first-time homebuyer loans:

• Fannie Mae and Freddie Mac have both improved their support for the single-family affordable lending market over the past eleven years, but historically over past periods, such as 1993-2003, 1996-2003, and 1999-2003, they have lagged the overall conventional conforming market in providing affordable loans to lower-income borrowers and underserved areas. This finding is based on HUD's analysis of GSE and HMDA data and on numerous studies by academics and research organizations.
• The GSEs have shown different patterns of mortgage purchases. Except for two years (1999 and 2000), Fannie Mae has performed better than Freddie Mac since 1993 on all three goals-qualifying categories—low-mod, special affordable, and underserved areas. As a result, the percentage of Freddie Mac's purchases benefiting historically underserved families and their neighborhoods has been less than the corresponding shares of total market originations, while Fannie Mae's purchases have been somewhat closer to the patterns of originations in the primary market.
• The above patterns can be seen by the following percentage shares of home purchase loans that qualified for the three housing goals between 1996 and 2003:

	Special affordable (percent)	Low-Mod (percent)	Underserved areas (percent)
Freddie Mac	13.2	40.3	22.0
Fannie Mae	14.1	42.2	24.0
Market (w/o B&C)	15.9	43.6	25.7

• During 2001-2003, Fannie Mae improved its performance enough to lead the special affordable and low-moderate income markets, although it continued to lag the underserved areas market. During 2001-2003, Freddie Mac lagged the conventional conforming market on all three goals-qualifying categories; see Figure A.2 for the low- and moderate-income shares for Fannie Mae, Freddie Mac and the market.

• Both Fannie Mae and Freddie lag the conventional conforming market in funding first-time homebuyers, and by a rather wide margin. Between 1999 and 2001, first-time homebuyers accounted for 27 percent of each GSE's purchases of home loans, compared with 38 percent for home loans originated in the conventional conforming market.
• The GSEs also account for a very small share of the market for important groups such as minority first-time homebuyers. Considering the total mortgage market (both government and conventional loans), it is estimated that the GSEs purchased only 14 percent of loans originated between 1999 and 2001 for African-American and Hispanic first-time homebuyers, or one-third of their share (42 percent) of all home purchase loans originated during that period. Considering the conventional conforming market and the same time period, it is estimated that the GSEs purchased only 31 percent of loans originated for African-American and Hispanic first-time homebuyers, or approximately one-half of their share (57 percent) of all home purchase loans in that market.

To summarize, the Department's analysis suggests that, except for Fannie Mae's recent performance on the special affordable and low-moderate categories, the GSEs have not been leading the single-family-owner market in purchasing goals-qualifying and first-time homebuyer loans. Freddie Mac, in participation, continues to lag the market on all categories considered. There is room for Freddie Mac, as well as Fannie Mae, to further improve their performance in purchasing affordable loans in the underserved portion of the market, particularly in the minority first-time homebuyer market. Evidence suggests that there is a significant population of potential homebuyers who might respond well to aggressive outreach by the GSEs—immigrants and minorities, in particular, are expected to be a major source of future homebuyers. Furthermore, studies indicate the existence of a large untapped pool of potential homeowners among the rental population. Indeed, the GSEs' recent experience with new outreach and affordable housing initiatives is important confirmation of this potential. To move the GSEs into a leadership position, the Department is establishing three subgoals for home purchase loans that qualify for the three housing goals. The low- and moderate-income subgoal is discussed in Section I.3 below.

c. Overall Market Shares

This appendix also included an analysis of the GSEs' role in the overall (owner and rental) conventional conforming mortgage market. While GSE mortgage purchases represented 55 percent of total dwelling units financed between 1999 and 2002, they represented smaller shares of the three goals-qualifying markets: 48 percent of housing units financed for both low- and moderate-income families and properties located in underserved areas; and 41 percent of units financed for the very-low-income and other families that qualify as special affordable. (See Figure A.3.) In other words, the GSEs accounted for approximately 50 percent or less of the single-family and multifamily units financed in the goals-qualifying markets. This market share analysis suggests that there is room for the GSEs to increase their purchases in these goals-qualifying markets.

The market analysis also examined the GSEs' presence in the owner-occupied home purchase mortgage and rental property sectors of the mortgage market: single-family owner (a 61 percent share for the GSEs between 1999 and 2002) and single-family rental and multifamily rental (a combined rental share of 37 percent). The GSEs, and particularly Freddie Mac, have historically played a smaller role in the market financing rental properties, as compared with their role in the owner market. Fannie Mae and Freddie Mac have recently increased their purchases of these mortgages, but their purchases totaled only 37 percent of the rental units that received financing between 1999 and 2002.^[342] A further increased presence by Fannie Mae and Freddie Mac would bring lower interest rates and liquidity to this market, as well as improve their housing goals performance.

d. The GSEs' Purchases of Multifamily Mortgages

Fannie Mae and, especially, Freddie Mac have rapidly expanded their presence in the multifamily mortgage market in the period since the passage of FHEFSSA. The Senate report on this legislation in 1992 referred to the GSEs' activities in the multifamily arena as “troubling,” citing Freddie Mac's September 1990 suspension of its purchases of new multifamily mortgages and criticism of Fannie Mae for “creaming” the market.^[343]

Freddie Mac has successfully rebuilt its multifamily acquisition program, as shown by the increase in its purchases of multifamily mortgages: from $27 million in 1992 to $3 billion in 1997 and then to approximately $7 billion during the next three years (1998 to 2000), before rising further to $11.9 billion in 2001, $13.3 billion in 2002, and $21.6 billion in 2003. Multifamily properties accounted for 10.3 percent of all dwelling units (both owner and rental) financed by Freddie Mac during 2003. Concerns regarding Freddie Mac's multifamily capabilities no longer constrain their performance with regard to low- and moderate-income families.

Fannie Mae never withdrew from the multifamily market, but it has also stepped up its activities in this area substantially, with multifamily purchases rising from $3.0 billion in 1992 to $9.4 billion in 1999, $18.7 billion in 2001, $18.3 billion in 2002, and $33.3 billion in 2003. Multifamily units as a share of all dwelling units (both owner and rental) financed by Fannie Mae varied in the 10-13 percent range between 1999 and 2001, before falling to 7.3 percent during heavy refinancing year of 2002 and 8 percent in 2003.

The increased purchases of multifamily mortgages by Fannie Mae and Freddie Mac have major implications for the Low- and Moderate-Income Housing Goal, since a very high percentage of multifamily units have rents which are affordable to low- and moderate-income families. However, the potential of the GSEs to lead the multifamily mortgage industry has not been fully developed. As reported earlier in Tables A.30 and A.31b, the GSEs' purchases between 1999 and 2002 accounted for 35-41 percent of the multifamily units that received financing during this period. Certainly there are ample opportunities and room for expansion of the GSEs' share of the multifamily mortgage market. The GSEs' size and market position between loan originators and mortgage investors makes them the logical institutions to identify and promote needed innovations and to establish standards that will improve market efficiency. As their role in the multifamily market continues to grow, the GSEs will have the knowledge and market presence to push simultaneously for standardization and for programmatic flexibility to meet special needs and circumstances, with the ultimate goal of increasing the availability and reducing the cost of financing for affordable and other multifamily rental properties.

3. Ability to Lead the Single-Family-Owner Market: A Low- and Moderate-Income Subgoal

As discussed in Section E, the Department is proposing to establish a subgoal of 45 percent for each GSE's purchases of home purchase loans for low- and moderate-income families in the single-family-owner market of metropolitan areas for 2005, with the subgoal rising to 46 percent in 2006 and 47 percent in 2007 and 2008. The purpose of this subgoal is to encourage the GSEs to improve their acquisitions of home purchase loans for lower-income families and first-time homebuyers who are expected to enter the homeownership market over the next few years. If the GSEs meet this goal, they will be leading the primary market by approximately one percentage point in 2005 and by three percentage points in 2007 and 2008, based on the income characteristics of home purchase loans reported in HMDA. Between 2002 and 2003, HMDA data show that low- and moderate-income families accounted for an (unweighted) average of 44.1 percent of single-family-owner loans originated in the conventional conforming market of metropolitan areas. (The market and GSE data reported in this paragraph are based on “projected” data that account for new Census geography and the new OMB metropolitan area definitions; see Table A.17b.) Loans in the B&C portion of the subprime market are not included in these averages. To reach the 45-percent (47-percent) subgoal, Freddie Mac would have to improve its performance by 0.8 (2.8) percentage points over its 2003 performance. Fannie Mae would have to keep up its high level (47.5 percent) of performance during 2003. The approach taken is for the GSEs to obtain their leadership position by staged increases in the low-mod subgoal; this will enable the GSEs to take new initiatives in a correspondingly staged manner to achieve the new subgoal each year. Thus, the increases in the low-mod subgoal are sequenced so that the GSEs can gain experience as they improve and move toward the new higher subgoal targets.

As explained in Section E.9, the subgoal applies only to the GSEs' purchases in metropolitan areas because reliable market data for non-metropolitan areas are not available from HMDA. The Department is also setting home purchase subgoals for the other two goals-qualifying categories, as follows: 17-18 percent for special affordable loans and 32-34 percent for underserved area loans (also called Geographically Targeted loans).

The Department considered the following factors when setting the subgoal for low- and moderate-income loans.

(a) The GSEs have the ability to lead the market. The GSEs have the ability to lead the primary market for single-family-owner loans, which is the “bread-and-butter” of their business. They both have substantial experience in this market, which means there are no issues as whether or not the GSEs have yet penetrated the market, as there are with the single-family rental and multifamily markets. Both GSEs have not only been operating in the owner market for years, they have been the dominant players in that market, funding 61 percent of the single-family-owner mortgages financed between 1999 and 2002. As discussed in Section G, their underwriting guidelines are industry standards and their automated mortgage systems are widely used throughout the mortgage industry. Through their new downpayment and subprime products, and their various partnership initiatives, the GSEs have shown that they have the capacity to reach out to lower-income families seeking to buy a home. Both Fannie Mae and Freddie Mac have the staff expertise and financial resources to make the extra effort to lead the primary market in funding single-family-owner mortgages for low- and moderate-income mortgages, as well for special affordable and undeserved area mortgages.

(b) GSEs' Performance Relative to the Market. Even though the GSEs have had the ability to lead the home purchase market, their past average performance (1993-2003, 1996-2003, and 1999-2003) has been below market levels. During 2002 and 2003, Fannie Mae improved its performance enough to lead the low-mod market for home purchase loans, but Freddie Mac, although it also improved its performance during this recent period, continues to lag behind the primary market. The subgoals will ensure that Fannie Mae maintains and further improves its above-market performance and that Freddie Mac not only erases its current gap with the market but also takes a leadership position as well. With respect to the GSEs' historical performance, low- and moderate-income mortgages accounted for 40.3 (42.6) percent of Freddie Mac's purchases during 1996-2003 (1999-2003), for 42.2 (43.6) percent of Fannie Mae's purchases, and for 43.6 (44.1) percent of primary market originations (excluding B&C loans). The type of improvement needed for Freddie Mac to meet this new low-mod subgoal was demonstrated by Fannie Mae during 2001-2003, as Fannie Mae increased its low-mod purchases from 40.8 percent of its single-family-owner business in 2000 to 45.3 percent in 2002 and47.0 percent in 2003. (As noted above, Fannie Mae's 2003 performance was slightly higher at 47.5 percent when measured based on the new 2000 Census geography and new OMB definitions.)

(c) Disparities in Homeownership and Credit Access Remain. There remain troublesome disparities in our housing and mortgage markets, even after the “revolution in affordable lending” and the growth in homeownership that has taken place since the mid-1990s. The homeownership rate for African-American and Hispanic households remains 25 percentage points below that of white households. Minority families face many barriers in the mortgage market, such as lack of capital for down payment and lack of access to mainstream lenders (see above). Immigrants and minorities are projected to account for almost two-thirds of the growth in the number of new households over the next ten years. As emphasized throughout this Appendix, changing population demographics will result in a need for the primary and secondary mortgage markets to meet nontraditional credit needs, respond to diverse housing preferences and overcome information and other barriers that many immigrants and minorities face. The GSEs have to increase their efforts in helping these families because so far they have played a surprisingly small role in serving minority first-time homebuyers. It is estimated that the GSEs accounted for 46.5 percent of all (both government and conventional) home loans originated between 1999 and 2001; however, they accounted for only 14.3 percent of home loans originated for African-American and Hispanic first-time homebuyers. Within the conventional conforming market, it is estimated that the GSEs purchased only 20 percent of loans originated for African-American and Hispanic first-time homebuyers, even though they accounted for 57 percent of all home purchase loans in that market. A subgoal for home purchase loans should increase the GSEs' efforts in important sub-markets such as the one for minority first-time homebuyers.

(d) There are ample opportunities for the GSEs to improve their performance. Low- and moderate-income loans are available for the GSEs to purchase, which means they can improve their performance and lead the primary market in purchasing loans for borrowers with less-than-median income. Three indicators of this have already been discussed. First, Sections B and C of this appendix and Appendix D explain that the affordable lending market has shown an underlying strength over the past few years that are unlikely to vanish (without a significant increase in interest rates or a decline in the economy). The low-mod share of the home purchase market has averaged 43.6 percent since 1996 and annually has ranged from 42.1 percent to 44.8 percent. Second, the market share data reported in Table A.30 of Section G demonstrate that there are newly-originated low- and moderate-income loans available each year for the GSEs to purchase. As noted above, the GSEs have only a minimal presence in special sub-markets such the minority first-time homebuyer market, which suggests there are ample opportunities available for the GSEs to increase their purchases of loans for low- and moderate-income families. Finally, the GSEs' purchases under the subgoal are not limited to new mortgages that are originated in the current calendar year. The GSEs can purchase loans from the substantial, existing stock of affordable loans held in lenders' portfolios, after these loans have seasoned and the GSEs have had the opportunity to observe their payment performance. In fact, based on Fannie Mae's recent experience, the purchase of seasoned loans appears to be one useful strategy for purchasing goals-qualifying loans.

For the reasons given above, the Secretary believes that the GSEs can do more to raise the low- and moderate-income shares of their mortgages on these properties. This can be accomplished by building on various programs that the enterprises have already started, including (1) their partnership and outreach efforts, (2) their incorporation of greater flexibility into their underwriting guidelines, (3) their purchases of CRA loans, and (4) their targeting of important markets where they have had only a limited presence in the past, such as the market for minority first-time homebuyers. A wide variety of quantitative and qualitative indicators indicate that the GSEs' have the resources and financial strength to improve their affordable lending performance enough to lead the market for low- and moderate-income families. The recent experience of Fannie Mae indicates that the GSEs can lead the low- and moderate-income market.

4. Size of the Mortgage Market for Low- and Moderate-Income Families

As detailed in Appendix D, the low- and moderate-income mortgage market accounts for 51 to 56 percent of dwelling units financed by conventional conforming mortgages. In estimating the size of the market, HUD excluded the effects of the B&C market. HUD also used alternative assumptions about future economic and market affordability conditions that were less favorable than those that existed over the last five years. HUD is well aware of the volatility of mortgage markets and the possible impacts of changes in economic conditions on the GSEs' ability to meet the housing goals. Should conditions change such that the goals are no longer reasonable or feasible, the Department has the authority to revise the goals.

5. The Low- and Moderate-Income Housing Goal for 2005-2008.

The Low- and Moderate-Income Housing Goal is 52 percent of eligible units for 2005, 53 percent for 2006, 55 percent for 2007, and 56 percent for 2008. The market for the Low- and Moderate-Income Goal is estimated to be 51-56 percent. Under the new counting rules (i.e., 2000-Census income re-benchmarking and the new OMB metropolitan area definitions), Fannie Mae's low- and moderate-income performance is estimated to have been 46.3 percent in 1999, 51.2 percent in 2000, 48.7 percent in 2001, 47.9 percent in 2002, and 49.5 percent in 2003—for 2005, Fannie Mae would have to increase its performance by 3.3 percentage points over its average (unweighted) performance of 48.7 percent over these last five years, or by 0.8 percentage point over its previous peak performance (51.2 percent in 2000). By 2008, Fannie Mae's performance would have to increase by 6.3 percentage points over average 1999-2003 performance, and by 5.8 percentage points over its previous peak performance in 2000. Freddie Mac's performance is estimated to have been 46.0 percent in 1999, 50.2 percent in 2000, 47.0 percent in 2001, 44.6 percent in 2002, and 45.3 percent in 2003—for 2005, Freddie Mac would have to increase its performance by 5.3 percentage points over its average (unweighted) performance of 46.7 percent over these last five years, or by 1.8 percentage points over its previous peak performance (50.2 percent in 2000). By 2008, Freddie Mac's performance would have to increase by 9.3 percentage points over average 1999-2003 performance, and by 5.8 percentage points over its previous peak performance. However, the low- and moderate-income market is estimated to be 51-56 percent. Thus, the GSEs should be able to improve their performance enough to meet these goals of 52-56 percent.

The objective of the Low- and Moderate-Income Goal is to bring the GSEs' performance to the upper end of HUD's market range estimate for this goal (51-56 percent), consistent with the statutory criterion that HUD should consider the GSEs' ability to lead the market for each Goal. To enable the GSEs to achieve this leadership, the Department is proposing modest increases in the Low- and Moderate-Income Goal for 2005 which will increase further, year-by-year through 2008, to achieve the ultimate objective for the GSEs to lead the market under a range of foreseeable economic circumstances by 2008. Such a program of staged increases is consistent with the statutory requirement that HUD consider the past performance of the GSEs in setting the Goals. Staged annual increases in the Low- and Moderate-Income Goal will provide the enterprises with opportunity to adjust their business models and prudently try out business strategies, so as to meet the required 2008 level without compromising other business objectives and requirements.

Figure A.3 summarizes many of the points made in this section regarding opportunities for Fannie Mae and Freddie Mac to improve their overall performance on the Low- and Moderate-Income Goal. The GSEs' purchases provided financing for 26,118,927 (or 55 percent) of the 47,551,039 single-family and multifamily units that were financed in the conventional conforming market between 1999 and 2002. However, in the low- and moderate-income part of the market, the 12,608,215 units that were financed by GSE purchases represented only 48 percent of the 26,051,771 dwelling units that were financed in the market. Thus, there appears to be ample room for the GSEs to increase their purchases of loans that qualify for the Low- and Moderate-Income Goal. Examples of specific market segments that would particularly benefit from a more active secondary market have been provided throughout this appendix.

6. Conclusions

Having considered the projected mortgage market serving low- and moderate-income families, economic, housing and demographic conditions for 2005-08, and the GSEs' recent performance in purchasing mortgages for low- and moderate-income families, the Secretary has determined that the goals of 52 percent of eligible units financed in 2005, 53 percent in 2006, 55 percent in 2007, and 56 percent in 2008 are feasible. The Secretary is also establishing a subgoal of 45 percent for the GSEs' purchases of single-family-owner home purchase mortgages in metropolitan areas in 2005, increasing to 46 percent in 2006 and 47 percent in 2007 and 2008. The Secretary has considered the GSEs' ability to lead the industry as well as the GSEs' financial condition. The Secretary has determined that the proposed goals and the proposed subgoals are necessary and appropriate.

Appendix B—Departmental Considerations To Establish the Central Cities, Rural Areas, and Other Underserved Areas Goal

A. Introduction

1. Establishment of Goal

The Federal Housing Enterprises Financial Safety and Soundness Act of 1992 (FHEFSSA) requires the Secretary to establish an annual goal for the purchase of mortgages on housing located in central cities, rural areas, and other underserved areas (the “Underserved Areas Housing Goal”).

In establishing this annual housing goal, Section 1334 of FHEFSSA requires the Secretary to consider:

1. Urban and rural housing needs and the housing needs of underserved areas;

2. Economic, housing, and demographic conditions;

3. The performance and effort of the enterprises toward achieving the Underserved Areas Housing Goal in previous years;

4. The size of the conventional mortgage market for central cities, rural areas, and other underserved areas relative to the size of the overall conventional mortgage market;

5. The ability of the enterprises to lead the industry in making mortgage credit available throughout the United States, including central cities, rural areas, and other underserved areas; and

6. The need to maintain the sound financial condition of the enterprises.

Organization of Appendix. The remainder of Section A first defines the Underserved Areas Housing Goal for both metropolitan areas and nonmetropolitan areas. Sections B and C address the first two factors listed above, focusing on findings from the literature on access to mortgage credit in metropolitan areas (Section B) and in nonmetropolitan areas (Section C). Separate discussions are provided for metropolitan and nonmetropolitan (rural) areas because of differences in the underlying markets and the data available to measure them. Section D discusses the past performance of the GSEs on the Underserved Areas Housing Goal (the third factor) and Sections E-G report the Secretary's findings for the remaining factors. Section H presents the Department's rules relating to the definition of underserved areas in nonmetropolitan areas. Section I summarizes the Secretary's rationale for establishing a subgoal for single-family-owner home purchase mortgages and for setting the level for the Underserved Areas Housing Goal.

2. HUD's Underserved Areas Housing Goal

HUD's definition of the geographic areas targeted by this goal is basically the same as that used during 1996-2003. It is divided into a metropolitan component and a nonmetropolitan component. However, as explained below, switching to 2000 Census geography increases the number of census tracts defined as underserved, and this necessitates an adjustment of the goal level.

Metropolitan Areas. This rule provides that within metropolitan areas, mortgage purchases will count toward the goal when those mortgages finance properties that are located in census tracts where (1) median income of families in the tract does not exceed 90 percent of area (MSA) median income or (2) minorities comprise 30 percent or more of the residents and median income of families in the tract does not exceed 120 percent of area median income.

In this Rule, the underserved census tracts are defined in terms of the 2000 Census rather than the 1990 Census. As shown in Table B.1a, switching to 2000 Census data and re-specified MSA boundaries as of June 2003, increases the proportions of underserved census tracts, population, owner-occupied housing units, and population below the poverty line in metropolitan areas. The definition now covers 26,959 (51.3 percent) of the 52,585 census tracts in metropolitan areas, which include 48.7 percent of the population and 38.0 percent of the owner-occupied housing units in metropolitan areas.^[1] The 1990-based definition covered 21,587 (47.5 percent) of the 45,406 census tracts in metropolitan areas, which included 44.3 percent of the population and 33.7 percent of the owner-occupied units in metropolitan areas.

The census tracts included in HUD's definition of underserved areas exhibit low rates of mortgage access and distressed socioeconomic conditions. Between 1999 and 2002, the unweighted average mortgage denial rate in these tracts was 17.5 percent, almost double the average denial rate (9.3 percent) in excluded tracts. The underserved tracts include 75.3 percent of the number of persons below the poverty line in metropolitan areas.

HUD's establishment of this definition is based on a substantial number of studies of mortgage lending and mortgage credit flows conducted by academic researchers, community groups, the GSEs, HUD and other government agencies. As explained in the 2000 Rule, one finding stands out from the existing research literature on mortgage access for different types of neighborhoods: High-minority and low-income neighborhoods continue to have higher mortgage denial rates and lower mortgage origination rates than other neighborhoods. A neighborhood's minority composition and its level of income are highly correlated with access to mortgage credit.

Nonmetropolitan Areas. In nonmetropolitan areas, mortgage purchases count toward the Underserved Areas Housing Goal for properties which are located in counties where (1) median income of families in the county does not exceed 95 percent of the greater of (a) state nonmetropolitan median income or (b) nationwide nonmetropolitan median income, or (2) minorities comprise 30 percent or more of the residents and median income of families in the county does not exceed 120 percent of the greater of (a) state nonmetropolitan median income or (b) nationwide nonmetropolitan median income.

In 1995, two important factors influenced HUD's definition of nonmetropolitan underserved areas—lack of available data for measuring mortgage availability in rural areas and lenders' difficulty in operating mortgage programs at the census tract level in rural areas. Because of these factors, the 1995 Rule (as well as the 2000 Rule) used a more inclusive, county-based approach to designating underserved portions of rural areas. As discussed in a later section, HUD is now replacing the county-based definition with a tract-based definition.

As shown in Table B.1b, switching from 1990 to 2000 Census data and incorporating the June, 2003 specification of metropolitan areas causes a slight decrease in underserved proportions of counties, population, owner-occupied housing units, and poverty population in non-metropolitan areas. In terms of the 2000 Census geography and June 2003 metropolitan area specification, the definition covers 1,260 (61.4 percent) of the 2,052 counties in nonmetropolitan areas, which include 51.0 percent of the population, 50.7 percent of the owner-occupied housing units, and 64.3 percent of the population below the poverty level in non-metropolitan areas. The 1990-based definition covered 1,514 (65.5 percent) of the 2,311 counties in non-metropolitan areas, which included 54.6 percent of the population, 53.4 percent of the owner-occupied units, and 67.9 percent of the poor in non-metropolitan areas.^[2]

Data comparable to that in Table B.1b is presented in Table B.1c based on census tracts, rather than counties, in nonmetropolitan areas. As indicated, the tract-based definition includes 6,782 (54.9 percent) of the 12,359 nonmetropolitan census tracts in the country. These tracts contain 52.5 percent of the nonmetropolitan population (comparable to the 51.0 percent using a county-based definition) and 50.4 percent of owner-occupied housing units (close to the corresponding figure of 50.7 percent under the county-based approach). But the tract-based approach better targets families most in need, as shown, for example, by the fact that it includes 68.9 percent of the population in poverty, exceeding the corresponding figure of 64.3 percent under the county-based definition of nonmetropolitan underserved areas.

GSE Performance. Table B.1d shows the increases in the GSEs' overall goals performance under the more expansive geography of the 2000 Census. During 2000, Fannie Mae's performance would have been an estimated 37.5 percent if underserved areas were defined in terms of 2000 Census geography, compared with 31.0 percent under 1990 Census geography. These results for Fannie Mae (adjusted to be comparable with the 2000 figures) are 35.7 percent and 30.4 percent for 2001; 35.0 percent and 30.2 percent for 2002; and 34.1 percent and 29.2 percent for 2003. The corresponding figures for Freddie Mac are 34.1 percent and 29.2 percent for 2000 performance; 32.5 percent and 28.2 percent for 2001 performance; 32.4 percent and 28.0 percent for 2002 performance; and 31.6 percent and 27.7 percent for 2003 performance. (The 2001-03 housing goals percentages in the table are adjusted to exclude the effects of the bonus points and Freddie Mac's Temporary Adjustment Factor, which became applicable in 2001 for scoring of loans toward the housing goals.)

Goal and Subgoal Levels. The Department establishes the Underserved Areas Housing Goal as 37 percent of eligible units financed for 2005, 38 percent for 2006 and 2007, and 39 percent for 2008.

HUD is establishing a subgoal of 32 percent for the share of each GSE's total single-family-owner mortgage purchases that finance single-family-owner properties located in underserved census tracts of metropolitan areas for 2005, with this subgoal rising to 33 percent for 2006 and 2007 and 34 percent in 2008. In this case, subgoal performance for a particular calendar year would be calculated for each GSE by dividing (a) the number of mortgages purchased by the GSE that finance single-family-owner properties located in underserved areas (i.e., census tracts) of metropolitan areas by (b) the number of mortgages purchased by the GSE that finance single-family-owner properties located in metropolitan areas. As explained in Section H, the purpose of this subgoal is to encourage the GSEs to lead the primary market in funding mortgages in underserved census tracts.

B. Consideration of Factors 1 and 2 in Metropolitan Areas: The Housing Needs of Underserved Urban Areas and Housing, Economic, and Demographic Conditions in Underserved Urban Areas

This section discusses differential access to mortgage funding in urban areas and summarizes available evidence on identifying those neighborhoods that have historically experienced problems gaining access to credit. Section B.1 provides an overview of the problem of unequal access to mortgage funding, focusing on discrimination and other housing problems faced by minority families and the communities where they live. Section B.2 examines mortgage access at the neighborhood level and discusses in some detail the rationale for the Underserved Areas Housing Goal in metropolitan areas. The most thorough studies available provide strong evidence that low-income and high-minority census tracts are underserved by the mortgage market. Section B.3 presents recent statistics on the credit characteristics and socioeconomic characteristics of underserved areas under HUD's definition. Readers are referred to the expansive literature on this issue, which is reviewed in some detail in Appendix B of HUD's 2000 Rule. This section focuses on some of the main studies and their findings.

Three main points are made in this section:

• Both borrowers and neighborhoods can be identified as currently being underserved by the nation's housing and mortgage markets. Appendix A provided evidence of racial disparities in the sale and rental of housing and in the provision of mortgage credit. Partly as a result of this, the homeownership rate for minorities is substantially below that for whites.
• The existence of substantial neighborhood disparities in mortgage credit is well documented for metropolitan areas. Research has demonstrated that census tracts with lower incomes and higher shares of minority population consistently have poorer access to mortgage credit, with higher mortgage denial rates and lower origination rates for mortgages. Thus, the income and minority composition of an area is a good measure of whether that area is being underserved by the mortgage market.
• Research supports a targeted neighborhood-based definition of underservice. Studies conclude that characteristics of mortgage loan applicants and the neighborhood where the property is located are the major determinants of mortgage denial rates and origination rates.

Once these characteristics are accounted for, other influences, such as location in a central city, play only a minor role in explaining disparities in mortgage lending.^[3]

1. Discrimination in the Mortgage and Housing Markets—An Overview

The nation's housing and mortgage markets are highly efficient systems, where most homebuyers can put down relatively small amounts of cash and obtain long-term funding at relatively small spreads above the lender's borrowing costs, even though transactions costs are still too high and too bundled. Unfortunately, this highly efficient financing system does not work everywhere or for everyone. Studies have shown that access to credit often depends on improper evaluation of characteristics of the mortgage applicant and the neighborhood in which the applicant wishes to buy. In addition, though racial discrimination has become less blatant in the home purchase market, studies have shown that it is still widespread in more subtle forms. Partly as a result of these factors, the homeownership rate for minorities is substantially below that of whites. Appendix A provided an overview of the homeownership gaps and lending disparities faced by minorities. This section briefly reviews evidence on lending discrimination as well as a recent HUD-sponsored study of discrimination in the housing market.

Mortgage Denial Rates. A quick look at mortgage denial rates reported by Home Mortgage Disclosure Act (HMDA) data reveals that in 2002 minority denial rates were higher than those for white loan applicants. For lower-income borrowers, the denial rate for African Americans applying for conventional loans was 2.1 times the denial rate for white borrowers, while for higher-income borrowers, the denial rate for African Americans was 2.7 times the rate for white borrowers.^[4]

Differentials in denial rates, such as those reported above, are frequently used to demonstrate the problems that minorities face obtaining access to mortgage credit. However, an important question is the degree to which variations in denial rates reflect lender bias against certain kinds of borrowers relative to the degree to which they reflect the credit quality of potential borrowers (as indicated by applicants' available assets, credit rating, employment history, etc.). Without fully accounting for the creditworthiness of the borrower, racial differences in denial rates cannot be attributed to lender bias. Some studies of credit disparities have attempted to control for credit risk factors that might influence a lender's decision to approve a loan.

Boston Fed Study. The best example of accounting for credit risk is the study of mortgage denial rates by researchers at the Federal Reserve Bank of Boston.^[5] This landmark study found that racial differentials in mortgage denial rates cannot be fully explained by differences in credit risk. To control for credit risk, the Boston Fed researchers included 38 borrower and loan variables indicated by lenders to be critical to loan decisions. For example, the Boston Fed study included a measure of the borrower's credit history, which is a variable not included in other studies. The Boston Fed study found that minorities' higher denial rates could not be explained fully by income and credit risk factors. The denial rate for African Americans and Hispanics was 17 percent, compared with 11 percent for Whites with similar characteristics. That is, African Americans and Hispanics were about 60 percent more likely to be denied credit than Whites, even after controlling for credit risk characteristics such as credit history, employment stability, liquid assets, self-employment, age, and family status and composition. Although almost all highly-qualified applicants were approved, differential treatment was observed among borrowers with more marginal qualifications. That is, highly-qualified borrowers of all races seemed to be treated equally, but in cases where there was some flaw in the application, white applicants seemed to be given the benefit of the doubt more frequently than minority applicants. A subsequent refinement of the data used by the Federal Reserve Bank of Boston confirmed the findings of that study.^[6]

The Boston Fed study, as well as reassessments of that study by other researchers, concluded that the effect of borrower race on mortgage rejections persists even after controlling for legitimate determinants of lenders' credit decisions.^[7] Thus, these studies imply that variations in mortgage denial rates, such as those reported above, are not determined entirely by borrower risk, but reflect discrimination in the housing finance system. However, the independent race effect identified in these studies is still difficult to interpret. In addition to lender bias, access to credit can be limited by loan characteristics that reduce profitability ^[8] and by underwriting standards that have disparate effects on minority and lower-income borrowers and their neighborhoods.^[9]

Paired-Testing Studies. As discussed in Appendix A, paired testing studies of the pre-qualification process have supported the findings of the Boston Fed study. Based on a review of paired tests conducted by the National Fair Housing Alliance, The Urban Institute concluded that differential treatment discrimination at the pre-application level occurred at significant levels in at least some cities. Minorities were less likely to receive information about loan products, received less time and information from loan officers, and were quoted higher interest rates in most of the cities where tests were conducted.^[10] Another Urban Institute study used the paired testing methodology to examine the pre-application process in Los Angeles and Chicago. African Americans and Hispanics faced a significant risk of unequal treatment when they visited mainstream mortgage lending institutions to make pre-application inquiries.^[11]

Sales and Rental Markets. In 2002, HUD released its third Housing Discrimination Study (HDS) in the sale and rental of housing. The study, entitled Discrimination in Metropolitan Housing Markets: National Results from Phase I of the Housing Discrimination Study (HDS), was conducted by the Urban Institute.^[12] The results of this HDS were based on 4,600 paired tests of minority and non-minority home seekers conducted during 2000 in 23 metropolitan areas nationwide. The report showed large decreases between 1989 and 2000 in the level of discrimination experienced by Hispanics and African Americans seeking to buy a home. There has also been a modest decrease in discrimination toward African Americans seeking to rent a unit. This downward trend, however, has not been seen for Hispanic renters, who now are more likely to experience discrimination in their housing search than are African American renters. But while generally down since 1989, the report found that housing discrimination still exists at unacceptable levels. The greatest share of discrimination for Hispanic and African American home seekers can still be attributed to being told units are unavailable when they are available to non-Hispanic whites and being shown and told about fewer units than a comparable non-minority. Although discrimination is down on most areas for African American and Hispanic homebuyers, there remain worrisome upward trends of discrimination in the areas of geographic steering for African Americans and, relative to non-Hispanic whites, the amount of help agents provide to Hispanics with obtaining financing. On the rental side, Hispanics are more likely in 2000 than in 1989 to be quoted a higher rent than their white counterpart for the same unit.

Another HUD-sponsored study asked respondents to a nationwide survey if they “thought” they had ever been discriminated against when trying to buy or rent a house or an apartment.^[13] While the responses were subjective, they are consistent with the findings of the HDS. African Americans and Hispanics were considerably more likely than whites to say they have suffered discrimination—24 percent of African Americans and 22 percent of Hispanics perceived discrimination, compared to only 13 percent of whites.

Segregation in Urban Areas. Discrimination, while not the only cause, contributes to the pervasive level of segregation that persists between African Americans and Whites in our urban areas. The Census Bureau recently released one of the most exhaustive studies of residential segregation ever undertaken, entitled Racial and Ethnic Residential Segregation in the United States: 1980-2000.^[14] The Census Bureau found that the United States was still very much racially divided. While African Americans have made modest strides, they remain the most highly segregated racial group. The authors said that residential segregation likely results from a variety of factors, including choices people make about where they want to live, restrictions on their choices, or lack of information. The fact that many mainstream lenders do not operate in segregated areas makes it even more difficult for minorities to obtain access to reasonable-priced mortgage credit.^[15] Section C.8 of Appendix A cited several studies showing that these inner city neighborhoods are often served mainly by subprime lenders. In addition, there is evidence that denial rates are higher in minority neighborhoods regardless of the race of the applicant. The next section explores the issue of credit availability in neighborhoods in more detail.

2. Evidence About Access to Credit in Urban Neighborhoods—An Overview

HUD's Underserved Areas Housing Goal focuses on low-income and high-minority neighborhoods that are characterized by high loan application denial rates and low loan origination rates. As explained in Section B.3 below, the mortgage denial rate during 2001 in census tracts defined as underserved by HUD was twice the denial rate in excluded (or “served”) tracts. In addition to such simple denial rate comparisons, there is a substantial economics literature justifying the targeted neighborhood definition that HUD has used to define underserved areas. Appendix B of the 1995 and 2000 GSE Rules reviewed that literature in some detail; thus, this section simply provides an overview of the main studies supporting the need to improve credit access to low-income and high-minority neighborhoods. Readers not interested in this overview may want to proceed to Section B.3, which examines the credit and socioeconomic characterizes of the census tracts included in HUD's underserved area definition.

As explained in HUD's 2000 Rule, the viability of neighborhoods—whether urban, rural, or suburban—depends on the access of their residents to mortgage capital to purchase and improve their homes. While neighborhood problems are caused by a wide range of factors, including substantial inequalities in the distribution of the nation's income and wealth, there is increasing agreement that imperfections in the nation's housing and mortgage markets are hastening the decline of distressed neighborhoods. Disparate denial of credit based on geographic criteria can lead to disinvestment and neighborhood decline. Discrimination and other factors, such as inflexible and restrictive underwriting guidelines, limit access to mortgage credit and leave potential borrowers in certain areas underserved.

Data on mortgage credit flows are far from perfect, and issues regarding the identification of areas with inadequate access to credit are both complex and controversial. For this reason, it is essential to define “underserved areas” as accurately as possible based on existing data and evidence. There are three sets of studies that provide the rationale for the Department's definition of underserved areas: (1) Studies examining racial discrimination against individual mortgage applicants; (2) studies that test whether mortgage redlining exists at the neighborhood level; and (3) studies that support HUD's targeted approach to measuring areas that are underserved by the mortgage market. In combination, these studies provide strong support for the definition of underserved areas chosen by HUD. The main studies of discrimination against individuals have already been summarized in Section B.1 above. Thus, this section focuses on the neighborhood-based studies in (2) and (3). As noted above, this brief overview of these studies draws from Appendix B of the 1995 GSE Rule; readers are referred there for a more detailed treatment of earlier studies of the issues discussed below.

a. Controlling for Neighborhood Risk and Tests of the Redlining Hypothesis

In its deliberations leading up to FHEFSSA, Congress was concerned about geographic redlining—the refusal of lenders to make loans in certain neighborhoods regardless of the creditworthiness of individual applicants. During the 1980s and early 1990s, a number of studies using HMDA data (such as that reported in Tables B.2 and B.3, below) attempted to test for the existence of mortgage redlining. Consistent with the redlining hypothesis, these studies found lower volumes of loans going to low-income and high-minority neighborhoods.^[16] However, such analyses were criticized because they did not distinguish between demand, risk, and supply effects ^[17] —that is, they did not determine whether loan volume was low because families in high-minority and low-income areas were unable to afford homeownership and therefore were not applying for mortgage loans, or because borrowers in these areas were more likely to default on their mortgage obligations, or because lenders refused to make loans to creditworthy borrowers in these areas.^[18] , ^[19]

More Comprehensive Tests of the Redlining Hypothesis. Recent statistical studies have sought to test the redlining hypothesis by more completely controlling for differences in neighborhood risk and demand. In these studies, the explanatory power of neighborhood race is reduced to the extent that the effects of neighborhood risk and demand are accounted for; thus, they do not support claims of racially induced mortgage redlining. Many of these studies find that the race of the individual borrower is more important than the racial composition of the neighborhood. However, these studies cannot reach definitive conclusions about redlining because segregation in inner cities makes it difficult to distinguish the impacts of geographic redlining from the effects of individual discrimination. The following are two good examples of these studies.

Holmes and Horvitz examined variations in conventional mortgage originations across census tracts in Houston.^[20] Their model explaining census-tract variations in mortgage originations included the following types of explanatory variables: (a) The economic viability of the loan, (b) characteristics of properties in and residents of the tract (e.g., house value, income, age distribution and education level), (c) measures of demand (e.g., recent movers into the tract and change in owner-occupied units between 1980 and 1990), (d) measures of credit risk (defaults on government-insured loans and change in tract house values between 1980 and 1990), and (e) the racial composition of the tract, as a test for the existence of racial redlining. Most of the neighborhood risk and demand variables were significant determinants of the flow of conventional loans in Houston. The coefficients of the racial composition variables were insignificant, which led Holmes and Horvitz to conclude that allegations of redlining in the Houston market could not be supported.

Schill and Wachter include several individual borrower and neighborhood characteristics to explain mortgage acceptance rates in Philadelphia and Boston.^[21] They found that the applicant race variables—whether the applicant was African American or Hispanic—showed significant negative effects on the probability that a loan would be accepted. Schill and Wachter stated that this finding does not provide evidence of individual race discrimination because applicant race is most likely serving as a proxy for credit risk variables omitted from their model (e.g., credit history, wealth and liquid assets). Schill and Wachter find that when their neighborhood risk proxies are included in the model along with the individual loan variables, the percentage of the census tract that was African American became insignificant. Thus, similarly to Holmes and Horvitz, Schill and Wachter stated that “once the set of independent variables is expanded to include measures that act as proxies for neighborhood risk, the results do not reveal a pattern of redlining.”^[22]

Other Redlining Studies. To highlight the methodological problems of single-equation studies of mortgage redlining, Fred Phillips-Patrick and Clifford Rossi developed a simultaneous equation model of the demand and supply of mortgages, which they estimated for the Washington, DC metropolitan area.^[23] Phillips-Patrick and Rossi found that the supply of mortgages is negatively associated with the racial composition of the neighborhood, which led them to conclude that the results of single-equation models (such as the one estimated by Holmes and Horvitz) are not reliable indicators of redlining or its absence. However, Phillips-Patrick and Rossi noted that even their simultaneous equations model does not provide definitive evidence of redlining because important underwriting variables (such as credit history), which are omitted from their model, may be correlated with neighborhood race.

A few studies of neighborhood redlining have attempted to control for the credit history of the borrower, which is the main omitted variable in the redlining studies reviewed so far. Samuel Myers, Jr. and Tsze Chan, who studied mortgage rejections in the state of New Jersey in 1990, developed a proxy for bad credit based on the reasons that lenders give in their HMDA reports for denying a loan.^[24] They found that 70 percent of the gap in rejection rates could not be explained by differences in Black and white borrower characteristics, loan characteristics, neighborhoods or bad credit. Myers and Chan concluded that the unexplained Black-white gap in rejection rates is a result of discrimination. With respect to the racial composition of the census tract, they found that Blacks are more likely to be denied loans in racially integrated or predominantly-white neighborhoods than in predominantly-Black neighborhoods. They concluded that middle-class Blacks seeking to move out of the inner city would face problems of discrimination in the suburbs.^[25]

Geoffrey Tootell has authored two papers on neighborhood redlining based on the mortgage rejection data from the Boston Fed study.^[26] Tootell's studies are important because they include a direct measure of borrower credit history, as well as the other underwriting, borrower, and neighborhood characteristics that are included in the Boston Fed data base; thus, his work does not have the problem of omitted variables to the same extent as previous redlining studies.^[27] Tootell found that lenders in the Boston area did not appear to be redlining neighborhoods based on the racial composition of the census tract or the average income in the tract. Consistent with the Boston Fed and Schill and Wachter studies, Tootell found that it is the race of the applicant that mostly affects the mortgage lending decision; the location of the applicant's property appears to be far less relevant. However, he did find that the decision to require private mortgage insurance (PMI) depends on the racial composition of the neighborhood. Tootell suggested that, rather than redline themselves, mortgage lenders may rely on private mortgage insurers to screen applications from minority neighborhoods. Tootell also noted that this indirect form of redlining would increase the price paid by applicants from minority areas that are approved by private mortgage insurers.

In a 1999 paper, Stephen Ross and Geoffrey Tootell used the Boston Fed data base to take a closer look at both lender redlining and the role of private mortgage insurance (PMI) in neighborhood lending.^[28] They had two main findings. First, mortgage applications for properties in low-income neighborhoods were more likely to be denied if the applicant did not apply for PMI. Ross and Tootell concluded that their study provides the first direct evidence based on complete underwriting data that some mortgage applications may have been denied based on neighborhood characteristics that legally should not be considered in the underwriting process. Second, mortgage applicants were often forced to apply for PMI when the housing units were in low-income neighborhoods. Ross and Tootell concluded that lenders appeared to be responding to CRA by favoring low-income tracts once PMI has been received, and this effect counteracts the high denial rates for applications without PMI in low-income tracts.

Studies of Information Externalities. Another group of studies related to redlining and the credit problems facing low-income and minority neighborhoods focus on the “thin” mortgage markets in these neighborhoods and the implications of lenders not having enough information about the collateral and other characteristics of these neighborhoods. The low numbers of house sales and mortgages originated in low-income and high-minority neighborhoods result in individual lenders perceiving these neighborhoods to be more risky. It is argued that lenders do not have enough historical information to project the expected default performance of loans in low-income and high-minority neighborhoods, which increases their uncertainty about investing in these areas.

This recent group of studies that focus on economies of scale in the collection of information about neighborhood characteristics has implications for the identification of underserved areas and understanding the problems of mortgage access in low-income and minority neighborhoods. William Lang and Leonard Nakamura argue that individual home sale transactions generate information which reduce lenders' uncertainty about property values, resulting in greater availability of mortgage financing.^[29] Conversely, appraisals in neighborhoods where transactions occur infrequently will tend to be more imprecise, resulting in greater uncertainty to lenders regarding collateral quality, and more reluctance by them in approving mortgage loans in neighborhoods with thin markets. As a consequence, “prejudicial practices of the past may lead to continued differentials in lending behavior.”

If low-income or minority tracts have experienced relatively few recent transactions, the resulting lack of information available to lenders will result in higher denial rates and more difficulty in obtaining mortgage financing, independently of the level of credit risk in these neighborhoods. A number of empirical studies have found evidence consistent with the notion that mortgage credit is more difficult to obtain in areas with relatively few recent sales transactions. Some of these studies have also found that low transactions volume may contribute to disparities in the availability of mortgage credit by neighborhood income and minority composition. Paul Calem found that, in low-minority tracts, higher mortgage loan approval rates were associated with recent sales transactions volume, consistent with the Lang and Nakamura hypothesis.^[30] While this effect was not found in high-minority tracts, he concludes that “informational returns to scale” contribute to disparities in the availability of mortgage credit between low-minority and high-minority areas. Empirical research by David Ling and Susan Wachter found that recent tract-level sales transaction volume does significantly contribute to mortgage loan acceptance rates in Dade County, Florida, also consistent with the Lang and Nakamura hypothesis.^[31]

Robert Avery, Patricia Beeson, and Mark Sniderman found significant evidence of economies associated with the scale of operation of individual lenders in a neighborhood.^[32] They concluded that “The inability to exploit these economies of scale is found to explain a substantial portion of the higher denial rates observed in low-income and minority neighborhoods, where the markets are generally thin.” Low-income and minority neighborhoods often suffer from low transactions volume, and low transactions volume represents a barrier to the availability of mortgage credit by making mortgage lenders more reluctant to approve and originate mortgage loans in these areas.

b. Geographic Dimensions of Underserved Areas—Targeted Versus Broad Approaches

HUD's definition of metropolitan underserved areas is a targeted neighborhood definition, rather than a broad definition that would encompass entire cities. It also focuses on those neighborhoods experiencing the most severe credit problems, rather than neighborhoods experiencing only moderate difficulty obtaining credit. During the regulatory process leading to the 1995 rule, some argued that underserved areas under this goal should be defined to include all parts of all central cities, as defined by OMB. HUD concluded that such broad definitions were not a good proxy for mortgage credit problems—to use them would allow the GSEs to focus on wealthier parts of cities, rather than on neighborhoods experiencing credit problems. Appendix B of the 1995 and 2000 Rules reviewed findings from academic researchers that support defining underserved areas in terms of the minority and/or income characteristics of census tracts, rather than in terms of a broad definition such as all parts of all central cities. This section briefly reviews two of the studies. The targeted nature of HUD's definition is also examined in Section B.3 below, which describes the credit and socioeconomic characteristics of underserved census tracts.

Shear, Berkovec, Dougherty, and Nothaft conducted an analysis of mortgage flows and application acceptance rates in 32 metropolitan areas that supports a targeted definition of underserved areas.^[33] They found: (a) Low-income census tracts and tracts with high concentrations of African American and Hispanic families had lower rates of mortgage applications, originations, and acceptance rates; and (b) once census tract influences were accounted for, central city location had only a minimal effect on credit flows. These authors recognized that it is difficult to interpret their estimated minority effects—the effects may indicate lender discrimination, supply and demand effects not included in their model but correlated with minority status, or some combination of these factors. Still, they conclude that income and minority status are better indicators of areas with special needs than central city location.

Avery, Beeson, and Sniderman of the Federal Reserve Bank of Cleveland specifically addressed the issue of underserved areas in the context of the GSE legislation.^[34] Their study examined variations in application rates and denial rates for all individuals and census tracts included in the 1990 and 1991 HMDA data base. These authors found that the individual applicant's race exerts a strong influence on mortgage application and denial rates. African American applicants, in particular, had unexplainably high denial rates. Once individual applicant and other neighborhood characteristics were controlled for, overall denial rates for purchase and refinance loans were only slightly higher in minority census tracts than non-minority census tracts. For white applicants, on the other hand, denial rates were significantly higher in minority tracts. That is, minorities had higher denial rates wherever they attempted to borrow, but whites faced higher denials when they attempt to borrow in minority neighborhoods. In addition, Avery et al. found that home improvement loans had significantly higher denial rates in minority neighborhoods. Given the very strong effect of the individual applicant's race on denial rates, the authors noted that since minorities tend to live in segregated communities, a policy of targeting minority neighborhoods may be warranted. They also found that the median income of the census tract had strong effects on both application and denial rates for purchase and refinance loans, even after other variables were accounted for. Avery, Beeson and Sniderman concluded that a tract-level definition is a more effective way to define underserved areas than using the list of OMB-designated central cities as a proxy.

c. Conclusions from the Economics Literature about Urban Underserved Areas

The implications of studies by HUD and others for defining underserved areas can be summarized briefly. First, the existence of large geographic disparities in mortgage credit is well documented. Low-income and high-minority neighborhoods receive substantially less credit than other neighborhoods and fit the definition of being underserved by the nation's credit markets.

Second, researchers are testing models that more fully account for the various risk, demand, and supply factors that determine the flow of credit to urban neighborhoods. The studies by Holmes and Horvitz, Schill and Wachter, and Tootell are examples of this research. Their attempts to test the redlining hypothesis show the analytical insights that can be gained by more rigorous modeling of this issue. However, the fact that urban areas are highly segregated means that the various loan, applicant, and neighborhood characteristics currently being used to explain credit flows are often highly correlated with each other, which makes it difficult to reach definitive conclusions about the relative importance of any single variable such as neighborhood racial composition. Thus, their results are inconclusive, and the need continues for further research on the underlying determinants of geographic disparities in mortgage lending.^[35]

Finally, much research strongly supports a targeted definition of underserved areas. Studies by Shear, et al. and Avery, Beeson, and Sniderman conclude that characteristics of both the applicant and the neighborhood where the property is located are the major determinants of mortgage denials and origination rates—once these characteristics are controlled for, other influences such as central city location play only a minor role in explaining disparities in mortgage lending.

HUD recognizes that the mortgage origination and denial rates forming the basis for the research mentioned in the preceding paragraph, as well as for HUD's definition of underserved areas, are the result of the interaction of individual risk, demand and supply factors that analysts have yet to fully disentangle and interpret. The need continues for further research addressing this problem.

3. Characteristics of HUD's Underserved Areas

a. Credit Characteristics

HMDA data provide information on the disposition of mortgage loan applications (originated, approved but not accepted by the borrower, denied, withdrawn, or not completed) in metropolitan areas. HMDA data include the census tract location of the property being financed and the race and income of the loan applicant(s). Therefore, this is a rich data base for analyzing mortgage activity in urban neighborhoods. HUD's analysis using HMDA data for 2003 shows that high-minority and low-income census tracts have both relatively high loan application denial rates and relatively low loan origination rates.

Table B.2 presents mortgage denial and origination rates by the minority composition and median income of census tracts in metropolitan areas. Two patterns are clear:

• Census tracts with higher percentages of minority residents have higher mortgage denial rates and lower mortgage origination rates than all-white or substantially-white tracts. For example, in 2003 the denial rate for census tracts that are over 90 percent minority (20.6 percent) was 2.3 times that for census tracts with less than 10 percent minority (9.0 percent).
• Census tracts with lower incomes have higher denial rates and lower origination rates than higher income tracts. For example, in 2003 mortgage denial rates declined from 23.2 percent to 7.2 percent as tract income increased from less than 40 percent of area median income to more than 150 percent of area median income.

• Table B.3 illustrates the interaction between tract minority composition and tract income by aggregating the data in Table B.2 into nine minority and income combinations. The low-minority (less than 30 percent minority), high-income (over 120 percent of area median) group had a denial rate of 7.2 percent and an origination rate of 32.4 loans per 100 owner occupants in 2003. The high-minority (over 50 percent), low-income (under 90 percent of area median) group had a denial rate of 19.3 percent and an origination rate of only 17.8 loans per 100 owner occupants. The other groupings fall between these two extremes.

The advantages of HUD's underserved area definition can be seen by examining the minority-income combinations highlighted in Table B.3. The sharp differences in denial rates and origination rates between the underserved and remaining served categories illustrate that HUD's definition delineates areas that have significantly less success in receiving mortgage credit. In 2003 underserved areas had over one and a three-fourths times the average denial rate of served areas (15.9 percent versus 8.9 percent) and two-thirds the average origination rate per 100 owner occupants (20.1 versus 29.1). HUD's definition does not include high-income (over 120 percent of area median) census tracts even if they meet the minority threshold. The average denial rate (10.3 percent) for high-income tracts with a minority share of population over 30 percent is much less than the denial rate (15.9 percent) in underserved areas as defined by HUD.

Figure B.1 compares underserved and served areas within central cities and suburbs. First, Figure B.1 shows that HUD's definition targets central city neighborhoods that are experiencing problems obtaining mortgage credit. The 16.8 percent denial rate in these neighborhoods in 2003 was almost twice the 8.9 percent denial rate in the remaining areas of central cities. A broad, inclusive definition of “central city” that includes all areas of all central cities would include these “remaining” portions of cities. Figure B.1 shows that these areas, which account for approximately 36 percent of the population in central cities, appear to be well served by the mortgage market. As a whole, they are not experiencing problems obtaining mortgage credit.

Second, Figure B.1 shows that HUD's definition also targets underserved census tracts in the suburbs as well as in central cities. The average denial rate in underserved suburban areas (14.8 percent) is 1.7 times that in the remaining served areas of the suburbs (8.7 percent), and is almost as large as the average denial rate (16.8 percent) in underserved central city tracts. Low-income and high-minority suburban tracts appear to have credit problems similar to their central city counterparts. These suburban tracts, which account for 34 percent of the suburban population, are included in HUD's definition of other underserved areas.

b. Socioeconomic Characteristics

The targeted nature of HUD's definition can be seen from the data presented in Table B.4, which show that families living in tracts within metropolitan areas that are underserved based on HUD's definition experience much more economic and social distress than families living in served areas. For example, the poverty rate in underserved census tracts is 18.5 percent, or over three times the poverty rate (5.7 percent) in served census tracts. The unemployment rate and the high-school dropout rate are also higher in underserved areas. In addition, there are nearly three times more female-headed households with children in underserved areas (30.0 percent) than in served areas (13.2 percent). Three-fourths of units in served areas are owner-occupied, while only one-half of units in underserved areas are owner-occupied.

C. Consideration of Factors 1 and 2 in Nonmetropolitan Areas: The Housing Needs of Underserved Rural Areas and the Housing, Economic, and Demographic Conditions in Underserved Rural Areas

Based on discussions with rural lenders in 1995, the definition of underserved rural areas was established at the county level, since such lenders usually do not make distinctions on a census tract basis. A nonmetropolitan county is classified as an underserved area if median income of families in the county does not exceed 95 percent of the greater of state nonmetropolitan or national nonmetropolitan median income, or minorities comprise 30 percent or more of the residents and the median income of families in the county does not exceed 120 percent of the greater of state nonmetropolitan or national nonmetropolitan median income. For nonmetropolitan areas the median income component of the underserved definition is broader than that used for metropolitan areas. While tract income is compared with area income for metropolitan areas, in rural counties income is compared with the greater of state nonmetropolitan income and national nonmetropolitan income. This is based on HUD's analysis of 1990 census data, which indicated that comparing county nonmetropolitan income only to state nonmetropolitan income would lead to the exclusion of many lower-income low-minority counties from the definition, especially in Appalachia. Based on 1990 census geography, underserved counties account for 57 percent (8,091 of 14,419) of the census tracts and 54 percent of the population in rural areas. By comparison, the definition of metropolitan underserved areas encompassed 47 percent of metropolitan census tracts and 44 percent of metropolitan residents.

The purchasing of loans from underserved areas by the GSEs is intended to induce greater homeownership among moderate, low, very low income, and poor families and minorities. For various reasons, including creditworthiness and lending discrimination, these groups experience greater difficulty in securing loans under fair and reasonable terms and in buying decent and affordable housing, and it is for them that the geographic goals were designed. The geographic goals, then, are meant to target places where these “underserved” populations live in order to stimulate local mortgage lending and, it is hoped, the availability of credit to those families who reside there who, otherwise, will have difficulty securing credit. This section addresses the basic question of whether and the extent to which HUD's definition of underservice in nonmetropolitan areas effectively targets areas that encompass large populations of socially and economically disadvantaged families.

Table B.5 shows data on demographic and socioeconomic conditions of underserved and served nonmetropolitan areas based on HUD's definition applied at the county level using Census 2000 data. (A later section considers the effects of applying the definition of the census tract level.) Several variables are used to describe area demographic and socioeconomic conditions.

On the national level, a few key results show that the 1995 definition of underservice captures a potentially disadvantaged segment of the population. In examining the minority composition, one can see that the percentage of African Americans, Hispanics/Latinos, and total minority population is higher in underserved nonmetropolitan areas as compared to served nonmetropolitan areas. Overall, the minority population of underserved areas is 25.8 percent as compared with 9.3 percent in served areas. Other supporting results include median family income, poverty rate, unemployment rate, school dropout rate, and in-migration rate. Specifically we find:

• Median income is approximately $10,000 less in underserved areas than in served areas. This represents an average gap of 25 percent.
• Poverty in underserved areas is twice the rate in served areas (14.5 vs. 7.5 percent).
• Unemployment is 7.3 percent in underserved areas and 5.2 percent in served areas.
• The school dropout rate is 28.1 percent in underserved areas and 18.7 percent in served areas.
• Migration into underserved areas is somewhat lower than in served areas: 7.4 vs. 8.0 percent.

Table B.5 also includes data on homeownership rates, housing affordability, housing quality, and overcrowding. On several of these dimensions, housing conditions and needs in underserved areas are not substantially worse than in served areas. Although housing quality and crowding appear to be marginally worse in underserved areas, homeownership in the two areas is about the same and owning a home actually appears to be more affordable in underserved areas than in served areas. Specific findings include the following:

• Homeownership is slightly higher in underserved than in served nonmetropolitan counties: 74.3 percent vs. 73.7 percent. Removing manufactured homes lowers ownership rates slightly, because ownership of such homes is relatively high, but this does not affect the basic result.
• Owner-occupied and rental vacancy rates are both somewhat higher in underserved areas.
• Median housing unit values are significantly lower in underserved areas: $67,358 vs. $88,099.
• The value of a housing affordability index for owner-occupied housing is slightly higher in underserved areas.^[36] On average, median income is 1.83 times higher than income required to qualify to buy a home of median value in underserved areas. The comparable factor for served areas is 1.78.

• Rental affordability is approximately the same in underserved and served areas.
• While nearly all housing in served and underserved areas have complete plumbing and kitchens, the percentage of units with incomplete facilities in underserved is twice the percentage in served areas.
• Crowded units are a small share of all housing in nonmetropolitan areas, but the rate is higher for underserved areas: 4.3 vs. 2.3 percent.

Mikesell ^[37] found using the 1995 American Housing Survey that while the rate of homeownership in nonmetropolitan areas is higher than metropolitan areas, the quality of housing is lower as compared to metropolitan areas. Results based on the 2000 Census show that the homeownership rate for nonmetropolitan areas was 74 percent (73 percent without manufactured homes), and for metropolitan areas it was 64 percent, but both metropolitan and nonmetropolitan areas had approximately 97.5 percent of units with complete plumbing and 99 percent with complete kitchens.

D. Factor 3: Previous Performance and Effort of the GSEs in Connection With the Central Cities, Rural Areas and Other Underserved Areas Goal

Section D.1 reports the past performance of each GSE with regard to the Underserved Areas Housing Goal. Section D.2 then examines the role that the GSEs are playing in funding single-family mortgages in underserved urban neighborhoods based on HUD's analysis of GSE and HMDA data. That section also discusses an underserved area subgoal for home purchase loans. Section D.3 concludes this section with an analysis of the GSEs' purchases in rural (nonmetropolitan) areas.

The increased coverage of the Underserved Areas Housing goal due to switching to 2000 census geography is discussed throughout this section.

1. Past Performance of the GSEs

This section discusses each GSE's performance under the Underserved Areas Housing Goal over the 1996-2003 period.^[38] As explained in Appendix A, the data presented are “official HUD results” which, in some cases, differ from goal performance reported by the GSEs in the Annual Housing Activities Reports (AHARs) that they submit to the Department.

The main finding of this section is that Fannie Mae surpassed the Department's Underserved Areas Housing Goals for each of the seven years during this period. Freddie Mac surpassed the goal in six of the seven years, falling slightly short in 2002. Specifically:

• The goal was set at 21 percent for 1996; Fannie Mae's performance was 28.1 percent and Freddie Mac's performance was 25.0 percent.
• The goal was set at 24 percent for 1997-2000. Fannie Mae's performance was 28.8 percent in 1997, 27.0 percent in 1998, 26.8 percent in 1999, and 31.0 percent in 2000; and Freddie Mac's performance was 26.3 percent in 1997, 26.1 percent in 1998, 27.5 percent in 1999, and 29.2 percent in 2000.
• In the October 2000 rule, the underserved areas goal was set at 31 percent for 2001-03. As of January 1, 2001, several changes in counting requirements came into effect for the undeserved areas goal, as follows: “bonus points” (double credit) for purchases of goal-qualifying mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on 2-4 unit owner-occupied properties; a “temporary adjustment factor” (1.20 units credit, subsequently increased by Congress to 1.35 units credit) for Freddie Mac's purchases of goal-qualifying mortgages on large (more than 50-unit) multifamily properties; and eligibility for purchases of certain qualifying government-backed loans to receive goal credit. These changes are explained below. Fannie Mae's performance was 32.6 percent in 2001, 32.4 percent in 2002, and 32.1 percent in 2003; and Freddie Mac's performance was 31.7 percent in 2001, slightly less than 31 percent in 2002, and 32.7 percent in 2003, thus Fannie Mae surpassed this higher goal in all three years and Freddie Mac surpassed the goal in 2001 and 2003, but fell slightly short in 2002. This section discusses the October 2000 counting rule changes in detail below, and provides data on what goal performance would have been in 2001-03 without these changes.^[39]

a. Performance on the Underserved Areas Housing Goal in 1996-2003

HUD's December 1995 rule specified that in 1996 at least 21 percent of the number of units financed by each of the GSEs that were eligible to count toward the Underserved Areas Goal should qualify as units in properties located in underserved areas, and at least 24 percent should qualify in 1997-2000. HUD's October 2000 rule made various changes in the goal counting rules, as discussed below, and increased the Underserved Areas Goal to 31 percent for 2001-03.

Table B.6 shows performance on the underserved areas goal over the 1996-2003 period, based on HUD's analysis. The table shows that Fannie Mae surpassed the goals by 7.1 percentage points and 4.8 percentage points in 1996 and 1997, respectively, while Freddie Mac surpassed the goals by narrower margins, 4.0 and 2.3 percentage points. In 1998 Fannie Mae's performance fell by 1.8 percentage points, while Freddie Mac's performance fell only slightly, by 0.2 percentage point. Freddie Mac showed a gain in performance to 27.5 percent in 1999, exceeding its previous high by 1.2 percentage points. Fannie Mae's performance in 1999 was 26.8 percent, which, for the first time, slightly lagged Freddie Mac's performance in that year.

Both GSEs exhibited sharp gains in goal performance in 2000—Fannie Mae's performance increased by 4.2 percentage points, to a record level of 31.0 percent, while Freddie Mac's performance increased somewhat less, by 1.7 percentage points, which also led to a record level of 29.2 percent. Fannie Mae's performance was 32.6 percent in 2001, 32.4 percent in 2002, and 32.1 percent in 2003; Freddie Mac's performance was 31.7 percent in 2001, slightly less than 31 percent in 2002, and 32.7 percent in 2003. However, as discussed below, using consistent accounting rules for 2000-03, under one method each GSE's performance in 2001-03 was below its performance in 2000.

Fannie Mae's performance on the underserved areas goal surpassed Freddie Mac's in every year through 1998. This pattern was reversed in 1999, as Freddie Mac surpassed Fannie Mae in goal performance for the first time, though by only 0.7 percentage point. This improved relative performance of Freddie Mac was due to its increased purchases of multifamily loans, as it re-entered that market, and to increases in the goal-qualifying shares of its single-family mortgage purchases. However, Fannie Mae's performance once again exceeded Freddie Mac's performance in 2000, 31.0 percent to 29.2 percent. Fannie Mae's official performance also exceeded Freddie Mac's official performance in 2001-02, despite the fact that Freddie Mac benefited from a difference in the counting rules applicable to the two GSEs as enacted by Congress; if the same counting rules were applied to both GSEs, Fannie Mae's performance would have exceeded Freddie Mac's performance. In fact, Freddie Mac would have just attained the goal, at 31.4 percent in 2003, and fallen short of the goal in 2001 and 2002.

b. Changes in the Goal Counting Rules for 2001-03

Several changes in the counting rules underlying the calculation of underserved areas goal performance took effect beginning in 2001. These also applied to the low- and moderate-income goal and are discussed in Appendix A; only brief summaries of those changes are given here:^[40] Bonus points for multifamily and single-family rental properties. Each qualifying unit in a small multifamily property counted as two units in the numerator in calculating performance on all of the goals for 2001-03. And, above a threshold equal to 60 percent of the average number of qualifying rental units financed in owner-occupied properties over the preceding five years, each unit in a 2-4 unit owner-occupied property also counted as two units in the numerator in calculating goal performance.

Freddie Mac's Temporary Adjustment Factor. Freddie Mac received a “Temporary Adjustment Factor” of 1.35 units of credit for each qualifying unit financed in “large” multifamily properties (i.e., those with 51 or more units) in the numerator in calculating its performance on the housing goals for 2001-03.^[41] This factor did not apply to units in large multifamily properties in underserved areas whose mortgages were financed by Fannie Mae during this period.

Purchases of certain government-backed loans. Prior to 2001, purchases of government-backed loans were not taken into account in determining performance on the GSEs' low- and moderate-income and underserved area housing goals. As discussed in Appendix A, the 2000 rule established eligibility for FHA-insured home equity conversion mortgages (HECMs) for mortgagors in underserved areas, purchases of mortgages on properties on tribal lands insured under FHA's Section 248 program or HUD's Section 184 program, and purchases of mortgages under the Rural Housing Service's Single Family Housing Guaranteed Loan Program to count toward the underserved area goal.

c. Effects of Changes in the Counting Rules on Goal Performance

Because of the changes in the underserved areas goal counting rules that took effect in 2001, direct comparisons between official goal performance in 2000 and 2001-03 are somewhat of an “apples-to-oranges comparison.” For this reason, the Department has calculated what performance would have been in 2000 under the 2001-03 rules; this may compared with official performance in 2001-03—an “apples-to-apples comparison.” HUD has also calculated what performance would have been in 2001-03 under the 1996-2000 rules; this may be compared with official performance in 2000—an “oranges-to-oranges comparison.” These comparisons are presented in Table B.7a.

Specifically, Table B.7a shows performance under the underserved areas goal in three ways. Baseline A represents the counting rules in effect in 1996-2000. Baseline B incorporates the one minor technical change in counting rules pertaining to the underserved areas goal—eligibility of certain government-backed loans for goals credit. Baseline C incorporates in addition to that technical change the bonus points and, for Freddie Mac, the temporary adjustment factor. Boldface figures under Baseline A for 1999-2000 and under Baseline C for 2001-02 indicate official goal percentages based on the counting rules in effect in those years—e.g., for Freddie Mac, 27.5 percent in 1999, 29.2 percent in 2000, 31.7 percent in 2001, slightly less than 31 percent in 2002, and 32.7 percent in 2003.

Performance on the Underserved Areas Goal under 1996-2000 Counting Rules Plus Technical Changes. If the “Baseline B” counting approach had been in effect in 2000-03 and the GSEs” had purchased the same mortgages that they actually did purchase in those years, Fannie Mae would have just matched the underserved areas goal in 2000 and fallen short in 2001-03, while Freddie Mac would have fallen short of the goal in all four years, 2000-03. Specifically, Fannie Mae's performance would have been 31.0 percent in 2000, 30.4 percent in 2001, 30.2 percent in 2002, and 29.2 percent in 2003. Freddie Mac's performance would have been 29.2 percent in 2000, 28.2 percent in 2001, 28.0 percent in 2002, and 27.7 percent in 2003.

Performance on the Underserved Areas Goal under 2001-2003 Counting Rules. If the 2001-03 counting rules had been in effect in 2000-02 and the GSEs had purchased the same mortgages that they actually did purchase in those years (i.e., abstracting from any behavioral effects of “bonus points,” for example), both GSEs would have surpassed the underserved areas goal in all four years, and both GSEs' performance figures would have increased from 2000 to 2002. Specifically, Fannie Mae's “Baseline C” performance would have been 32.3 percent in 2000, 32.6 percent in 2001, 32.4 percent in 2002, and 32.1 percent in 2003. Freddie Mac's performance would have been 31.4 percent in 2000, 31.7 percent in 2001, slightly less than 31.0 percent in 2002, and 32.7 percent in 2003. Measured on this consistent basis, then, Fannie Mae's performance increased by 0.3 percentage point in 2001, fell by 0.7 percentage points in 2002, and increased by 1.3 percentage points in 2003. These increases were the effect of increased purchases of mortgages eligible to receive bonus points between 2000 and 2001-03.

Details of Effects of Changes in Counting Rules on Goal Performance in 2001. As discussed above, counting rule changes that took effect in 2001 had significant impacts on the performance of both GSEs on the underserved areas goal in that year—2.2 percentage points for Fannie Mae, and 3.5 percentage points for Freddie Mac. This section breaks down the effects of these changes on goal performance for both GSEs; results are shown in Table B.7a along with figures for other years.

Freddie Mac. The largest impact of the counting rule changes on Freddie Mac's goal performance was due to bonus points for purchases of mortgages on small multifamily properties; this added 1.3 percentage points to goal performance in 2001, 0.5 percentage point in 2002, and 2.9 percentage points in 2003, as shown in Table B.7a. The application of the temporary adjustment factor for purchases of mortgages on large multifamily properties enacted by Congress added 0.9 percentage points to goal performance in 2002 and 1.3 percentage points in 2003. Bonus points for purchase of mortgages on owner-occupied 2-4 unit rental properties also added 1.1 percentage points to performance in 2001, 1.6 percentage points in 2002, and 0.9 percentage point in 2003. Credit for purchases of qualifying government-backed loans played a minor role in determining Freddie Mac's goal performance.

Fannie Mae. The temporary adjustment factor which applied to Freddie Mac's goal performance did not apply to Fannie Mae, thus overall counting rule changes had less impact on its performance than on Freddie Mac's performance in 2001-03. The largest impact of the counting rule changes on Fannie Mae's goal performance was due to the application of bonus points for purchases of mortgages on owner-occupied 2-4 unit rental properties, which added 1.7 percentage points to performance in 2001, 1.8 percentage points in 2002, and 1.7 percentage points in 2003, and for purchases of mortgages on small multifamily properties, which added 0.5 percentage point to performance in 2001, 0.8 percentage point in 2002, and 1.2 percentage points in 2003. Credit for purchases of qualifying government-backed loans also played a minor role in determining Fannie Mae's goal performance.

d. Bonus Point Incentives for the GSEs' Purchases in Underserved Areas

The Department established “bonus points” for 2001-03 to encourage the GSEs to step up their activity in two segments of the mortgage market—the small (5-50 unit) multifamily mortgage market, and the market for mortgages on 2-4 unit properties where 1 unit is owner-occupied and 1-3 units are occupied by renters.

Bonus points for small multifamily properties. Each unit financed in a small multifamily property that qualified for any of the housing goals was counted as two units in the denominator (and one unit in the numerator) in calculating goal performance for that goal.

Fannie Mae financed 37,389 units in small multifamily properties in 2001 that were eligible for the underserved areas goal, an increase of more than 400 percent from the 7,196 units financed in 2000. Further increases were recorded in 2002, to 77,382 units, and in 2003, to 230,290 units. As explained in Appendix A, small multifamily properties also accounted for a greater share of Fannie Mae's multifamily business in 2001—7.4 percent of total multifamily units financed, up from 2.5 percent in 2000, with this share rising to 16.8 percent in 2002 and 28.9 percent in 2003. However, HUD's Housing Goals 2000 Final Rule cited a Residential Finance Survey finding that small multifamily properties account for 37 percent of total units in multifamily mortgaged properties, thus Fannie Mae is still somewhat less active in this market than in the market for large multifamily properties.^[42]

Within the small multifamily market, there was some evidence that Fannie Mae targeted properties in underserved areas to a greater extent in 2001 than in 2000. That is, 56 percent of Fannie Mae's small multifamily units qualified for the underserved areas goal in 2000, but this rose to 64 percent in 2001. The share of Fannie Mae's small multifamily units that qualified for the underserved areas goal was 65 percent in 2002 and 50 percent in 2003.

Freddie Mac financed 50,211 units in small multifamily properties in 2001 that were eligible for the underserved areas goal, an increase of more than 1500 percent from the small base of 2,985 units financed in 2000. Financing of such units actually fell in 2002, to 22,195 units, but rebounded to 181,126 units in 2003. Small multifamily properties also accounted for a significantly greater share of Freddie Mac's multifamily business in 2001—16.1 percent of total multifamily units financed, up from 1.8 percent in 2000, with this share amounting to 7.1 percent in 2002 and 30.5 percent in 2003.

Within the small multifamily market, there was some evidence that Freddie Mac targeted properties in underserved areas to a greater extent in 2001 than in 2000. That is, 61 percent of Freddie Mac's small multifamily units qualified for the underserved areas goal in 2000; this rose to 86 percent in 2001. The share of Freddie Mac's small multifamily units that qualified for the underserved areas goal was 88 percent in 2002 and 87 percent in 2003.

Bonus points for single-family rental properties. Above a threshold, each unit financed in a 2-4 unit property with at least one owner-occupied unit (referred to as “OO24s” below) that qualified for any of the housing goals was counted as two units in the denominator (and one unit in the numerator) in calculating goal performance for that goal in 2001-03. The threshold was equal to 60 percent of the average number of such qualifying units over the previous five years. For example, Fannie Mae financed an average of 47,100 underserved area units in these types of properties between 1996 and 2000, and 105,946 such units in 2001. Thus in 2001 Fannie Mae received 77,688 bonus points in this area in 2001—that is, 105,946 minus 60 percent of 47,100. So 183,629 units were entered in the numerator for these properties in calculating underserved area goal performance.

Single-family rental bonus points thus encouraged the GSEs to play a larger role in this market, and also to purchase mortgages on such properties in which large shares of the units qualify for the housing goals. As for small multifamily bonus points, some evidence on the effects of such bonus points on the GSEs' operations may be gleaned from the data provided to HUD by the GSEs for 2001-2003.

Fannie Mae financed 177,872 units in OO24s in 2001 that were eligible for the underserved areas goal, an increase of 116 percent from the 82,464 units financed in 2000. Further increases were recorded in 2002, to 231,581 units, and in 2003, to 353,916 units. However, as a result of the refinance boom Fannie Mae's total single-family business increased at approximately the same rate as its OO24 business in 2001-03, thus the share of its business accounted for by OO24s was the same in 2001 as in 2000—4 percent, with this share also amounting to 4 percent in 2002 and 2003.

Within the OO24 market, there was no evidence that Fannie Mae targeted affordable properties to a greater extent in 2001 than in 2000. That is, approximately 60 percent of Fannie Mae's OO24 units qualified for the underserved area goal in both 2000 and 2001. The share of Fannie Mae's OO24 units that qualified for the underserved areas goal was 62 percent in 2002 and 60 percent in 2003.

Freddie Mac financed 96,983 units in OO24s in 2001 that were eligible for the underserved areas goal, an increase of 91 percent from the 50,868 units financed in 2000. Further increases were recorded in 2002, to 146,502 units, and in 2003, to 154,924 units. However, with the refinance boom, Freddie Mac's total single-family business increased at approximately the same rate as its OO24 business in 2001-03, thus the share of its business accounted for by OO24s was the same in 2001 as in 2000—3 percent, with this share amounting to 3.7 percent in 2002 and 3.1 percent in 2003.

As for Fannie Mae, within the OO24 market there was no evidence that Freddie Mac targeted affordable properties to a greater extent in 2001 than in 2000. That is, 60 percent of Fannie Mae's OO24 units qualified for the underserved areas goal in both 2000 and 2001. The share of Freddie Mac's OO24 units that qualified for the underserved areas goal was 61 percent in 2002 and 50 percent in 2003.

e. Effects of 2000 Census on Scoring of Loans Toward the Underserved Areas Housing Goal

Background. Scoring of housing units under the Underserved Areas Housing Goal is based on decennial census data used to identify underserved areas, as follows: For properties in MSAs scoring is based on the median income of the census tract where the property is located, the median income of the MSA, and the percentage minority population in the census tract where the property is located. For properties located outside of MSAs scoring is based on the median income of the county, the median income of the non-metropolitan portion of the State in which the property is located or of the non-metropolitan portion of the United States, whichever has the larger median income, and the percentage minority population in the county where the property is located. Thus, scoring loans under the Underserved Areas Housing Goal requires decennial census data on median incomes for metropolitan census tracts, MSAs, non-metropolitan counties, the non-metropolitan portions of States, and the non-metropolitan portion of the United States. The determination has been based on 1990 census data through 2004, and beginning in 2005 will be based on 2000 census data.^[43, 44] Under this rule, the basis for the determination outside of MSAs will change from counties to census tracts beginning in 2005.

2005 Procedure. Relative to the above procedure, Underserved Areas Housing Goals performance percentages for loans purchased by the GSEs in and after 2005 will be affected by three factors. First, 2000 census data on median incomes and minority populations replace 1990 census data. Second, the Office of Management and Budget in June, 2003, respecified MSA boundaries based on analysis of 2000 census data. Third, the Department's re-specification of the Underserved Areas goal in terms of census tracts rather than counties in non-metropolitan areas will come into effect.^[45] Thus, for properties located outside of MSAs the basis of determination for non-metropolitan areas will be changed for properties located outside of MSAs to: The median income of the census tract where the property is located; the median income of the non-metropolitan portion of the State in which the property is located or of the non-metropolitan portion of the United States, whichever is larger; and the percentage minority population in the census tract where the property is located.

Analysis. HUD used 2000 census data to generate underserved area designations for census tracts as defined for the 2000 census with 2003 MSA designations. Because Fannie Mae and Freddie Mac geocoded the mortgages they purchased prior to 2003 based on census tract boundaries as established for the 1990 census, GSE mortgages purchased prior to 2003 can be directly identified as being from a served or underserved area only where the property is located in a 1990-defined census tract whose area consists entirely of whole 2000-defined census tracts, or portions of such tracts, which are all designated either as served or as underserved. In the situation where the area of a 1990-defined census tract includes whole 2000-defined census tracts, or portions of such tracts, some of which are served and some underserved, HUD calculated an “underservice factor” defined as the underserved percentage of the 1990-defined tract's population, based on population data from the 2000 census.^[46] These factors were used in estimating underservice percentages for aggregated GSE purchases in and before 2003 based on the 2000 census.

The resulting underserved areas file was used to re-score loans purchased by the GSEs between 1999 and 2003, and was used further in estimating the share of loans originated in metropolitan areas that would be eligible to score toward the Underserved Areas Housing Goal, from HMDA data. The results of the retrospective GSE analysis are provided in Table B.7b The results of the GSE-HMDA comparative analysis are presented in the next section.

Table B.7b shows four sets of estimates for each GSE, based respectively on the counting rules in place in 2001-2003 (but disregarding the bonus points and Temporary Adjustment Factor), on shifting from 1990 to 2000 census data on median incomes and minority concentrations, on the further addition 2003 MSA specification, and finally on shifting from counties to tracts as the basis for scoring loans in non-metropolitan areas.

2. GSEs' Mortgage Purchases in Metropolitan Neighborhoods

Metropolitan areas accounted for about 85 percent of total GSE purchases under the Underserved Areas Housing Goal in 2001 and 2002. This section uses HMDA and GSE data for metropolitan areas to examine the neighborhood characteristics of the GSEs' mortgage purchases. In subsection 2.a, the GSEs' performance in underserved neighborhoods is compared with the overall market. This section therefore expands on the discussion in Appendix A, which compared the GSEs' funding of affordable loans with the overall conventional conforming market. A subgoal that the Department is establishing for each GSE's acquisitions of home purchase loans financing properties in the underserved census tracts of metropolitan areas is also discussed subsection 2a. In subsection 2.b., the characteristics of the GSEs' purchases within underserved areas are compared with those for their purchases in served areas.

a. Comparisons With the Primary Market

Market Comparisons Based on 1990 Census Geography. Section E.8-10 in Appendix A provided detailed information on the GSEs' funding of mortgages for properties located in underserved neighborhoods for the years 1993 to 2003. To take advantage of historical data going back to 1993, these comparisons were first made using 1990 Census tract geography. The findings with respect to the GSEs' funding of underserved neighborhoods are similar to those reported in Appendix A regarding the GSEs' overall affordable lending performance in the single-family-owner market. While both GSEs improved their performance, they historically lagged the conventional conforming market in providing affordable loans to underserved neighborhoods. The two GSEs themselves engaged in very different patterns of funding—Freddie Mac was less likely than Fannie Mae to fund home loans in underserved neighborhoods, as the following percentage shares for home purchase loans indicate:

Year	Freddie Mac (percent)	Fannie Mae (percent)	Market (w/o B&C) (percent)
1996-2003	22.0	24.0	25.7
1999-2003	23.1	24.7	26.2
2001-2003	24.1	26.0	26.4

Between 1996 and 2003, 22.0 percent of Freddie Mac's purchases financed properties in underserved neighborhoods, compared with 24.0 percent of Fannie Mae's purchases and 25.7 percent of home purchase loans originated in the conventional conforming market (excluding B&C loans). Thus, Freddie Mac performed at only 86 percent of the market (22.0 divided by 25.7), while Fannie Mae performed at 93 percent of the market. Freddie Mac's recent performance has been slightly closer to the market. Over the past three years (2001 to 2003), Freddie Mac performed at 91 percent of the market (24.1 percent for Freddie Mac compared at 26.4 percent for the market). (See Tables A.13 to A.16 in Appendix A for complete data going back to 1993.)

Fannie Mae has funded underserved areas at a higher level than Freddie Mac, as indicated above. And during 2001 and 2003, Fannie Mae average performance was only slightly below the market. In 2003, the share of Fannie Mae's purchases going to underserved areas was 26.8 percent, compared with a market level of 27.6 percent. Like Freddie Mac, Fannie Mae's longer-term performance (since 1993 or 1996) as well as its recent average performance (1999 to 2003) has consistently been below market levels. Still, it is encouraging that Fannie Mae significantly improved its 2001-2003 performance and closed its gap with the market during the first three years of HUD's higher housing goal levels.

Market Comparisons Based on 2000 Census Geography. As explained in Section A.2 of this appendix, HUD will be defining underserved areas based on 2000 Census data beginning in 2005. The number of census tracts in metropolitan areas covered by HUD's definition will increase from 21,587 tracts (based on 1990 Census) to 26,959 tracts (based on 2000 Census and new OMB metropolitan area specifications). The increase in the number of tracts defined as underserved means that both GSE performance and the market estimates will be higher than reported above. This section provides an analysis of the performance of the GSEs in the single-family-owner market based on 2000 census tract geography. For the years 1999, 2000, 2001, and 2002, HUD used the apportionment technique to re-allocate 1990-based GSE and HMDA data into census tracts as defined by the 2000 Census. GSE and HMDA data for 2003 were already expressed in terms of 2000 Census geography.

The main results are provided in Table B.8, which compares the GSEs to the market using both the 1990 Census geography and the 2000 Census geography. Switching to the 2000-based tracts increases the underserved area share of market originations by about five percentage points. Between 1999 and 2003, 31.4 percent of home purchase mortgages (without B&C loans) were originated in underserved tracts based on 2000 geography, compared with 26.2 percent based on 1990 geography—a differential of 5.2 percentage points. As also shown in Table B.8, the underserved areas share of Fannie Mae's purchases rises by 5.3 percentage points, and the underserved areas share of Freddie Mac's purchases rises by 5.2 percentage points. Thus, the conclusions reported above and in Appendix A about the GSEs' performance relative to the market about remain the same when the analysis is conducted based on 2000 Census geography.

It is interesting to repeat the earlier 1990-based analysis of home purchase loans but this time based on the 2000 Census geography. The following results are obtained for home purchase loans from Table B.8:

Year	Freddie Mac (percent)	Fannie Mae (percent)	Market (w/o B&C) (percent)
1999	25.6	25.3	30.2
2000	27.3	29.0	31.7
2001	27.3	29.8	30.7
2002	31.7	32.3	31.8
2003	29.0	32.0	32.5
1996-2003 (estimate)	27.2	29.3	30.9
1999-2003 (average)	28.3	30.0	31.4
2001-2003 (average)	29.4	31.4	31.7

Between 1999 and 2003, 28.3 percent of Freddie Mac's purchases and 30.0 percent of Fannie Mae's purchases financed properties in underserved neighborhoods, compared with 31.4 percent home purchase loans originated in the conventional conforming market (excluding B&C loans). Thus, Freddie Mac performed at 90 percent of the market level, while Fannie Mae performed at 96 percent of the market level—both results similar to those reported above for underserved areas based on 1990 Census geography. The 2000 Census data show that the Fannie Mae has been much closer to the market during the recent 2001-2003 period. The share of Fannie Mae's purchases going to underserved areas was 31.4 during 2001-2003, which placed it close to the market level of 31.7 percent. However, the 2000-based results show that, like Freddie Mac, Fannie Mae's longer-term performance (since 1996) as well as its recent average performance (1999 to 2003) have consistently been below market levels. (Note that the 1996-2003 averages reported above are estimated by adding the following 2000-Census versus 1990-Census differentials calculated for 1999-2003: 5.2 percentage points for Freddie Mac, 5.3 for Fannie Mae, and 5.2 for the market.)

Underserved Area Subgoal for Home Purchase Loans. The Department is establishing a subgoal of 32 percent for each GSE's acquisitions of home purchase loans financing single-family-owner properties located in the underserved census tracts of metropolitan areas for 2005, with this subgoal rising to 33 percent for 2006 and 2007, and to 34 percent in 2008. If the GSEs meet the 2008 subgoal, they will be leading the primary market by over two percentage points, based on historical data. This home purchase subgoal will encourage the GSEs to provide additional credit and capital to urban neighborhoods that historically have not been adequately served by the mortgage industry—but in the future may be the very neighborhoods where the growing population of immigrants and minorities choose to live. As detailed in Section I.5 of this appendix, there are four specific reasons for establishing this subgoal: (1) The GSEs have the expertise, resources, and ability to lead the single-family-owner market, which is their “bread and butter” business; (2) the GSEs have been lagging the primary market in underserved areas, not leading it; (3) the GSEs can help reduce troublesome neighborhood disparities in access to mortgage credit; and (4) there are ample opportunities for the GSEs to expand their purchases in low-income and high-minority neighborhoods. Sections E.9 and G of Appendix A provide additional information on the opportunities for an enhanced GSE role in underserved area segment of the home purchase market and on the ability of the GSEs to lead that market.

As discussed above, underserved areas accounted for an average of approximately 31.5 percent of home purchase loans originated in the conventional conforming market of metropolitan areas (computed over 1999-2003 or over 2001-2003). To reach the 34-percent subgoal for 2008, both GSEs will have to improve over their earlier peak performances—Freddie Mac by 2.3 percentage points over its previous peak performance of 31.7 percent in 2002, and Fannie Mae by 1.7 percentage points over its previous peak performance of 32.3 percent in 2003. To meet the 2008 subgoal, Freddie Mac will have to improve by 2.6 percentage points over its 2002-2003 average (unweighted) performance of 30.4 percent, while Fannie Mae will have to improve by 1.8 percentage points over its 2002-2003 average performance of 32.2 percent.

The subgoal applies only to the GSEs' purchases in metropolitan areas because the HMDA-based market benchmark is only available for metropolitan areas. HMDA data for non-metropolitan counties are not reliable enough to serve as a market benchmark. The Department is also setting home purchase subgoals for the other two goals-qualifying categories, as explained in Appendices A and C.

b. Characteristics of GSEs' Purchases of Mortgages on Properties in Metropolitan Underserved Areas

Several characteristics of loans purchased in 2003 by the GSEs in metropolitan underserved areas are presented in Table B.9. As shown, borrowers in underserved areas are more likely than borrowers in served areas to be first-time homebuyers, all female, all male and younger than 40. And, as expected, borrowers in underserved areas are more likely to have below-median income and to be members of minority groups. For example, first-time homebuyers make up 6.7 percent of the GSEs' mortgage purchases in underserved areas and 4.2 percent of their business in served areas. In underserved areas, 53.7 percent of borrowers had incomes below the area median, compared with 36.4 percent of borrowers in served areas.

Minorities' share of the GSEs' mortgage purchases in underserved areas (33.2 percent) was greater than two times their share in served areas (13.9 percent). And the pattern was even more pronounced for African Americans and Hispanics, who accounted for 23.1 percent of the GSEs' business in underserved areas, but only 7.0 percent of their purchases in served areas.

Other similarities in Fannie Mae and Freddie Mac purchases in served and underserved areas include the following. The GSEs are slightly more likely to purchase refinance loans in served areas than in underserved areas; mortgage purchases with loan-to-value ratios below 80 percent are more likely to be in underserved than in served areas; and seasoned mortgage purchases are more likely to be in underserved than in served areas.

3. GSE Mortgage Purchases in Nonmetropolitan Areas

There are numerous studies that have evaluated the impact of the GSEs' purchases on metropolitan areas, but few address the impact on nonmetropolitan areas; therefore, our understanding of the GSEs and the nonmetropolitan markets is very limited.

A study of the GSE market share in underserved counties ^[47] found that location has a role in the accessibility of credit for some people in nonmetropolitan areas (low income, minority, and first-time homebuyers). West North Central counties (Minnesota, Missouri, South Dakota, Iowa, Kansas, Nebraska, and North Dakota) have much lower GSE activity than all other geographic regions, suggesting that the 1995 definition of underservice does not capture the specific characteristics of this region, leading to limited GSE activity.

Additionally, The Urban Institute prepared a report for HUD that investigated the factors influencing GSE activity in nonmetropolitan areas.^[48] The authors found that Fannie Mae and Freddie Mac have increased their lending to nonmetropolitan areas since 1993; however, there are still weak areas in terms of the percentage of affordable loans being offered.^[49] They also established that GSE underwriting criteria was not a major barrier in nonmetropolitan areas.

In nonmetropolitan areas, the financial market is often made up of locally owned community banks, manufactured home lenders, and subprime lenders. Industry representatives contacted by the Urban Institute researchers assessed that the barriers nonmetropolitan lenders faced were in the areas of availability of sales comparables, technology, and the type and number of lenders in the area. They also believed that for the GSEs' market share to improve in underserved nonmetropolitan areas, the GSEs would have to begin to build relationships with the community lenders and provide education/training on how to sell loans directly to the GSEs rather than using intermediaries.

a. Effects of 2000 Census Geography

In order to compare served and underserved areas, either in terms of GSE performance or socioeconomic characteristics, it is first necessary to update current geographic (county) designations, which reflect 1990 census median income and minority population data, to reflect newly available 2000 census data. Table B.10 shows the impact on 2000, 2001, and 2002 GSE purchases. These are reported for total GSE purchases and separately for Fannie Mae and Freddie Mac. As above, the results also are shown separately for counties that change classification and those that do not. This analysis is limited to nonmetropolitan areas based on both the pre- and post-June, 2003 OMB metropolitan area designations.

Applying 2000 census median income and minority population data results in a slight drop in the proportion of counties that are classified as underserved. Out of a total of 2,493 counties, 1,514 (65.5 percent) are underserved based on 1990 data, and 1,260 (61.4 percent) based on 2000 data. This small net change disguises a somewhat larger shift of counties, as about 11.2 percent of currently underserved counties are reclassified as served counties and 4.6 percent of currently served counties are reclassified as underserved.

Comparing underserved and served nonmetropolitan areas in Table B.10, it is apparent that underserved nonmetropolitan areas make up a larger percentage of nonmetropolitan areas as a whole than do served nonmetropolitan areas, as shown by the number of counties (1,260 for underserved (61.4%); 792 for served (38.6%)). These relationships hold true also for the number of households (9.5 million for underserved (50.5%); 9.3 million for served (49.5%)), and the population (24.9 million for underserved (51%); 23.9 million for served (49%)) as shown in Table B.5.

Table B.10 shows that Fannie Mae's performance in 2002 (40.2 percent) was somewhat higher than Freddie Mac's (36.3 percent). This gap widens slightly (1.8 percent) in applying 2000 census income and minority data and 2003 metropolitan area definitions.

b. Characteristics of GSEs' Purchases of Mortgages on Properties in Nonmetropolitan Underserved Areas

Nonmetropolitan mortgage purchases made up 12.6 percent of the GSEs' total mortgage purchases in 2003. Mortgages in underserved counties made up 38.6 percent of the GSEs' business in nonmetropolitan areas.^[50]

Unlike the underserved areas definition for metropolitan areas, which is based on census tracts, the rural underserved areas definition is based on counties. Rural lenders argued that they identified mortgages by the counties in which they were located rather than the census tracts; and therefore, census tracts were not an operational concept in rural areas. Market data on trends in mortgage lending for metropolitan areas are provided by HMDA; however, no comparable data source exists for rural mortgage markets. The absence of rural market data is a constraint for evaluating credit gaps in rural mortgage lending and for defining underserved areas.

One concern is whether the broad definition overlooks differences in borrower characteristics in served and underserved counties that should be included. Table B.11 compares borrower and loan characteristics for the GSEs' mortgage purchases in served and underserved areas.

Fannie Mae is slightly more likely and Freddie Mac is less likely to purchase loans for first-time homebuyers in underserved areas than in served areas. Mortgages to first-time homebuyers accounted for 4.3 percent of Fannie Mae's mortgage purchases in served counties, compared with 4.6 percent of its purchases in underserved counties. For Freddie Mac the corresponding figures are 3.4 percent in served counties and 3.3 percent in underserved counties.

The GSEs are more likely to purchase mortgages for high-income borrowers in underserved than in served counties. Surprisingly, borrowers in served counties were more likely to have incomes below the median than in underserved counties (39.6 percent compared to 35.4 percent). These findings lend some support to the claim that, in rural underserved counties, the GSEs purchase mortgages for borrowers that probably encounter few obstacles in obtaining mortgage credit.

The following similarities in Fannie Mae and Freddie Mac purchases in served and underserved counties in nonmetropolitan areas mirror those found for the GSEs in served and underserved census tracts in metropolitan areas. The GSEs are slightly more likely to purchase refinance loans in served than in underserved counties; mortgage purchases with loan-to-value ratios below 80 percent are more likely to be in underserved than in served counties; and seasoned mortgage purchases are more likely to be in underserved than in served counties.

E. Factor 4: Size of the Conventional Conforming Mortgage Market for Underserved Areas

HUD estimates that underserved areas account for 35-39 percent of the conventional conforming mortgage market. The analysis underlying this estimate is detailed in Appendix D.

F. Factor 5: Ability to Lead the Industry

This factor is the same as the fifth factor considered under the goal for mortgage purchases on housing for low- and moderate-income families. Accordingly, see Section G of Appendix A for a discussion of this factor, as well as Section I.5 of this Appendix, which describes the home purchase subgoal which is designed to place the GSEs in a leadership role in the underserved market.

G. Factor 6: Need to Maintain the Sound Financial Condition of the Enterprises

HUD has undertaken a separate, detailed economic analysis of this rule, which includes consideration of (a) the financial returns that the GSEs earn on loans in underserved areas and (b) the financial safety and soundness implications of the housing goals. Based on this economic analysis and reviewed by the Office of Federal Housing Enterprise Oversight, HUD concludes that the goals raise minimal, if any, safety and soundness concerns.

H. Defining Nonmetropolitan Underserved Areas

1. Whether to Adopt a Tract-Based Definition of Underserved Areas

The current county-based definition for targeting GSE purchases to underserved nonmetropolitan areas was adopted in 1995 over alternative narrower definitions, such as census tracts, despite the use of census tracts in metropolitan areas. In the 1995 Final Rule, HUD found the merits of a county-based system of targeting outweighed a tract-based system. Now, with seven years of experience under a county-based system, the release of Census 2000 data, and improvements in information technology and systems, HUD can reexamine whether to switch to census tracts for defining underserved nonmetropolitan areas. This section compares impacts of the potential shift in definition for both served and underserved populations as determined by tract-based and county-based definitions using a number of common industry variables as focal points for analysis.

The rationale for choosing counties in 1995 rested primarily on perceived shortcomings of census tracts.^[51] In particular, rural lenders did not perceive their market areas in terms of census tracts, but rather, in terms of counties. Another concern was a perceived lack of reliability in geocoding 1990 census tracts. At the same time, HUD found merit in using a tract-based geography for nonmetropolitan areas. Because tracts encompass more homogeneous populations than counties, they permit more precise targeting of underserved populations. In other words, more homogeneous geographic areas increase the potential for targeting the GSE mortgage purchases into areas where borrowers are more likely to face obstacles and other challenges in securing mortgage credit.

The criteria used for this analysis include the following:

7. Do tracts provide a sharper delineation of served and underserved areas? Specifically, are underserved nonmetropolitan populations more clearly differentiated by adopting tracts vs. counties? Could service to the underserved nonmetropolitan populations be more comprehensive under tract-based definitions?

8. What is the impact on GSE purchasing patterns if underserved areas are defined by tract?

9. Applying the current criteria for identifying underserved areas to tracts would result in reclassifying approximately 23 percent of all tracts, with 28 percent of tracts in served counties being redesignated as underserved and 19 percent of tracts in underserved counties being redesignated as served. Overall, roughly the same percentage of families (and population) would be reclassified. However, because underserved tracts are somewhat less densely populated than served tracts, the corresponding proportions of families that shift from served and underserved counties are closer: 25 vs. 21 percent.

a. Do Census Tracts Allow a Sharper Delineation of Served and Underserved Areas?

This section compares the differences in housing need and economic, demographic, and housing conditions in served and underserved nonmetropolitan areas classified on, respectively, counties and tracts. Additionally, the “efficiency” with which counties and tracts cover the target populations is compared. That is, does tract-based targeting do a better job of capturing lower income households and excluding higher income households than county-based targeting?

Table B.12 presents several indicators of socioeconomic and housing condition in served and underserved areas under both a tract-based and a county-based definition. In addition, served and underserved counties are subdivided into their served and underserved tract components. This allows a closer examination of the population and housing characteristics of the tracts that are reclassified (i.e., served to underserved or visa versa) under tract-based targeting. Thus, area characteristics of housing need and housing, economic, and demographic conditions can be compared, for the following four groups of tracts: (1) Tracts in served counties that would remain “served” classified as tracts; (2) tracts that remain “underserved”; (3) tracts that shift from served to underserved; and (4) tracts that shift from underserved to served. In addition, we provide counts of tracts falling into each of these groups. If a tract-based classification of underserved areas improves geographic targeting, the regrouping of tracts would be more similar to one another than to the other tracts in their respective counties: e.g., formerly underserved areas that become served should be more similar to tracts that were and remain served than to underserved (unchanged).

Socioeconomic and Demographic Conditions. Table B.12 shows that in important socioeconomic and demographic characteristics, tract-based targeting would more effectively distinguish underserved populations. Median family income, poverty, unemployment, school dropout rates, and minority population all exhibit greater differences between served and underserved areas using tracts. For example, the difference in median income between served and underserved counties is $9,579, or alternatively, between served and underserved tracts, the difference is $12,744. Similarly, there is a 7-percentage point gap in poverty rates (7.5 vs. 14.5 percent poverty) using counties, which widens to 8.6 percentage points (6.6 vs. 15.3 percent) using tracts. Minority population also is captured somewhat better with tracts, with the served/underserved gap increasing from 16.5 to 17.3 percentage points. In all cases, the levels of the indicators for underserved areas move in a direction consistent with targeting lower income households and areas with higher minority populations.

The 4-way breakdown of served and underserved counties reveals some significant differences between the two component groups. In most respects, “underserved tracts” (i.e., those meeting the underserved criteria), whether located in an underserved or served county, are more alike than they are like served tracts. Using median income again to illustrate, the effect of reclassifying areas by tract characteristics is to put together two groups of underserved tracts: Tracts that were in previously underserved counties and are not reclassified and tracts that were in served counties but meet the underserved criteria. A new group of served tracts is similarly formed. In both cases, the difference in median incomes of the constituent groups is about $3,500. In contrast, the served and underserved counties now encompass “served” and “underserved” groups of tracts whose respective median incomes differ by almost $11,000. Combined with the fact that a fairly large number of tracts are affected overall (i.e., switch), these results support an assessment that counties are relatively crude for targeting underserved populations.

Housing Needs and Conditions. Table B.12 shows that tract-based targeting would produce modest gains in focusing GSE purchases on areas with relatively greater housing needs and conditions as measured by low owner-occupancy, higher vacancy rates, and crowding. For each of these indicators, measured need increases in underserved areas and the gap between served and underserved areas widens when tracts are used to classify areas. Most notably, the percent of owner-occupied housing units switches from being higher in underserved than served counties to being significantly lower among underserved tracts. With a shift to tracts overall ownership drops in underserved areas, from 74 to 72 percent, and increases in served areas from 74 to 77 percent. In contrast, the homeownership rate for tracts located in served counties that would be deemed underserved if judged separately is only 65 percent. In fact, this rate is much lower even than underserved tracts in underserved counties. Shifting these tracts from served to underserved largely accounts for the switching of homeownership rates.

Results for other indicators of housing need and conditions are less clear-cut. No definitive patterns are apparent for two, admittedly weak, measures of housing quality—units with complete plumbing and units with complete kitchen facilities, as well as for crowding. Purchase affordability, as measured by the ratio of median housing value to the income necessary to qualify for a loan for the median valued unit, is higher in underserved areas than in served areas. However, the measure of purchase affordability presented here is influenced by many market and other economic factors, some of which do not relate to housing need. For example, a low affordability ratio may reflect abundant supply, but it may also reflect low demand stemming from, e.g., limited availability of credit or high interest rates.

Coverage Efficiency. The coverage efficiency index measures the effect of adopting tract-based targeting. This index can be used to indicate how well underserved areas encompass populations deemed to be underserved (“sensitivity”) and to exclude populations that are deemed to be served (“specificity”). The index is computed for median income as the difference in two percentages: (1) The proportion of all families in nonmetropolitan areas that meet the applicable income threshold who live in underserved tracts minus (2) the proportion of all families in nonmetropolitan areas that do not meet the applicable underserved income threshold who live in underserved areas. This difference can range from 1 (perfect) to −1 (bad; perverse). For example, a coverage efficiency index equal to 1 implies that every family in need is living in an underserved area while there are no families who are not in need living in an underserved area; a coverage efficiency index equal to −1 implies that none of the families in need live in an underserved area, or equivalently, all families in underserved areas are not in need.

Comparing coverage efficiency for counties and tracts indicates that tracts do a better job; capturing a higher percentage of nonmetropolitan families whose income falls below the applicable income threshold and excluding more families whose income exceeds the threshold.^[52] Overall, the efficiency index rises from 0.22 to 0.274.

Given income thresholds that are not far away from median income in most places and the degree of income variation even with census tract boundaries, it should not come as a great surprise that neither the levels of coverage efficiency (0.22-0.27) nor improvement produced in applying tracts (5 percentage points) are not more dramatic. Nevertheless, tracts do produce better tracking of lower income, very low income, and minority families.

b. Does GSE Performance Vary Between Served and Underserved Tracts Within Underserved Counties?

A similar analytical approach is used to examine how a shift to tracts would impact GSE purchases. Having applied income and minority thresholds from the 2000 census and updating census tract geography, Table B.13 compares, respectively, 2000, 2001, and 2002 GSE purchases for served and underserved counties and tracts and also for the served and underserved tracts within county boundaries. On net there would be somewhat more tracts classified as underserved under a tract-based system than currently: 6,782 vs. 6,414. As noted above, however, 23.1 percent of all tracts are reclassified. Moving to tracts also would have a significant effect on the relative performance of the GSEs. In 2002, Fannie Mae's performance would drop 2.1 percentage points to 35.4 percent, while Freddie Mac's performance would increase by 0.9 percent to 32.7 percent.

Differences between qualifying purchases of single-family and multifamily loans are further increased when assessed at the tract level. Performance for single-family loans drops 0.7 percentage points to 35.2, but for multifamily increases by 2.5 percentage points to 46.8. These changes dramatically compound the results observed in updating to 2000 census data, resulting in a widening of the single- and multifamily performance difference from the current level of 7.0 percentage points to 11.6 percentage points.

2. Alternative Definitions of Underservice

The current definition of underservice in nonmetropolitan areas was established in 1995 to be relatively broad, encompassing nearly twice as many underserved as served counties and somewhat more than half of the total nonmetropolitan population. This was done primarily to ensure that certain areas with low incomes and/or high minority populations, which might not be considered underserved in comparison to the rest of their State, would nevertheless be identified as underserved from a national perspective. This section summarizes a new analysis, based on 2000 census data, to evaluate the extent to which the current definition focuses GSE purchasing activity toward stimulating mortgage lending in areas with populations having greatest housing need. Alternative definitions of underservice are considered as follows: (1) Variations of the current thresholds; (2) applying only the State median income level for qualifying underserved counties and tracts; and (3) establishing different thresholds in micropolitan and “outside of core” nonmetropolitan areas. In each case the objective is to assess how redesignating served and underserved areas would affect relative conditions and needs and GSE purchasing performance. In distinguishing micropolitan and “outside of core” areas, it is of interest to determine whether it would be appropriate to establish different thresholds for underservice. The overarching criterion for evaluating and comparing definitions is their ability to serve very low-income, low-income and moderate-income households, households in poverty, first-time homebuyers, minorities, and households in remote locations.^[53]

In the current definition, areas are classified as underserved if either the minority population share is greater than 30 percent and median income is less than 120 percent of the greater of State nonmetropolitan or national nonmetropolitan median income; or area median income is less than or equal to 95 percent of the greater of State nonmetropolitan or national nonmetropolitan median income. The greater of State nonmetropolitan or national median income is termed the “reference income.” Denoting the current thresholds as “30/120/95,” the following set of alternative thresholds are evaluated:

• 30/120/95 vs. 30/120/90 vs. 30/120/80—to examine the effect of lowering the general income threshold from 95 percent to 90 percent to 80 percent.
• 30/120/95 vs. 30/110/95 vs. 30/110/80—to examine the effect of lowering both the minority (from 120% to 110%) and general income (from 95% to 80%) thresholds; and
• 30/120/95 vs. 50/120/95—to examine the effect of increasing the minority population threshold that must be attained before applying the minority income threshold.

For each alternative, indicators of socioeconomic and housing conditions are calculated for served and underserved areas for each alternative and compare the results to the current definition. Of particular interest is whether certain thresholds of minority population and median income capture the differences in housing needs and conditions between served and underserved areas better than others. The “coverage efficiency” of each alternative relative to households below the poverty line, below 50, 70, and 95 percent of area reference income, and below the alternative income level(s) used to define underservice, is also presented. GSE purchasing activity is also examined for each alternative definition, specifically, the percentage of eligible loans that qualify towards the goal for underserved areas defined by different thresholds. Each analysis is conducted both with counties and tracts as the geographic unit.

County Results. The main effect of lowering the general income threshold from 95 to 90 to 80 percent of the reference income is to roughly halve the number of counties and population residing in underserved areas. Under the current definition, 11.6 million people reside in underserved areas as opposed to fewer than 10 million in served areas. With a general income threshold of 80 percent, 5.7 million would be left in underserved areas. A 90 percent threshold would produce a shift of approximately half this amount.

In terms of social, economic, demographic, and housing characteristics, lowering the income threshold from 95 to 80 percent would have the following notable consequences:

• Minority population in underserved areas would increase from 12.4 to 20.8 percent with no significant change in served areas.
• Median income would fall in both served and underserved areas with the difference remaining nearly constant at $10,000.
• Poverty, unemployment, school drop out rates all would be higher in both served and underserved areas. The gap would increase for each of these characteristics.
• Migration into underserved areas (from other States) would be relatively lower than into served areas with an 80 percent income threshold.
• Indicators of homeownership would decline somewhat in underserved areas relative to served areas. For all units, for example, ownership would decline from 74.3 to 72.9 percent in underserved areas and increase from 73.5 to 74.3 percent in served areas.
• Median housing values would fall in both served and underserved areas with a significant narrowing in the gap from approximately $25,000 to $19,000 at an 80 percent median income threshold.
• Housing affordability would decline in underserved areas, becoming nearly equal with affordability in served areas at 80 percent.
• Crowding would be higher in underserved areas, absolutely and relative to served areas. Thus, more narrowly defined underserved areas would more strongly manifest conditions and needs associated with underservice: lower income, higher poverty, higher minority populations, lower homeownership, lower affordability, more crowding, etc. However, served areas would expand to encompass significant numbers of these same underserved and target populations.

Use of the coverage efficiency index highlights one of the tradeoffs between using a low median income threshold versus a high median income threshold in redefining underservice. Coverage efficiency based on all variables examined, including “underserved,” poor, very low income, low income and even moderate income families, declines sharply as the income threshold is lowered from 95 to 80 percent, becoming negative for most groups. Coverage for the “underserved” cohort declines from 22.0 to −1.0 percent, and for families with up to 95 percent of reference income, it declines from 17.2 to −10.0 percent. These changes result from losing almost half of the families in target income ranges without any appreciable gain in specificity, i.e., shrinking the proportion of people living in underserved counties with incomes above the respective target levels. Similar patterns are observed for families with below 70 percent of reference income, below 50 percent of reference income, and families in poverty.

The second set of comparisons builds on the first set by lowering the income threshold applicable to areas with relatively high minority populations (30 percent) from 120 to 110 percent in addition to the general threshold. This change further shrinks, albeit, only marginally, the size and population of underserved areas. Minority underserved populations would be smaller and socioeconomic and housing conditions would be worse. Not surprisingly, coverage efficiencies and GSE purchase performance levels also would decline across the board, although the marginal effects of reducing the minority income threshold are quite small. The 30/110/80 alternative is the narrowest definition examined and produces the biggest losses in efficiency and GSE performance.

The third variation of the current definition is an increase in the minority population threshold from 30 to 50 percent. Thus, if an area does not qualify as underserved against the general income threshold of 95 percent it could still qualify if its population is 50 percent minority and median income is less than or equal to 120 percent of the reference income level.

Relatively few counties qualify solely under the current minority thresholds. Raising the population threshold would trim this number by an additional 73 counties (457 tracts). Not surprisingly, the percent minority in underserved areas would decrease. However, the areas being redesignated as served are apparently somewhat above average in terms of socioeconomic and housing conditions in underserved areas and below-average in terms of conditions in served areas. Coverage efficiencies for all cohorts would be lower than for the current definition of underservice and GSE performance overall would be approximately 90 percent of the current level.

Using the State median income, alone, as the general reference income would reduce the number underserved counties relative to the current definition, and, although there would still be more underserved counties (1,274 vs. 1,064), the underserved population actually would become smaller than the served population. The effect of this alternative on differences in housing conditions and needs between served and underserved areas is generally small and ambiguous, but overall, results in less contrast. Consistent with the results for other alternatives, applying a State median income standard, alone, would result in lower coverage efficiency across all target groups.

Census Tract Results. As discussed above, the adoption of a tract-based system would result in greater coverage efficiency of underserved populations and sharper distinctions in the socioeconomic, demographic and housing characteristics of served and underserved areas. That is, tracts more effectively carve out areas that exhibit characteristics that are associated with underservice, such as low income, large minority populations and low homeownership. The converse is true for served areas. In analysis at the tract level, these patterns tend to be maintained quite consistently. A tract-based system would improve the power to differentiate underserved and served populations. According to virtually every indicator of socioeconomic, demographic, and housing conditions, applying State median income, alone, with a tract-based geography would produce superior differentiation to the current county-based definition. In terms of coverage efficiency, we again see improvement with tracts, but not enough to offset the loss of eliminating the national median income threshold. For the underserved population, for example, coverage efficiency would be 16.9 percent with tracts, still below 22 percent under the current definition.^[54]

I. Determination of the Underserved Areas Housing Goal

The annual goal for each GSE's purchases of mortgages financing housing for properties located in geographically targeted areas (central cities, rural areas, and other underserved areas) is 37 percent of eligible units financed in 2005, 38 percent in 2006 and 2007, and 39 percent in 2008. The 2008 goal will remain in effect in subsequent years, unless changed by the Secretary prior to that time. The goal of 37 percent for 2005 is larger than the goal of 31 percent for 2001-03 mainly because, compared with the 1990 Census, the 2000 Census includes a larger number of census tracts that meet HUD's definition of underserved area. The new 37 percent-39 percent goals are commensurate with recent market share estimates of 37-39 percent for 1999-2002, presented in Appendix D.

In addition, an Underserved Areas Housing Subgoal of 32 percent is established for the GSEs' acquisitions of single-family-owner home purchase loans in metropolitan areas in 2005, with the subgoal rising to 33 percent in 2006 and in 2007 and 34 percent in 2008. The subgoal is designed to encourage the GSEs to lead the primary market in providing mortgage credit in underserved areas.

This section summarizes the Secretary's consideration of the six statutory factors that led to the Underserved Area Housing Goal and the subgoal for home purchase loans in metropolitan areas. This section discusses the Secretary's rationale for defining underserved areas and it compares the characteristics of such areas and untargeted areas. The section draws heavily from earlier sections which have reported findings from HUD's analyses of mortgage credit needs as well as findings from other research studies investigating access to mortgage credit.

1. Housing and Credit Disparities in Metropolitan Areas

To identify areas underserved by the mortgage market, HUD focused on two traditional measures used in a number of studies based on HMDA data: Application denial rates and mortgage origination rates per 100 owner-occupied units. Tables B.2 and B.3 in Section B of this Appendix presented detailed data on denial and origination rates by the racial composition and median income of census tracts for metropolitan areas. Aggregating this data is useful in order to examine denial and origination rates for broader groupings of census tracts: ^[55]

Minority composition (percent)	Denial rate (percent)	Orig. rate
0-30	9.6	26.7
30-50	12.4	26.9
50-100	17.2	20.8

Tract income	Denial rate (percent)	Orig. rate
Less than 90% of AMI	16.9	18.1
90-120%	11.3	25.4
Greater than 120%	7.8	32.7

Two points stand out. First, high-minority census tracts have higher denial rates and lower origination rates than low-minority tracts. Specifically, tracts that are over 50 percent minority have nearly twice the denial rate and two-thirds the origination rate of tracts that are under 30 percent minority.^[56] Second, census tracts with lower incomes have higher denial rates and lower origination rates than higher income tracts. Tracts with income less than 90 percent of area median income have over twice the denial rate and almost half of the origination rate of tracts with income over 120 percent of area median income.

In both the 1995 and the 2000 GSE Rules, HUD's research determined that “underserved areas” could best be characterized in metropolitan areas as census tracts where: (1) Median income of families in the tract does not exceed 90 percent of area (MSA) median income or (2) minorities comprise 30 percent or more of the residents and median income of families in the tract does not exceed 120 percent of area median income. The earlier analysis was based on 1990 Census data. HUD has now conducted the same analysis using 2000 Census data and has determined that the above definition continues to be a good proxy for underserved areas in metropolitan areas. The income and minority cutoffs produce sharp differentials in denial and origination rates between underserved areas and adequately served areas. For example, in 2003 the mortgage denial rate in underserved areas (15.9 percent) was over one and three-fourths times that in adequately served areas (8.9 percent).

These minority population and income thresholds apply in the suburbs as well as in central cities. The average denial rate in underserved suburban areas (14.8 percent) is 1.7 times that in the remaining served areas of the suburbs (8.7 percent), and is almost as large as the average denial rate (16.8 percent) in underserved central city tracts. Low-income and high-minority suburban tracts appear to have credit problems similar to their central city counterparts. Thus HUD uses the same definition of underserved areas throughout metropolitan areas—there is no need to define such areas differently in central cities and in the suburbs.

This definition of metropolitan underserved areas based on 2000 Census geography includes 26,316 of the 51,040 census tracts in metropolitan areas, covering 49.2 percent of the metropolitan population in 2000. (By contrast, the 1990-based definition included 21,587 of the 45,406 census tracts in metropolitan areas, covering 44.3 percent of the metropolitan population in 1990.) The 2000-based definition includes 75.7 percent of the population living in poverty in metropolitan areas. The unemployment rate in underserved areas is more than twice that in served areas, and owner units comprise only 51.6 percent of total dwelling units in underserved tracts, versus 75.9 percent of total units in served tracts. As shown in Table B.14, this definition covers most of the population in several distressed central cities including Bridgeport (100 percent), Newark (99 percent), and Detroit (93 percent). The nation's five largest cities also contain large concentrations of their population in underserved areas: New York (68 percent), Los Angeles (72 percent), Chicago (75 percent), Houston (73 percent), and Phoenix (50 percent).

2. Identifying Underserved Portions of Nonmetropolitan Areas

Based on an exploration of alternative numerical criteria for identifying underserved nonmetropolitan areas using 2000 census data, HUD has concluded that the current definition of underservice is broad but efficacious and that any narrower definition of underservice would not serve congressional intent under FHEFSSA. Narrowing the definition of underservice potentially could promote more intense purchasing in needier communities, but this seems unlikely. On the contrary, the greatest marginal impact of GSE purchasing could be in the very areas that would be excluded under the alternatives.

Research comparing a tract-based system for defining underserved areas with the current county-based system, using 2000 census data, indicates that a tract-based system would result in more effective geographic targeting of GSE purchases. Although the total number of tracts designated as served and underserved areas would change very little, 23 percent of all tracts would be reclassified, reassigning approximately equal numbers of families from served to underserved and from underserved to served.

The main effect of the reclassification is to align tracts into more homogeneous and distinct groups as measured by differences in key socioeconomic and demographic characteristics such as median family income, poverty, unemployment, school dropouts, and minority population. As a result of reclassification, underserved areas stand out more as areas of lower income and economic activity and somewhat larger minority populations.

Tract-based targeting would potentially focus GSE purchases in areas with relatively greater housing needs and conditions as measured by owner-occupancy, vacancy rates, and crowding. For each of these indicators, measured need increases in underserved areas and the gap between served and underserved areas widens when tracts are used to classify areas. Most notably, homeownership would be significantly lower in underserved areas relative to served areas under a tract-based system. Currently, and contrary to expectations, homeownership actually is slightly greater in underserved areas. Driving this reversal is the fact that tracts in served counties that would be reclassified as underserved tracts have an ownership rate of just 65 percent, which is much lower even than in the underserved tracts in underserved counties, where ownership is 73 percent. Meanwhile, the served tracts in served and underserved counties have the same ownership rate of 77 percent, which is significantly higher than in underserved areas.

Two groups of measures of housing conditions—housing quality and affordability—exhibit less clear-cut results from applying tracts. However, we conclude that these results are consistent with the ambiguous patterns discussed in chapter 4 above and do not undermine the overall conclusion that basing geographic targeting on tracts would more sharply define areas with greater housing need and adverse housing conditions.

Not surprisingly, the results from analyzing housing, socioeconomic, and demographic characteristics are further reinforced in finding that a tract-based system would better capture underserved populations and exclude served populations from geographic targeting. Defining underserved families as those in any area whose income was less than 95 percent of the reference income (or in areas with a minority population of 30 percent or more, families with incomes below 120 percent of the reference income) the use of more refined tract geography results in a 5 percentage point increase in the coverage efficiency index, from 22 to 27 percent. This reflects two improvements under a tract system: Underserved areas would capture more of the nonmetropolitan “underserved” families (62 vs. 65 percent) and fewer “served” families (decreasing from 40 to 37 percent of families in underserved areas).

3. Past Performance of the GSEs

Goals Performance. In the October 2000 rule, the underserved areas goal was set at 31 percent for 2001-03. Effective on January 1, 2001, several changes in counting requirements came into effect for the undeserved areas goal, as follows: (a) “Bonus points” (double credit) for purchases of mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on 2-4 unit owner-occupied properties; (b) a “temporary adjustment factor” (1.35 units credit) for Freddie Mac's purchases of mortgages on large (more than 50 unit) multifamily properties; and (c) eligibility for purchases of certain qualifying government-backed loans to receive goal credit. Under these counting rules, as shown in Table B.7a and Figure B.2, Fannie Mae's performance was 32.6 percent in 2001, 32.4 percent in 2002, and 32.1 percent in 2003, while Freddie Mac's performance was 31.7 percent in 2001, slightly less than 31 percent in 2002, and 32.7 percent in 2003; thus Fannie Mae surpassed the goal of 31 percent in all thee years, while Freddie Mac fell slightly short of the goal in 2002.

Counting requirements (a) and (b) expired at the end of 2003, while (c) will remain in effect. If this counting approach—without the bonus points and the “temporary adjustment factor”—had been in effect in 2000-03, and the GSEs’ had purchased the same mortgages that they actually did purchase in both years, then Fannie Mae's performance would have been 31.0 percent in 2000, 30.4 percent in 2001, 30.1 percent in 2002, and 29.2 percent in 2003. Freddie Mac's performance would have been 29.2 percent in 2000, 28.2 percent in 2001, 28.0 percent in 2002, and 27.7 percent in 2003. Therefore, Fannie Mae would have just matched the underserved areas goal of 30 percent in 2000 and fallen short in 2001-03, while Freddie Mac would have fallen short of the goal in all four years, 2000-2003.

The above performance figures are for underserved areas (census tracts in metropolitan areas and counties in non-metropolitan areas) defined in terms of 1990 Census geography. Switching to 2000 Census data increases the coverage of underserved areas, which increases the share of the GSEs' purchases in underserved areas by approximately 5 percentage points. Based on 2000 Census geography, and excluding counting requirements (a) and (b) then Fannie Mae's performance would have been 38.1 percent in 2000, 36.6 percent in 2001, 35.9 percent in 2002, and 34.1 percent in 2003, as shown in Table B.7b. Freddie Mac's performance would have been 35.1 percent in 2000, 33.5 percent in 2001, 33.3 percent in 2002, and 31.6 percent in 2003.

Single-Family-Owner Home Purchase Mortgages. Sections E.9 of Appendix A and D.2 of this appendix compared the GSEs' funding of home purchase loans in underserved areas with originations by lenders in primary market. To take advantage of HMDA and GSE data going back to 1993, the analysis was conducted using 1990 Census tract geography. While both GSEs have improved their performance since 1993, they have both lagged the conventional conforming market in providing affordable loans to underserved areas. The 1990-based analysis shows that the two GSEs have engaged in very different patterns of funding—Freddie Mac has been much less likely than Fannie Mae to fund home loans in underserved neighborhoods. HUD will begin defining underserved areas based on 2000 Census geography and new OMB definitions of metropolitan areas in 2005, the first year of the rule. As noted above, the 2000-based definition of underserved areas includes 5,372 more census tracts in metropolitan areas than the 1990-based definition, which means the GSE-market comparisons need to be updated to incorporate tract designations from the 2000 Census. Therefore, for the years 1999, 2000, 2001, and 2002, HUD used various apportionment techniques to re-allocate 1990-based GSE and HMDA data into census tracts as defined by the 2000 Census. The GSE and HMDA data for 2003 were already based on 2000 geography, so no apportionment was needed for that year. Switching to the 2000-based tracts increases the underserved area share of market originations by 5.7 percentage points. Between 1999 and 2002, 31.4 percent of mortgage originations (without B&C loans) were originated in underserved tracts based on 2000 geography, compared with 26.2 percent based on 1990 geography. As shown in Table B.8 of Section D.2, the underserved areas share of each GSE's purchases also rises by approximately five percentage points. Thus, conclusions about the GSEs' performance relative to the market are similar whether the analysis is conducted in terms of 2000 Census geography or 1990 Census geography.

The analysis for home purchase loans based on 2000 Census geography will be summarized here (see Section D.2 of this appendix for a similar analysis using 1990-based geography):

Year	Freddie Mac (percent)	Fannie Mae (percent)	Market (w/o B&C) (percent)
1999	25.6	25.3	30.2
2000	27.3	29.0	31.7
2001	27.3	29.8	30.7
2002	31.7	32.3	31.8
2003	29.0	32.0	32.5
1996-2003 (estimate)	27.2	29.3	30.9
1999-2003 (average)	28.3	30.0	31.4
2001-2003 (average)	29.4	31.4	31.7

Between 1999 and 2003, 28.3 percent of Freddie Mac's purchases and 30.0 percent of Fannie Mae's purchases financed properties in underserved neighborhoods, compared with 31.4 percent home purchase loans originated in the conventional conforming market (excluding B&C loans). Thus, Freddie Mac performed at 90 percent of the market level, while Fannie Mae performed at 96 percent of the market level—both results similar to those reported above for underserved areas based on 1990 Census geography. The 2000 Census data show that Fannie Mae has been much closer to the market during the recent 2001-2003 period. The share of Fannie Mae's purchases going to underserved areas was 31.4 percent during 2001-2003, which placed it closer to the market level of 31.7 percent. However, the 2000-based results show that, like Freddie Mac, Fannie Mae's longer-term performance (since 1996) as well as its recent average performance (1999 to 2003) have consistently been below market levels. But, it is encouraging that Fannie Mae significantly improved its performance relative to the market during the first two years of HUD's higher housing goal levels. (See Section D.2 for the method of estimating the 1996-2003 average results.)

4. Ability To Lead the Single-Family-Owner Market: A Subgoal for Underserved Areas

The Secretary believes the GSEs can play a leadership role in underserved markets. Thus, as discussed in Section D.2, the Department is establishing a subgoal of 32 percent for each GSE's acquisitions of home purchase loans for single-family-owner properties located in the underserved census tracts of metropolitan areas in 2005, rising to 33 percent in 2006 and 2007 and to 34 percent in 2008. If the GSEs meet the 2008 subgoal, they will be leading the primary market by over two percentage points. As discussed above, underserved areas accounted for an average of approximately 31.5 percent of home purchase loans originated in the conventional conforming market of metropolitan areas (computed over 1999-2003 or over 2001-2003). To reach the subgoal for 2008, both GSEs will have to improve over their earlier peak performances—Freddie Mac by 2.3 percentage points over its previous peak performance of 31.7 percent in 2002, and Fannie Mae by 1.7 percentage points over its previous peak performance of 32.3 percent in 2003. To meet the 2008 subgoal, Freddie Mac will have to improve by 2.6 percentage points over its 2002-2003 average (unweighted) performance of 30.4 percent, while Fannie Mae will have to improve by 1.8 percentage points over its 2002-2003 average performance of 32.2 percent.

The subgoal applies only to the GSEs' purchases in metropolitan areas because the HMDA-based market benchmark is only available for metropolitan areas. HMDA data for nonmetropolitan counties are not reliable enough to serve as a market benchmark. The Department is also setting home purchase subgoals for the other two goals-qualifying categories, as explained in Appendices A and C.

The approach taken is for the GSEs to obtain their leadership position by staged increases in the underserved areas subgoal; this will enable the GSEs to take new initiatives in a correspondingly staged manner to achieve the new subgoal each year. Thus, the increases in the underserved areas subgoal are sequenced so that the GSEs can gain experience as they improve and move toward the new higher subgoal targets.

Appendix A discusses in some detail the factors that the Department considered when setting the subgoal for low- and moderate-income loans. Several of the considerations were general in nature—for example, related to the GSEs' overall ability to lead the single-family-owner market—while others were specific to the low-mod subgoal. Because the reader can refer to Appendix A, this appendix provides a briefer discussion of the more general factors. The specific considerations that led to the subgoal for underserved areas can be organized around the following four topics:

(1) The GSEs have the ability to lead the market. As discussed in Appendix A, the GSEs have the ability to lead the primary market for single-family-owner loans, which is their core business. Both GSEs have been dominant players in the home purchase market for years, funding 57 percent of the single-family-owner mortgages financed between 1999 and 2002. Through their many new product offerings and their various partnership initiatives, the GSEs have shown that they have the capacity to operate in underserved neighborhoods. They also have the staff expertise and financial resources to make the extra effort to lead the primary market in funding single-family-owner mortgages in underserved areas.

(2) The GSEs have lagged the market. Even though they have the ability to lead the market, they have not done so, as discussed above. Fannie Mae demonstrated the type of improvement needed to meet this new underserved area subgoal during 2001 and 2002. During 2001, underserved area loans declined as a percentage of primary market originations (from 31.7 to 30.7 percent), but they increased as a percentage of Fannie Mae's purchases (from 29.0 to 29.8 percent); and during 2002, they increased further as a percentage of Fannie Mae's purchases (from 29.8 to 32.3 percent), placing Fannie Mae at the market level.

(3) There are disparities among neighborhoods in access to mortgage credit. There remain troublesome neighborhood disparities in our mortgage markets, even after the substantial growth in conventional lending to low-income and minority neighborhoods that accompanied the so-called “revolution in affordable lending”. There is growing evidence that inner city neighborhoods are not being adequately served by mainstream lenders. Some have concluded that a dual mortgage market has developed in our nation's financing system, with conventional mainstream lenders serving white families living in the suburbs and FHA and subprime lenders serving minority families concentrated in inner city neighborhoods.^[57] In addition to the unavailability of mainstream lenders, families living in these often highly-segregated neighborhoods face many additional hurdles, such as lack of cash for a down payment, credit problems, and discrimination. Immigrants and minorities, who disproportionately live in underserved areas, are projected to account for almost two-thirds of the growth in the number of new households over the next ten years. To meet the diverse and unique needs of these families, the GSEs must continue adjusting their underwriting guidelines and offering new products so that they can better serve these areas and hopefully attract more mainstream lenders into our inner city neighborhoods.

(4) There are ample opportunities for the GSEs to improve their performance. Mortgages are available for the GSEs to purchase in underserved areas. They can improve their performance and lead the primary market in purchasing loans in these low-income and high-minority neighborhoods. The underserved areas share of the home purchase market has consistently been around 31 percent since 1995 (and 32 percent in the last two years), which suggests a degree of underlying strength in the market. According to the market share data reported in Table A.30 of Appendix A, the GSEs have been purchasing 48 percent of new originations in underserved areas, which means there are plenty of purchase opportunities left for them in the non-GSE portion of that market. In addition, the GSEs' purchases under the subgoal are not limited to new mortgages that are originated in the current calendar year. The GSEs can purchase loans from the substantial, existing stock of affordable loans held in lenders' portfolios, after these loans have seasoned and the GSEs have had the opportunity to observe their track record. In fact, both GSEs have often purchased seasoned loans that were used to finance properties in underserved areas (see Table A.11 in Appendix A).

To summarize, although single-family-owner mortgages comprise the “bread-and-butter” of their business, the GSEs have lagged behind the primary market in financing properties in underserved areas. For the reasons given above, the Secretary believes that the GSEs can do more to raise the share of their home loan purchases in underserved areas. This can be accomplished by building on efforts that the enterprises have already started, including their new affordable lending products, their many partnership efforts, their outreach to inner city neighborhoods, their incorporation of greater flexibility into their underwriting guidelines, and their purchases of CRA loans. A wide variety of quantitative and qualitative indicators indicate that the GSEs have the resources and financial strength to improve their affordable lending performance enough to lead the market in underserved areas.

5. Size of the Mortgage Market for Underserved Areas

As detailed in Appendix D, the market for mortgages in underserved areas is projected to account for 35-39 percent of dwelling units financed by conventional conforming mortgages; in estimating the size of the market, HUD used alternative assumptions about future economic and market conditions that were less favorable than those that existed over the last five years. HUD is well aware of the volatility of mortgage markets and the possible impacts on the GSEs' ability to meet the housing goals. Should conditions change such that the goals are no longer reasonable or feasible, the Secretary has the authority to revise the goals.

6. The Underserved Areas Housing Goal for 2005-2008

The Underserved Areas Housing Goal for 2005 is 37 percent of eligible purchases, rising to 38 percent in 2006 and 2007 and 39 percent in 2008. Five percent of the six percentage point increase in 2005 simply reflects the expanded coverage of HUD's definition in the 2000 Census tract data. The bonus points for small multifamily properties and owner-occupied 2-4 units, as well as Freddie Mac's Temporary Adjustment Factor, will no longer be in effect for goal counting purposes. It is recognized that neither GSE would have met the 37 percent target for 2005 in the past three years, and only Fannie Mae would have met this goal in 2000. Specifically, Fannie Mae's performance is projected to have been 37.5 percent in 2000, 35.7 percent in 2001, 35.0 percent in 2002, and 34.1 percent in 2003 under a 2000-based underserved area goal. On this basis, Freddie Mac's performance is projected to have been 34.1 percent in 2000, 32.5 percent in 2001, 32.4 percent in 2002, and 31.7 percent in 2003. However, GSE goal performance in 2001-03 was reduced by the heavy refinance wave of this period.

The objective of HUD's Underserved Areas Housing Goal is to bring the GSEs' performance to the upper end of HUD's market range estimate for this goal (35-39 percent), consistent with the statutory criterion that HUD should consider the GSEs' ability to lead the market for each Goal. To enable the GSEs to achieve this leadership, the Department is modestly increasing the Underserved Areas Housing Goal for 2005 which will increase further through 2008, to achieve the ultimate objective for the GSEs to lead the market under a range of foreseeable economic circumstances by 2008. Such a program of staged increases is consistent with the statutory requirement that HUD consider the past performance of the GSEs in setting the Goals. Staged increases in the Underserved Areas Housing Goal will provide the enterprises with opportunity to adjust their business models and prudently try out business strategies, so as to meet the required 2008 level without compromising other business objectives and requirements.

The analysis of this section implies that there are many opportunities for Fannie Mae and Freddie Mac to improve their overall performance on the Underserved Areas Housing Goal. The GSEs provided financing for 55 percent of the single-family and multifamily units that were financed in the conventional conforming market between 1999 and 2002. However, in the underserved areas portion of the market, the GSE purchases represented only 48 percent of the dwelling units that were financed in the market. Thus, there appears to be ample room for the GSEs to increase their purchases of loans that qualify for the Underserved Areas Housing Goal. In addition, there are several market segments that would benefit from a greater secondary market role by the GSEs, and many of these market segments are concentrated in underserved areas.

7. Conclusions

Having considered the projected mortgage market serving low- and moderate-income families, economic, housing and demographic conditions for 2005-08, and the GSEs' recent performance in purchasing mortgages in underserved areas the Secretary has determined that the annual goal of 37 percent of eligible units financed in 2005, 38 percent in 2006 and 2007, and 39 percent in 2008 is feasible. The Secretary has also established a subgoal of 32 percent for the GSEs' purchases of single-family-owner mortgages in metropolitan areas for 2005, rising to 33 percent in 2006 and 2007 and 34 percent in 2008. The Secretary has considered the GSEs' ability to lead the industry as well as the GSEs' financial condition. The Secretary has determined that the goals and subgoals are necessary and appropriate.

Appendix C—Departmental Considerations To Establish the Special Affordable Housing Goal

A. Introduction

1. Establishment of the Goal

The Federal Housing Enterprises Financial Safety and Soundness Act of 1992 (FHEFSSA) requires the Secretary to establish a special annual goal designed to adjust the purchase by each GSE of mortgages on rental and owner-occupied housing to meet the unaddressed needs of, and affordable to, low-income families in low-income areas and very-low-income families (the Special Affordable Housing Goal).

In establishing the Special Affordable Housing Goal, FHEFSSA requires the Secretary to consider:

1. Data submitted to the Secretary in connection with the Special Affordable Housing Goal for previous years;

2. The performance and efforts of the GSEs toward achieving the Special Affordable Housing Goal in previous years;

3. National housing needs of targeted families;

4. The ability of the GSEs to lead the industry in making mortgage credit available for low-income and very-low-income families; and

5. The need to maintain the sound financial condition of the enterprises.

2. The Goal and Subgoals

Special Affordable Housing Goal. The rule provides that the Special Affordable Housing Goal will be 22 percent in 2005, 23 percent in 2006, 25 percent in 2007, and 27 percent in 2008.

Units That Count Toward the Goal. Units that count toward the Special Affordable Housing Goal include units occupied by low-income owners and renters in low-income areas, and very low-income owners and renters. Other low-income rental units in multifamily properties count toward the goal where at least 20 percent of the units in the property are affordable to families whose incomes are 50 percent of area median income or less, or where at least 40 percent of the units are affordable to families whose incomes are 60 percent of area median income or less.

Multifamily Subgoal. HUD has established a special affordable subgoal for GSE purchases of multifamily mortgages. This subgoal is expressed in terms of a minimum annual dollar volume of multifamily mortgage purchases for units qualifying for the goal, rather than as a percentage of total units financed, as for the three housing goals. Both GSEs have consistently surpassed the multifamily subgoal since its establishment in 1996. The rule increases the subgoal such that, of the total Special Affordable mortgage purchases each year, each GSE must purchase special affordable multifamily mortgages in dollar amount equal to at least 1 percent of its combined (i.e., single-family and multifamily) annual average mortgage purchases over the 2000-2002 period. The level of this subgoal is $5.49 billion per year for Fannie Mae and $3.92 billion per year for Freddie Mac.

Single-Family-Owner Home Purchase Subgoal. The Department is establishing a subgoal of 17 percent for the share of each GSE's purchases of single-family-owner home purchase mortgages that qualify as special affordable and are originated in metropolitan areas in 2005 and 2006, with the subgoal rising to 18 percent in 2007 and 2008.

B. Consideration of the Factors

In considering the factors under FHEFSSA to establish the Special Affordable Housing Goal, HUD relied upon data gathered from the American Housing Survey through 2001, the Census Bureau's 1991 and 2001 Residential Finance Surveys, the 1990 and 2000 Censuses of Population and Housing, Home Mortgage Disclosure Act (HMDA) data for 1992 through 2003, and annual loan-level data from the GSEs on their mortgage purchases through 2003. Appendix D discusses in detail how these data resources were used and how the size of the conventional conforming market for this goal was estimated.

The remainder of Section C discusses the factors listed above, and Section D provides the Secretary's rationale for establishing the Special Affordable Housing Goal.

Factors 1 and 2. Data submitted to the Secretary in connection with the Special Affordable Housing Goal for previous years, and the performance and efforts of the enterprises toward achieving the Special Affordable Housing Goal in previous years.

The discussions of these two factors have been combined because they overlap to a significant degree.

This section discusses each GSE's performance under the Special Affordable Housing Goal over the 1996-2003 period.^[1] As explained in Appendix A, the data presented are “official HUD results” which, in some cases, differ from goal performance reported by the GSEs in the Annual Housing Activities Reports (AHARs) that they submit to the Department.

The main finding of this section is that both Fannie Mae and Freddie Mac surpassed the Department's Special Affordable Housing Goals for each of the seven years during this period. Specifically:

• The goal was set at 12 percent for 1996; Fannie Mae's performance was 15.4 percent and Freddie Mac's performance was 14.0 percent.
• The goal was set at 14 percent for 1997-2000. Freddie Mac's performance was 15.2 percent in 1997, 15.9 percent in 1998, 17.2 percent in 1999, and 20.7 percent in 2000; and Fannie Mae's performance was 17.0 percent in 1997, 14.3 percent in 1998, 17.6 percent in 1999, and 19.2 percent in 2000.
• In HUD's Housing Goals 2000 Final Rule, the special affordable goal was set at 20 percent for 2001-03. As of January 1, 2001, several changes in counting requirements took effect for the special affordable goal, as follows: “bonus points” (double credit) for purchases of goal-qualifying mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on 2-4 unit owner-occupied properties; a “temporary adjustment factor” (1.20 units credit, subsequently increased by Congress to 1.35 units credit) for Freddie Mac's purchases of goal-qualifying mortgages on large (more than 50-unit) multifamily properties; changes in the treatment of missing data; a procedure for the use of imputed or proxy rents for determining goal credit for multifamily mortgages; and changes regarding the “recycling” of funds by loan originators. These changes are explained below. Fannie Mae's performance was 21.6 percent in 2001, 21.4 percent in 2002, and 21.2 percent in 2003. Freddie Mac's performance was 22.6 percent in 2001, 20.4 percent in 2002, and 21.4 percent in 2003. Both GSEs surpassed this higher goal in all years. This section discusses the October 2000 counting rule changes in detail and provides data on what goal performance would have been in 2001-03 without these changes.^[2]

In addition, HUD has established a special affordable subgoal for GSE purchases of multifamily mortgages. This subgoal is expressed in terms of a minimum annual dollar volume of multifamily mortgage purchases for units qualifying for the goal, rather than as a percentage of total units financed, as for the three housing goals. As discussed below, both GSEs surpassed the multifamily subgoal in each of these years.

a. Performance on the Special Affordable Housing Goal in 1996-2003

HUD's Housing Goals 1995 Final Rule specified that in 1996 at least 12 percent of the number of units financed by each of the GSEs that were eligible to count toward the Special Affordable Housing Goal should qualify for the goal (that is, be for very low-income families or low-income families in low-income areas), and at least 14 percent should qualify in 1997-2000. HUD's October 2000 rule made various changes in the goal counting rules, as discussed below, and increased the Special Affordable Housing Goal to 20 percent for 2001-03.

In the December 1995 rule, the minimum special affordable multifamily subgoals for 1996-2000 were set at 0.8 percent of the total dollar volume of each GSE's mortgage purchases in 1994, or $1.29 billion annually for Fannie Mae and $0.99 billion annually for Freddie Mac. These subgoals were increased for 2001-03 in the October 2000 rule, to $2.85 billion annually for Fannie Mae and $2.11 billion annually for Freddie Mac, or 1.0 percent of the average dollar volume of each GSE's mortgage purchases over the 1997-99 period.

Table C.1 and Figure C.1 show performance on the special affordable goal and the special affordable multifamily subgoal over the 1996-2003 period, based on HUD's analysis. The table shows that Fannie Mae surpassed the goals by 3.4 percentage points and 3.0 percentage points in 1996 and 1997, respectively, while Freddie Mac surpassed the goals by narrower margins, 2.0 and 1.2 percentage points. In 1998 Fannie Mae's performance fell by 2.7 percentage points, while Freddie Mac's performance continued to rise, by 0.7 percentage point, thus for the first time Freddie Mac outperformed Fannie Mae on this goal. Freddie Mac showed a gain in performance to 17.2 percent in 1999, while Fannie Mae exhibited an even greater gain, to 17.6 percent

Both GSEs exhibited sharp gains in goal performance in 2000—Fannie Mae's performance increased by 1.6 percentage points, to a record level of 19.2 percent, while Freddie Mac's performance increased even more, by 3.5 percentage points, which also led to a record level of 20.7 percent. Fannie Mae's performance was 21.6 percent in 2001, 21.4 percent in 2002 and 21.3 percent in 2003; Freddie Mac's performance was 22.6 percent in 2001, 20.4 percent in 2002, and 21.4 percent in 2003. However, as discussed below, using consistent accounting rules for 2000-03, each GSE's Special Affordable Housing Goal performance fell in every year from 2001 through 2003—in total, by 2 percentage points for Fannie Mae and 3.2 percentage points for Freddie Mac.

With regard to the special affordable multifamily subgoal, Fannie Mae's purchases have exceeded the subgoal by wide margins in all years, with performance ranging from 184 percent of the goal in 1996 to 315 percent of the goal in 1999. Fannie Mae's subgoal was more than doubled in the October 2000 rule, to a minimum of $2.85 billion in each year from 2001 through 2003, but its qualifying purchases amounted to $7.36 billion, or 258 percent of the goal, in 2001, and $7.57 billion, or 260 percent of the goal, in 2002; and $12.10 billion, or 425 percent of the subgoal, in 2003.

Freddie Mac has also exceeded its special affordable multifamily subgoals in every year, albeit by smaller margins than Fannie Mae. In 1996 Freddie Mac's special affordable multifamily mortgage purchases amounted to $1.06 billion, or 107 percent of the goal. This ratio rose to 122 percent in 1997, and exceeded 200 percent for each year from 1998 through 2000. Freddie Mac's subgoal was more than doubled in the October 2000 rule, to a minimum of $2.11 in each year from 2001 through 2003, but its qualifying purchases amounted to $4.65 billion, or 220 percent of the goal, in 2001; $5.22 billion, or 247 percent of the goal, in 2002; and $8.79 billion, or 417 percent of the subgoal, in 2003.

The official figures for Freddie Mac's special affordable goal performance presented above differ from the corresponding figures presented by Freddie Mac in its Annual Housing Activity Reports to HUD by 0.1-0.2 percentage point for 1996-2000, reflecting minor differences in the application of counting rules. The official figures for special affordable goal performance by both GSEs are the same as those submitted by the enterprises for both GSEs for 2001, and for Fannie Mae for 2002. However, for 1996-2000, HUD's official special affordable goal performance figures for Fannie Mae were approximately 1-3 percentage points lower than the corresponding figures reported by the enterprise. This was due to differences between HUD and Fannie Mae in the application of counting requirements applicable to purchases of portfolios of seasoned loans, based on a statutory requirement that the proceeds of such GSE purchases by the loan sellers should be “recycled” in order for the GSE to receive Special Affordable goal credit.^[3] This discrepancy did not persist in 2001-02 because of a change in counting requirements, described below. And for 2002, HUD's official goal performance figure was 20.4 percent, somewhat below the figure of 20.6 percent submitted to the Department by Freddie Mac. For 2003, official performance on this goal for both GSEs was somewhat greater than that reported by the GSEs—official performance was 21.2 percent for Fannie Mae (as compared with 20.9 percent reported by Fannie Mae to the Department) and 21.4 percent for Freddie Mac (as compared with 20.3 percent reported by Freddie Mac to the Department).

Fannie Mae's performance on the Special Affordable Housing Goal surpassed Freddie Mac's in 1996-97. This pattern was reversed in 1998, as Freddie Mac surpassed Fannie Mae in goal performance for the first time, though by only 0.2 percentage point. This improved relative performance of Freddie Mac was due to its increased purchases of multifamily loans, as it re-entered that market, and to increases in the goal-qualifying shares of its single-family mortgage purchases. However, Fannie Mae again surpassed Freddie Mac in special affordable goal performance in 1999, 17.6 percent to 17.2 percent; Freddie Mac regained the lead in 2000, 20.7 percent to 19.2 percent. Freddie Mac's official performance also exceeded Fannie Mae's official performance in 2001, but this reflected a difference in the counting rules applicable to the two GSEs that was enacted by Congress; if the same counting rules were applied to both GSEs, Fannie Mae's performance would have exceeded Freddie Mac's performance, by 21.6 percent to 21.1 percent.

In 2002, Freddie Mac's performance on the special affordable goal was below Fannie Mae's performance (21.4 percent), even though Freddie Mac had the advantage of the Temporary Adjustment Factor (TAF), which did not apply to performance by Fannie Mae. Freddie Mac's performance would have trailed Fannie Mae's without this factor, and in fact Freddie Mac would have fallen short of the goal, at 19.3 percent. In 2003, Freddie Mac's performance (21.4 percent) slightly exceeded Fannie Mae's performance (21.2 percent), but this resulted from application of the TAF to Freddie Mac's performance—without this, Freddie Mac's performance would have been 20.2 percent, barely in excess of the 20 percent goal.

b. Changes in the Goal Counting Rules for 2001-03

Several changes in the counting rules underlying the calculation of special affordable goal performance took effect beginning in 2001. Most of these also applied to the low- and moderate-income goal and are discussed in Appendix A; only brief summaries of those changes are given here:

• Bonus points for multifamily and single-family rental properties. Each qualifying unit in a small multifamily property counted as two units in the numerator in calculating special affordable goal performance on all of the goals for 2001-03. And, above a threshold equal to 60 percent of the average number of qualifying rental units financed in owner-occupied properties over the preceding five years, each qualifying unit in a 2-4 unit owner-occupied property also counted as two units in the numerator in calculating goal performance.
• Freddie Mac's Temporary Adjustment Factor. Freddie Mac received a “Temporary Adjustment Factor” of 1.35 units of credit for each qualifying unit financed in “large” multifamily properties (i.e., those with 51 or more units) in the numerator in calculating special affordable goal performance for 2001-03.^[4] This factor did not apply to special affordable units in large multifamily properties whose mortgages were financed by Fannie Mae during this period.

• Missing data for single-family properties. The GSEs may exclude loans with missing borrower income from the denominator if the property is located in a below-median income census tract, subject to a ceiling of 1 percent of total owner-occupied units financed. The enterprises are also allowed to exclude single-family rental units with missing rental information from the denominator in calculating performance for the special affordable goal.
• Missing data and proxy rents for multifamily properties. If rent is missing for multifamily units, the GSEs may apply “proxy rents,” up to a ceiling of 5 percent of total multifamily units financed, in determining whether such units qualify for the special affordable goal. If such proxy rents cannot be estimated, these multifamily units are excluded from the denominator in calculating performance under these goals.
• Change in “recycling” requirements. Under Section 1333(b)(1)(B) of FHEFSSA, if a GSE acquires a portfolio of mortgages originated in a previous year (that is, seasoned mortgages) that qualify under the Special Affordable Housing goal, the seller must be “engaged in a specific program to use the proceeds of such sales to originate additional loans that meet such goal” and such purchases or refinancings must “support additional lending for housing that otherwise qualifies under such goal” in order to receive credit toward the goal. This has been referred to as the “recycling requirement.” The 2000 rule both clarified the conditions under which HUD would regard these statutory conditions to be satisfied and established certain categories of lenders that would be presumed to meet the recycling requirements. These included BIF-insured and SAIF-insured depository institutions that are regularly in the business of mortgage lending and which are subject to, and have received at least a satisfactory Community Reinvestment Act performance evaluation rating under specified conditions.^[5]

c. Effects of Changes in the Counting Rules on Goal Performance

Because of the changes in special affordable goal counting rules that took effect in 2001, direct comparisons between official goal performance in 2000 and 2001-03 are somewhat of an “apples-to-oranges comparison.” For this reason, the Department has calculated what performance would have been in 2000 under the 2001-03 rules; this may be compared with official performance in 2001-03—an “apples-to-apples comparison.” HUD has also calculated what performance would have been in 2001-03 under the 1996-2000 rules; this may be compared with official performance in 2000—an “oranges-to-oranges comparison.” These comparisons are presented in Table C.2.

Specifically, Table C.2 shows performance under the special affordable goal in three ways. Baseline A presents performance under the counting rules in effect for 1996-2000. Baseline B incorporates the technical changes in counting rules—changes in the treatment of missing data (including use of proxy rents), and changes in procedures related to the “recycling” requirement. Baseline C incorporates in addition to the technical changes the bonus points and, for Freddie Mac, the temporary adjustment factor. Baseline B corresponds to the counting approach used in this rule to take effect in 2005. Boldface figures under Baseline A for 1999-2000 and under Baseline C for 2001-03 indicate official goal performance based on the counting rules in effect in those years—e.g., for Freddie Mac, 17.2 percent in 1999, 20.7 percent in 2000, 22.6 percent in 2001, 20.4 percent in 2002 and 21.4 percent in 2003.

• Performance on the Special Affordable Housing Goal under 1996-2000 Counting Rules Plus Technical Changes. If the “Baseline B” counting approach had been in effect in 2000-03 and the GSEs” had purchased the same mortgages that they actually did purchase in those years, Fannie Mae would have surpassed the special affordable goal in both 2000 and 2001, but not in 2002 or 2003, while Freddie Mac would have surpassed the goal in 2000 but fallen short in 2001-2003. Specifically, Fannie Mae's performance would have been 21.4 percent in 2000, 20.2 percent in 2001, 19.9 percent in 2002, and 19.3 percent in 2003. Freddie Mac's performance would have been 21.0 percent in 2000, 19.3 percent in 2001, 18.1 percent in 2002, and 17.8 percent in 2003.
• Performance on the Special Affordable Housing Goal under 2001-2003 Counting Rules. If the 2001-03 counting rules had been in effect in 2000-03 and the GSEs' had purchased the same mortgages that they actually did purchase in that year (i.e., abstracting from any behavioral effects of “bonus points,” for example), both GSEs would have substantially surpassed the special affordable goal in all four years, but both GSEs' performance figures would have deteriorated somewhat between 2000 and 2003. Specifically, Fannie Mae's “Baseline C” performance would have been 22.2 percent in 2000, 21.6 percent in 2001, 21.4 percent in 2002, and 21.2 percent in 2003. Freddie Mac's performance would have been 23.4 percent in 2000, 22.6 percent in 2001, 20.4 percent in 2002 and 21.4 percent in 2003. Measured on this consistent basis, then, Fannie Mae's performance fell by 0.9 percentage point between 2000 and 2003. Freddie Mac's “Baseline C” performance fell by 2.0 percentage points between 2000 and 2003. These reductions were primarily due to 2001-03 being years of heavy refinance activity.

Details of Effects of Changes in Counting Rules on Goal Performance in 2001-03. As discussed above, counting rule changes that took effect in 2001 had significant impacts on the performance of both GSEs on the special affordable goal in 2001—3.0 percentage points for Fannie Mae and 3.5 percentage points for Freddie Mac. This section breaks down the effects of these changes on goal performance for both GSEs; results are shown in Table C.2.

• Freddie Mac. The largest impact of the counting rule changes on Freddie Mac's goal performance was due to the application of the temporary adjustment factor for purchases of mortgages on large multifamily properties, as enacted by Congress; this added 1.4 percentage points to goal performance in 2001, as shown in Table C.2. Bonus points for purchases of mortgages on small multifamily properties added 1.1 percentage points to performance, and bonus points for purchase of mortgages on owner-occupied 2-4 unit rental properties added 0.7 percentage point to performance. The remaining impact (0.2 percentage point) was due to technical changes in counting rules—primarily, the exclusion of single-family units with missing information from the denominator in calculating goal performance. Changes in the Department's counting rules related to “recycling” did not play a role in Freddie Mac's performance on the special affordable goal. These same patterns also generally appeared in 2002. But in 2003 bonus points for financing special affordable unit in small multifamily properties had a greater impact on performance that the temporary adjustment factor.
• Fannie Mae. The temporary adjustment factor applied to Freddie Mac's goal performance, but not to Fannie Mae's performance, thus counting rule changes had less impact on its performance than on Freddie Mac's performance in 2001-03. The largest impacts of the counting rule changes on Fannie Mae's goal performance in 2001 were due to the application of bonus points for purchases of mortgages on owner-occupied 2-4 unit rental properties, which added 0.9 percentage point to performance; bonus points for purchases of mortgages on small multifamily properties, which added 0.4 percentage point to performance; and technical changes, which added 1.6 percentage points to performance—the latter included the change in the Department's rules regarding “recycling” and the exclusion of single-family units with missing information from the denominator in calculating goal performance.^[6] The use of proxy rents for multifamily properties played a minor role in determining Fannie Mae's special affordable goal performance. These same patterns also generally appeared in 2002 and 2003.

d. Bonus Points for the Special Affordable Housing Goal

As discussed above and in Appendix A, the Department established “bonus points” to encourage the GSEs to step up their activity in 2001-03 in two segments of the mortgage market—the small (5-50 unit) multifamily mortgage market, and the market for mortgages on 2-4 unit properties where 1 unit is owner-occupied and 1-3 units are occupied by renters. Bonus points did not apply to purchases of mortgages for owner-occupied 1-unit properties, for investor-owned 1-4 unit properties, and for large (> 50-unit) properties, although as also discussed above, a “temporary adjustment factor” applied to Freddie Mac's purchases of goal-qualifying mortgages on large multifamily properties.

Bonus points for small multifamily properties. Each unit financed in a small multifamily property that qualified for any of the housing goals was counted as two units in the numerator (and one unit in the denominator) in calculating goal performance for that goal. For example, if a GSE financed a mortgage on a 40-unit property in which 10 of the units qualified for the special affordable goal, 20 units would be entered in the numerator and 40 units in the denominator for this property in calculating goal performance.

Fannie Mae financed 37,449 units in small multifamily properties in 2001 that were eligible for the special affordable goal, 58,277 such units in 2002, and 214,619 such units in 2003—this compares with only 7,196 such units financed in 2000. Small multifamily properties also accounted for a greater share of Fannie Mae's multifamily business in 2001-03—7.4 percent of total multifamily units financed in 2001, 13.2 percent in 2002, and 28.4 percent in 2003, up from 2.5 percent in 2000. However, HUD's 2000 rule reported information from the 1991 Residential Finance Survey that small multifamily properties accounted for 37 percent of all multifamily units, thus Fannie Mae was still less active in this market than in the market for large multifamily properties. Within the small multifamily market, there was no evidence that Fannie Mae targeted affordable properties to a greater extent in 2001-03 than in 2000. That is, 61 percent of Fannie Mae's small multifamily units qualified for the special affordable goal in 2000; this fell to 46 percent in 2001, 52 percent in 2002, and 42 percent in 2003.

Freddie Mac financed 50,299 units in small multifamily properties in 2001 that were eligible for the special affordable goal, 22,255 such units in 2002, and 177,561 such units in 2003, as compared with only 2,996 such units financed in 2000. Small multifamily properties also accounted for a significantly greater share of Freddie Mac's multifamily business in 2001-03—16.0 percent of total multifamily units financed in 2001, 7.5 percent in 2002, and 30.0 percent in 2003, up from 1.8 percent in 2000.

Within the small multifamily market, there was some evidence that Freddie Mac targeted affordable properties to a greater extent in 2001 than in 2000. That is, 55 percent of Freddie Mac's small multifamily units qualified for the special affordable goal in 2000; this rose to 73 percent in 2001, but declined to 60 percent in 2002 and 54 percent in 2003.

In summary, then, there is evidence that bonus points for small multifamily properties had an impact on Fannie Mae's role in this market in 2001-03 and an even larger impact on Freddie Mac's role in this market. In addition, Fannie Mae has announced a program to increase its role in this market further in future years.^[7]

Bonus points for single-family rental properties. Above a threshold, each unit financed in a 2-4 unit property with at least one owner-occupied unit (referred to as “OO24s” below) that qualified for any of the housing goals was counted as two units in the numerator (and one unit in the denominator) in calculating goal performance for that goal in 2001-03. The threshold was equal to 60 percent of the average number of such qualifying units over the previous five years. For example, Fannie Mae financed an average of 24,780 special affordable units in these types of properties between 1996 and 2000, and 55,118 such units in 2001. Thus Fannie Mae received 40,250 bonus points in this area in 2001—that is, 55,118 minus 60 percent of 24,780. So 95,368 units were entered in the numerator for these properties in calculating special affordable goal performance.

Fannie Mae financed 176,369 units in OO24s that were eligible for the special affordable goal in 2001, 229,827 such units in 2002, and 355,994 such units in 2003, as compared with 77,985 such units financed in 2000. However, Fannie Mae's total single-family business increased at approximately the same rate as its OO24 business over the 2001-03 period, thus the share of this business accounted for by OO24s was the same in 2001-03 as in 2000—4 percent.

Within the OO24 market, there was no evidence that Fannie Mae targeted special affordable properties to a greater extent in 2001-03 than in 2000. That is, approximately 30 percent of Fannie Mae's OO24 units qualified for the special affordable goal in each of these years.

Freddie Mac financed 96,204 units in OO24s that were eligible for the special affordable goal in 2001, 146,242 such units in 2002, and 154,535 such units in 2003, as compared with 49,993 such units financed in 2000. However, Freddie Mac's total single-family business increased at approximately the same rate as its OO24 business between 2000 and 2002, thus the share of this business accounted for by OO24s was the same in 2002 as in 2000—4 percent. And its overall single-family business increased more rapidly than its OO24 business in 2003, thus OO24 units accounted for 3 percent of all single-family units last year.

As for Fannie Mae, within the OO24 market there was no evidence that Freddie Mac targeted special affordable properties to a greater extent in 2001-03 than in 2000. That is, approximately 32-36 percent of Freddie Mac's OO24 units qualified for the special affordable goal in each of these four years.

e. Effects of 2000 Census on Scoring of Loans Toward the Special Affordable Housing Goal

Background. Scoring of housing units under the Special Affordable Housing Goal is based on data for mortgagors' incomes for owner-occupied units, rents for rental units, area median incomes, and, for units that are in the low-income but not the very low-income range, decennial census data used to determine whether the median income for the area where the property is located is in the low-income range. Specifically, for single-family owner-occupied units scoring is based on.

• The mortgagors' income at the time of mortgage origination
• The median income of an area specified in the same way as for the Low- and Moderate-Income Housing Goal, that is: (i) For properties located in Metropolitan Statistical Areas (MSAs) the area is the MSA; and (ii) for properties located outside of MSAs, the area is the county or the non-metropolitan portion of the State in which the property is located, whichever has the larger median income, as of the year of mortgage origination (which may be for the current year or a prior year).
• Also, if the property is located in a Metropolitan Statistical Area (MSA), the determination for purposes of the Special Affordable Housing Goal involves data on median income of the MSA; or if the property is located elsewhere, the median income of the county or the non-metropolitan portion of the State in which the property is located, whichever is larger, as of the most recent decennial census.

Analogous specifications to those detailed in Appendix A for the Low- and Moderate-Income Housing Goal are applied in the case of the Special Affordable Housing Goal for rental units in single-family properties with rent data available (assuming no income data available for actual or prospective tenants), for rental units in multifamily properties where rent data are available, and for rental units in multifamily properties where rent data are not available.

Thus, scoring loans under the Special Affordable Housing Goal requires a data series showing annual median incomes for MSAs, non-metropolitan counties, and the non-metropolitan portions of states; decennial census data on median incomes for census tracts; and decennial census data on median incomes for MSAs, non-metropolitan counties, and the non-metropolitan portions of States.^[8]

For scoring loans purchased by the GSEs year-by-year from 1993 through 2003, area median income estimates produced by HUD's Economic and Market Analysis Division were used. The same median income data series described in Appendix A for the Low- and Moderate-Income Goal was used. The determination of low-income areas was based on 1990 census data.

2005 Procedure. Relative to the above procedure, scoring of loans purchased by the GSEs in and after 2005 will be affected by two factors—first, re-benchmarking of area median incomes to the 2000 census as described in Appendix A, with a shift from 1990 to 2000 census data for identifying low-income areas, and second, the Office of Management and Budget's June, 2003, re-specification of MSA boundaries based on analysis of 2000 census data.^[9]

Analysis. For purposes of specifying the level of the Special Affordable Housing Goal, the HUD estimates of area median incomes for MSAs, non-metropolitan counties, and the non-metropolitan parts of States, as described in Appendix A, were used in conjunction with the data identifying low-income areas based on the 2000 census, to re-score loans purchased by the GSEs between 1999 and 2003. The same data series were used further in estimating the share of loans originated in metropolitan areas that would be eligible to score toward the Special Affordable Housing Goal, from HMDA data. The results of the retrospective GSE analysis are provided in Table C.3. The results of the GSE-HMDA comparative analysis are presented in the next section.

Table C.3 shows three sets of estimates for each GSE, based respectively on the counting rules in place in 2001-2003 (but disregarding the bonus points and Temporary Adjustment Factor), on the addition of 2000 census re-benchmarking and low-income areas, and finally on the further addition of 2003 MSA specification.

f. The GSEs' Multifamily Special Affordable Purchases

Since 1996 each GSE has been subject to an annual dollar-based subgoal for Special Affordable multifamily mortgage purchases, as discussed above. This subgoal was established for 1996-2000 as 0.8 percent of the total dollar volume of single-family and multifamily mortgages purchased by the respective GSE in 1994. Thus Fannie Mae's subgoal was $1.29 billion per year and Freddie Mac's subgoal was $988 million per year during that period. Fannie Mae surpassed the subgoal by $1.08 billion, $1.90 billion, $2.24 billion, $2.77 billion, and $2.50 billion in 1996, 1997, 1998, 1999, and 2000 respectively, while Freddie Mac exceeded the subgoal by $18 million, $220 million, $1.70 billion, $1.27 billion, and $1.41 billion.

The subgoal was established for 2001-03 as 1.0 percent of the average annual volume of each GSE's total mortgage purchases over the 1997-99 period. Thus Fannie Mae's subgoal was established as $2.85 billion per year and Freddie Mac's as $2.11 billion per year. In 2001 Fannie Mae exceeded its subgoal by $4.51 billion and Freddie Mac exceeded its subgoal by $2.54 billion. In 2002, Fannie Mae exceeded its subgoal by $4.72 billion and Freddie Mac exceeded its subgoal by $3.11 billion. Both GSEs exceeded their subgoals in 2003 by wide margins—Fannie Mae, with special affordable multifamily purchases of $12.11 billion (goal of $2.85 billion), and Freddie Mac, with purchases of $8.79 billion (goal of $2.11 billion.) Those subgoals are also in effect for 2004. Table C.1 includes figures on subgoal performance, and they are depicted graphically in Figure C.2.

g. Characteristics of the GSEs' Special Affordable Purchases

The following analysis presents information on the composition of the GSEs' Special Affordable purchases according to area income, unit affordability, tenure of unit and property type (single-or multifamily).

Tables C.4 and C.5 show that each GSE's reliance on multifamily housing units to meet the special affordable goal has been variable from year to year since 1996. Fannie Mae's multifamily purchases were at 37.7 percent in 1996,28.8 percent in 2001, and 20.0 percent in 2002, with a high of 44.0 percent in 1997 and a low of 19.6 percent in 2003. Freddie Mac's multifamily purchases represented 29.4 percent of all purchases qualifying for the goal in 1996, 27.0 percent in 2001, and 20.4 percent in 2002, with a high of 31.5 percent in 1997 and a low of 20.4 percent in 2002. The two GSEs' purchase percentages for single-family owner properties exhibited a similar variability through this entire period, as did their purchases of mortgages financing single-family rental units from 1996 through 2003. Both GSEs' high points for mortgages financing single-family rental units occurred in 2002: Fannie Mae's purchase percentage was 20.0 percent while Freddie Mac's was 18.1 percent.

Tables C.4 and C.5 also show the allocation of units qualifying for the goal as related to the family income and area median income criteria in the goal definition. Very-low-income families (shown in the two leftmost columns in the tables) accounted for 83.4 percent of Fannie Mae's units qualifying under the goal in 1997, rising to 85.2 percent in 1999. For Freddie Mac, very-low-income families accounted for 81.9 percent of units qualifying under the goal in 1997, rising to 84.9 percent in 1999. In contrast, mortgage purchases from low-income areas (shown in the first and third columns in the tables) accounted for 33.7 percent of Fannie Mae's units qualifying under the goal in 1997, compared to 35.5 percent in 2001. The corresponding percentages for Freddie Mac were 38.3 percent in 1997 and 35.5 percent in 2001. Thus given the definition of special affordable housing in terms of household and area income characteristics, both GSEs have consistently relied substantially more on low-income characteristics of households than low-income characteristics of census tracts to meet this goal.

h. The GSEs' Performance Relative to the Market

Section E.9 in Appendix A uses HMDA data and GSE loan-level data for home purchase mortgages on single-family-owner properties in metropolitan areas to compare the GSEs' performance in special affordable lending to the performance of depositories and other lenders in the conventional conforming market. (See Tables A.13 to A.16 in Appendix A.). There were two main findings with respect to the special affordable category. First, Freddie Mac and Fannie Mae have historically lagged depositories and the overall market in providing mortgage funds for special affordable borrowers over periods, such as 1993-2003, 1996-2003 and 1999-2003. Between 1993 and 2003, 12.2 percent of Freddie Mac's mortgage purchases were for special affordable borrowers, 13.3 percent of Fannie Mae's purchases, 15.4 percent of loans originated by depositories, and 15.5 percent of loans originated in the conventional conforming market (without estimated B&C loans). For the recent years, the GSE-market comparisons are as follows:

Year (in percent)	Freddie Mac (in percent)	Fannie Mae (in percent)	Market (w/o B&C) (in percent)
1999	12.8	12.5	17.0
2000	14.7	13.3	16.6
2001	14.4	14.9	15.6
2002	15.8	16.3	16.1
2003	15.6	17.1	15.9
1996-2003	13.2	14.1	15.9
1999-2003	14.7	15.1	16.2
2001-2003	15.2	16.2	15.9

During the period between 1999 and 2003, the GSEs' performance was slightly over 90 percent of the market—special affordable loans accounted for 15.1 percent of Fannie Mae's purchases, 14.7 percent of Freddie Mac's purchases, and 16.2 percent of loans originated in the conforming market.

Second, while both GSEs have improved their performance over the past few years, Fannie Mae has been made more progress than Freddie Mac in erasing its gap with the market. During the first three years (2001, 2002, and 2003) of HUD's new housing goal targets, the average share of Fannie Mae's purchases going to special affordable loans was 16.2 percent, which was above the market average of 15.9 percent. The share of Freddie Mac's purchases going to special affordable loans was 15.2 percent during this period.

Section G in Appendix A discusses the role of the GSEs both in the overall special affordable market and in the different segments (single-family owner, single-family rental, and multifamily rental) of the special affordable market. The GSEs' special affordable purchases accounted for 41 percent of all special affordable owner and rental units that were financed in the conventional conforming market between 1999 and 2002. The GSEs' 41-percent share of the special affordable market was three-fourths of their 55-percent share of the overall market. Even in the owner market, where the GSEs account for 61 percent of the market, their share of the special affordable market was only 52 percent during this period. While the GSEs improved their market shares during 2001-2003, this analysis shows that there is room and ample opportunities for the GSEs, and particularly Freddie Mac, to improve their performance in purchasing affordable loans at the lower-income end of the market. Section C.3 of this appendix discusses a home purchase subgoal designed to place the GSEs in such a leadership position in the special affordable single-family-owner market.

Factor 3. National Housing Needs of Low-Income Families in Low-Income Areas and Very-Low-Income Families

This discussion concentrates on very-low-income families with the greatest needs. It complements Section C of Appendix A, which presents detailed analyses of housing problems and demographic trends for lower-income families which are relevant to the issue addressed in this part of Appendix C.

Data from the American Housing Survey demonstrate that housing problems and needs for affordable housing continue to be more pressing in the lowest-income categories than among moderate-income families, as established in HUD's analysis for the 1995 and 2000 Final Rules. Table C.6 displays figures on several types of housing problems—high housing costs relative to income, physical housing defects, and crowding-for both owners and renters. Figures are presented for households experiencing multiple (two or more) of these problems as well as households experiencing a severe degree of either cost burden or physical problems. Housing problems in 2001 continued to be much more frequent for the lowest-income groups.^[10] Incidence of problems is shown for households in the income range covered by the special affordable goal, as well as for higher income households.

This analysis shows that priority problems of severe cost burden or severely inadequate housing are noticeably concentrated among renters and owners with incomes below 60 percent of area median income: 30.5 percent of renter households and 34.9 percent of owner households had priority problems. In contrast, in the next higher income range, up to 80 percent of area median income, 2.5 percent of renter households and 7.3 percent of owner households had priority problems. The table demonstrates the significance of affordability problems: Sixty-five percent of very-low-income renter families had rent burden over 30 percent of income; 35 percent had rent burden over 50 percent of income. Thirteen percent had moderately or severely inadequate housing; 6 percent lived in crowded conditions, defined as more than one person per room.

Factor 4. The Ability of the Enterprises To Lead the Industry in Making Mortgage Credit Available for Low-Income and Very-Low-Income Families

The discussion of the ability of Fannie Mae and Freddie Mac to lead the industry in Section G of Appendix A is relevant to this factor—the GSEs' roles in the owner and rental markets, their role in establishing widely-applied underwriting standards, their role in the development of new technology for mortgage origination, their strong staff resources, and their financial strength. Additional analyses of the potential ability of the enterprises to lead the industry in the low- and very-low-income market appears below in Section D, which explains the Department's rationale for the home purchase subgoal for Special Affordable loans.

Factor 5. The Need to Maintain the Sound Financial Condition of the GSEs

HUD has undertaken a separate, detailed economic analysis of this final rule, which includes consideration of (a) the financial returns that the GSEs earn on special affordable loans and (b) the financial safety and soundness implications of the housing goals. Based on this economic analysis, HUD concludes that the housing goals in this final rule raise minimal, if any, safety and soundness concerns.

C. Determination of the Special Affordable Housing Goal

Several considerations, many of which are reviewed in Appendixes A and B and in previous sections of this Appendix, led to the determination of the Special Affordable Housing Goal, the multifamily special affordable subgoal, and the special affordable subgoal for home purchase loans on single-family-owner properties in metropolitan areas.

1. Severe Housing Problems

The data presented in Section C.3 demonstrate that housing problems and needs for affordable housing are much more pressing in the lowest-income categories than among moderate-income families. The high incidence of severe problems among the lowest-income renters reflects severe shortages of units affordable to those renters. At incomes below 60 percent of area median, 34.7 percent of renters and 21.6 percent of owners paid more than 50 percent of their income for housing. In this same income range, 65.6 percent of renters and 42.4 percent of owners paid more than 30 percent of their income for housing. In addition, 31.5 percent of renters and 23.8 percent of owners exhibited “priority problems”, meaning housing costs over 50 percent of income or severely inadequate housing. Homeownership gaps and other disparities in the housing and mortgage markets discussed in Section H of Appendix A also apply to Special Affordable housing and mortgages.

2. GSE Performance and the Market

a. The GSEs' Special Affordable Housing Goals Performance

In the October 2000 rule, the special affordable goal was set at 20 percent for 2001-03. Effective on January 1, 2001, several changes in counting requirements came into effect for the special affordable goal, as follows: (a)“bonus points” (double credit) for purchases of mortgages on small (5-50 unit) multifamily properties and, above a threshold level, mortgages on 2-4 unit owner-occupied properties; (b) a “temporary adjustment factor” (1.35 unit credit) for Freddie Mac's purchases of mortgages on large (more than 50 unit) multifamily properties; (c) changes in the treatment of missing data; (d) a procedure for the use of imputed or proxy rents for determining goal credit for multifamily mortgages; and (e) changes regarding the “recycling” of funds by loan originators.

Counting requirements (a) and (b) expired at the end of 2003 while (c)-(e) will remain in effect after that. If this counting approach—without the bonus points and the “temporary adjustment factor”—had been in effect in 2000-2003, and the GSEs' had purchased the same mortgages that they actually did purchase in both years, then Fannie Mae's performance would have been 21.4 percent in 2000, 20.2 percent in 2001, 19.9 percent in 2002, and 19.4 percent in 2003. Freddie Mac's performance would have been 21.0 percent in 2000, 19.3 percent in 2001, 18.1 percent in 2002, and 17.8 percent in 2003. Fannie Mae would have surpassed the special affordable goal in both 2000 and 2001, but not in 2002 or 2003. Freddie Mac would have surpassed the goal in 2000 but fallen short in 2001-03.

The above performance figures are for the special affordable goal defined in terms of 1990 Census geography. Switching to 2000 Census data slightly increases the coverage of special affordable goal, which increases the special affordable share of the GSEs' purchases by up to one percentage point. Based on 2000 Census geography and adding 2003 MSAs, and excluding counting requirements (a) and (b), then Fannie Mae's performance would have been 21.7 percent in 2000, 20.1 percent in 2001, 19.4 percent in 2002, and 20.8 percent in 2003. Freddie Mac's performance would have been 20.8 percent in 2000, 19.1 percent in 2001, 17.3 percent in 2002 and 19.0 percent in 2003. See Table C.3.

b. Single-Family Market Comparisons in Metropolitan Areas

The Special Affordable Housing Goal is designed, in part, to ensure that the GSEs maintain a consistent focus on serving the very low-income portion of the housing market where housing needs are greatest. Section C compared the GSEs' performance in special affordable lending to the performance of depositories and other lenders in the conventional conforming market for single-family home loans. The analysis showed that while both GSEs have been improved their performance, their past average performance (1993-2003, 1996-2003, and 1999-2003) has been below market levels. During 2002 and 2003, Fannie Mae improved its performance enough to lead the special affordable market for home purchase loans, but Freddie Mac, although it also improved its performance during this recent period, continues to lag behind the primary market. Between 1999 and 2003, special affordable borrowers accounted for 15.1 percent of the home loans purchased by Fannie Mae, 14.7 percent of Freddie Mac's purchases, 16.2 percent of home loans originated by depositories, and 16.2 percent of all home loans originated in the conventional conforming market (without B&C loans). As noted above, while both GSEs have improved their performance over the past few years, Fannie Mae has made more progress than Freddie Mac in closing its gap with the market. During 2003, the share of Fannie Mae's purchases going to special affordable loans was 17.1 percent, which was 1.2 percentage points above the market average of 15.9 percent. The share of Freddie Mac's purchases going to special affordable loans had improved to 15.6 percent by 2003. (See Figure C.3.)

3. Ability To Lead the Single-Family Owner Market: A Special Affordable Subgoal

The Secretary believes the GSEs can play a leadership role in the special affordable market. Thus, the Department is establishing a subgoal of 17 percent for each GSE's purchases of home purchase loans for special affordable families in the single-family-owner market of metropolitan areas for 2005 and 2006, rising to 18 percent during 2007 and 2008. The purpose of this subgoal is to encourage the GSEs to improve their purchases of mortgages for very-low-income and minority first-time homebuyers who are expected to enter the housing market over the next few years. If the GSEs meet the 18-percent subgoal, they will be leading the primary market by approximately two percentage points, based on the income characteristics of home purchase loans reported in HMDA. HMDA data show that special affordable families accounted for an average of 16.2 (15.9) percent of single-family-owner loans originated in the conventional conforming market of metropolitan areas between 1999 and 2003 (2001 and 2003). Loans in the B&C portion of the subprime market are not included in these averages. As explained in Appendix D, HUD also projected special affordable shares for the market for 1999 to 2002 using the new 2000 Census geography and the new OMB specifications. For special affordable loans, the 2000-based Census data resulted in special affordable shares for the market and the GSEs that were similar to the 1990-based special affordable shares reported in Section C of this appendix.

To reach the 18-percent subgoal for 2008, Freddie Mac would have to improve its performance by 2.4 percentage points over its special affordable share of 15.6 percent in 2003. Fannie Mae would have to improve its performance by 0.9 percentage point over its market-leading special affordable share of 17.1 percent in 2003. The approach taken is for the GSEs to obtain their leadership position by staged increases in the special affordable subgoal; this will enable the GSEs to take new initiatives in a correspondingly staged manner to achieve the new subgoal each year. Thus, the increases in the special affordable subgoal are sequenced so that the GSEs can gain experience as they improve and move toward the new higher subgoal targets.

The subgoal applies only to the GSEs' purchases in metropolitan areas because the HMDA-based market benchmark is only available for metropolitan areas. HMDA data for non-metropolitan counties are not reliable enough to serve as a market benchmark. The Department is also setting home purchase subgoals for the other two goals-qualifying categories, as explained in Appendices A and B. Sections E.9 and G of Appendix A provide additional information on the opportunities for an enhanced GSE role in the special affordable segment of the home purchase market and on the ability of the GSEs to lead that market.

The preamble and Appendix A discuss in some detail the factors that the Department considered when setting the subgoal for low- and moderate-income loans. Several of the considerations were general in nature—for example, related to the GSEs' overall ability to lead the single-family-owner market—while others were specific to the low-mod subgoal. Because the reader can refer to Appendix A, this appendix provides a briefer discussion of the more general factors. The specific considerations that led to the subgoal for special affordable loans can be organized around the following four topics:

(1) The GSEs have the ability to lead the market. As discussed in Appendix A, the GSEs have the ability to lead the primary market for single-family-owner loans, which is their “bread-and-butter” business. Both GSEs have been dominant players in the home purchase market for years, funding 61 percent of the single-family-owner mortgages financed between 1999 and 2002. Through their many new product offerings and their various partnership initiatives, the GSEs have shown that they have the capacity to reach out to very-low-income and other special affordable borrowers. They also have the staff expertise and financial resources to make the extra effort to lead the primary market in funding single-family-owner mortgages for special affordable borrowers.

(b) GSEs' Performance Relative to the Market. Even though the GSEs have had the ability to lead the home purchase market, their past average performance (1993-2003, 1996-2003, and 1999-2003) has been below market levels. During 2003, Fannie Mae improved its performance enough to lead the special affordable market for home purchase loans, but Freddie Mac, although it also has improved its performance, continues to lag behind the primary market. The subgoals will ensure that Fannie Mae maintains and further improves its above-market performance and that Freddie Mac not only erases its current gap with the market but also takes a leadership position as well. With respect to the GSEs' historical performance, special affordable mortgages accounted for 13.2 (14.7) percent of Freddie Mac's purchases during 1996-2003 (1999-2003), for 14.1 (15.1) percent of Fannie Mae's purchases, and for 15.9 (16.2) percent of primary market originations (excluding B&C loans). The type of improvement needed for Freddie Mac to meet this new special affordable subgoal was demonstrated by Fannie Mae during 2001-2003, as Fannie Mae increased its special affordable performance from 14.9 percent of its single-family-owner business in 2001 to 16.3 percent in 2002 to 17.1 percent in 2003.

(3) Disparities in Homeownership and Credit Access Remain. There remain troublesome disparities in our housing and mortgage markets, even after the “revolution in affordable lending” and the growth in homeownership that has taken place since the mid-1990s. The homeownership rate for African-American and Hispanic households remains 25 percentage points below that of white households. Minority families face many barriers in the mortgage market, such as lack of capital for down payment and lack of access to mainstream lenders (see above). Immigrants and minorities—many of whose very-low-income levels will qualify them as special affordable—are projected to account for almost two-thirds of the growth in the number of new households over the next ten years. As emphasized in Appendix A, changing population demographics will result in a need for the primary and secondary mortgage markets to meet nontraditional credit needs, respond to diverse housing preferences, and overcome information and other barriers that many immigrants and minorities face. The GSEs have to increase their efforts in helping special affordable families—but so far they have played a surprisingly small role in serving minority first-time homebuyers. It is estimated that the GSEs accounted for 46.5 percent of all (both government and conventional) home loans originated between 1999 and 2001; however, they accounted for only 14.3 percent of home loans originated for African-American and Hispanic first-time homebuyers. A subgoal for special affordable home purchase loans should increase the GSEs' efforts in important sub-markets such as the one for minority first-time homebuyers.

(4) There are ample opportunities for the GSEs to improve their performance. Special affordable mortgages are available for the GSEs to purchase, which means they can improve their performance and lead the primary market in purchasing loans for these very-low-income borrowers. Sections B, C, and I of Appendix A and Section H of Appendix D explain that the special affordable lending market has shown an underlying strength over the past few years that is unlikely to vanish (without a significant increase in interest rates or a decline in the economy). The special affordable share of the home purchase market has averaged approximately 16 percent since 1996 and annually has been in the 15-17 percent range. Second, the market share data reported in Table A.30 of Appendix A demonstrate that there are newly originated loans available each year for the GSEs to purchase. The GSEs' purchases of single-family owner loans represented 61 percent of all single-family-owner loans originated between 1999 and 2002, compared with 52 percent of the special affordable loans that were originated during this period. Thus, half of the special affordable conforming market is not touched by the GSEs. As noted above, the situation is even more extreme for special sub-markets such the minority first-time homebuyer market where the GSEs have only a minimal presence. Between 1999 and 2001, the GSEs purchased only 33 percent of conventional conforming loans originated for minority first-time homebuyers, even though they purchased 57 percent of all home loans originated in the conventional conforming market during that period. But also important, the GSEs' purchases under the subgoal are not limited to new mortgages that are originated in the current calendar year. The GSEs can purchase loans from the substantial, existing stock of special affordable loans held in lenders' portfolios, after these loans have seasoned and the GSEs have had the opportunity to observe their payment performance. In fact, based on Fannie Mae's recent experience, the purchase of seasoned loans appears to be one useful strategy for purchasing goals-qualifying loans.

For the reasons given above, the Secretary believes that the GSEs can do more to raise the special affordable shares of the home loans they purchase on single-family-owner properties. This can be accomplished by building on efforts that the enterprises have already started, including their new affordable lending products aimed at special groups such as first-time homebuyers, their many partnership efforts, their outreach to inner city neighborhoods, their incorporation of greater flexibility into their underwriting guidelines, and their purchases of seasoned CRA loans. A wide variety of quantitative and qualitative indicators indicate that the GSEs' have the resources and financial strength to improve their special affordable performance enough to lead the market.

4. Size of the Overall Special Affordable Mortgage Market

As detailed in Appendix D, single-family and multifamily special affordable mortgages are estimated to account for 23-27 percent of the dwelling units financed by conventional conforming mortgages; in estimating the size of the market, HUD used alternative assumptions about future economic and market affordability conditions that were less favorable than those that existed over the past several years. HUD is well aware of the volatility of mortgage markets and the possible impacts on the GSEs' ability to meet the housing goals. Should conditions change such that the goals are no longer reasonable or feasible, the Secretary has the authority to revise the goals.

5. The Special Affordable Housing Goal for 2005-2008

The Special Affordable Housing Goal for 2005 is 22 percent of eligible purchases, a two percentage point increase over the current goal of 20 percent, with the goal rising to 23 percent in 2006, 25 percent in 2007, and 27 percent in 2008. The bonus points for small multifamily properties and owner-occupied 2-4 unit properties, as well as Freddie Mac's Temporary Adjustment Factor, will no longer be in effect for goal counting purposes. It is recognized that neither GSE would have met the 22-percent target for 2005 in the past three years. Under the new counting rules, Fannie Mae's special affordable performance is estimated to have been 18.6 percent in 1999, 21.7 percent in 2000, 20.1 percent in 2001, 19.4 percent in 2002, and 20.8 percent in 2003. Fannie Mae would have to increase its performance in 2005 by 1.9 percentage points over its average (unweighted) performance of 20.1 percent over these last five years. By 2008 this increase relative to average 1999-2003 performance would be 6.9 percentage points. Freddie Mac's performance is projected to have been 17.4 percent in 1999, 20.8 percent in 2000, 19.1 percent in 2001, 17.3 percent in 2002, and 19.0 percent in 2003. Freddie Mac would have to increase its performance in 2005 by 3.3 percentage points over its average (unweighted) performance of 18.7 percent over these last five years. By 2008 this increase relative to average 1999-2002 performance would be 8.3 percentage points. However, GSE goal performance in 2001-03 was reduced by the heavy refinance wave of this period.

The objective of HUD's Special Affordable Goal is to bring the GSEs' performance to the upper end of HUD's market range estimate for this goal (23-27 percent), consistent with the statutory criterion that HUD should consider the GSEs' ability to lead the market for each Goal. To enable the GSEs to achieve this leadership, the Department is establishing modest increases in the Special Affordable Goal for 2005, which will increase year-by-year through 2008, to achieve the ultimate objective for the GSEs to lead the market under a range of foreseeable economic circumstances by 2008. Such a program of staged increases is consistent with the statutory requirement that HUD consider the past performance of the GSEs in setting the Goals. Staged annual increases in the Special Affordable Goal will provide the enterprises with opportunity to adjust their business models and prudently try out business strategies, so as to meet the required 2008 level without compromising other business objectives and requirements.

Section C compared the GSEs' role in the overall market with their role in the special affordable market. The GSEs' purchases provided financing for 26,118,927 dwelling units, which represented 55 percent of the 47,551,039 single-family and multifamily units that were financed in the conventional conforming market between 1999 and 2002. However, in the special affordable part of the market, the 5,103,186 units that were financed by GSE purchases represented only 41 percent of the 12,413,759 dwelling units that were financed in the market. Thus, there appears to be ample room for the GSEs to improve their performance in the special affordable market. In addition, there are several market segments (e.g., first-time homebuyers) that would benefit from a greater secondary market role by the GSEs, and special affordable borrowers are concentrated in these markets.

6. Multifamily Special Affordable Subgoals

Based on the GSEs' past performance on the special affordable multifamily subgoals, and on the outlook for the multifamily mortgage market, HUD is establishing that these subgoals be retained and increased for the 2005-2008 period. Unlike the overall goals, which are expressed in terms of minimum goal-qualifying percentages of total units financed, these subgoals for 2001-03 and in prior years have been expressed in terms of minimum dollar volumes of goal-qualifying multifamily mortgage purchases. Specifically, each GSE's special affordable multifamily subgoal is currently equal to 1.0 percent of its average total (single-family plus multifamily) mortgage volume over the 1997-99 period. Under this formulation, in October 2000 the subgoals were set at $2.85 billion per year for Fannie Mae and $2.11 billion per year for Freddie Mac, in each of calendar years 2001 through 2003. These represented increases from the goals for 1996-2000, which were $1.29 billion annually for Fannie Mae and $0.99 billion annually for Freddie Mac. These subgoals are also in effect for 2004.

HUD's Determination. The multifamily mortgage market and both GSEs' multifamily transactions volume grew significantly over the 1993-2003 period, indicating that both enterprises have provided increasing support for the multifamily market, and that they have the ability to continue to provide further support for the market.

Specifically, Fannie Mae's total eligible multifamily mortgage purchase volume increased from $4.6 billion in 1993 to $12.5 billion in 1998, and then jumped sharply to $18.7 billion in 2001 and $18.3 billion in 2002, and $33.3 billion in 2003. Its special affordable multifamily mortgage purchases followed a similar path, rising from $1.7 billion in 1993 to $3.5 billion in 1998 and $4.1 billion in 1999, and also jumping sharply to $7.4 billion in 2001 and $7.6 billion in 2002 and $12.2 billion in 2003. As a result of its strong performance, Fannie Mae's purchases have been at least twice its minimum subgoal in every year since 1997—247 percent of the subgoal in that year, 274 percent in 1998, 315 percent in 1999, 294 percent in 2000, and, under the new higher subgoal level, 258 percent in 2001, 266 percent in 2002, and 426 percent in 2003.

Freddie Mac's total eligible multifamily mortgage purchase volume increased even more sharply, from $0.2 billion in 1993 to $6.6 billion in 1998, and then jumped sharply in 2001 to $11.8 billion and $13.3 billion in 2002, and $21.5 billion in 2003. Its special affordable multifamily mortgage purchases followed a similar path, rising from $0.1 billion in 1993 to $2.7 billion in 1998, and also jumping sharply to $4.6 billion in 2001 and $5.2 billion in 2002, and $8.8 billion in 2003. As a result of its strong performance, Freddie Mac's purchases have also been at least twice its minimum subgoal in every year since 1998—272 percent of the subgoal in that year, 229 percent in 1999, 243 percent in 2000, and, under the new higher subgoal level, 220 percent in 2001, 247 percent in 2002, and 417 percent in 2003.

The Special Affordable Housing Multifamily Subgoals set forth in this rule are reasonable and appropriate based on the Department's analysis of this market. The Department's decision to retain the multifamily subgoal is based on the fact that HUD's analysis indicates that multifamily housing still serves the housing needs of lower-income families and families in low-income areas to a greater extent than single-family housing. By retaining the multifamily subgoal, the Department ensures that the GSEs continue their activity in this market, and that they achieve at least a minimum level of special affordable multifamily mortgage purchases that are affordable to lower-income families. The Department establishes each GSE's special affordable multifamily subgoal as 1.0 percent of its average annual dollar volume of total (single-family and multifamily) mortgage purchases over the 2000-2002 period. In dollar terms, the Department's subgoal is $5.49 billion per year in special affordable multifamily mortgage purchases for Fannie Mae, and $3.92 billion per year in special affordable multifamily mortgage purchases for Freddie Mac. These subgoals would be less than actual special affordable multifamily mortgage purchase volume in 2001-2003 for both GSEs; thus the Department believes that they would be feasible for the 2005-2008 period.

Some commenters advocated increasing the special affordable multifamily subgoals from the levels in the rule. In light of the high levels of such purchases by both GSEs in 2003, HUD considered raising these subgoals, but decided not to do so because HUD believes that the overall special affordable goals established in this final rule will provide sufficient incentives for the GSEs to play a major role in the special affordable multifamily mortgage market, and that in all likelihood they will continue to exceed these subgoals by significant margins for 2005-08.

7. Conclusion

HUD has determined that the Special Affordable Housing Goal in this rule addresses national housing needs within the income categories specified for this goal, while accounting for the GSEs' past performance in purchasing mortgages meeting the needs of very-low-income families and low-income families in low-income areas. HUD has also considered the size of the conventional mortgage market serving very-low-income families and low-income families in low-income areas. Moreover, HUD has considered the GSEs' ability to lead the industry as well as their financial condition. HUD has determined that a Special Affordable Housing Goal of 22 percent in 2005, 23 percent in 2006, 25 percent in 2007, and 27 percent in 2008 is both necessary and achievable. HUD has also determined that a multifamily special affordable subgoal for 2005-2008 set at 1.0 percent of the average of each GSE's respective dollar volume of combined (single-family and multifamily) 1999-2002 mortgage purchases in is both necessary and achievable. Finally, HUD is establishing a subgoal of 17 percent for the GSEs' purchases of single-family-owner mortgages that qualify for the special affordable goal and are originated in metropolitan areas, for 2005, with this subgoal remaining at 17 percent in 2006, then rising to 18 percent in both 2007 and 2008. The Secretary has considered the GSEs' ability to lead the industry as well as the GSEs' financial condition. The Secretary has determined that the goals, the multifamily subgoals, and the single-family-owner subgoals are necessary and appropriate.

Appendix D—Estimating the Size of the Conventional Conforming Market for Each Housing Goal

A. Introduction

In establishing the three housing goals, the Secretary is required to assess, among a number of factors, the size of the conventional market for each goal. This appendix explains HUD's methodology for estimating the size of the conventional market for each of the three housing goals. Following this overview, Section B summarizes the main components of HUD's market-share model and identifies those parameters that have a large effect on the relative market shares. Sections C and D discuss two particularly important market parameters, the size of the multifamily market and the share of the single-family mortgage market accounted for by single-family rental properties. Section E provides a more systematic presentation of the model's equations and main assumptions. Sections F, G, and H report HUD's estimates for the Low- and Moderate-Income Goal, the Underserved Areas Goal, and the Special Affordable Housing Goal, respectively.

HUD received numerous comments on the proposed rule relating to its market methodology and the size of its market ranges for each of the three goals. These comments, and HUD's responses to them, are discussed throughout this appendix.

In developing this final rule, HUD has followed the same basic approach that it followed in the last two GSE final rules and the recent GSE proposed rule. HUD has carefully reviewed existing information on mortgage activity in order to understand the weakness of various data sources and has conducted sensitivity analyses to show the effects of alternative parameter assumptions. HUD is well aware of uncertainties with some of the data and much of this appendix is spent discussing the effects of alternative assumptions about data parameters and presenting the results of an extensive set of sensitivity analyses, many of the latter being directly related to comments received on the proposed rule.

In an earlier critique of HUD's market share model, Blackley and Follain (1995, 1996) concluded that conceptually HUD had chosen a reasonable approach to determining the size of the mortgage market that qualifies for each of the three housing goals.^[1] Blackley and Follain correctly note that the challenge lies in getting accurate estimates of the model's parameters. In their comments on the 2000 Proposed GSE Rule, both Fannie Mae and Freddie Mac stated that HUD's market share model (outlined in Section B below) was a reasonable approach for estimating the goals-qualifying (low-mod, special affordable, and underserved areas) shares of the mortgage market. Freddie Mac stated:

We believe the Department takes the correct approach in the Final rule by examining several different data sets, using alternative methodologies, and conducting sensitivity analysis. We applaud the Department's general approach for addressing the empirical challenges.^[2]

* * *

Similarly, Fannie Mae stated that “HUD has developed a reasonable model for assessing the size of the affordable housing market.” ^[3]

However, in their comments on the proposed rule, both GSEs criticized HUD's implementation of its market methodology.^[4] As noted above, their major criticisms and HUD's responses to their criticisms can be found throughout this appendix. HUD recognizes that there is no single, perfect data set for estimating the size of the affordable lending market and that available data bases on different sectors of the market must be combined in order to implement its market share model (as outlined in Section B below). As this appendix will show, HUD has carefully combined various mortgage market data bases in a manner which draws on the strength of each in order to implement its market methodology and to arrive at a reasonable range of estimates for the three goals-qualifying shares of the mortgage market. In this appendix, HUD demonstrates the robustness of its market estimates by reporting the results of numerous sensitivity analyses that examine a range of assumptions about the relative importance of the rental and owner markets and the goals-qualifying shares of the owner portion of the mortgage market.

This appendix reviews in some detail HUD's efforts to combine information from several mortgage market databases to obtain reasonable values for the model's parameters. The next section provides an overview of HUD's market share model.

B. Overview of HUD's Market Share Methodology ⁵

1. Definition of Market Share

The size of the market for each housing goal is one of the factors that the Secretary is required to consider when setting the level of each housing goal.^[6] Using the Low- and Moderate-Income Housing Goal as an example, the market share in a particular year is defined as follows:

Low- and Moderate-Income Share of Market: The number of dwelling units financed by the primary mortgage market in a particular calendar year that are occupied by (or affordable to, in the case of rental units) families with incomes equal to or less than the area median income divided by the total number of dwelling units financed in the conforming conventional primary mortgage market.

There are three important aspects to this definition. First, the market is defined in terms of “dwelling units” rather than, for example, “value of mortgages” or “number of properties.” Second, the units are “financed” units rather than the entire stock of all mortgaged dwelling units; that is, the market-share concept is based on the mortgage flow in a particular year, which will be smaller than total outstanding mortgage debt. Third, the low- and moderate-income market is expressed relative to the overall conforming conventional market, which is the relevant market for the GSEs.^[7] The low- and moderate-income market is defined as a percentage of the conforming market; this percentage approach maintains consistency with the method for computing each GSE's performance under the Low- and Moderate-Income Goal (that is, the number of low- and moderate-income dwelling units financed by GSE mortgage purchases relative to the overall number of dwelling units financed by GSE mortgage purchases).

2. Three-Step Procedure

Ideally, computing the low- and moderate-income market share would be straightforward, consisting of three steps:

Step 1: Projecting the market shares of the four major property types included in the conventional conforming mortgage market, i.e.—

(a) Single-family owner-occupied dwelling units (SF-O units);

(b) Rental units in 2-4 unit properties where the owner occupies one unit (SF 2-4 units); ^[8]

(c) Rental units in one-to-four unit investor-owned properties (SF Investor units); and,

(d) Rental units in multifamily (5 or more units) properties (MF units).^[9]

Step 2: Projecting the “goal percentage” for each of the above four property types (for example, the “Low- and Moderate-Income Goal percentage for single-family owner-occupied properties” is the percentage of those dwelling units financed by mortgages in a particular year that are occupied by households with incomes below the area median).

Step 3: Multiplying the four percentages in (2) by their corresponding market shares in (1), and summing the results to arrive at an estimate of the overall share of dwelling units financed by mortgages that are occupied by low- and moderate-income families.

The four property types are analyzed separately because of their differences in low- and moderate-income occupancy. Rental properties have substantially higher percentages of low- and moderate-income occupants than owner-occupied properties. This can be seen in the top portion of Table D.1, which illustrates Step 3's basic formula for calculating the size of the low- and moderate-income market.^[10] In this example, low- and moderate-income dwelling units are estimated to account for 53.9 percent of the total number of dwelling units financed in the conforming mortgage market.

To examine the other housing goals, the “goal percentages” in Step 2 would be changed and the new “goal percentages” would be multiplied by Step 1's property distribution, which remains constant. For example, the Underserved Areas Goal ^[11] would be derived as illustrated in the bottom portion of Table D.1. In this example, units eligible under the Underserved Areas Goal are estimated to account for 31.4 percent of the total number of dwelling units financed in the conforming mortgage market.^[12]

3. Data Issues

Unfortunately, complete and consistent mortgage data are not readily available for carrying out the above three steps. A single data set for calculating either the property shares or the housing goal percentages does not exist. However, there are several major data bases that provide a wealth of useful information on the mortgage market. HUD combined information from the following sources: the Home Mortgage Disclosure Act (HMDA) reports, the American Housing Survey (AHS), HUD's Survey of Mortgage Lending Activity (SMLA), the Census Bureau's AHS-based Property Owners and Managers Survey (POMS), and the Census Bureau's recent 2001 Residential Finance Survey (RFS). In addition, information on the mortgage market was obtained from the Mortgage Bankers Association, Fannie Mae, Freddie Mac and other organizations.

Property Shares. To derive the property shares, HUD started with forecasts of single-family mortgage originations (expressed in dollars). These forecasts, which are available from the GSEs and industry groups such as the Mortgage Bankers Association, do not provide information on conforming mortgages, on owner versus renter mortgages, or on the number of units financed. Thus, to estimate the number of single-family units financed in the conforming conventional market, HUD had to project certain market parameters based on its judgment about the reliability of different data sources. Sections D and E report HUD's findings related to the single-family market.

Total market originations are obtained by adding multifamily originations to the single-family estimate. Because of the wide range of estimates available, the size of the multifamily mortgage market turned out to be one of the most controversial issues raised during the initial rule-making process during 1995; this was also an issue that the GSEs focused on in their comments on the 2000 final rule and their comments on the 2004 proposed GSE rule. Because most renters qualify under the Low- and Moderate-Income Goal, the chosen market size for multifamily can have a substantial effect on the overall estimate of the low- and moderate-income market (as well as on the estimate of the special affordable market). Thus, it is important to consider estimates of the size of the multifamily market in some detail, as Section C does. In addition, given the uncertainty surrounding estimates of the multifamily mortgage market, it is important to consider a range of market estimates, as Sections F-H do.

Goal Percentages. To derive the goal percentages for each property type, HUD relied heavily on HMDA, AHS, POMS and RFS data. For single-family-owner originations, HMDA provides comprehensive information on borrower incomes and census tract locations for metropolitan areas. Unfortunately, it provides no information on the incomes of renters living in mortgaged properties (either single-family or multifamily) or on the rents (and therefore the affordability) of rental units in mortgaged properties. The AHS, however, does provide a wealth of information on rents and the affordability of the outstanding stock of single-family and multifamily rental properties. An important issue here concerns whether rent data for the stock of rental properties can serve as a proxy for rents on newly-mortgaged rental properties. During the 2000 rule-making process, POMS data were used to examine the rents of newly-mortgaged rental properties; thus, the POMS data supplements the AHS data. The recently released RFS provides information on property shares (e.g., the relative importance of rental versus owner properties) and several other important parameters in HUD's market model. The data base issues as well as other technical issues related to the goal percentages (such as the need to consider a range of mortgage market environments) are discussed in Sections F, G, and H, which present the market share estimates for the Low- and Moderate-Income Goal, the Underserved Areas Goal, and the Special Affordable Goal, respectively.

4. Conclusions

HUD is using the same basic methodology for estimating market shares that it used in its 1995 and 2000 final rules and its 2004 proposed rule. As demonstrated in the remainder of this appendix, HUD has attempted to reduce the range of uncertainty around its market estimates by carefully reviewing all known major mortgage data sources, by considering comments on the 2004 proposed rule, and by conducting numerous sensitivity analyses to show the effects of alternative assumptions. Sections C, D, and E report findings related to the property share distributions called for in Step 1, while Sections F, G, and H report findings related to the goal-specific market parameters called for in Step 2. These latter sections also report the overall market estimates for each housing goal calculated in Step 3.

In considering the levels of the goals, HUD carefully examined comments by the GSEs and others on the methodology used to establish the market share for each of the goals. Based on that thorough evaluation, as well as HUD's additional analysis for this final rule, HUD concludes that its basic methodology is a reasonable and valid approach to estimating market shares. As in the past, HUD recognizes the uncertainty regarding some of these estimates, which has led the Department to undertake a number of sensitivity and other analyses to reduce this uncertainty and also to provide a range of market estimates (rather than precise point estimates) for each of the housing goals.

C. Size of the Conventional Multifamily Mortgage Market ¹³

This Section C differs from the version published in the May 3, 2004, Proposed Rule in the following ways: The estimates from the “HUD New” and “Flow of Funds” methods discussed below in parts 2 and 3 have been updated through 2003, and responses to comments received on those methods have been added to those sections. The part titled “Most Likely Range” has been revised in light of the 2003 estimates and comments received. The discussion of “Loan Amount per Unit,” part 5, has been revised in response to comments and to newly available data from the GSEs and the 2003 American Housing Survey. The multifamily mix discussion, part 6, has been revised in accordance with other changes. Section C.7 has been added on the multifamily mix as estimated from the newly released 2001 Residential Finance Survey (RFS). Lastly, Section C.8 discusses the multifamily mixes that will be examined in HUD's projection model for 2005-2008. Other than these changes and minor editorial corrections, the text in this section is identical to that in the Proposed Rule published May 3, 2004. Changes to Tables D.2 through D.5 are noted in the text and table notes. The old Table D.5 is now D.5a and Tables D.5b and D.5c have been added.

This section provides estimates of (a) the annual dollar volume of conventional multifamily mortgage originations and (b) the annual average loan amount per unit financed. The estimates build on research reported in the Final Rule on HUD's Regulation of Fannie Mae and Freddie Mac as published in the Federal Register on October 31, 2000, especially in Appendix D. That material from the 2000 Rule will not be repeated here but will be referenced or summarized where appropriate.

This section uses the information on dollar volume of multifamily originations and average loan amounts to estimate the number of multifamily units financed each year as a percentage share of the total (both single-family and multifamily) number of dwelling units financed each year. This percentage share, called the “multifamily mix”, is an important parameter in HUD's projection model of the mortgage market for 2005-08 (see Section C.8 below)

Estimating this “multifamily mix” is important because relative to its share of the overall housing market, the multifamily rental sector has disproportionate importance for the housing goals established for Fannie Mae and Freddie Mac. This is because most multifamily rental units are occupied by households with low or moderate incomes. Between 1999 and 2002, for example, the GSEs purchased mortgages on approximately 26.1 million housing units, of which only 9.5 percent were multifamily rental units. However, of the GSEs' purchases qualifying as mortgages on low- and moderate-income housing during this period, 18 percent of the units financed were multifamily rental units. Of the GSEs' purchases qualifying as special affordable mortgages during this period, 25 percent of the units financed were multifamily rental units.

The methods used in the 2000 Rule for estimating the size of the multifamily mortgage market and related variables were the product of extensive research by HUD and review by interested parties. The approach here is first to extend those estimates through 2002 using the same methods as in the 2000 Rule, and then to present alternative methods, along with commentary.

1. Data Sources

The data sources available for estimating the size of the multifamily mortgage market are more limited in scope and timeliness than was the case for the 2000 Rule. Among the key sources described in detail in the 2000 Rule, the following are now less useful:

Survey of Mortgage Lending Activity. This survey has been discontinued; estimates are available only through 1997.

Residential Finance Survey: The 1991 Residential Finance Survey (RFS) is now 13 years out of date. (See Section C.7 for results from the 2001 RFS.)

Urban Institute Statistical Model: This model, developed in 1995 and calibrated using data from 1975-1990, is now even further removed from its calibration period and probably captures current market conditions less accurately.

Estimates from the GSEs: As part of their comments on the proposed 2000 Rule, Fannie Mae and Freddie Mac shared with HUD their own estimates of the size of the multifamily mortgage market.

Fortunately, several key sources are available with the timeliness and quality comparable to the sources used during development of the 2000 Rule. These sources are: The Home Mortgage Disclosure Act (HMDA); activity reports submitted to HUD and the Office of Federal Enterprise Oversight (OFHEO) by Fannie Mae and Freddie Mac; non-GSE mortgage-backed security issuance from the Commercial Mortgage Alert database; and multifamily mortgage activity by life insurance companies, as estimated by the American Council of Life Insurers (ACLI). For background information on each of these sources, readers are referred to Appendix D of the 2000 Rule.

2. Estimates Based on “HUD New” Methodology

In the 2000 Rule, HUD developed a new methodology for estimating aggregate multifamily conventional loan originations. The method, here labeled “HUD New”, was developed to make full use of the available data, and in particular the four sources listed above, which encompass most of the multifamily mortgage market.

The advantages of HUD New are that it provides reasonably complete coverage of the market, produces those estimates within nine months of the end of the year, generally includes only current originations and avoids double counting. The main disadvantage of HUD New is that it produces a lower bound estimate. Some loan originators are missed, including pension funds, government entities at the federal, state, and local levels, real estate investment trusts, and some mortgage bankers. Also, excluded are loans made by private individuals and partnerships. In addition to these exclusions, estimates from the covered lenders require some judgmental adjustments to conform to the definitions and time intervals of HUD New.

Despite these limitations, HUD New is one sound way to estimate the size of the multifamily conventional mortgage market. Although the method requires unavoidable judgment calls on which analysts may differ, sensitivity analysis can be performed to show the effects of different multifamily origination volumes on the goals qualifying market estimates (see Sections F-H). Due to the reasonableness of the HUD New approach, the value of maintaining continuity in estimation methods, and the fact that no data has become available in the past few years that would argue for modifying HUD New, it is used here for the baseline estimate of the size of the conventional multifamily mortgage market in 2000, 2001, 2002 and 2003.

The estimates from HUD New are presented in Table D.2. This table is the counterpart of Table D.5 in the 2000 Rule. The historical years have two columns each, one for the estimates presented in the 2000 Rule and one for estimates independently produced as part of this research. Footnotes to the table provide more complete descriptions of the components. Additional background on the calculations is provided in the 2000 Rule (Appendix D, Section C).

The revisions to the historical estimates (i.e., those in the 2000 Final Rule) result from both revisions to some of the input data and recalculations. For the years 1995 through 1998, the revisions are small for the estimates of total originations. The only one of note is a 5 percent upward revision to the estimate for 1995, prompted by a recalculation of the entry for life insurance companies. The revision to 1999 is larger, and results mostly from the substitution of the actual HMDA results for that year for the projected value used in the 2000 Rule. Surprisingly, the revised estimate for 2000 based on complete data for that year only varies slightly from the projection made at the time of the 2000 Rule. Most of the historical estimates produced in 2000 can be replicated or closely approximated, including those for Fannie and Freddie, CMBS, HMDA, and life insurance companies. The replicability of the CMBS figures is especially important, in light of all the selection criteria and hand calculations required to generate those estimates from the CMBS database. (In the 2000 Rule, the estimates for Freddie Mac and CMBS originations in 1997 appear to have been switched, and the revised estimates make this correction.)

The revised figures for 1999 and 2000 indicate that total conventional originations dropped 8 percent in 1999 from 1998's very strong level and another 13 percent in 2000. However, the HUD New estimate indicates that total conventional originations then jumped 40 percent in 2001 and further increased 15 percent in 2002. Judging from Survey of Mortgage Lending Activity estimates since 1970, the 2002 number is a new record high. For 2002, most of the increased volume is due to increases by HMDA lenders and life insurance companies.

One possible concern is that the significant increase in the HMDA number in 2002 was caused by the FFIEC relaxing its eligibility requirements between 2001 and 2002. This concern turns out to be unfounded. The FFIEC actually raised its eligibility requirements. The level of assets required by FFIEC to be reported to HMDA increased from $31 million in 2001 to $32 million in 2002. In addition, the number of HMDA reporters decreased from 7,771 in 2001 to 7,638 in 2002.

Compared with the version of Table D.2 in the Proposed Rule of May 3, 2004, the version here updates the estimates through 2003 and revises the 2001 and 2002 estimates slightly in response to newly available data. The data for 2003 point to a large, broad-based increase in the volume of multifamily lending. Total conventional originations, estimated at $89 billion, are up 32 percent from 2002, easily reaching a new record high. A large increase was observed in each of the five market segments listed in Table D.2.

Several organizations commented on the HUD New method. Fannie Mae says it involves double counting of originations. However, the one example they offer—between life insurance company data and CMBS data—should not be subject to double counting because securitizations by life insurance companies are deleted from the CMBS totals, as noted in Table D.2 and in documentation included in the 2000 Rule. Freddie Mac, through its contractor, uses an approach similar to HUD New but uses different data sources. Inadequate details are provided on the tabulations and judgments applied to evaluate the method. Lastly, MBA expresses a preference for the estimates provided by HUD New and says, without providing detail, that estimates developed by their consultants are similar to those presented in HUD New.

The comments received fail to note the point made repeatedly in the proposed rule text that the HUD New estimates are lower bounds on the volumes of originations. While HUD New is characterized in the proposed rule as providing “* * * the baseline estimate of the size of the conventional multifamily mortgage market * * *”, other language in the rule makes clear that “baseline” is used in the sense of “starting point.” For example, the proposed rule also states that “* * * unavoidable gaps in coverage make the resulting HUD New figures lower-bound estimates of actual originations rather than best ‘point’ estimates” (p. 24450).

3. An Alternative Method

The HUD New method makes use of all the available sources of data on individual origination sources in attempting to estimate total conventional mortgage originations. However, as discussed in the 2000 Rule and summarized above, unavoidable gaps in coverage make the resulting HUD New figures lower-bound estimates of actual originations rather than best “point” estimates. In addition, even for those loans that are available, certain assumptions must be made to convert the available data into estimates corresponding to the desired definition and time periods. An alternative to the bottom-up approach of HUD New avoids some of the data problems. The Federal Reserve's Flow of Funds accounts provide the most complete and timely set of estimates of multifamily mortgage credit. The Flow of Funds statistics refer to net changes in credit outstanding rather than gross originations. Specifically, balance sheet estimates of mortgage assets of lenders are used to produce estimated changes in holdings of mortgages over time. An alternative label for the resulting time series is “net change in mortgage debt outstanding.”

The historical relationship between gross originations and net change can be used to estimate recent origination volume. Separate information on FHA multifamily activity can be used to convert the total originations to estimates of only conventional originations. The Flow of Funds method that is described in this section will be called “FoF-based.”

Flow of Funds estimates of mortgage debt outstanding are based on data from sources of varying accuracy and timeliness. Bank and thrift institution holdings, taken from regulatory filings, are by all accounts highly accurate, as are those from the government sponsored agencies and direct Federal government holdings. The private MBS data and the life insurance company figures, both taken from Wall Street sources, are also thought to be reasonably accurate. Less accurate are the estimates of loans made by private individuals and certain institutions, for which comprehensive data on loans outstanding is provided only once every ten years, through the Residential Finance Survey. Fortunately, the depository institutions, GSEs, and mortgage-backed securities account for the bulk of all holdings of mortgage debt (approximately 72 percent, according to the Flow of Funds estimates for year-end 2001). Thus, most of the Flow of Funds data are from highly accurate sources.

The net change in mortgage debt outstanding in any year is the lower bound on originations. This is because the net change is defined as originations less the sum of principal repayments and charge offs. Historically loan originations have exceeded the net change by a considerable margin in both the multifamily and single-family markets. There are several reasons why the relationship of originations to net change differs between the multifamily and single-family sectors, but the basic principles apply to both sectors.

Table D.3 presents the annual estimates from the Flow of Funds. Also shown are the estimates of multifamily conventional originations as published in Table D.10 from the 2000 rule, and FHA originations from HUD administrative records.

The ratio of mortgage originations to net change should be positively correlated with the proportion of total originations that are refinancings, for which the net change in mortgage debt would be expected to be low relative to that on loans taken out in connection with a property acquisition. (This is the pattern observed in the single-family mortgage market.) Refinancings, in turn, would be expected to be prevalent relative to purchase loans at times when interest rates are low relative to their recent past.

The historical evidence generally supports this expectation regarding the relationship of originations to net lending. As shown in Table D.3, total originations have been highest relative to net change when interest rates have been low relative to their recent past. [Note: Columns A1, B1, C1, and D1 are the figures appearing in the Proposed Rule version of this table. Owing to extensive revisions to the input data, new columns with the revised inputs and calculated values have been added to facilitate comparisons. These revised figures appear in Columns A2, B2, C2, and D2.] The ten-year Treasury yield, a common benchmark for pricing multifamily mortgages, has generally trended down since 1990. The early 1990s were all marked by high originations relative to net change, and these were also years in which interest rates were particularly low relative to their trailing five-year averages. In 1996 and 1997, by contrast, originations were less high relative to net change, and these were years in which interest rates were only slightly lower than their five-year trailing averages. In estimating conventional originations for 1999-2002, the 1998 experience is a useful benchmark. That year, total originations exceeded the net change by about 80 percent, as shown in Table D.3. There was also a big drop in interest rates in 1998 relative to the recent past, providing an incentive for refinancings. As shown in the table, interest rates rose slightly in 1999 and again in 2000, presumably diminishing the incentive to refinance. Nonetheless, the net change in mortgage debt was higher in 1999 and 2000 than it had been in 1998.

Putting all this together, it seems that the appropriate ratio of total originations to net change to apply to 1999 and 2000 would be below that of 1998 and of most other years of the 1990s. Applying a ratio of 1.5 to the net change estimates in 1999 and 2000 results in a total originations estimate of approximately $56 billion. Subtracting the $4 billion in FHA originations results in estimates of $52 billion for conventional originations in each year. A subjective confidence band around this point estimate is at least +/−$2 billion.

Turning to the estimate for 2001, the first thing to note is that net change in mortgage debt jumped to $48 billion from $37 billion of the previous two years. The second thing to note is that interest rates fell by nearly a percentage point in 2001 relative to their past average. For both of these reasons, total originations in 2001 would be expected to have been higher than in 1999 or 2000. How much higher is a subjective judgment, but 1.5 would seem an appropriate multiple to apply to the net change number in 2001. This is the same multiple as in 1999 and 2000, despite the added refinancing incentive in 2001. By the beginning of 2001, there were relatively few properties “at risk” of refinancing. Many presumably had refinanced in one of the preceding years, and lock-out provisions, yield maintenance agreements, and other loan conditions may have kept these properties from coming in for refinancings. Also, there may have been some short-run capacity problems in the multifamily loan origination industry in 2001 that further curtailed volume.

Applying the 1.5 multiple to 2001's net change of $48 billion yields a total originations estimate of $71 billion. Subtracting FHA business results in a conventional originations estimate of $67 billion, to which a subjective confidence band of at least ±$2 billion appears warranted.

As seen in Table D.3, the Flow of Funds methodology indicates that total conventional originations decreased 6.5% between 2001 and 2002. In 2002, the net change in mortgage debt decreased slightly to $44 billion. Using the 1.5 multiple for 2002's net change of $44.2 billion yields a total originations estimate of $67 billion. Subtracting $4.5 billion of FHA business results in a conventional originations estimate of $62 billion.

This Flow of Funds estimate is over $5 billion less than the estimate from HUD New. This is surprising given that the HUD New method is supposed to serve as a lower boundary on the size of the multifamily market, while the Flow of Funds method is designed to produce a higher “point” estimate of the actual size of the market.

Like the estimates for HUD New, those for the Flow of Funds method have been revised and updated through 2003 to incorporate new data. As with HUD New, the Flow of Funds method suggests a large increase in conventional mortgage lending in 2003. The estimate for conventional originations in 2003 is $75 billion, up 29 percent from the revised estimate for 2002. In percentage terms, the increase in 2003 almost matches that of the HUD New method's estimates of Table D.2.

The originations estimates for earlier years, and especially 2000-2002, have been revised downward in response to revisions by the Federal Reserve to the Flow of Funds accounts and by an update to HUD's FHA estimate for 2002. The downward revision was largest for 2000, for which year the new figure of $44 billion of conventional originations is $8 billion less than the earlier estimate.

The big increase in estimated originations in 2003 is largely the result of the Federal Reserve's estimate of a large increase that year in net change in mortgage debt outstanding, shown in column A2 of Table D.3 The increase in 2003 in the Flow of Funds accounts is likely to be fairly accurate, because almost all of it is attributable to holder types for which the Fed has reliable statistics, specifically depository institutions and GSE mortgage securities. As in 1999-2002, in 2003 the net change was converted into total originations by applying a multiplier of 1.5, under the assumption that the continued decline in interest rates provided even stronger incentives for refinancing. As shown in the last columns of Table D.3, ten-year Treasury yields in 2003 averaged about 60 basis points below those of 2002, and approximately 130 basis points below the average of the previous five years.

Comments on the Flow of Funds method for estimating multifamily originations focused on the approach to converting net change into loan originations. Fannie Mae argued that it was preferable to convert by applying a liquidation rate to the stock of mortgage debt and deriving originations as net change plus estimated liquidations. A trade organization noted the historical instability of the ratio of originations to net change and argued that the “HUD New” approach to estimating originations was superior. Freddie Mac and its consultant, while not commenting directly on the Flow of Funds method, expressed a preference for a modified version of HUD New, as described in the previous part of this section.

The most recent data suggest that originations may in fact have been higher than estimated in the Flow of Funds approach and that the 1.5 multiplier used to convert net change into originations is too low. The reason is that in both 2002 and 2003, the 1.5 multiplier results in estimated conventional originations that are less than those produced by the HUD New method. As discussed earlier, HUD New provides a lower bound estimate. Fannie Mae's lower estimates of originations in recent years, relative to those in the proposed rule, result from the liquidation rate used in the calculation, which is that from Fannie Mae's own portfolio. But Fannie Mae's liquidation rate would be expected to fall below the market wide average, because Fannie Mae's multifamily business has been growing more rapidly than the market overall, and as a result its loans presumably on average are “younger” and consequently less likely to prepay or be retired than are the loans in the market as a whole. Lastly, regarding the historical instability of the ratio of originations to net change noted by a trade organization, Table D.3 of the proposed rule also presented the annual difference between originations and net change, which is considerably more stable. The differences corresponding to the 1.5 multiplier for the past several years are, as shown in D.3, below the historical averages. This is additional evidence that the 1.5 multiplier is perhaps too low.

4. Most Likely Range

In the 2000 Rule, estimates of conventional multifamily loan originations from various sources and methods were evaluated in determining the most likely range of annual originations. Those estimates were summarized in Table D.10 in the 2000 Rule. Some of the estimates from that table are reproduced below, in Table D.4, along with updates and estimates from the Flow of Funds method.

Both HUD New (column #4 in Table D.4) and FoF-based (column #9) indicate a surge in lending activity in 2001. Some corroboration of this jump is provided by other indicators, flawed though they may be. HMDA has well-documented coverage problems with multifamily loans, but it is noteworthy that HMDA-estimated conventional originations stayed in the same general range ($26 to $31 billion) in 1998-2000 before jumping to $36 billion in 2001. The composite of 1.25 times HMDA originations plus life insurance commitments, described in the 2000 Rule and updated here in column #5, also follows this basic path. Similarly, aggregate GSE multifamily purchases and securitizations stayed in the same general level in 1998-2000, before jumping in 2001, although this trend reflects changes in both market size and GSE market share. FHA originations (not shown) also rose substantially in 2001, but this too may indicate more than just market size trends.

Column #11 of Table D.4 gives the likely ranges of originations for each of the years. These are based on the estimates from all sources and interpretations of their strengths and weaknesses. In 1999, the $4 billion upward revision to the HUD New estimate from the preliminary figure reported in the 2000 Rule, together with the higher estimate produced by the FoF-based method, justify an upward revision to the $45-$48 range estimated in the 2000 Rule. The revised range is set at $50-54 billion. In 2000, HUD New (revised and extended version) suggests that originations were somewhat lower than in 1999, but FoF-based has originations holding at $52 billion. Balancing these conflicting indicators, a range of $48-$52 billion is selected for 2000. Finally, all indicators point to a substantial pickup in 2001, and the range that seems to fit best with those indicators is $65-$69 billion.

In 2002, the various methods of estimation give a mixed picture. HUD New indicates a surge in lending activity in 2002, while the flow of funds method shows a decrease in lending activity. Other methods also show divergent trends. The composite of 1.25 times HMDA originations plus life insurance commitments also shows a significant increase between 2001 and 2002. On the other hand, aggregate GSE multifamily purchases and securitizations showed a slight decrease between 2001 and 2002. FHA originations (not shown) also decreased slightly in 2002.

While this is a subjective judgment, 1.5 may not be the appropriate multiple to apply to net mortgage debt outstanding in the flow of funds model in 2002. The difference between the flow of funds estimate and the HUD estimate cannot be reconciled without adjusting the FoF multiple. Given the low interest rates in 2002, and a refinancing boom in the single-family mortgage market, it could be that the multifamily market also had a significant amount of refinancing activity. In such a case, there could be an increase in the size of the multifamily market without a corresponding increase in net mortgage debt outstanding. A higher multiple would need to be applied to the Flow of Funds model to compensate for the increase in multifamily refinancings.

Due to data limitations, the above remains a speculation. The largest increase in multifamily volume came from HMDA reporting lenders. The HMDA data do not allow for the separation of multifamily purchase originations from refinancings. Other data sources need to be explored to determine if an adjustment to the FoF-based model is appropriate.

Both HUD New and the FoF-based method indicate a large increase in conventional multifamily loan originations in 2003. But the FoF estimates for each of the previous four years have been revised downward in light of revised input data. According to these updated and revised estimates, conventional multifamily originations by HUD New have exceeded the estimates of FoF in two of the past five years, and in the other three years FoF exceeded HUD New by only narrow margins. Because HUD New produces lower bound estimates of originations, whereas FoF is intended to provide best point estimates, the Department concludes that the 1.5 multiplier applied in the FoF method is too low, and as a result the FoF estimates understate originations in the past several years. In light of this probable underestimate of the multiplier, and after consideration of comments received, the Department believes that the likely ranges of conventional originations for 2002 and earlier years as published in the May, 2004, Proposed Rule continue to be reasonable estimates, although likely on the conservative side. As for 2003, the estimates from HUD New and FoF indicate a substantially higher likely range, which the Department has set at $85 billion to $100 billion. As explained in Section C.6 below, HUD will conduct sensitivity analyses in Sections F-H showing the effects of different multifamily mixes on the historical estimates of the goals-qualifying shares of the mortgage market.

5. Loan Amount per Unit

In determining the size of the conventional multifamily mortgage market for purposes of the GSE rules, the measure of market size is the annual number of conventionally financed multifamily rental housing units. The number of units is derived by dividing the aggregate annual originations by an estimate of the average loan amount per housing unit financed. For this reason, accuracy in the estimate of loan amount per unit is as important as accuracy in the dollar estimate of aggregate conventional originations. A 10 percent error in either will result in a 10 percent error in the estimate of market size.

The 2000 Rule used estimates of loan amount per unit drawn from various sources. As summarized in Table D.9 of the 2000 Rule and the accompanying text, the estimates for 1993-1998 were taken from the GSEs and for 1999 from CMBS data. “Unpaid Principal Balance” or UPB—a balance sheet measure which for current year loan originations will differ little from the initial loan amount—is used to calculate aggregate originations of loans bought or securitized by the GSEs or pooled into non-GSE mortgage-backed securities. The figures from Table D.9 of the 2000 Rule are reproduced below in Table D.5a, along with updated estimates from all three sources for 2000, 2001 and 2002. The estimates that are new since the 2000 Rule appear in italics.

Several options are available for developing estimates for 2000, 2001 and 2002. The first is to use the UPB (unpaid principal balance) per unit estimates from the GSEs. These estimates, taken from the Fannie Mae and Freddie Mac annual activity reports to HUD, are as follows, computed as in the 2000 Rule as a unit-weighted average of the unpaid principal balance (UPB) per multifamily unit in Fannie Mae's and Freddie Mac's portfolios:

1997	$27,266
1998	31,041
1999	35,038
2000	37,208
2001	37,258
2002	39,787

The figure for 2002 is approximately 46 percent higher than in 1997. Both Fannie Mae and Freddie Mac's portfolios generate estimates of between $39,000 and $40,000 for 2002.

Several alternative approaches to estimating loan amount per unit are available. The first is to base the estimate on CMBS data, as was done for 1999 in the 2000 Rulemaking. As shown in the last column of Table D.5, the estimates of UPB/unit from this source are somewhat below those of the GSEs and indicate less increase since the late 1990s.

In the first 10 months of 2002, CMBS properties showed a UPB/unit of $37,038, a nearly 14 percent jump over the previous year. Although slightly below the UPB/unit for the GSEs, the CMBS numbers are closer to the GSE calculations than in previous years.

Another approach is to move the 1999 estimate of UPB/unit forward by some justifiable index. The 2001 estimates use the change in average rent on multifamily rental units from the American Housing Survey. Because AHS data are not available for 2002, the 2002 estimate uses the consumer price index for rent of primary residence. Both AHS and CPI rent estimates are listed below:

Year	Median	Mean	CPI
1999	$550	$592	177.5
2001	590	647	192.1
2002	N/A	N/A	199.7

There is some variation between the two measures. In the AHS, median rent rose 7.3 percent over this two-year period, and mean rent increased 9.3 percent. Meanwhile, the CPI showed an increase of 8.2 percent. In 2001, using the AHS produces an estimate of $34,000. The CPI yields a smaller estimate for 2001; applying the 8.2 percent increase from the CPI results in a 2001 estimate of $33,200. Since the AHS data are unavailable in 2002, the CPI provides a 2002 estimate of approximately $35,000.

In 2001, the rent-adjusted 1999 estimate was in between the estimates from the CMBS and GSE data, and was a fair estimate of the actual size of the market. In 2002, however, the rent-adjusted number is below both the CMBS and GSE calculations. The rent-adjusted number could be underestimating the 2002 UPB/unit. Either the CMBS or GSE calculations, or an average of the various methods could be used. Sections F-H will report the results of sensitivity analyses showing the effects of the different multifamily mortgage estimates and different per unit amounts on the goals-qualifying shares for the year 2002. Under the various estimates, the multifamily mix (defined below) for 2002 varies from 9.5 percent-11 percent.

Since the proposed rule was issued by the Department, data for 2003 have become available that permit updates of some of the sources of UPB/unit estimates. The GSEs' experience, shown in the bottom row of Table D.5a, was mixed. Fannie Mae's UPB/unit increased about 4 percent from 2002, but Freddie Mac's dropped 9 percent. The volume-weighted average UPB/unit for the GSEs in 2003 was $39,082, off about 2 percent from the 2002 average of $39,787 shown in the text table above.

The most recent rent estimates from the American Housing Survey also suggest limited or negative recent growth in UPB/unit. The median and mean rents for 2003 that correspond to those in the table above are $609 and $671. Given the logic of this method as described in the proposed rule, it seemed most appropriate to use the percent increase in AHS rents from 1999 to 2003 to update the 1999 UPB/unit ($30,719) to a 2003 figure. Using the 13.3 percent increase in mean rent between 1999 and 2003 (the increase in median was only 10.7 percent) and moving the baseline UPB/unit from 1999 forward to 2003 by this proportion brings the 2003 UPB/unit to $34,805. That is the number appearing in Table D.5a. For comparison, the CPI rent index rose 15.8 percent between 1999 and 2003.

In commenting on HUD's UPB/unit estimates for 2000-2002, as published in the May 2004 Proposed Rule, both Fannie Mae and Freddie Mac expressed the view that the estimates were too low. They cited both their own experience and other evidence and argued that HUD's reliance on CMBS and rent data, and switching of benchmark years, resulted in UPB/unit estimates that were substantially below the actual market averages.

In reviewing the comments and in light of these new data HUD has concluded that the estimates in the proposed rule likely were too low. The more difficult determination is where to set the estimates. The Department has not revised its estimate of UPB/unit for 2002 and earlier years, because of this uncertainty. The situation is similar to that discussed in the previous part of this section in discussing the likely range of conventional multifamily originations, where the new data lead the Department to think the Flow of Funds estimates may be too low, but no adjustments were made to the likely range as reported in Table D.4. If adjustments were made to the historical estimates of originations and UPB per unit, the revisions would be at least partially offsetting, with little net effect on the historical estimates of number of multifamily units financed. As for 2003, weighing all available information, the Department has set the UPB/unit at $39,082, the weighted average of the GSEs' actual UPB/unit for that year. As explained in the next section, goals-qualifying estimates for 1995-2002 are reported in Sections F-H that include multifamily mixes approximately two-three percentage points lower that the multifamily mixes suggested by the most likely range of multifamily dollar estimates and the UPB/unit estimates.

6. Multifamily Mix During the 1990s

This section uses the information on dollar volume of multifamily originations (Table D.4) and average loan amounts (Table D.5a) to estimate the number of multifamily units financed each year as a percentage share of the total (both single-family and multifamily) number of dwelling units financed each year. Because of the high goals-qualifying shares of multifamily housing, the multifamily mix is an important parameter in HUD's projection model for the overall market; other things equal, a higher multifamily mix (or conversely, a lower share of single-family loans) leads to a higher estimate of goals-qualifying loans in the overall mortgage market. This percentage share, or “multifamily mix”, is reported in the last two columns of Table D.4 for the years 1991 to 2002.^[14] The “minimum” (“maximum”) multifamily mix figure reflects the low (upper) end of the “likely range” of multifamily dollar originations, also reported in Table D.4. Because they will be compared with other estimates of the MF mix, these “likely range” data are reproduced in the first two columns of Table D.5b.

Table D.5b includes several averages of the MF mix for different time periods between 1991 and 2002. Based on the “likely range” of annual conventional multifamily origination volume, multifamily units have represented 15.3 percent (the average of the “minimum” figures) to 16.6 percent (the average of the “maximum” figures) of units financed each year between 1991 and 2002. Considering the mid-points of the “likely range”, the multifamily mix averaged 15.9 percent during this period. Notice that the multifamily mix is lower during years of heavy refinancing when single-family originations dominate the mortgage market; the multifamily mix was only 13-14 percent during 1993, 1998, and 2001, and 11 percent (or less) during 2002.^[15] As discussed in Sections F-H, record single-family originations ($3.8 trillion) during 2003 likely resulted in that year having a lower multifamily mix than any of the years between 1991 and 2002. Sensitivity analyses are conducted to show the effects of multifamily mixes less than the previous lows of 11 percent in 1992 and 2002.

As discussed earlier, several commented that HUD had understated the UPB/unit, which caused HUD to overstate the share of newly-mortgage multifamily dwelling units. Section C.5 explains that HUD's UPB/unit estimates for recent years are likely too low but that could be offset by low estimates of originations. To allow for different views about the volume of mortgage originations and the UPB/unit, Sections F-H will conduct sensitivity analyses with lower multifamily mixes than suggested by the mid-points of the likely ranges in Table D.5b. The third column of Table D.5b lists the “mid-point” MF mixes while the fourth column of Table D.5b lists the lower MF mixes used in Sections F-H. Over the 1995-2002 period, the average MF mix ranged from 13.9 percent (the lower MF mix approach) to 16.2 percent (the mid-point MF mix approach).^[16] Over the more recent period, the averages have ranged from 12.6 percent to 14.5 percent for 1999-2002, from 15.1 percent to 17.5 percent for recent home purchase years, and from 11.2 percent to 12.9 percent for the refinance years of 1998, 2001, and 2002.

The impact of the lower MF mix on the UPB/unit assumption can be illustrated for the case of 2001, which assumed a loan-amount-per-unit figure of $34,000. Reducing the MF mix from 13.5 percent to 12.0 percent is consistent with increasing the UPB/unit from $34,000 to $39,075 (holding constant mortgage originations at $67 billion). Of course, the lower MF mix of 12.0 percent is consistent with a lower volume of mortgage originations if the initial UPB/unit of $34,000 is retained.

Fannie Mae (op.cit., page I-29) developed three sets of UPB-per-unit figures for 1997 to 2002; below Fannie Mae's estimates are compared with the UPB-per-unit figures that result from HUD's model that uses the lower MF mixes.

	Fannie Mae's Estimates			HUD's Lower MF Mix
	High	Low	Baseline	Model
1997	$35,063	$28,488	$31,776	$33,582
1998	40,155	32,626	36,390	37,492
1999	42,430	33,992	38,211	36,260
2000	45,797	37,210	41,504	38,142
2001	48,363	39,295	43,829	39,075
2002	53,507	43,474	48,491	44,009
Average	44,219	35,847	40,033	38,093

Three points stand out. First, there is a rather large differential between Fannie Mae's Low and High UPB-per-unit figures, reflecting the lack of available data. Second, HUD's UPB-per-unit estimates based on its lower MF mix model are in between Fannie Mae's Low and Baseline estimates. Third, the differentials between HUD's and Fannie Mae's Baseline estimates are largest during the two heavy refinance years of 2001 and 2002.

HUD's conducting its market share analysis with the lower MF mixes (as well as with the mid-point MF mixes) recognizes different views about the size of the mortgage market and the UPB/unit. This does not mean that the HUD's range of MF mixes includes estimates as low as those suggested by ICF (Freddie Mac's contractor) and Fannie Mae.

ICF's estimates of multifamily shares for the 1994-2002 were lower than those that HUD used (as reported in Table D.5b). ICF's Best Estimates and Lower Bound Estimates were as follows:^[17]

	Best estimates (percent)	Lower bound estimates (percent)
1994	17.2	14.0
1995	16.5	14.0
1996	13.7	11.5
1997	14.4	12.3
1998	11.3	9.9
1999	12.3	10.7
2000	13.8	11.7
2001	10.8	9.0
2002	10.2	8.5

Various averages of ICF's Best Estimates are calculated in Table D.4b. Over the 1995-2002 period, ICF's Best Estimates averaged 12.9 percent, while HUD's mid-point estimates averaged 16.2 percent and HUD's lower MF mix estimates averaged 13.9 percent. Thus, the average of ICF's Best Estimates is slightly lower (one percentage point) than the average of HUD's lower MF mixes. Over the more recent 1999-2002 period, ICF's Best Estimates averaged 11.8 percent, while HUD's mid-point estimates averaged 14.5 percent and its lower MF mix estimates averaged 12.6 percent.

ICF also produces lower bound estimates of the multifamily share of the market (see above list for 1994 to 2002). ICF's lower bound estimates for the multifamily mix averaged 11.3 percent between 1994 and 2002. It is interesting that ICF's lower bound estimates are in some cases either similar or less than the multifamily shares of Fannie Mae's business. The multifamily share of Fannie Mae's business was 9.9 percent in 1999 (versus ICF's lower bound estimate for the market of 10.7 percent), 13.3 percent in 2000 (versus ICF's lower bound of 11.7 percent), and 10.9 percent in 2001 (versus ICF's lower bound market estimate of 9.0 percent). Even though these Fannie Mae data include both their seasoned and current-year purchases, it is surprising that ICF's market estimates would be similar or less than Fannie Mae's multifamily shares, given that Fannie Mae purchased practically no small (less-than-50-unit-property) multifamily loans during this period.

In its comments, Fannie Mae also provided various historical estimates of the MF mix (see its Appendix I, pages I-29 and I-30). First, without giving the details of its analysis, Fannie Mae asserts that “Fannie Mae's analysis shows an average multifamily share of 10.2 percent for the 1997-2002 period, compared with HUD's 14 to 15 percent range” (page I-30). Fannie Mae's estimate of 10.2-percent is below ICF's Best Estimate (12.1 percent), HUD's lower MF mix estimate (13.1 percent), and HUD's mid-point MF mix estimate (15.2 percent). (See Table D.5b.) Fannie Mae's estimate of 10.2 percent is practically the same as ICF's Lower Bound Estimate, which averaged 10.4 percent between 1997 and 2002; of course, this raises the same issue mentioned above with respect to ICF's Lower Bound Estimates.

Fannie Mae also provided various estimates of UPB per unit (see above) and applied its “Low UPB per Unit Assumption” and its “High UPB per Unit Assumption” to HUD's likely range of MF mortgage originations (as reported in column 11 of Table D.4). For the period 1997-2002, Fannie Mae obtained: (A) a range of 12.7-13.8 percent using its “Low UPB per Unit Assumption” and (B) a range of 10.5-11.5 percent using its “High UPB per Unit Assumption.” (See Fannie Mae's Table I.6 on page I-30.) Fannie Mae's (A) results are similar to HUD's lower MF mix estimates, which averaged 13.1 percent over the 1997-2002 period; its (B) results are slightly higher than ICF's Lower Bound Estimates, which averaged 10.4 percent over the 1997-2002 period.

Finally, Fannie Mae notes that its baseline analysis shows that the multifamily share dropped to 5.6 percent in 2003 and that HUD's MF assumptions (e.g., 13.5 percent) clearly overstate typical multifamily shares and therefore the likely market opportunity for the GSEs (page I-30). HUD recognizes that the MF mix will be lower during heavy refinance years such as 2003, making it more difficult for the GSEs to achieve the housing goals; HUD's Advance Notice of Proposed Rulemaking (described in the Preamble) seeks proposals on how to treat heavy refinance years in the goals determination process. The range of MF mixes (13.5-15.0-16.0 percent) in HUD's projection model apply to a home purchase environment, not a heavy refinance environment.

As discussed in Section C.8 below, HUD will continue to use a 15 percent MF Mix as its baseline. In their comments on the proposed rule, both Fannie Mae and Freddie Mac expressed the view that HUD's 15 percent baseline estimate of the multifamily share of the conventional mortgage market was too high. As described earlier in this section, those organizations argued that HUD's estimates of multifamily loan originations were too high, that HUD's estimates of multifamily UPB/unit were too low, and that these two errors together combined to produce an estimate multifamily market share that was one to four percentage points too high. A trade organization reached similar conclusions in their comments on the multifamily mix.

The Department has carefully considered these comments and the analysis supporting them. But HUD's conclusion is that the 15.0-percent baseline multifamily mix appropriately reflected the estimates and analysis appearing in the May 2004 Proposed Rule. The Department's responses to critiques of the individual components of the multifamily mix calculation appear earlier in this section. In addition, the Department's confidence that a 15 percent estimate for multifamily's share of conventionally financed is not too high is bolstered by data from the newly released 2001 Residential Finance Survey (RFS). As discussed in the next section, the RFS indicates a long-run market share for multifamily that is considerably higher than 15 percent. After presenting the RFS results, Section C.8 will return to the discussion of the baseline MF mix used in HUD's projection model.

7. Evidence on the Multifamily Mix from the 2001 Residential Finance Survey

Subsequent to the Department issuing the proposed rule in May, 2004, the Census Bureau released the 2001 Residential Finance Survey (RFS). The RFS provides new information on the size and composition of the residential mortgage market. As noted by Fannie Mae, Freddie Mac, and other organizations commenting on the draft rule, the RFS is an important and unique data source of data, because it is designed to provide comprehensive, nationally representative estimates on the volume and characteristics of single-family and multifamily mortgage loans and the properties they finance. Some organizations urged that the Rule not be finalized until data from the RFS has been analyzed.

The RFS data suggest a mortgage market somewhat different in size and composition from that estimated by most analysts based on partial data. Beginning with multifamily lending, the multifamily mortgage market is considerably larger than most analysts have thought, according to the RFS. For example, the RFS estimate of total mortgage debt outstanding on properties with five of more housing units is $608 billion dollars. The only other comprehensive estimate comes from the Federal Reserve Board's “Flow of Funds” accounts, which draw on data from multiple sources and on judgments by the Fed staff. The Flow of Funds estimate of multifamily debt outstanding as of 2002Q2 (the quarter most comparable to reporting dates of RFS respondents) was only $457 billion. In other words, the RFS estimates a stock of multifamily mortgage debt 32 percent larger than Federal Reserve.

As with debt outstanding, multifamily loan originations in the RFS exceed most other estimates. Over the period 1998-2001, annual originations averaged $66 billion according to the RFS, and conventional originations (total less FHA insured) averaged $61 billion. HUD's estimates of conventional multifamily originations for these years, as summarized in Table D.2 of the proposed rule, averaged only $56 billion. In commenting on the proposed rule, Fannie Mae and Freddie Mac offered estimates of market size considerably below these.^[18]

The single-family mortgage estimates from the 2001 RFS, like the multifamily estimates, are at odds with those from some other sources. For example, total mortgage debt on 1-to-4 family residences, according to the RFS, was $5.032 trillion, whereas the Flow of Funds estimate for 2002Q1 was a much higher $6.546 billion.

In summary, the RFS estimates a somewhat smaller residential mortgage market than the Flow of Funds—19 percent smaller as measured by total debt outstanding. Furthermore, multifamily debt is a much larger part of the total residential market in the RFS than in the Flow of Funds.

The RFS also records the number of housing units at each surveyed property, providing an opportunity to measure directly the number of housing units financed instead of relying on indirect methods. The RFS estimates indicate that, as with debt outstanding, the mix of mortgage lending by the measure of units financed is more heavily multifamily than previously thought. This is shown in Table D.5c, where units financed are presented for the loan origination years 2000 and 2001. These are the years for which the estimates are least likely to be biased by refinancing between the loan origination date and the survey. The estimates for 2001 are incomplete, because approximately 10 percent of the survey respondents reported as of dates prior to December 31, 2001 and loans subsequently originated on those properties would not be included. This undercount should affect single-family and multifamily reporting about proportionally, with little effect on the market share calculations.

By the housing goals' metric of number of conventionally financed, conforming housing units, the 2001 Residential Finance Survey indicates a multifamily market share substantially above the pre-RFS estimates of HUD and GSEs. As detailed in Table D.5b, the multifamily share estimated for 2001 is 0.197, or 19.7%, and the share for 2000 is a striking 0.254, or 25.4%. These high figures are particularly noteworthy because the year 2001 was marked by high levels of refinancings, which have been viewed as boosting single-family lending proportionally more than multifamily. HUD's estimate of the multifamily share for 2000, for example, was only 13%-14%, as derived elsewhere in this rule.

There are several reasons for accepting the RFS estimates as an accurate portrayal of the residential mortgage market. First, the estimates are generated from a national representative sample of properties as drawn by experts at the U.S. Census Bureau. Second, the survey forms were designed in consultation with industry experts. Third, participation in the survey was mandatory, because it was conducted in conjunction with the 2000 Census. And fourth, data processing and editing at the Census Bureau prior to public release of census and survey results is meticulous.

Nonetheless, for the specific reasons noted, results from the RFS should be interpreted cautiously. First, loan originations for any year will be understated, because the RFS will record only those loans still outstanding as of the late 2001 or early 2002 survey date. Loans originated in, for example, 1998, will be recorded only if those loans have not been refinanced, repaid, or charged off prior to the RFS survey date. For this reason, the RFS unit count and especially the market share estimates for 2001 are more reliable than those for 2000 and earlier years. Second, some of the results of the RFS are substantially at odds with other evidence and industry perceptions, as noted already. Another example of a surprising RFS finding is the time path of multifamily loan originations. According to the RFS, originations were roughly 50 percent greater in 1998-1999 than in 2000-2001, whereas most other evidence points to originations in 2000-2001 that at least equaled, and likely exceeded, the volume of 1998-1999.

Lastly, in response to user feedback and its own data checks, the Census Bureau has revised the RFS estimates three times since the initial data release in early July 2004. The possibility remains that additional errors will be found and that the resulting revisions to the data will significantly change the RFS portrayal of the multifamily mortgage market. HUD will continue its analysis of the RFS as new versions are released.

On balance, the Department views the RFS as providing strong additional evidence that the Department's baseline multifamily mix percentage of 15% is not an overestimate. The RFS data, weighed alone, would have that percentage set much higher.

8. Multifamily Mix in HUD's Model—Further Discussion

As noted above, the “multifamily mix” is the number of multifamily units financed each year as a percentage share of the total (both single-family and multifamily) number of dwelling units financed each year. Because of the high goals-qualifying shares of multifamily housing, the multifamily mix is an important parameter in HUD's projection model for the overall market; other things equal, a higher multifamily mix (or conversely, a lower share of single-family loans) leads to a higher estimate of goals-qualifying loans in the overall mortgage market.

The multifamily share of the conforming conventional market (or “multifamily mix”) is utilized below as part of HUD's analysis of the share of units financed each year meeting each of the housing goals. The proposed rule considered multifamily mixes of 13.5 percent, 15.0 percent, and 16.5 percent, as well as even lower multifamily mixes for heavy refinance environments such as 2001-03. The 15.0 percent level was considered as the baseline based on analysis of multifamily shares during home purchase environments of the 1990s. In the market sections below, HUD continues to focus on the baseline 15.0 percent but also considers a range of estimates, including those provided by commenters on the proposed rule. Comments by Fannie Mae and ICF are summarized below.

In its projection model, Fannie Mae uses a multifamily mix of 12.3 percent (see Table 1.6 on page 11). As noted in Section C.6 above, Fannie Mae estimated an average multifamily mix of only 10.2 percent over the 1997-2002 period. Fannie Mae notes that HUD's 13.5-16.5 range is “well above the range of estimates suggested by an examination of all available data and is inconsistent with the current weak fundamentals in the multifamily market.” (Fannie Mae, p. 15) Fannie Mae's views about the future mortgage market were discussed on pages I-14 to I-17 in its Appendix I (“Comments on HUD's Analysis of the Statutory Factors”) to its comments. As discussed earlier, Fannie Mae's somewhat pessimistic views about the future market were driven by the current high vacancy rates for multifamily properties and the fact that the high-renter age group (the so-called “echo boom” aged 20-34) will not begin to increase until after 2007. Fannie Mae also emphasized that the recent spike in multifamily originations (beginning in 2001) means that a large portion of today's holders of multifamily mortgages have already refinanced and therefore will have only limited ability and incentive to refinance over the next several years, due to yield maintenance provisions on their existing multifamily mortgages. According to Fannie Mae, these loans will not begin to exit their yield maintenance periods until sometime between 2008 and 2010, with the result being that the 2005-2008 period appears to have relatively limited prospects for multifamily refinancing. Fannie Mae notes that single-family lending is not subject to these constraints and is more likely to undergo modest refinance waves as a result of interest rate fluctuations. Based on its analysis, Fannie Mae concludes that a multifamily share of 12.3 percent is “consistent with reasonable estimates” of the multifamily market (Fannie Mae Appendix, Table I.15, p. I-42).

Based on its analysis of the multifamily market, ICF, Freddie Mac's contractor, offered higher projections of the MF mix. Specifically, ICF provided the following estimates of the multifamily mix during the projection period, 2005-08, as follows:

	ICF MF Mix (percent)
2005	13.7
2006	14.5
2007	14.7
2008	13.9
average	14.2

Thus, ICF's 14.2-percent average estimate is a little less than HUD's baseline (15.0 percent), standing at the mid-point of HUD's 13.5 and 15.0 figures. For a discussion of ICF's methodology for estimating the multifamily mix, and their actual use of their estimated multifamily mixes in projecting overall market estimates for the three housing goal categories, see pages 126-140 of their technical appendix, entitled “Analysis of the Proportion of the Mortgage Market that Meets the GSEs” Affordable Housing Goals: Issues of Variability and Uncertainty: Technical Appendix” (July 15, 2004). According to ICF, they projected the number of multifamily (MF) units based on the existing number of units likely to be refinanced (rollover) and the expected number of MF units that would be added to the housing stock (new completions). The amount of rollover was estimated as the average of the number of units financed 8, 9, and 10 years ago. ICF used these time periods because 10-year balloon mortgages are the most common MF mortgages, and MF loans typically include a yield maintenance period to limit prepayments.^[19] In their basic report, they state that they view the above estimates from their MF projection model as “our core, or our most likely forecast for 2005 through 2008” (ICF Report, p. 40). While they state that “our [ICF] multifamily projections for 2005 through 2008 have a sound empirical basis owing to the nature of multifamily mortgages and new multifamily construction,” ICF also reminds readers of the uncertainty of its MF projections when it states “while we believe the core range is the best and most likely estimate of the future market, we [ICF] recognize that it is possible that the actual outcomes may be outside this range, either higher or lower” (ICF Report, p. 40). The ICF basic report is entitled “Analysis of the Proportion of the Mortgage Market that Meets the GSEs” Affordable Housing Goals: Issues of Variability and Uncertainty: Technical Appendix” (July 15, 2004). Because the basic report and the appendix are paginated differently, they will be referenced separately—ICF's basic report will be referred to as the “ICF Report”, while their appendix will be referred to as the “ICF Appendix”.

As discussed earlier, the 2001 RFS provides higher estimates of the MF mix for 1999-2001 than either Fannie Mae or ICF. The RFS data suggest that 15.0 percent is a reasonable baseline, particularly for a home purchase environment. Thus, the market analysis of the housing goals in Sections F-H will continue to use 15.0 percent as the baseline MF mix. To reflect the uncertainty with the MF data, market projections will also be provided for alternative MF mixes of 12.25 percent (approximating Fannie Mae's projection of 12.3 percent), 13.5 percent (the low-end projection for a home purchase environment used in HUD's 2004 proposed rule), 14.25 percent (approximating the 12.2 percent average of ICF's best projections of MF mixes between 2005 and 2008), and 16.0 percent (a half percentage point below the high-end projection for a home purchase environment used in HUD's 2004 proposed rule). Based on ICF's best projection and HUD's analysis of the 2001 RFS, the bottom end of the range probably should not go below 13.5 percent for a home purchase environment. However, results are provided for the 12.25 percent in order to show the sensitivity of the market sizing to the assumption made by Fannie Mae in its analysis. Of course, it is recognized that the multifamily mix will be significantly lower during heavy refinancing periods such as 2001-2003. Therefore, additional sensitivity analyses will be conducted to show the effects of even lower multifamily mixes. But as explained in the Preamble of this Final Rule, in its goals scoring, HUD will reduce refinance loans so they account for not more than 40 percent of combined home purchase and refinance loans. This addresses the problem of a low MF mix during a heavy refinancing period reducing the ability of the GSEs to meet the new goal targets.

D. Single-Family Owner and Rental Mortgage Market Shares

1. Available Data on Investor Share

As explained later, HUD's market model will also use projections of mortgage originations on single-family (1-4 unit) properties. Current mortgage origination data combine mortgage originations for the three different types of single-family properties: Owner-occupied, one-unit properties (SF-O); 2-4 unit rental properties (SF 2-4); and 1-4 unit rental properties owned by investors (SF-Investor). The fact that the goal percentages are much higher for the two rental categories argues strongly for disaggregating single-family mortgage originations by property type. This section discusses available data for estimating the relative size of the single-family rental mortgage market.

The Residential Finance Survey (RFS) and HMDA are the data sources for estimating the relative size of the single-family rental market. The 2001 RFS provides mortgage origination estimates for each of the three single-family property types, as it includes mortgages originated during 2001, as well as surviving mortgages that were originated in earlier years such as 1999 and 2000. HMDA divides newly-originated single-family mortgages into two property types:^[20]

(1) Owner-occupied originations, which include both SF-O and SF 2-4.

(2) Non-owner-occupied mortgage originations, which include SF Investor.

The percentage distributions of single-family mortgages from HMDA and the 2001 RFS are provided in Table D.6a and D.6b. HMDA data will be discussed first. Because HMDA combines the first two categories (SF-O and SF 2-4), the comparisons between the data bases must necessarily focus on the SF investor category. The following points stand out from Table D.6.a:

• The investor share of all single-family loans has ranged from 5.7 percent (1993) to 9.1 percent (2000), with an average of 7.8 percent. Over the more recent 1999-2003 period, the investor share has averaged 8.3 percent.
• The investor share is much higher for home purchase loans than for refinance loans. The investor share of home purchase loans averaged 9.6 percent between 1993 and 2003, as compared with a 6.8 percent average for refinance loans.
• The investor share for home purchase loans recently increased, rising from slightly above 9.0 percent during 1999 to around 10.0 percent during 2000-2001 to 12.0-13.0 percent during 2002 and 2003. The average investor share for home purchase loans was 11.2 percent between 1999 and 2003.
• In its comments, Fannie Mae noted that HUD should deduct subprime loans from investor loans. As shown in the middle portion of Table D.6a, deducting investor subprime loans reduces the overall investor share by approximately one-half percentage point (e.g., 1999-2003 average is reduced from 8.3 percent to 7.7 percent).^[21]

• HMDA data for metropolitan areas (bottom portion of Table D.6.a) show a slightly lower investor share than HMDA data for both metropolitan and non-metropolitan areas (top portion of Table D.6a). Between 1993 and 2003, the investor share in metropolitan areas averaged 7.5 percent, as compared with 7.8 percent for the U.S. as a whole. During the more recent 1999-2003 period, the differential was slightly higher, 7.8 percent versus 8.3 percent.

Table D.6b provides information on investor loans from the 2001 RFS. During 2001, investors accounted for 13.4 percent of all new single-family mortgages. Similar to the pattern in HMDA, the RFS-reported investor share of home purchase loans (15.7 percent) was higher than the investor share (9.0 percent) of refinance loans (see Table D.6b). The RFS-based investor shares were similar for single-family mortgages originated in earlier years that had also survived (i.e., not prepaid) until the time of the RFS survey in 2001; for example, the investor share was 13.0 percent for surviving 1999 mortgages and 14.0 percent for surviving year 2000 mortgages.

For comparison purposes, Table D.6c provides investor shares of the single-family mortgages purchased by the GSEs. Between 1999 and 2003, the investor share of Fannie Mae's single-family mortgage purchases ranged from 4.2 percent (1999) to 7.8 percent (2000). Freddie Mac's investor share has been lower, ranging from 3.0 percent (2003) to 4.8 percent (2000). The low figure for 2003 was due to the heavy refinancing of owner loans in that year.

The RFS investor share of 13.4 percent in 2001 is substantially larger than the corresponding HMDA investor share of 7.8 percent. In their comments on the 2004 proposed rule, as well as in their comments on HUD's earlier 1995 and 2000 GSE rules, the GSEs have argued that HUD should use the HMDA-reported SF investor share. In its 1995 and 2000 rules and the 2004 proposed GSE rule, HUD's baseline model assumed a 10 percent share for the SF investor group—only slightly higher than the HMDA-based estimates; alternative models assuming 8 percent and 12 percent were also considered. At that time, HUD argued that its baseline projection of 10 percent was probably quite conservative; however, given the uncertainty around the data, it was difficult to draw firm conclusions about the size of the single-family investor market, which necessitated that HUD conduct sensitivity analyses using investor shares (e.g., 8 percent) less than 10 percent. HUD's argument that its 10 percent baseline work was probably conservative was based on earlier work by Blackley and Follain. It is interesting to briefly review their work because they focused on the differences between RFS and HMDA data.

2. Blackley and Follain Analysis of Investor Market Share

As mentioned, during the 1995 rule-making, HUD asked the Urban Institute to analyze the differences between the RFS and HMDA investor shares and determine which was the more reasonable. The Urban Institute's analysis of this issue is contained in reports by Dixie Blackley and James Follain.^[22] Blackley and Follain provide reasons why HMDA should be adjusted upward as well as reasons why the RFS should be adjusted downward. They find that HMDA may understate the investor share of single-family mortgages because of “hidden investors” who falsely claim that a property is owner-occupied in order to more easily obtain mortgage financing. RFS may overstate the investor share of the market because units that are temporarily rented while the owner seeks another buyer may be counted as rental units in the RFS, even though rental status of such units may only be temporary. The RFS's investor share should be adjusted downward in part because the RFS assigns all vacant properties to the rental group, but some of these are likely intended for the owner market, especially among one-unit properties. Blackley and Follain's analysis of this issue suggests lowering the investor share from the 1991 RFS-reported investor share of 17.3 percent to about 14-15 percent.

Finally, Blackley and Follain note that a conservative estimate of the SF investor share is advisable because of the difficulty of measuring the magnitudes of the various effects that they analyzed. In their 1996 paper, they conclude that 12 percent is a reasonable estimate of the investor share of single-family mortgage originations.^[23] Blackley and Follain caution that uncertainty exists around this estimate because of inadequate data.

3. GSE Comments on SF Rental Shares in the Proposed Rule

Fannie Mae, Freddie Mac, and ICF thought that the investor share should be lower than the 10 percent used by HUD. While they agreed with HUD that the RFS provided the most accurate estimate of the true investor share of the market, they emphasized that lender reporting of investor loans to the GSEs was best proxied by HMDA data (which, of course, are based on lender reports). That is, the actual opportunities available to the GSEs in the SF investor market are best measured by data that lenders report based on information from actual loan applications. Based on this argument, they concluded that HUD's market sizing analysis should rely on HMDA data, not RFS data.

For example, Fannie Mae argued that the most valid measure of the single-family rental market is the same measure (lender-reported data to HMDA) against which the GSEs' performance is measured. Fannie Mae points out that that two (10 percent and 12 percent) of the three scenarios that HUD uses exceed the highest investor share ever reported in HMDA. Fannie rejects HUD's justification (the 1991 RFS and the Blackley-Follain analysis) for using the higher scenarios because the lender reporting to the GSEs is closer to HMDA data than to the reporting in the RFS. Fannie Mae argues that the 1995 Blackley and Follain analysis bolsters its case against the RFS measures. Fannie Mae notes that both HUD and Blackley and Follain conclude that there is a reporting bias in the HMDA data that is not present in the RFS. The bias is in part due to hidden investors. At the time of origination, the property may be owner-occupied or may be intended to be owner-occupied. In fact, the property may become rental shortly after origination. As a result, the RFS reports a more accurate higher percentage of rental housing because it is a snapshot of housing, not a collection of information at mortgage origination. Fannie Mae says HUD uses the RFS because it is the more accurate measure of the rental market at any moment in time. However, Fannie Mae argues that the same bias in HMDA also exists in its own reporting when it acquires mortgages. According to Fannie Mae, an apples to apples comparison would make sure that the GSE goals contain the same biases that the GSE reports contain, rather than no bias. Finally, Fannie says that even HMDA overstates the investor share of the single-family market because of second homes. Second homes are reported in HMDA as “not owner occupied” to determine investor status but are not goals eligible. Therefore, according to Fannie Mae, HUD's use of HMDA would overestimate the goals-eligible share of the single-family market. As a result of these data and methodology issues, Fannie believes HUD miscalculates the mix of units in the rental market and overstates the size of the goals-eligible portion of the rental market.

Similarly, Freddie Mac concluded that HUD overestimated the SF investor share of the market because it relied on the RFS rather than HMDA. Freddie Mac says investor-owners have an incentive to claim falsely they are owner-occupants because of higher underwriting standards and higher interest rates on investor-owner properties. According to Freddie Mac, these incentives likely result in HMDA's undercounting SF investor loans relative to the more accurate counts of investor loans from the RFS. Freddie Mac concludes as follows:

This undercounting [on the part of HMDA], however, is exactly what is desired when estimating the goal share available to the GSEs. Because the GSEs' information on their loans has the same “bias” as does the HMDA data. * * * The HMDA data, therefore, are more appropriate to estimating the market for goal setting than are the RFS data. (p.II-6)

Essentially, Freddie Mac concludes that HUD's market estimates should measure opportunities in the marketplace that are actually available to the GSEs. Such opportunities are best measured by lender-reported HMDA data, not the more accurate RFS data. ICF reaches a similar conclusion, as it states that “HMDA data, or its equivalent, are what the GSEs' performance will be measured against and is therefore the appropriate metric for estimating market goal shares” (ICF Report, p.20).

4. SF Investor Shares in the Final Rule

In this final Rule, HUD has switched to a HMDA-based system and provides overall market share estimates for a range of single-family investor shares. For each year between 1993 and 2003, the top-right-hand-side portion of Table D.6a shows the projected investor share in a “home purchase environment” assuming a refinance share of 35 percent, 40 percent, and 45 percent. Refinance shares greater than 35 percent are included here because single-family investor loans typically have higher refinance shares than single-family-owner loans. As shown in Table D.6a, the average 1993-2003, HMDA-based investor share would have been 8.5 (8.4) percent if the investor refinance share had been 40 (45) percent during this period. During the more recent 1999-2003 period, which was characterized by particularly high HMDA-reported investor shares for home purchase loans, the average investor share would have been 9.4 (9.2) percent if the investor refinance share had been 40 (45) percent during this period. As noted earlier, the HMDA-reported investor shares for metropolitan areas are slightly lower than those for the entire U.S. As shown in the bottom-right-hand portion of Table D.6a, the average 1999-2003, HMDA-based investor share for metropolitan areas would have been 8.9 (8.7) percent if the investor refinance share had been 40 (45) percent during this period.

The above analysis suggests that the HMDA-reported investor share of a future home purchase market will probably be around 8.5-9.0 percent, or possibly higher if the recent figures for home purchase loans hold up (in this case, around 9.5 percent). Thus, HUD's analysis of market shares in Sections F-H will report overall market estimates for a range of SF investor shares—8.0 percent, 8.5 percent, 9.0 percent, and 10.5 percent.

5. Single-Family Market in Terms of Unit Shares

The market share estimates for the housing goals need to be expressed as percentages of units rather than as percentages of mortgages. Since a SF 2-4 and a SF-investor mortgage finances more than one dwelling unit, adjustments reflecting units-per-mortgage have to be made in order to arrive at the distribution of newly-financed single-family dwelling units. From HMDA, one can obtain the share of investor mortgages (those reported in Table D.6a) and the share of owner mortgages (obtained by subtracting the share of investor mortgages from 100 percent). HMDA does not disaggregate the SF-owner (SF-O) mortgage category into its two components: SF-O 1-Unit mortgages and SF-O 2-4 mortgages. To arrive at shares of SF financed dwelling units, two sets of adjustments have to be made to the HMDA data.

First, the owner-occupied HMDA data have to be disaggregated between SF-O 1-Unit and SF 2-4 mortgages. HUD's 2004 proposed GSE rule assumed that SF 2-4 mortgages accounted for 2.0 percent of all single-family mortgages. Based on the 2001 RFS data, this percentage is reduced to about 1.6 percent in this Final Rule. In 2001, the RFS shows the following distribution across the three single-family mortgage types: (a) 85.1 percent for SF-O 1-Unit mortgages; (b) 1.5 percent for SF-O 2-4 mortgages; and (c) 13.4 percent for SF-Investor mortgages (see Table D.6b). Thus, according to 2001 RFS data, SF 2-4 mortgages represent 1.73 percent of all single-family-owner mortgages (obtained by dividing (b) by the sum of (a) and (b)). In the market projection models, the SF-investor mortgage share is assumed to be lower than the RFS-reported figure of 13.4 percent. If the SF-investor share is 8.5 percent, then the SF-O share is 91.5 percent, which is split as follows: 1.58 percent for SF-O 2-4 mortgages (obtained by multiplying 0.0173 by 91.5 percent) and 89.92 percent for SF-O 1-Unit mortgages (obtained by subtracting 1.58 percent for the overall SF-O share of 91.5 percent). Thus, in this scenario, the distribution across SF mortgage types would be as follows: (d) 89.92 percent for SF-O 1-Unit mortgages; (b) 1.58 percent for SF-O 2-4 mortgages; and (c) 8.50 percent for SF-Investor mortgages. Table D.6d shows the distribution of SF mortgages under the various assumptions assumed in Sections F-H. For comparison purposes, the SF-O 2-4 shares for the GSEs are reported in Table D.6c. The 1999 to 2003 shares for Fannie Mae are approximately 2.0 percent while those for Freddie Mac are approximately 1.5 percent. Thus, the Fannie Mae shares are consistent with the 2.0 percent assumption used in the 2004 proposed rule while the Freddie Mac shares are consistent with the 1.6 percent assumption used in this Final Rule. Sensitivity analyses in Sections F-G will show the effects of using the 2.0 percent assumption (as compared with the 1.6 percent baseline).

Second, the resulting mortgage-based distributions have to be shifted to unit-based distributions by applying the unit-per-mortgage assumptions. The 2004 proposed GSE rule assumed the following: 2.25 units per SF 2-4 property and 1.35 units per SF investor property. Based on RFS data, these numbers are reduced slightly to the following: 2.2 units per SF 2-4 property and 1.3 units per SF investor property. These figures are based on 1999-2001 averages from the RFS. The corresponding 2001 figures from the RFS were 2.1 and 1.4, respectively. As shown in Table D.6d, the GSE data has consistently been around the figures in the 2004 proposed GSE rule, which were 2.25 and 1.35, respectively. Thus, it was decided to use the 1999-01 RFS averages which drop each units-per-mortgage figure by 0.05. Sensitivity analysis shows that this issue (whether to use the 1999-01 combination of 2.2/1.3 or to use the 2001 combination of 2.1/1.4) has little impact on the market sizing results.

Based on these calculations, the percentage distribution of newly-mortgaged single-family dwelling units was derived for each of the various estimates of the investor share of single-family mortgages. The results are presented in Table D.6e for investor percentage shares of 8.0, 8.5, 9.0, and 9.5. Three points should be made about these data.

First, notice that the rental categories represent a larger share of the unit-based market than they did of the mortgage-based market reported earlier. For example, when the SF-investor category represents 8.5 percent of all SF mortgages, it represents 10.6 percent of all SF units financed. This, of course, follows directly from applying the loan-per-unit expansion factors.

Second, notice that the “All SF-Rental Units” column highlights the share of the single-family mortgage market accounted for by all single-family rental units, for both SF-O 2-4 properties and SF-Investor properties. For example, when the investor mortgage share is 8.5 percent, single-family rental units (in SF 2-4 properties as well as in SF investor properties) account for 12.4 percent of all newly-mortgaged SF units. This single-family rental share compares with 15.1 percent under the baseline assumptions of the 2004 proposed GSE Rule; the 15.1 percent figure is reported in Table D.6b of the 2004 proposed GSE rule. If the single-family investor share is 9.0 (9.5) percent, then single-family rental units account for account for 13.0 (13.6) percent of all newly-mortgaged SF units.

ICF projected that SF rental units would account for 12.0 percent of all single-family-financed units during the 2005-2008 projection period (ICF Appendix, p.126). Under the units-per-mortgage and SF-O 2-4 share assumptions that ICF was using (2.25 for SF-O 2-4 and 1.35 for SF-Investor and a 2.0 percent share for SF-O 2-4 mortgages), ICF's 12-percent assumption for single-family rental units translates back to an investor mortgage share of 7.5 percent.^[24]

In its projections, Fannie Mae assumes 8.0 percent for the investor share of mortgages, a figure Fannie Mae says is consistent with HMDA data (Fannie Mae Appendix I, Table 1.11, p. I-38). Under the 2001 RFS assumptions (see above), this translates into a single-family rental share (on a units basis) of 11.8 percent. Under the units-per-loan and SF-O 2-4 assumptions of the proposed rule, this translates into a single-family rental share (expressed on a units basis) of 12.7 percent.

Third, if the investor mortgage share were 13 percent (the 2001 figure from the RFS), single-family rental units would account for over 17 percent of all newly-mortgaged single-family units.

The unit distributions reported for the GSEs in Table D.6f will be discussed in the next section.

E. HUD's Market Share Model

This section integrates findings from the previous two sections about the size of the multifamily mortgage market and the relative distribution of single-family owner and rental mortgages into a single model of the mortgage market. The section provides the basic equations for HUD's market share model and identifies the remaining parameters that must be estimated.

The output of this section is a unit-based distribution for the four property types discussed in Section B.^[25] Sections F-H will apply goal percentages to this property distribution in order to determine the size of the mortgage market for each of the three housing goals.

1. Basic Equations for Determining Units Financed in the Mortgage Market

The model first estimates the number of dwelling units financed by conventional conforming mortgage originations for each of the four property types. It then determines each property type's share of the total number of dwelling units financed.

a. Single-Family Units

This section estimates the number of single-family units that will be financed in the conventional conforming market, where single-family units (SF-UNITS) are defined as:

SF-UNITS = SF-O + SF 2-4 + SF-INVESTOR

First, the dollar volume of conventional conforming single-family mortgages (CCSFM$) is derived as follows:

(1) CCSFM$ = CONV% * CONF% * SFORIG$

where

CONV% = conventional mortgage originations as a percent of total mortgage originations; estimated to be 88%.^[26]

CONF% = conforming mortgage originations (measured in dollars) as a percent of conventional single-family originations; forecasted to be 80% by industry.

SFORIG$ = dollar volume of single-family one-to-four unit mortgages; $1,700 billion is used here as a starting assumption to reflect market conditions during the years 2005-2008.^[27] While alternative assumptions will be examined, it must be emphasized that the important concept for deriving the goal-qualifying market shares is the relative importance of single-family versus multifamily mortgage originations (the “multifamily mix” discussed in Section C) rather than the total dollar volume of single-family originations considered in isolation.

Substituting these values into (1) yields an estimate for the conventional conforming market (CCSFM$) of $1,197 billion.

Second, the number of conventional conforming single-family mortgages (CCSFM#) is derived as follows:

(2) CCSFM# = (CCSFM$ * (1-REFI)/PSFLOAN$) + (CCSFM$ * REFI)/RSFLOAN$)

where

REFI= the refinance rate, assumed to be 35 percent for the baseline.^[28]

PSFLOAN$ = the average conventional conforming purchase mortgage amount for single-family properties; estimated to be $146,000.^[29]

RSFLOAN$ = the average conventional conforming refinance mortgage amount for single-family properties; estimated to be $131,000.^[30]

Substituting these values into (2) yields an estimate of 8.5 million mortgages.

Third, the total number of single-family mortgages is divided among the three single-family property types. Using the 89.9/1.6/8.5 percentage distribution for single-family mortgages (see Section D), the following results are obtained:

(3a) SF-OM# = 0.899 * CCSFM# = number of owner-occupied, one-unit mortgages = 7.642 million.

(3b) SF-2-4M# = 0.016 * CCSFM# = number of owner-occupied, two-to-four unit mortgages = 0.136 million.

(3c) SF-INVM# = 0.085 * CCSFM# = number of one-to-four unit investor mortgages = 0.723 million.

Fourth, the number of dwelling units financed for the three single-family property types is derived as follows:

(4a) SF-O = SF-OM# + SF-2-4M# = number of owner-occupied dwelling units financed = 7.778 million.

(4b) SF 2-4 = 1.2 * SF-2-4M# = number of rental units in 2-4 properties where an owner occupies one of the units = 0.163 million.^[31]

(4c) SF-INVESTOR= 1.3 * SF-INVM# = number of single-family investor dwelling units financed = 0.940 million.

Fifth, summing equations 4a-4c gives the projected number of newly-mortgaged single-family units (SF-UNITS):

(5) SF-UNITS = SF-O + SF 2-4 + SF-INVESTOR = 8.915 million

b. Multifamily Units

The number of multifamily dwelling units (MF-UNITS) financed by conventional conforming multifamily originations is calculated by the following series of equations:

(5a) TOTAL = SF-UNITS + MF-UNITS

(5b) MF-UNITS = MF-MIX * TOTAL = MF-MIX * (SF-UNITS + MF-UNITS) = [MF-MIX/(1 − MF-MIX)] * SF-UNITS where

MF-MIX = the “multifamily mix”, or the percentage of all newly-mortgaged dwelling units that are multifamily; as discussed in Section C, alternative estimates of the multifamily market will be included in the analysis. As explained in Section C above, the baseline model assumes a multifamily mix of 15 percent; results are also presented in the basic market tables of Sections F-H for a higher (16.0 percent) multifamily mix and for lower (12.25 percent, 13.5 percent and 14.25 percent) multifamily mixes. In addition, further sensitivity analyses are reported in those sections for even lower multifamily mixes that could occur during periods of heavy single-family refinancing activity.

Assuming a multifamily mix of 15 percent and solving (5b) yields the following:

(5c) MF-UNITS = [0.15/0.85] * SF-UNITS = 0.176 * SF-UNITS = 1.6 million.

c. Total Units Financed

The total number of dwelling units financed by the conventional conforming mortgage market (TOTAL) can be expressed in three useful ways:

(6a) TOTAL = SF-UNITS + MF-UNITS = 10.6 million (or more precisely, 10,632,145 units)

(6b) TOTAL = SF-O + SF 2-4 + SF-INVESTOR + MF- UNITS

(6c) TOTAL = SF-O + SF-RENTAL + MF-UNITS where SF-RENTAL equals SF-2-4 plus SF-INVESTOR

2. Dwelling Unit Distributions by Property Type

The next step is to express the number of dwelling units financed for each property type as a percentage of the total number of units financed by conventional conforming mortgage originations.^[32]

The projections used above in equations (1)-(6) produce the following distributions of financed units by property type:

	% Share
SF-O	74.5
SF 2-4	1.5
SF INVESTOR	9.0
MF-UNITS	15.0
Total	100.0
or
SF-O	74.5
SF-RENTER	10.5
MF-UNITS	15.0
Total	100.0

Sections C and D discussed alternative projections for the mix of multifamily originations and the investor share of single-family mortgages. This appendix will report results for multifamily mixes of 13.5 percent, 15.0 percent, and 16.0 percent but sensitivity analyses for two other multifamily mix assumptions (e.g., the 12.3 percent assumption used by Fannie Mae and the 14.2 assumption used by ICF) will also be reported. Under the baseline 15.0 percent multifamily mix, the newly-mortgaged unit distribution would be 74.5 percent for Single-Family Owner, 10.5 percent for Single-Family Renter, and 15.5 percent for Multifamily-Units. The analysis in sections F-H will focus on goals-qualifying market shares for this property distribution as well as the ones noted above.

As discussed in Section D, the basic tables providing the goals-qualifying market estimates in this appendix will report results for the following investor shares of single-family mortgages—8 percent, 8.5 percent, 9.0 percent, and 9.5 percent. For reasons discussed in Section D, these investor mortgage shares are lower than the range (8.0 percent, 10.0 percent, and 12.0 percent) considered in the 2004 proposed GSE rule. The middle values (8.5 percent and 9.0 percent) are probably the ones that should be considered as “baseline” projections; the above example used a mortgage share of 8.5 percent, but 9.0 percent could also have been used to characterize a home purchase environment. However, HUD recognizes the uncertainty of projecting origination volume in markets such as single-family investor properties; therefore, the analysis in Sections F-H considers market assumptions other than these baseline assumptions.

Table D.7 reports the unit-based distributions produced by HUD's market share model for different combinations of these projections. Unit-based distributions are reported for each combination of a multifamily mix (12.25, 13.5, 14.25, 15.0, and 16.0) and investor mortgage share (8.0, 8.5, 9.0, and 9.5). The effects of the different projections can best be seen by examining the owner category which varies by 4.8 percentage points, from a low of 72.6 percent (multifamily mix of 16.0 percent coupled with an investor mortgage share of 9.5 percent) to a high of 77.4 percent (multifamily mix of 12.25 percent coupled with an investor mortgage share of 8.0 percent). The overall rental share is also highlighted in Table D.7, varying from 22.6 percent to 27.4 percent.

The baseline projection of newly-mortgaged units in the 2004 proposed GSE rule was 72.2 percent for owner units, 12.8 percent for single-family rental units, and 15.0 percent for multifamily units. In this Final Rule, the baseline projection is 74.5 percent for owner units, 10.5 percent for single-family rental units, and 15.0 percent for multifamily units, if an investor mortgage share of 8.5 percent is used. If an investor share of 9.0 percent is used, then the baseline projection is 74.0 percent for owner units, 11.0 percent for single-family rental units, and 15.0 percent for multifamily units. Either way, compared with the 2004 proposed GSE rule, the rental share of financed dwelling units has dropped by approximately two percentage points due to the lower HMDA-based investor shares used in the Final Rule.

The unit distribution in ICF's projection model for 2005-2008 averaged 75.5 percent for owner units, 10.3 percent for single-family rental units, and 14.2 percent for multifamily units, which produces an overall rental share of 24.5 percent, a figure closed to those reported above (25.5-26.0 percent). The unit distribution used by Fannie Mae was approximately 77.4 percent for owner units, 10.4 percent for single-family rental units, and 12.3 percent for multifamily units, which produces an overall rental share of 22.6 percent,^[33] a figure less than used by ICF (24.5 percent) or HUD (25.0-26.0 percent). Notice that Fannie Mae and ICF assume similar single-family rental shares (about 10.3 percent), but ICF assumes a larger multifamily mix than Fannie Mae (14.2 percent versus 12.3 percent). HUD's single-family rental shares (10.5-11.0 percent) are slightly higher than the shares (about 10.3 percent) used by ICF and Fannie Mae. HUD's multifamily baseline share (15.0 percent) is slightly higher than the average (14.2 percent) of ICF's best estimate, and significantly higher than Fannie Mae's assumed multifamily mix (12.3).

As discussed in Sections C and D, the Residential Finance Survey is the only mortgage data source that provides unit-based property distributions directly comparable to those reported below. Based on RFS data for 2001, HUD estimated that, of total dwelling units in properties financed by recently acquired conventional conforming mortgages, 68.3 percent were owner-occupied units, 16.5 percent were single-family rental units, and 15.2 percent were multifamily rental units. Thus, the RFS presents a much lower owner share than does HUD's, ICF's, or Fannie Mae's models. See Sections C and D for further discussion of the RFS.

Finally, it is interesting to compare the above market-based distributions of financed units with the distributions of units financed by mortgages purchased by Fannie Mae and Freddie Mac. As shown in Table D.6f, the 1993-2003 averages (unweighted) for Fannie Mae were 81.0 percent for owner units, 9.0 percent for single-family rental units, and 10.0 percent for multifamily units, which produces an overall rental share of 19.0 percent. During the year 2000, Fannie Mae's overall rental share did reach a peak of 24.1 percent. Freddie Mac's rental shares have been markedly lower than Fannie Mae's. The 1993-2003 averages (unweighted) for Freddie Mac were 84.3 percent for owner units, 6.3 percent for single-family rental units, and 9.3 percent for multifamily units, which produces an overall rental share of 15.7 percent.^[34] Freddie Mac's rental share did peak at 17.5 percent in 2000. Still, it is clear that the market-based distributions project much higher rental shares than Freddie Mac has been purchasing. For example, the HUD projection of a 25-percent rental share is over nine percentage points higher than Freddie Mac's 1999-2003 average rental share (15.7 percent) and over seven percentage points higher than Freddie Mac's peak rental share (17.5 percent in 2000). The 31.7-percent rental share from the RFS is 16 percentage points higher than Freddie Mac's 1999-2003 average rental share (15.7 percent) and over 14 percentage points higher than Freddie Mac's peak rental share (17.5 percent in 2000).

F. Size of the Conventional Conforming Mortgage Market Serving Low- and Moderate-Income Families

This section estimates the size of the low- and moderate-income market by applying low- and moderate-income percentages to the property shares given in Table D.7. This section essentially accomplishes Steps 2 and 3 of the three-step procedure discussed in Section B.2.

Technical issues and data adjustments related to the low- and moderate-income percentages for owners and renters are discussed in the first two subsections. Then, estimates of the size of the low- and moderate-income market are presented along with several sensitivity analyses. Based on these analyses, HUD concludes that 51-56 percent is a reasonable estimate of the mortgage market's low- and moderate-income share for the four years (2005-2008) when the new goals will be in effect.

1. Low- and Moderate-Income Percentage for Single-Family-Owner Mortgages

a. HMDA Data

The most important determinant of the low- and moderate-income share of the mortgage market is the income distribution of single-family borrowers. HMDA reports annual income data for families who live in metropolitan areas and purchase a home or refinance their existing mortgage.^[35] The data cover conventional mortgages below the conforming loan limit, which was $322,700 in 2003. Table D.8a gives the percentage of mortgages originated for low- and moderate-income families for the years 1992-2003. Data are presented for home purchase, refinance, and all single-family-owner loans. The discussion below will often focus on home purchase loans because they typically account for the majority of all single-family-owner mortgages.^[36] For each year, a low- and moderate-income percentage is also reported for the conforming market without B&C loans.

Table D.8a also reports similar data for very-low-income families (that is, families with incomes less than 60 percent of area median income). As discussed in Section H, very-low-income families are the main component of the special affordable mortgage market.

Two trends in the income data should be mentioned—one related to the growth in the market's funding of low- and moderate-income families during the 1990s (and particularly the growth since 1998 which was the last year analyzed in HUD's 2000 GSE Rule); and the other related to changes in the borrower income distributions for refinance and home purchase mortgages. Throughout this appendix, “low- and moderate-income” will often be referred to as “low-mod'.

Recent Trends in the Market Share for Lower Income Borrowers. First, focus on the percentages in Table D.8a for the total (both home purchase and refinance) conforming market. After averaging about 30 percent during 1992-93, the percentage of borrowers with less than area median income jumped to 41.0 percent in 1994, and remained above 40 percent through 2003. Over the ten-year period, 1994 to 2003, the low-mod share of the total market averaged 42.9 percent (or 42.2 percent if B&C loans are excluded from the market totals).^[37] The share of the market accounted for by very-low-income borrowers followed a similar trend, increasing from 6-7 percent in 1992-93 to about 12 percent in 1994 and averaging 13.2 percent during the 1994-to-2003 period (or 12.7 percent if B&C loans are excluded).

Next, consider the percentages for home purchase loans. The share of the home loan market accounted for by less-than-median-income borrowers increased from 34.4 percent in 1992 to 44.7 percent in 2003. Within the 1994-to-2002 period, the low-mod share of the home purchase market averaged 44.4 percent between 1999 and 2003, compared with 42.1 percent between 1994 and 1998. Similarly, the very-low-income share of the home purchase market was also higher during the 1999-to-2002 period than during the 1994-to-1998 period (14.1 percent versus 12.6 percent). Note that within the more recent period, the low-mod share for home purchase loans was particularly high during 1999 (45.2 percent) and 2000 (44.3 percent) before falling slightly in 2001 (43.2 percent), only to rebound again in 2002 (44.8 percent) and 2003 (44.7 percent). As shown in Table D.8a, the low-mod shares do not change much when B&C home loans are excluded from the market definition; this is because B&C loans are mainly refinance loans.

It appears that the affordable lending market for home purchase loans is even stronger today than when HUD wrote the 2000 Rule, which covered market data through 1998. The very-low-income and low-mod percentages were higher during 1999 to 2003 than they were during the earlier period. In addition, when HUD wrote the 2000 Rule, there had been five years (1994-98) of solid affordable lending for lower-income borrowers. Now, with five additional years of data for 1999-2003, there have been ten years of strong affordable lending.

Of course, it is recognized that lending patterns could change with sharp changes in interest rates and the economy. However, the fact that lending to low-income families has remained at a high level for ten years demonstrates that the market has changed in fundamental ways from the mortgage market of the early 1990s. The numerous innovative products and outreach programs that the industry has developed to attract lower-income families into the homeownership and mortgage markets appear to be working and there is no reason to believe that they will not continue to assist in closing troubling homeownership gaps that exist today. As explained in Appendix A, the demand for homeownership on the part of minorities, immigrants and non-traditional borrowers should help to maintain activity in the affordable portion of the mortgage market. Thus, while economic recession or higher interest rates would likely reduce the low- and moderate-income share of mortgage originations, there is evidence that the low-mod market might not return to the low levels of the early 1990s. There is also evidence that the affordable lending market increased slightly since 1998, although it is recognized that this could be due to the recent period of historically low interest rates.

Refinance Mortgages. In the 2000 Rule, HUD's market projection model assumed that low-mod borrowers represented a smaller share of refinance mortgages than they do of home purchase mortgages. However, as shown in Table D.8a, the income characteristics of borrowers refinancing mortgages seem to depend on the overall level of refinancing in the market. During the refinancing wave of 1992 and 1993, refinancing borrowers had much higher incomes than borrowers purchasing homes. For example, during 1993 low- and moderate-income borrowers accounted for 29.3 percent of refinance mortgages, compared to 38.9 percent of home purchase borrowers. While this same pattern was exhibited during the two recent refinancing periods (1998 and 2001-2002-2003), the differentials were much smaller—during 2001-2002-2003 (1998), low-mod borrowers accounted for 41.5 (39.7) percent of refinance loans, compared with 44.2 (43.0) percent of home purchase loans. However, the refinance effect was still evident, as can be seen by the almost ten percentage point drop in the low-mod percentage for refinance loans between 2000 (a low refinance year) and 2001 (a high refinance year).

On the other hand, for recent years characterized by a low level of refinancing, the low-mod share of refinance mortgages has been about the same or even greater than that of home purchase mortgages. As shown in Table D.8a, there was little difference in the very-low-income and low-mod shares of refinance and home purchase loans during 1995 and 1996. In 1997, 1999, and 2000, the two lower-income shares (i.e., very-low-income and low-mod shares) of refinance mortgages were significantly higher than the lower-income shares of home purchase loans. To a certain extent, this pattern was influenced by the growth of subprime loans, which are mainly refinance loans. If B&C loans are excluded from the market definition, the home purchase and refinance percentages are approximately the same in 1997 and 1999, as well as in 1995 and 1996. (See Table D.8a.) Even after excluding all subprime loans from the market definition in 1997 and 1999, the very-low-income and low-mod shares for refinance loans are only slightly less (about one percentage point) than those for home purchase loans.

The year 2000 stands out because of the extremely high lower-income shares for refinance loans. In that year, the low-mod (very-low-income) share of refinance loans was 7.0 (4.4) percentage points higher than the low-mod (very-low-income) share of home purchase loans; this differential is reduced to 5.4 (3.3) percent if B&C loans are excluded from the market definition (see Table D.8a). The differential for 2000 is reduced further to 2.8 (1.5) percent if all subprime loans (both A-minus and B&C) are excluded from the market definition (not reported). While the projection model (explained below) for years 2005-08 will input low-mod percentages for the entire conforming market, the model will exclude the effects of B&C loans. Sensitivity analyses will also be conducted showing the effects on the overall market estimates of excluding all subprime loans as well as other loan categories such as manufactured housing loans.

2000 Census Data and New OMB Metropolitan Area Definitions. Going forward, HUD will be re-benchmarking its median incomes for metropolitan areas and non-metropolitan counties based on 2000 Census median incomes, and will be incorporating the effects of the new OMB metropolitan area definitions. Thus, under the new housing goals, the GSEs' performance will be scored based on 2000 Census data and new OMB definitions of metropolitan areas (labeled “CBSA definitions”). One issue concerns whether the new data and the new definitions will result in lower or higher low-mod percentages relative to historical low-mod percentages based on the 1990 Census and earlier OMB definitions of metropolitan areas (labeled “MSA definitions”). HUD projected the effects of these two changes on the low- and moderate-income shares of the single-family-owner market for the years 1999-2003. The middle portion of Table D.8b reports low-mod shares for single-family-owner loans under the MSA and CBSA approaches for the years 1999-2003. Except for 2003, the low-mod shares for both home purchase and total SFO loans are lower under the new CBSA approach than under the old MSA approach. Because the results for 1999-2002 differed from the results for 2003, these two periods are considered separately. Under the historical data, the average low-mod share of the conventional conforming market was 44.4 percent for home purchase loans (unweighted average of 1999-2002 percentages in Table D.8a); the corresponding average with the projected data was 43.2 percent, yielding a differential of 1.2 percentage points. For total (both home purchase and refinance) loans, the average low-mod share of the conventional conforming market based on historical data was 44.6 percent (unweighted average of 1999-2002 percentages); the corresponding average with the projected data was 43.4 percent, again yielding a differential of 1.2 percentage points, with the same pattern exhibited for the annual differentials.^[38] It appears that the low-mod share for single-family-owners in the conventional conforming market will be at least one percentage point less due to the re-benchmarking of area median incomes and the new OMB definitions of metropolitan areas.

Based on the above analysis of 1999-2002 data, it would appear the low-mod share of the conventional conforming market is about one percentage point less when based on projected data, as compared with historical data. However, the data for 2003 suggest a different picture. As shown in Table D.8b, the 2003 CBSA-based low-mod share for home purchase loans is 45.8 percent, which is 1.1 percentage points higher than the corresponding MSA-based percentage of 44.7 percent. Similarly, the CBSA-based percentage is 1.1 percentage point higher when all owner loans are considered. Thus, the more recent 2003 data suggest that the GSEs will be scored higher than they have historically been scored.

Table A.18 in Appendix A reported similar MSA and CBSA data for home purchase loans acquired by Fannie Mae and Freddie Mac. Again, the low-mod shares for the GSEs' purchases of both home purchase and total SFO loans were lower under the new CBSA approach than under the old MSA approach for 1999-2002, but not for 2003. The proposed GSE rule accounted for the 1999-2002 discrepancy by reducing the overall low-mod estimates by one percentage point. Given the 2003 results, which show higher low-mod shares under the new CBSA approach, that procedure is questionable. This Final Rule follows a different procedure. The actual CBSA-based low-mod shares for owners (reported in Table D.8b) are incorporated directly into the analysis.

The projection model will initially assume that refinancing is 35 percent of the single-family mortgage market; this will be followed by projection models that reflect heavy refinance environments. Given the volatility of refinance rates from year to year, it is important to conduct sensitivity tests using different refinance rates. However, as explained in the preamble, HUD has included a provision in this Final Rule that eliminates the negative effects of heavy refinancing periods on the GSEs' goals performance.

b. Manufactured Housing Loans

Because manufactured housing loans are such an important source of affordable housing, they are included in the mortgage market definition in this appendix—or at least that portion of the manufactured housing market located in metropolitan areas is included, as HMDA doesn't adequately cover non-metropolitan areas. The GSEs have questioned HUD's including these loans in its market estimates; therefore, following the same procedure used in the 2000 Rule and the 2004 proposed GSE Rule, this Appendix will report the effects of excluding manufactured home loans from the market estimates. As explained later, the effect of manufactured housing on HUD's metropolitan area market estimate for each of the three housing goals is approximately one percentage point or less.

As discussed in Appendix A, the manufactured housing market increased rapidly during the 1990s, as units placed in serviced increased from 174,000 in 1991 to 374,000 in 1998. However, due to various problems in the industry such as lax underwriting and repossessions, volume has declined in recent years, falling to 192,000 in 2001, to 172,000 in 2002, and to 135,000 in 2003. Still, the affordability of manufactured homes for lower-income families is demonstrated by their average price of $48,800 in 2001, a fraction of the median price for new ($175,000) and existing ($147,800) homes. Many households live in manufactured housing because they simply cannot afford site-built homes, for which the construction costs per square foot are much higher.

Although manufactured home loans cannot be identified in the HMDA data, Randy Scheessele at HUD identified 21 lenders that primarily originated manufactured home loans in 2001 and likely account for most of these loans in the HMDA data for metropolitan areas.^[39] HMDA data on home loans originated by these lenders indicate that: ^[40]

• A very high percentage of these loans—75 percent in 2001—would qualify for the Low- and Moderate-Income Goal,
• A substantial percentage of these loans—42 percent in 2001—would qualify for the Special Affordable Goal, and
• Almost half of these loans—47 percent in 2001—would qualify for the Underserved Areas Goal (defined in terms of the 1990 Census data).^[41]

Thus an enhanced presence in this market by the GSEs would benefit many lower-income families. It would also contribute to their presence in underserved rural areas, especially in the South.

2. Low- and Moderate-Income Percentage for Renter Mortgages

Following the 2000 Rule, measures of the rent affordability of the single-family rental and the multifamily rental markets are obtained from the American Housing Survey (AHS) and the Property Owners and Managers Survey (POMS). As explained below, the AHS provides rent information for the stock of rental properties while the POMS provides rent information for flow of mortgages financing that stock. As discussed below, the AHS and POMS data provide very similar estimates of the low- and moderate-income share of the rental market.

a. American Housing Survey Data

The American Housing Survey does not include data on mortgages for rental properties; rather, it includes data on the characteristics of the existing rental housing stock and recently completed rental properties. Current data on the income of prospective or actual tenants has also not been readily available for rental properties. Where such income information is not available, FHEFSSA provides that the rent of a unit can be used to determine the affordability of that unit and whether it qualifies for the Low- and Moderate-Income Goal. A unit qualifies for the Low- and Moderate-Income Goal if the rent does not exceed 30 percent of the local area median income (with appropriate adjustments for family size as measured by the number of bedrooms). Thus, the GSEs' performance under the housing goals is measured in terms of the affordability of the rental dwelling units that are financed by mortgages that the GSEs purchase; the income of the occupants of these rental units is not considered in the calculation of goal performance. For this reason, it is appropriate to base estimates of market size on rent affordability data rather than on renter income data.

A rental unit is considered to be “affordable” to low- and moderate-income families, and thus qualifies for the Low- and Moderate-Income Goal, if that unit's rent is equal to or less than 30 percent of area median income. Table D.14 of Appendix D in HUD's 2000 Rule reported AHS data on the affordability of the rental housing stock for the survey years between 1985 and 1997. The 1997 AHS showed that for 1-4 unit unsubsidized single-family rental properties, 94 percent of all units and of units constructed in the preceding three years had gross rent (contract rent plus the cost of all utilities) less than or equal to 30 percent of area median income. For multifamily unsubsidized rental properties, the corresponding figure was 92 percent. The AHS data for the other survey years were similar to the 1997 data.

b. Property Owners and Managers Survey (POMS)

As discussed in the 2000 GSE Rule, there were concerns about using AHS data on rents from the outstanding rental stock to proxy rents for newly mortgaged rental units. HUD investigated that issue further using the POMS.

POMS Methodology. The affordability of multifamily and single-family rental housing backing mortgages originated in 1993-1995 was calculated using internal Census Bureau files from the American Housing Survey-National Sample (AHS) from 1995 and the Property Owners and Managers Survey from 1995-1996. The POMS survey was conducted on the same units included in the AHS survey, and provides supplemental information such as the origination year of the mortgage loan, if any, recorded against the property included in the AHS survey. Monthly housing cost data (including rent and utilities), number of bedrooms, and metropolitan area (MSA) location data were obtained from the AHS file.

In cases where units in the AHS were not occupied, the AHS typically provides rents, either by obtaining this information from property owners or through the use of imputation techniques. Estimated monthly housing costs on vacant units were therefore calculated as the sum of AHS rent and utility costs estimated using utility allowances published by HUD as part of its regulation of the GSEs. Observations where neither monthly housing cost nor monthly rent was available were omitted, as were observations where MSA could not be determined. Units with no cash rent and subsidized housing units were also omitted. Because of the shortage of observations with 1995 originations, POMS data on year of mortgage origination were utilized to restrict the sample to properties mortgaged during 1993-1995. POMS weights were then applied to estimate population statistics. Affordability calculations were made using 1993-95 area median incomes calculated by HUD.

POMS Results. The rent affordability estimates from POMS of the affordability of newly-mortgaged rental properties are quite consistent with the AHS data on the affordability of the rental stock (discussed above). Ninety-six (96) percent of single-family rental properties with new mortgages between 1993 and 1995 were affordable to low- and moderate-income families, as were 96 percent of newly-mortgaged multifamily properties. Thus, these percentages for newly-mortgaged properties from the POMS are similar to those from the AHS for the rental stock.

Further Results and Comments. The baseline projection from HUD's market share model assumes that 90 percent of newly-mortgaged, single-family rental and multifamily units are affordable to low- and moderate-income families.^[42] As noted above, the analysis of AHS and POMS data from the mid-1990s supports the use of a 90 percent low-mod figure, and also supports using rental stock data from the AHS as a proxy for the affordability characteristics of new mortgages financing rental properties. Updating these results using the 2001 and 2003 AHS produced similar (over 90 percent) low-mod estimates for both the single-family rental stock and the multifamily rental stock. For example, using ICF's assumptions for an AHS analysis (see ICF Appendix, p. 45), the 2003 AHS showed that 94 (93) percent of single-family (multifamily) rental units would qualify as being affordable to low- and moderate-income families. While ICF used 90 percent for multifamily, ICF concluded that 87.5 percent should be used for single-family rentals. HUD's updated analysis of the AHS, which is explained in more detail in Section H below, does not support using ICF's 87.5 percent assumption, except for sensitivity analysis. Since single-family rental units account for approximately 10 percent all financed units in both ICF's and HUD's market share models, the effect on the overall low-mod goal of using 87.5 percent instead of 90.0 percent would be only 0.25 percentage point. (the 2.5 percentage point low-mod differential multiplied by the 0.10 property share for single-family rental properties).

Based on its analysis of the AHS (see Fannie Mae Appendix, I-31-I-32), Fannie Mae concluded that the low-mod shares for both single-family and multifamily properties had fallen from 90 percent in 1997 to 86 percent in 2001. In its analysis, Fannie Mae provides a weight of 0.07 to the low-mod share (74.8 percent) of recently-constructed single-family rental units in the AHS, and the residual 0.93 weight to the low-mod share (91.8 percent) of the remaining existing units in the AHS. While Fannie Mae appears to use a low-mod share of 86 percent for single-family rentals in its market sizing models, applying these weights to the 2001 AHS data (reported by Fannie Mae in Table I.7 on p. I-32) yields approximately 90 percent for the low-mod share of single-family rental properties. Similarly, for multifamily properties, Fannie Mae provides a weight of 0.11 to the low-mod share (75.3 percent) of recently-constructed multifamily rental units in the AHS, and the residual 0.89 weight to the low-mod share (91.3 percent) of the remaining existing units in the AHS. Again, while Fannie Mae appears to use a low-mod share of 86 percent for multifamily rentals in its market sizing models, applying the above weights to the 2001 AHS data also yields approximately 90 percent for the low-mod share of multifamily rental properties. Since single-family and multifamily rental units combined account for about 25 percent of all financed units in the market sizing models, the effect on the overall low-mod share of using 86 percent instead of 90 percent would be about one percentage point. (the 4.0 percentage point low-mod differential multiplied by the 0.25 property share for single-family and multifamily rental properties).^[43] Fannie Mae expressed particular concern with HUD's Case 3, which assumed an even higher 95.0 percent low-mod share for rental properties; HUD has reduced this assumption to 92.5 percent in the Case 3 analysis below. HUD's Case 2 will also consider a low-mod percentage of 87.5 percent.

The low-mod characteristics of the GSEs' own purchases can also be examined. Between 1999 and 2003, 86.4 percent of Fannie Mae's single-family rental purchases qualified as low-mod, as did 87.3 percent of Freddie Mac's purchases. During the same period, 90.7 percent of Fannie Mae's multifamily rental purchases qualified as low-mod, as did 92.6 percent of Freddie Mac's purchases. One issue discussed below concerns the impact on the GSEs' low-mod performance of switching to 2000 Census data and the new OMB metropolitan area definitions. The above GSE percentages were recalculated after applying the new data and new OMB definitions back to 1999. Similar low-mod results were obtained for both single-family and multifamily rentals. Thus, the 2000 Census data and the new OMB metropolitan area definitions will have no impact on the low-mod scoring of the GSEs' rental purchases.

Most of ICF's and the GSEs' concerns about HUD's estimates of the affordability of rental housing properties related to the sizing of the special affordable market. Therefore, more detail treatment of these issues will be provided in Section H below.

3. Size of the Low- and Moderate-Income Mortgage Market

This section provides estimates of the size of the low- and moderate-income mortgage market. Subsection 3.a presents new estimates of the low-mod market while Subsection 3.b reports the sensitivity of the new estimates to changes in assumptions about economic and mortgage market conditions.

a. Estimates of the Low- and Moderate-Income Market

This section provides HUD's estimates for the size of the low- and moderate-income mortgage market that will serve as a proxy for the four-year period (2005-2008) when the new housing goals will be in effect. The estimates are compared with recent experience in the low-mod market since 1999. As discussed in Sections C and D, market estimates will be presented for different combinations of the investor mortgage share (8.8, 8.5, 9.0, and 9.5) and the multifamily mix (12.25, 13.5, 14.25, 15.0, and 16.0). This range reflects uncertainty about the data and future conditions in these rental markets. As discussed in Section C, HUD continues to use a multifamily (MF) mix of 15.0 percent as its baseline for a home purchase environment; this is strongly supported by RFS analysis. While results are reported for Fannie Mae's MF mix of 12.3 percent, HUD does not believe the MF mix will fall to that level in a home purchase environment; rather, the results are reported to gauge the effects on the market size of alternative assumptions supported by Fannie Mae. Three alternative sets of projections about rental property low- and moderate-income percentages are given in Table D.9. Case 1 projections represent the baseline and intermediate case; for example, it assumes that the low-mod share of rental loans is 90 percent. Case 1 will be the focus of the market analysis in this section. Case 2 assumes slightly lower goals-qualifying shares (e.g., an 85 percent low-mod share) for rental properties while Case 3 assumes slightly higher goals-qualifying shares (e.g., a 92.5 percent low-mod share).

Because single-family-owner units account for about 75 percent of all newly mortgaged dwelling units, the low- and moderate-income percentage for owners is the most important determinant of the total market estimate. Thus, Table D.10 provides market estimates for different low-mod percentages for the owner market as well as for different MF mix percentages and investor mortgage shares. In a home purchase environment, the most likely MF mix is 15.0 percent and the most likely investor mortgage share is in the 8.5-9.0 percent range. For simplicity, the combination of a 15.0-percent MF mix and a 8.5-percent investor share will be labeled the baseline when presenting the results below. Including a 9.0-percent investor mortgage share as the baseline would increase the low-mod market estimate by about 0.2-0.3 percentage point. The low-mod market estimates in Table D.10 exclude B&C loans, as explained below.

Table D.10 assumes a refinance rate of 35 percent, which means that the table reflects home purchase or low-refinancing environments. After presenting these results, market estimates reflecting heavy refinance environments will be presented. Because of the increase in single-family mortgages, the multifamily share of the mortgage market typically falls during a heavy refinance environment; therefore, several sensitivity analyses using lower multifamily mixes are examined below.

In the 2000 Rule, HUD assumed that the low-mod share of refinance loans was three percentage points lower than the low-mod share of borrowers purchasing a home. However, as discussed earlier, the low-mod share of refinance loans has equaled or been greater than the low-mod share of home purchase loans during recent home purchase environments such as 1995-97 or 1999-2000; thus, the assumption of a lower low-mod shares for refinance loans is initially dropped for this analysis but will be reintroduced during the sensitivity analysis and during the discussion of heavy refinance environments.

There are two ways to view the single-family-owner low-mod percentages reported in the first column of Table D.10. A first approach would be to view them as representing low-mod percentages of only the home purchase market. For example, a low-mod percentage for home purchase loans of 43 percent—combined with the assumption of an equal low-mod share for refinance loans (i.e., also 43 percent) and with the other model assumptions (such as a multifamily mix of 15 percent and an investor share of 8.5 percent)—produces an estimate of 54.6 percent for the low-mod share of the overall (owner and rental) market, excluding B&C loans. Thus, the reader can view Table D.10 as showing the overall low-mod market estimate once the reader specifies his or her views about the low-mod share of the single-family home purchase market (given the other model assumptions). In this case, if the reader believes that the low-mod share of refinance loans should be lower than that for home purchase loans, the reader simply has to multiply the differential amount by 0.35 (which is the refinance share of single-family-owner loans) and 0.745 (which is the single-family-owner share of all dwelling units in the model that assumes a 15 percent multifamily mix and 8.5 percent investor mortgage share). For example, applying the assumption in the 2000 Rule that the low-mod share is three percentage points lower for refinance loans would reduce the overall low-mod share of the market by 0.78 percentage points (3.0 times 0.35 times 0.745); if the low-mod share of refinance loans is one percentage point below that of home purchase loans, then the overall low-mod market estimate falls by 0.26 percentage point. In this manner, the reader can easily adjust the market estimates reported in Table D.10 to incorporate his or her own views about differences in the low-mod share of home purchase and refinance loans.

A second approach would be to view the low-mod percentages (in the first column of Table D.10) as representing low-mod shares for the overall single-family-owner market, including both home purchase and refinance loans. This approach does not specify separate low-mod percentages for home purchase and refinance loans, but rather focuses on the overall single-family-owner environments. Thus, it allows for mortgage market environments where the low-mod share of refinance loans is greater than the low-mod share for home purchase loans. For example, a low-mod percentage for single-family-owner loans of 47 percent would reflect the year 2000 environment, which had a low-mod home purchase percentage of 44.3 percent combined with a higher low-mod refinance percentage of 51.3 percent. Of course, the 47 percent low-mod share for the overall single-family-owner market could be consistent with other combinations of low-mod shares for home purchase and refinance loans. In this case, a 47 percent assumption for the overall single-family-owner market produces an estimate of 57.8 percent for the low-mod share of the overall (owner and rental) market, excluding B&C loans.

While both approaches will be discussed below, most of the discussion will focus on the first approach. It should be noted that several low-mod percentages of the owner market are given in Table D.10 to account for different perceptions of that market. Essentially, HUD's approach throughout this appendix is to provide several sensitivity analyses to illustrate the effects of different views about the goals-qualifying share of the single-family-owner market. This approach recognizes that there is some uncertainty in the data and that there can be different viewpoints about the various market definitions and other model parameters.

Market Estimates. Considering a 15.0-percent MF mix and a 8.5-percent investor mortgage share, the low-mod market estimates reported in Table D.10 are: 55.7 percent if the owner percentage is 44.4 percent (average home purchase share for 1999-2003); 56.2 percent if the owner percentage is 45 percent (home purchase share for 1999, 2002, and 2003); 55.4 percent if the owner percentage is 44 percent (home purchase share for 2000); 54.6 percent if the owner percentage is 43 percent (home purchase share for 1998 and 2001); and 53.8 percent if the owner percentage is 42 percent (home purchase average from 1994-97). Considering a range of 13.5-16.0 for the MF mix and a range of 8.5-9.0 for the investor mortgage share, the low-mod market estimates reported in Table D.10 are: 55.6-57.1 percent if the owner percentage is 45 percent; 54.8-56.1 percent if the owner percentage is 44 percent; 54.0-55.3 percent if the owner percentage is 43 percent; and 53.1-54.5 percent if the owner percentage is 42 percent. If the low-mod percent is at its 1999-2003 average (44.4 percent), the market range is 54.3-56.9 percent. If the low- and moderate income percentage for home purchase loans fell to 38 percent—or five percentage points from its 1994-2003 average level of 43 percent—then the overall market estimate would be about 51 percent. Thus, 51 percent is consistent with a rather significant decline in the low-mod share of the single-family home purchase market. Under the baseline projection, the home purchase percentage can fall as low as 36 percent—about four-fifths of the 1994-2003 average—and the low- and moderate-income market share would still be 49 percent.

Table D.8b reported so-called “CBSA-based” low-mod shares for single-family owner loans that reflect the new 2000 Census data and the new OMB metropolitan area definitions. Since these differed slightly from the historical “MSA-based” low-mod shares, it is useful to repeat the above analysis in terms of these new data, which will serve as the basis for scoring the GSEs' performance under the new housing goals. As shown in Table D.8b, the CBSA-based low-mod shares of home purchase loans averaged almost 44 percent between 1999 and 2003, suggesting an overall low-mod goal of 55.4 percent under the baseline, with a range from 54.8 percent to 56.1 percent. The CBSA-based measures of the low-mod share varied from approximately 42 percent (41.8 percent in 2001) to almost 46 percent (45.8 percent in 2003). Under baseline assumptions, an owner share of 42 percent translates into a 53.8 percent overall low-mod share while a 46 percent owner figure translates into a 57.0 percent low-mod share.

Case 2 (see Table D.9) considered a smaller low- and moderate-income percentage (85 percent) for both SF and MF rental properties, as compared with the baseline Case 1, which assumed 90 percent. Incorporating the Case 2 assumption reduces the low-mod market shares by about 1.3 percentage points. For example, if the SFO home purchase share is 45 percent, the overall low-mod market estimate is 54.9 percent under Case 2, as compared with 56.2 percent under Case 1 (see Table D.10). ICF considered a different option, as it reduced only the SF rental percentage from 90.0 percent to 87.5 percent. Since SF rental units account for about 10 percent of all financed units, this change reduces the overall low-mod market estimates by about 0.25 percentage points. As discussed earlier, the baseline Case 1 assumption of 90 percent is a reasonable approach for estimating the low-mod market shares.

Multifamily Mix. The volume of multifamily activity is also an important determinant of the size of the low- and moderate-income market. HUD is aware of the uncertainty surrounding projections of the multifamily market and consequently recognizes the need to conduct sensitivity analyses to determine the effects on the overall market estimate of different assumptions about the size of that market. Section C of this appendix provided HUD's rationale for its baseline MF mix of 15.0 percent and for its 13.5-16.0 percent range of MF mixes. Assuming a 13.5 percent multifamily mix reduces the overall low-mod market estimates by 0.6-0.7 percentage points compared with a 15 percent mix, and by 1.0-1.2 percentage points compared with a 16.0 percent mix. For example, when the low-mod share of the home purchase market is at 44 percent (its CBSA-based average for 1999-2003), the low-mod share of the overall market is 54.8 percent assuming a 13.5 percent multifamily mix, compared with 55.4 (56.8) percent assuming a 15 (16.0) percent multifamily mix.

As shown in Table D.10, ICF's MF mix of 14.2 percent produces results intermediate between HUD's 13.5 percent and 15.0 percent. Estimates of the low-mod market based on a MF mix of 14.2 percent are only 0.3-0.4 percentage points less than those based on a MF mix of 15.0 percent.

Fannie Mae's model combined an even lower MF mix of 12.3 percent with an investor mortgage share of 8.0 percent. If the low-mod share of home purchase loans is 44 percent (the average for 1999-2003), then the estimate for the overall low-mod market is 54.0 percent based on Fannie Mae's assumptions. In contrast, HUD's estimates (with a MF mix of 15.0 percent and 8.5-9.0 percent investor share) are 55.4-55.7 percent—about one and a half percentage points higher. If the low-mod share of home purchase loans is 45 percent (which is below the CBSA-based percentage of 45.8 for 2003), then Fannie Mae's assumptions result in a market estimate of 54.8 percent while HUD's assumptions (see previous sentence) result in market estimates of 56.2-56.5 percent.

Investor Mortgage Share. As shown in Table D.10, increasing the investor mortgage share by one percentage point from 8.0 percent to 9.0 percent increases the low-mod market estimate by approximately 0.5-0.6 percentage point. If the 10.0 percent baseline from the 2004 proposed GSE rule were used in this analysis, the market estimates would be approximately 0.6 (0.4) percentage points higher relative to the results reported in Table D.10 for a baseline of 8.5 (9.0) percent.

Examples of Home Purchase Years. The above projection results for a home purchase environment can be compared with actual results for the two most recent home purchase years, 1999 and 2000, as well as results from earlier home purchase years (1995-1997). According to the Mortgage Bankers Association of America, the refinance rate was 21 percent in 1995, 29 percent in 1996 and 1997, 34 percent in 1999, and 29 percent in 2000.

For 1999, the baseline model assumed a multifamily mix of 16.0 percent (see Section C) and a mortgage investor share of 8.2 percent (see Section D). Under these assumptions, the 1999 market estimate is 56.9 percent; if the 1999 MF mix was lower—for example, 15.0 (14.0) percent instead of 16.0 percent—then the estimate of the 1999 low-mod market share would be 56.4 (55.9) percent.

The 2004 proposed rule (Table D.9 in Appendix D) reported a higher baseline market estimate for 1999 of 58.2 percent, as compared with the 56.9 percent reported in the previous paragraph. The difference is largely due to the treatment of single-family rental mortgages. For example, using the proposed rule's 10-percent assumption for the mortgage investor share (instead of the lower 8.2 percent HMDA-based mortgage investor shares reported in the text) would increase the 1999 estimate to 57.7 percent, only 0.5 percentage points lower than the 58.2 percent reported in the proposed rule. Other minor changes that lower the market estimate included: (a) Further reducing the SF mortgage investor share by excluding B&C investor loans from the HMDA data (see Section C); (b) using 1.6 percent (instead of 2.0 percent) for the mortgage share of single-family 2-4 property owners; and (c) using slightly lower dwelling-units-per-mortgage assumptions for SF 2-4 properties (2.20 instead of 2.25) and for SF investor mortgages (1.30 instead of 1.35).

The above changes also affect the 1995-to-1997 estimates reported in Table D.9 of Appendix D of the proposed rule for the three home purchase environments prior to 1999. These estimates were 57.3 percent for both 1995 and 1996 and 57.5 percent for 1997, with an average of 57.4 percent.^[44] Given (a)-(c) in the previous paragraph and the fact that the HMDA-reported mortgage investor share was approximately eight percent during these three years (instead of the assumed 10 percent), these estimates should be reduced by about one percentage point, placing their average at 56.4. Allowing for a multifamily mix of three percentage points below the baseline estimates (similar to the approach used for 1999 and 2000 above) would drop the 1995-1997 low-mod estimates by approximately 1.4 percentage points.^[45] Thus, the 1995-1997 average would range from about 55.0 percent (with a MF mix of three percentage points below the baseline estimate) to 56.4 percent (with the baseline MF mix).^[46]

For 2000, the baseline model assumed a multifamily mix of 17.2 percent and a mortgage investor share of 9.1 percent. Under these assumptions, the 2000 low-mod market is estimated to be 57.9 percent. A lower MF mix—for example, 16.0 (15.0) percent instead of 17.2 percent—would reduce the estimated 2000 low-mod market share to 57.4 (57.0) percent. The baseline 57.9 percent estimate for 2000 is about one percentage point lower than the 59.1 percent share reported in Table D.9 of the proposed rule, mainly for the reasons discussed in the previous paragraph.

The above market estimates for 1999 and 2000 are slightly lower if the projected CBSA data are used instead of the historical 1990-based MSA data. The projected CBSA-based low-mod estimate was 56.2 percent for 1999, or 0.7 percentage points lower than the 56.9 percent estimate based on the historical MSA data. In this case, the low-mod estimate falls to 55.8 (55.4) percent if the MF mix is 15.0 (14.0) percent. Incorporating the CBSA data lowered the estimate for 2000 by 0.5 percentage points to 57.4 percent, and to 56.9 (56.5) percent if the MF mix is 16.0 (15.0) percent.

To summarize, the historical MSA-based low-mod share for all recent home purchase environments (1995-97 and 1999-2000) averaged from 55.6 percent (with a two- to three-percentage point lower MF mix than the baseline) to 56.8 percent (with the baseline MF mix). The averages (56.5 to 57.4) for the two most recent home purchase years, 1999 and 2000, were higher than those (55.0 to 56.4) for the earlier home purchase years, 1995-1997. When the data are expressed on a CBSA basis, the average low-mod shares for 1999 and 2000 decline slightly to 56.0 percent (with a two-percentage point lower MF mix than the baseline) and to 56.8 percent (with the baseline MF mix).

By comparison, ICF's best (lower bound) estimates for these home purchase years were 52 (49) percent for 1996, 55 (52-53) percent for 1997 and 1999, 56 (53) percent for 1995, and 57 (54) percent for 2000 (ICF Appendix, p. 66). Emphasizing the variability of these estimates, ICF also reported numerous other low-mod shares for these years, based on various simulations and assumptions. Some seem rather strange, or suggested that their analysis simply reduced the various input parameters to show that low estimates of the low-mod market could be the output. For example, ICF reports an overall market share of 46.9 percent share for 2000 (p. 66), which is about the same as the HMDA-reported single-family-owner percentage of 47.0 percent for 2000 (Table D.8a); it is difficult to imagine what scenario would result in the low-mod share of the rental market being in the less-than-fifty-percent range (although it is recognized that ICF was probably using an owner share less than 47 percent). ICF's report is full of such low estimates (e.g., 46.4 percent for 1996 on page 67, another 49.6 percent for 2000 on page 61) without any attempt to justify them, other than to argue that everything is variable and possible—an approach that is not very convincing if it produces a 46.9 percent low-mod share for the year 2000.

Heavy Refinancing Environments. The low-mod share of the market will decline during a period of heavy refinancing due to (a) a decline in the low-mod share of single-family refinance mortgages as middle- and upper-income borrowers dominate the refinance market; (b) a decline in the relative importance of the subprime market; and (c) a decline in the share of multifamily mortgages. For example, during 2002, the refinance share of low-mod loans was 41.8 percent (compared with 47-51 percent during the two home purchase years of 1999 and 2000); the subprime share of the single-family market was 8.6 percent (compared with 13 percent during 1999 and 2000); and the multifamily share of the market was 11 percent or less (compared with 16 percent or more during 1999 and 2000). Although there is some uncertainty with the data, the multifamily mix for 2003 could have been as low as 6 or 7 percent.

Table D.11 shows the impact on the low-mod market share under different assumptions about a refinancing environment. The table reports the results for a 65 percent refinance environment, which has been characteristic of recent (2002 and 2003) refinance waves. Refinancing environments are characterized by lower MF mixes because single owner properties dominate the market; therefore Table D.11 considers MF mixes from 6 to 12 percent. Most likely, a MF mix of 12-13 percent characterized 2001, 9-11 percent characterized 2002, and less than 7 percent characterized 2003; there is some uncertainty with these estimates, as discussed in Section C of this appendix. In a refinancing wave, the low-mod percent is typically lower for refinance loans than home purchase loans, as middle- and high-income borrowers take advantage of reduced interest rates. With respect to the low-mod characteristics of SF owner loans, two scenarios were considered: (A) Scenario A represents the average low-mod percentages for the last four refinance years (1998, 2001, 2002, and 2003)—43 percent for home purchase loans and 40 percent for refinance loans; and (B) Scenario B represents the average low-mod percentages for the two most recent refinance years (2002, and 2003)—44.5 percent for home purchase loans and 40.5 percent for refinance loans. Thus, there is a 3-4 percentage point differential between home purchase loans and refinance loans in a heavy refinancing environment. This analysis assumed an investor mortgage share of 8.0 percent (average for these refinancing years) and a subprime market share of 8.5 percent (instead of the 12-percent assumption in the baseline model).

Under Scenario A, the low-mod shares varied by approximately three percentage points, from 51.6 percent with a 12 percent MF mix to 48.9 percent with a 6 percent MF mix. Under Scenario B, the low-mod percentages are all 0.7 percent higher, and the pattern is from 52.3 percent with a MF mix of 12 percent to 49.6 percent with a MF mix of 6 percent. Notice that under Scenario B, the low-mod share remains in the 50-51 percent range even if the MF mix falls to 6-8 percent. These low-mod market shares are 4-7 percentage points lower than the low-mod shares reported in Table D.10 for HUD's baseline home purchase environment. In addition to higher-income borrowers dominating the single-family market, the share of the “goals rich” rental market declines in a refinancing wave, which tends to further reduce the low-mod share of market activity. The right-hand column of Table D.11 shows that the rental share falls to the 17-22 percent range, or 4-9 percentage points less that the almost 26-percent rental share in HUD's baseline model.

Model estimates were also made for the recent refinancing years of 1998, 2001, 2002, and 2003. The Mortgage Bankers Association of America estimated that the refinance rate was 50 percent in 1998, 55 percent in 2001, 59 percent in 2002, and 66 percent in 2003. The year 2003 stands out not only for its high rate of refinancing but also for the sheer volume of refinancing ($2.5 trillion), which led to record single-family mortgage originations ($3.8 trillion) that year.

For 1998, the baseline model assumed a multifamily mix of 14.0 percent (see Section C) and a mortgage investor share of 6.8 percent (see Section D). Under these assumptions, the 1998 market estimate is 51.9 percent. If the MF mix for 1998 had been 13.0 (12.0) percent, instead of the baseline of 14.0 percent, then the estimated low-mod market share for 1998 would be 51.3 (50.8) percent. For 2001, the baseline model assumed a multifamily mix of 13.5 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2001 market estimate is 53.4 percent. If the MF mix for 2001 had been 12.5 (12.0) percent, instead of the baseline of 13.5 percent, then the estimated low-mod market share for 2001 would be 52.9 (52.7) percent. For 2002, the baseline model assumed a multifamily mix of slightly over 11.0 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2002 low-mod market is estimated to be 53.2 percent.^[47] A lower MF mix—for example, 10.5 (9.5) percent instead of 11 percent—would reduce the estimated 2002 low-mod market share to 53.1 (52.5) percent.

Using the projected CBSA data (instead of the historical 1990-based MSA data) lowered the 2001 and 2002 low-mod estimates by approximately one percentage point. The 2001 market estimates are reduced to 52.3 percent (13.5 MF mix), 51.8 percent (12.5 MF mix), and 51.6 percent (12.0 MF mix). The 2002 market estimates are reduced to 52.1 percent (11.1 MF mix), 52.0 (10.5 MF mix), and 51.4 percent (9.5 MF mix).

By comparison, ICF's best estimates for these refinancing years are one or two percentage points lower than the above estimates: 49.7 percent for 1998, 51.1 percent for 2001, and 50.9 percent for 2002; because of the unavailability of 2003 HMDA data, no estimate was provided by ICF for that year. (See ICF Appendix, p. 60). ICF's lower bound estimates for these three years were in the 47-48 percent range. But as noted earlier, ICF also produces a number of even lower estimates without discussion of what circumstances might lead to them—examples include their 45.2 percent market estimate for 2001 when the SFO low-mod share was 42.3 percent (see Table D.8a) and their 44.9 percent estimate for 2002 when the SFO low-mod share was 42.7 percent. (See ICF Appendix, p. 66.)

For the years 1999 to 2002, Fannie Mae estimated a low-mod market share of 52-53 percent. (This is their estimate assuming no missing data; see their Table I.9, page I-34.) This compares with HUD's estimate of 53.7 percent to 54.5 percent. As discussed in Section C.6, Fannie Mae assumes a rather low MF mix (approximately 10 percent) in the model that generates its historical estimates.

Given that HUD did not receive 2003 HMDA data until August 2004, it was not possible to develop a complete projection model for 2003. Still, HUD developed some rough projections for 2003. Given the huge volume of single-family originations ($3.8 trillion), the 1998 MF mix was likely rather low. In fact, Fannie Mae estimates the MF mix dropped to five percent in 2003. Thus, the estimates of the low-mod market share for 2003 are presented for different assumptions about the MF mix, recognizing that firm data on the 2003 multifamily market are not available. Combining an investor mortgage share of 8.2 from HMDA (from HMDA) with different MF mixes produces the following estimates: 51.9 percent (MF mix of 8 percent); 51.4 percent (MF mix of 7 percent); and 51.0 percent (MF mix of 6.0 percent).

As shown by both the simulation results and by the actual experience during 1998 and 2001-2003, the low-mod share declines when refinances dominate the mortgage market. The above estimates place the low-mod average during these four years of heavy refinancing at 52 percent, with practically all of the estimates of annual low-mod shares varying between 51 and 53 percent. As noted above, the estimates for 2003 (around 51 percent) are somewhat speculative.

The various market estimates presented in Table D.10 for a home purchase environment and reported above for a refinance environment are not all equally likely. Most of them equal or exceed 51 percent. In the home purchase environment, estimates below 51 percent would require the low-mod share of the single-family-owner market for home purchase loans to drop to 38 percent, which would be five percentage points below the 1994-2003 average of 43 percent. Thus, 51 percent is consistent with a rather significant decline in the low-mod share of the single-family home purchase market. Sensitivity analyses of different refinance environments and model estimates for 1998, and 2001-2003 suggest that it would require a particularly heavy period of refinancing to fall below a 51-percent low-mod market share.

b. Economic Conditions and the Feasibility of the Low- and Moderate-Income Housing Goal

Commenters expressed a general concern that the market share estimates and the housing goals failed to recognize the volatility of housing markets and the existence of macroeconomic cycles. There was particular concern that the market shares and housing goals were based on a period of economic expansion accompanied by record low interest rates and high housing affordability. This section continues the discussion of these issues, noting that the Secretary can consider shifts in economic conditions when evaluating the performance of the GSEs on the goals, and noting further that the market share estimates can be examined in terms of less favorable market conditions than have existed during the 1993 to 2003 period. As also explained below, HUD is publishing in the Federal Register an Advance Notice of Proposed Rulemaking that advises the public of HUD's intention to consider by separate rulemaking a provision that recognizes and takes into consideration the impact of high volumes of refinance transactions on the GSEs' ability to achieve the housing goals in certain years, and solicits proposals on how such a provision should be structured and implemented.

Volatility of the Market. Changing economic conditions can affect the validity of HUD's market estimates as well as the feasibility of the GSEs' accomplishing the housing goals. The volatile nature of the mortgage market in the past few years suggests a degree of uncertainty around projections of the origination market. Large swings in refinancing, consumers switching between adjustable-rate mortgages and fixed-rate mortgages, and increased first-time homebuyer activity due to record low interest rates, have all characterized the mortgage market during the nineties. These conditions are beyond the control of the GSEs but they would affect their performance on the housing goals. A mortgage market dominated by heavy refinancing on the part of middle-income homeowners would reduce the GSEs' ability to reach a specific target on the Low- and Moderate-Income Goal, for example. A jump in interest rates would reduce the availability of very-low-income mortgages for the GSEs to purchase. But on the other hand, the next few years may be favorable to achieving the goals because of the high refinancing activity in 2001, 2002, and 2003. A period of low-to-moderate interest rates would sustain affordability levels without causing the rush to refinance seen earlier in 1998 and 2001-2003. A high percentage of potential refinancers have already done so, and are less likely to do so again. However, these same predictions were made after the 1998 refinance wave, which indicates the uncertainty of making predictions about the mortgage market.

Recent years have been characterized by record affordability conditions due to low interest rates and economic expansion. Thus, as Section F.3.a indicates, HUD also examined potential changes in the market shares under very different macroeconomic environments, including periods of recession, high interest rates, and heavy refinancing (accompanied by low interest rates). A recessionary environment would likely be characterized by a reduction in single-family activity (or an increase in the multifamily share of the market) and a reduction in the low-mod shares of the single-family-owner market. The home purchase percentage can fall as low as 36 percent—about four-fifths of the 1994-2003 average—and the low- and moderate-income market share would still be 49 percent. If the low-mod share of the owner market were reduced more modestly to 39 percent, the low-mod share for the overall market would fall to 51.5 percent, assuming a multifamily mix of 15.0 percent. (See Table D.10.)

As discussed in Appendix A, record low interest rates, a more diverse socioeconomic group of households seeking homeownership, and affordability initiatives of the private sector have encouraged first-time buyers and low-income borrowers to enter the market since the mid-1990s. Over the past eight years, the affordable lending market has demonstrated an underlying strength that suggests it will continue, particularly given demographic projections of increased minorities and immigrants in the mortgage market. However, a significant increase in interest rates over recent levels would reduce the presence of low-income families in the mortgage market and the availability of low-income mortgages for purchase by the GSEs. As noted above, the 51-56 percent range for the low-mod market share covers economic and market affordability conditions much less favorable than recent conditions of low interest rates and economic expansion. The low-mod share of the single-family home purchase market could fall to 38 percent, which is five percentage points lower than its 1995-2003 average level of 43 percent, and the low-mod market share would only be slightly below 51 percent. The above analysis of 1998 and the 2001-2003 period suggests that 51 percent is a reasonable minimum low-mod share for years of heavy refinancing.

Feasibility Determination. As stated in the 2000 Rule, HUD is well aware of the volatility of mortgage markets and the possible impacts on the GSEs' ability to meet the housing goals. FHEFSSA allows for changing market conditions.^[48] If HUD has set a goal for a given year and market conditions change dramatically during or prior to the year, making it infeasible for the GSE to attain the goal, HUD must determine “whether (taking into consideration market and economic conditions and the financial condition of the enterprise) the achievement of the housing goal was or is feasible.” This provision of FHEFSSA clearly allows for a finding by HUD that a goal was not feasible due to market conditions, and no subsequent actions would be taken. As HUD noted in both the 1995 and 2000 GSE Rules, it does not set the housing goals so that they can be met even under the worst of circumstances. Rather, as explained above, HUD has conducted numerous sensitivity analyses for economic and market affordability environments much more adverse than has existed in recent years. If macroeconomic conditions change even more dramatically, the levels of the goals can be revised to reflect the changed conditions. FHEFSSA and HUD recognize that conditions could change in ways that require revised expectations.

HUD received a number of public comments seeking a regulatory solution to the issue of the ability of the GSEs to meet the housing goals during a period when refinances of home mortgages constitute an unusually large share of the mortgage market. As explained in the Preamble, HUD is not addressing the refinance issue in this final rule. Elsewhere in this Federal Register, HUD is publishing an Advance Notice of Proposed Rulemaking that advises the public of HUD's intention to consider by separate rulemaking a provision that recognizes and takes into consideration the impact of high volumes of refinance transactions on the GSEs' ability to achieve the housing goals in certain years, and solicits proposals on how such a provision should be structured and implemented. HUD believes that it would benefit from further consideration and additional public input on this issue. HUD also notes (see above) that FHEFSSA provides a mechanism by which HUD can take into consideration market and economic conditions that may make the achievement of housing goals infeasible in a given year. (See 12 U.S.C. 1336(b)(e).)

c. Treatment of B&C Loans and Other Technical Market Issues

B&C Mortgages. As discussed in Appendix A, the market for subprime mortgages has experienced rapid growth over the past 6-7 years, rising from an estimated $65 billion in 1995 to $174 billion in 2001, $213 billion in 2002 and $332 billion in 2003.^[49] In terms of credit risk, subprime loans include a wide range of mortgage types. “A-minus” loans, which represent at least half of the subprime market, make up the least risky category.^[50] As discussed in Appendix A, the GSEs are involved in this market both through specific program offerings and through purchases of securities backed by subprime loans (including B&C loans as well as A-minus loans). The B&C loans experience much higher delinquency rates than A-minus loans.^[51]

The market estimates reported in Section F.3.a-b exclude the B&C portion of the subprime market; or conversely, they include the A-minus portion of the subprime market. This section explains how these “adjusted” market shares are calculated from “unadjusted” market shares that include B&C loans.

There are two possible approaches for adjusting for the effects of B&C owner loans in the projection model. First, readers could choose a single-family low-mod percentage (that is, one of the percentages in the first column in Table D.10) that they believe is adjusted for B&C loans and then obtain a rough estimate of the overall market estimate from the second to fourth columns corresponding to different multifamily mixes. For instance, if one believes the appropriate single-family-owner percentage adjusted for B&C loans (or adjusted for any other market sectors that the reader thinks appropriate) is 44 percent, then the low-mod market estimate is 55.4 percent assuming a multifamily mix of 15 percent. While intuitively appealing, such an approach would provide inaccurate results, as explained next.

Second, readers could choose a single-family-owner percentage directly from HMDA data that is unadjusted for B&C loans and then rely on HUD's methodology (described below) for excluding the effects of B&C loans. This is the approach taken in Table D.10. The advantage of the second approach is that HUD's methodology makes the appropriate adjustments to the various property shares (i.e., the owner versus rental percentages) that result from excluding single-family B&C loans from the analysis. According to HUD's methodology, dropping B&C loans would reduce the various low-mod market estimates by less than half of a percentage point. This minor effect is due to (a) the fact that the low-mod share of B&C loans is similar to that of the overall market; and (b) the offsetting effects of the increase in the rental market share when single-family B&C loans are dropped from the market totals.

As noted above, if one assumes the single-family-owner percentages in the first column of Table D.10 are unadjusted for B&C loans, then the overall low-mod market estimates must be adjusted to exclude these loans. The effects of deducting the B&C loans from the projection model can be illustrated using an example of a low-mod percentage of 44 percent for single-family-owner loans. Again, as explained earlier, this 44 percent figure could reflect a mortgage market environment where home purchase and refinance loans had similar low-mod percentages (i.e., 44 percent) or a mortgage market environment where home purchase and refinance loans had different low-mod market percentages that together resulted in a 44 percent average for the single-family-owner market.

As Table D.10 shows, a 44 percent low-mod share for owner mortgages translates into an overall low-mod market share of 55.4 percent. It is assumed that the subprime market accounts for 12 percent of all mortgages originated, which would be $204 billion based on $1,700 billion for the mortgage market. This $204 billion estimate for the subprime market is reduced by 20 percent to arrive at $163.2 billion for subprime loans that will be less than the conforming loan limit. Dividing this figure by the average loan amount for subprime loans gives 1,256,361 subprime loans in the conventional market. HMDA data indicate that six percent of these are SF investor loans (75,382) and the remaining ones are SF owner loans (1,180,979). Since this analysis retains half of subprime loans (i.e., the A-minus portion of that market), these figures are reduced by one-half to arrive at 590,489 owner B&C loans and 37,691 investor B&C loans. The investor loans are placed on a unit basis by multiplying by 1.3 (units per mortgage), yielding 48,998 financed dwelling units in the investor B&C market.

HMDA data was used to provide an estimate of the portion of the 590,489 owner B&C loans that would qualify for each of the housing goals. HMDA data does not identify subprime loans, much less divide them into their A-minus and B&C components. As explained in Appendix A, Randall Scheessele in HUD's Office of Policy Development and Research has identified almost 200 HMDA reporters that primarily originate subprime loans. Based on 1999-2002 HMDA data, the goals-qualifying percentages of loans originated by these subprime lenders were as follows: 58.6 percent qualified for the low-mod goal, 28.0 percent for the special affordable goal, and 52.0 percent for the underserved areas goal.^[52] Applying the goals-qualifying percentages to the 590,489 owner B&C loans gives the following estimates of B&C owner loans that qualified for each of the housing goals: Low-mod (346,027), special affordable (165,337), and underserved areas 614,109. The process for the smaller number (48,998) of investor B&C loans is similar. It is assumed that 90 percent (44,098) of these B&C rental units qualify for the low-mod goal, 58 percent (28,419) qualify for the special affordable goal, and 74 percent (36,259) qualify for the underserved areas goal (based on 2000 Census data).

Adjusting HUD's model to exclude B&C owner loans and B&C financed rental units involves subtracting the above eight figures—two for the overall owner and rental B&C market and six for B&C owner units and rental units that qualify for each of the three housing goals—from the corresponding figures estimated by HUD for the total single-family and multifamily market inclusive of B&C owner loans and B&C dwelling units. HUD's model projects that 10,478,681 single-family and multifamily units will be financed; of these, 5,842,313 (55.8 percent) qualified for the low-mod goal, 2,801,179 (26.7 percent) for the special affordable goal, and 3,983,005 (38.0 percent) for the underserved areas goal. Deducting the B&C owner and rental market estimates produces the following adjusted market estimates: A total market of 9,839,193, of which 5,452,188 (55.4 percent) qualified for the low-mod goal, 2,607,423 (26.5 percent) for the special affordable goal, and 3,639,692 (37.0 percent) for the 2000-based underserved areas goal.

The low-mod market share estimate exclusive of B&C loans (55.4 percent) is only slightly lower than the original market estimate (55.8 percent from above), as is also the special affordable market estimate (26.7 percent versus 26.5 percent). This occurs because the B&C owner loans that were dropped from the analysis have similar low-mod and special affordable percentages as the overall (both single-family and multifamily) market. For example, the low-mod share of B&C loans was projected to be 58.6 percent and HUD's market model (unadjusted for B&C loans) projected the overall low-mod share to be practically the same, 55.8 percent. Thus, dropping B&C owner loans from the market totals does not significantly reduce the overall low-mod share of the market. Because they qualify at such a high rate (e.g., 90 percent on low-mod), dropping B&C rental loans tends to reduce the market share estimates. However, they are relatively small in number—B&C owner loans dominate the results because they account for 92.3 percent (590,489 divided by 639,487) of the total B&C owner and rental units dropped from the market totals.

The situation is different for the underserved areas goal. Underserved areas account for 52.0 percent of the B&C owner loans, which is a higher percentage than the underserved area share of the overall market (38.0 percent). Thus, dropping the B&C owner loans (as well as the smaller number of B&C rental units) leads to a reduction in the underserved areas market share of 1.0 percentage points, from 38.0 percent to 37.0 percent. (If this analysis were conducted in terms of 1990-Census data, the one-percentage point reduction would be from about 33.0 percent to 32.0 percent.)

Dropping B&C loans from HUD's projection model changes the mix between rental and owner units in the final market estimate; rental units accounted for 26.7 percent of total units after dropping B&C loans compared with 25.6 percent before dropping B&C loans. Since practically all rental units qualify for the low-mod goal, their increased importance in the market partially offsets the negative effects on the goals-qualifying shares of any reductions in B&C owner loans. Thus, another way of explaining why the goals-qualifying market shares are not affected so much by dropping B&C owner loans is that the rental share of the overall market increases as the B&C owner units are dropped from the market. Since rental units have very high goals-qualifying percentages, their increased importance in the market partially offsets the negative effects on the goals-qualifying shares of any reductions in B&C owner loans. In fact, this rental mix effect would come into play with any reduction in owner units from HUD's model.

A similar analysis can be used to demonstrate the effects of deducting the remaining, A-minus portion of the subprime market from the market estimates. Of course, deducting A-minus loans as well as B&C loans is equivalent to deducting all subprime loans from the market. In the example given above (44 percent low-mod percentage for owners), deducting all subprime loans would further reduce the overall low-mod market estimate to 55.0 percent. Thus, the unadjusted low-mod market estimate is 55.8 percent, the estimate adjusted for B&C loans is 55.4 percent (reported in Table D.10), and the estimate adjusted for all subprime loans is 55.0 percent.

As discussed in the 2000 Rule, there are caveats that should be mentioned concerning the above adjustments for the B&C market. The adjustment for B&C loans depends on several estimates relating to the single-family mortgage market, derived from various sources. Different estimates of the size of the B&C market or the goals-qualifying shares of the B&C market could lead to different estimates of the goals-qualifying shares for the overall market. The goals-qualifying shares of the B&C market were based on HMDA data for selected lenders that primarily originate subprime loans; since these lenders are likely originating both A-minus and B&C loans, the goals-qualifying percentages used here may not be accurately measuring the goals-qualifying percentages for only B&C loans. The above technique of dropping B&C loans also assumes that the coverage of B&C and non-B&C loans in HMDA's metropolitan area data is the same; however, it is likely that HMDA coverage of non-B&C loans is higher than its coverage of B&C loans.^[53] Despite these caveats, it also appears that reasonably different estimates of the various market parameters would not likely change, in any significant way, the above estimates of the effects of excluding B&C loans in calculating the goals-qualifying shares of the market. As discussed in other sections, HUD provides a range of estimates for the goals-qualifying market shares to account for uncertainty related to the various parameters included in its projection model for the mortgage market.

Manufactured Housing Loans and Small Loans. HUD includes the effects of manufactured housing loans (at least those financing properties in metropolitan areas) in its market estimates. However, sensitivity analyses are conducted to determine the effects of excluding these loans. Excluding manufactured housing loans as well as small loans (loans less than $15,000) reduces the overall market estimates reported in Table D.10 by about one percentage point. This is estimated as follows. First, excluding these loans reduces the low-mod percentage for single-family-owner mortgages in metropolitan areas by about 1.9 percentage points, based on analysis of recent home purchase environments (1995-97 and 1999 and 2000). Multiplying this 1.9 percentage point differential by the property share (0.745) of single-family-owner units yields 1.4 percentage points, which serves as a proxy for the reduction in the overall low-mod market share due to dropping manufactured home loans from the market analysis. The actual reduction will be somewhat less because dropping manufactured home loans will increase the share of rental units, which increases the overall low-mod market share, thus partially offsetting the 1.4 percent reduction. The net effect is probably a reduction of about one percentage point.

The effects can be considered separately. Dropping only manufactured housing loans would reduce the market estimates by approximately three-quarters of a percentage point. ICF argued that loans with less than $15,000 should be excluded. The impact of doing this on the market estimates would be less than half a percentage point. ICF also considered scenarios where one-half of manufactured loans would be dropped, as well as small loans less than $15,000. The impact of doing this on the market estimates would be less than three-quarters of a percentage point.

The estimated reductions in goals-qualifying shares due to excluding manufactured housing would be even lower during the heavy refinance years such as 1998 and 2001-2003. It should also be mentioned that manufactured housing in non-metropolitan areas is not included in HUD's analysis due to lack of data; including that segment of the market would increase the goals-qualifying shares of the overall market. Thus, the analyses of manufactured housing reported above and throughout the this final rule pertain only to manufactured housing loans in metropolitan areas, as measured by loans originated by the 21 manufactured housing lenders identified by Randy Scheessele at HUD.

The above analyses of the effects of less affordable market conditions, different assumptions about the size of the rental market, and dropping different categories of loans from the market definition suggest that 51-56 percent is a reasonable range of estimates for the low- and moderate-income market. This range covers markets without B&C and allows for market environments that would be much less affordable than recent market conditions. The next section presents additional analyses related to market volatility and affordability conditions.

d. Conclusions About the Size of Low- and Moderate-Income Market

Based on the above findings as well as numerous sensitivity analyses, HUD concludes that 51-56 percent is a reasonable range of estimates of the mortgage market's low- and moderate-income share for the year 2005 and beyond. The range covers much more adverse economic and market affordability conditions than have existed recently, allows for different assumptions about the single-family and multifamily rental markets, and excludes the effects of B&C loans. HUD recognizes that shifts in economic conditions and refinancing could increase or decrease the size of the low- and moderate-income market during that period.

G. Size of the Conventional Conforming Market Serving Central Cities, Rural Areas, and Other Underserved Areas

The following discussion presents estimates of the size of the conventional conforming market for the Central City, Rural Areas, and other Underserved Areas Goal; this housing goal will also be referred to as the Underserved Areas Goal. The first three sections, which analyze historical data going back to the early 1990's, necessarily used 1990 Census geography to define underserved census tracts and underserved counties. The first two sections focus on underserved census tracts in metropolitan areas, as Section 1 presents underserved area percentages for different property types while Section 2 presents market estimates for metropolitan areas. Section 3 discusses B&C loans and rural areas. But as explained in Appendix B, HUD will be defining underserved areas based on 2000 Census geography beginning in 2005, the first year covered by this final rule. Therefore, Section 4 repeats much of the analyses in Sections 1-3 but in terms of 2000 Census geography, rather than 1990 Census geography

1. Underserved Areas Goal Shares by Property Type

For purposes of the Underserved Areas Goal, underserved areas in metropolitan areas are defined as census tracts with:

(a) Tract median income at or below 90 percent of the MSA median income; or

(b) A minority composition equal to 30 percent or more and a tract median income no more than 120 percent of MSA median income.

Owner Mortgages. The first set of numbers in Table D.12 are the percentages of single-family-owner mortgages that financed properties located in underserved census tracts of metropolitan areas between 1992 and 2003. There are several interesting patterns in these data. During 1999 and 2000, 28-30 percent of mortgages (both home purchase and refinance loans) financed properties located in these areas; this percentage fell to 25.7 percent in 2001, 25.0 percent in 2002, and 25.3 percent in 2003, figures that were slightly below the average (26.8 percent) between 1994 and 1998. In 1992 and 1993, the underserved areas share of single-family-owner mortgages was only 20 percent.

In most years, refinance loans are more likely than home purchase loans to finance properties located in underserved census tracts. Between 1994 and 2003, 27.3 percent of refinance loans were for properties in underserved areas, compared to 25.5 percent of home purchase loans. This 1.8 percentage point refinance-home-purchase differential is mostly due to the influence of subprime loans. Excluding B&C (all subprime) loans and considering the same time period, 26.1 (24.9) percent of refinance loans were for properties in underserved areas, compared to 25.1 (24.6) percent of home purchase loans. Thus, excluding B&C (subprime) loans reduces the differential from 1.8 percentage points to 1.0 (0.3) percentage point. In the year (2000) with the largest differential, excluding B&C (all subprime) loans reduced the refinance-home-purchase differential from 8.1 percent to 6.9 (5.7) percent; in this case, a significant differential remained after excluding B&C (subprime) loans. In the heavy refinance years of 1998, 2001, 2002, and 2003 underserved areas accounted for about 25 percent of total (both home purchase and refinance) owner loans.

The underserved areas share for home purchase loans has been in the 25-26 percent range since 1995, except for 2000 and 2002 when it increased to over 27 percent, and in 2003 when it increased to 28.5 percent. Considering all (both home purchase and refinance) loans during recent “home purchase” environments, the underserved areas share was a high 28-30 percent during 1999-2000, compared with a 27 percent average between 1995 and 1997; excluding B&C and other (i.e. A-minus) subprime loans places 1999 on par with the earlier years, with only the year 2000 showing a higher level of underserved area lending than occurred during 1995-97. These data indicate that the single-family-owner market in underserved areas has remained strong since the 2000 Rule was written. While it is recognized that economic and housing affordability conditions could change and reduce the size of the underserved areas market, it appears that the underserved market has certainly maintained itself at a high level over the past four years.

Renter Mortgages. The second and third sets of numbers in Table D.12 are the underserved area percentages for single-family rental mortgages and multifamily mortgages, respectively. Based on HMDA data for single-family, non-owner-occupied (i.e., investor) loans, the underserved area share of newly-mortgaged single-family rental mortgages has averaged about 44 percent (over nine or ten years). HMDA data also show that about half of newly-mortgaged multifamily rental units are located in underserved areas. HUD's baseline assumes that 42.5 percent of single-family investor loans and 48 percent of multifamily loans are located in underserved areas. The GSEs and ICF argued that HUD had overstated these underserved area percentages; Section G.4 below, which focuses on the 2000-based underserved area percentage, will discuss and respond to their concerns. Fannie Mae also said that subprime (or B&C) loans should be taken out of the SF investor loans. As shown in Table D.12, deducting B&C loans reduces the underserved area percentage for SF investor mortgages by almost one percentage point (the 1993-2003 unweighted average falls from 44.0 percent to 43.1 percent). HUD's model excludes B&C investor loans in the same manner it excludes B&C owner loans (see earlier explanation).

2. Market Estimates for Underserved Areas in Metropolitan Areas

Table D.13 reports HUD's estimates of the market share for underserved areas based on the projection model discussed earlier. The estimates in Table D.13 exclude the effects of B&C owner loans and B&C investor loans. The percentage of single-family-owner mortgages financing properties in underserved areas is the most important determinant of the overall market share for this goal. Therefore, Table D.13 reports market shares for different single-family-owner percentages ranging from 30 percent (2000 level) to 20 percent (1993 level) to 19 percent. Considering a 15.0-percent MF mix and a 8.5-percent investor mortgage share, the market share estimate is 31-32 percent if the overall (both home purchase and refinance) single-family-owner percentage for underserved areas is at its 1994-2003 HMDA average of 26.6 percent. The overall market share for underserved areas peaks at 35 percent when the single-family-owner percentage is at its 2000 level of 30 percent.

The analysis can also be conducted in terms of the home purchase percentages reported in Table D.13. Again, considering a 15.0-percent MF mix and an 8.5-percent investor mortgage share, the underserved area market estimates reported in Table D.13 are: 33.3 percent if the owner percentage is 28.5 percent (home purchase share for 2003); 32.1 if the owner percentage is 27 percent (home purchase share in 2000 and 2002 slightly above the 1999-2003 average home purchase share of 26.8 percent); 31.3 percent if the owner percentage is 26 percent (home purchase share for 1999 and 2001); and 30.5 percent if the owner percentage is 25 percent (home purchase average from 1994-98). This analysis assumes that the underserved areas share of refinance loans is the same as those listed above for home purchase loans. But, as Table D.12 shows, the underserved areas share of refinance loans tends to be higher than that for home purchase loans. And in the year 2000, the overall underserved areas share for owner loans reached 30 percent; as noted in the previous paragraph, the overall market estimate is 34.6 percent in this case. However, the next highest overall owner share is the 28.2 percent share in 1999, which yields a market estimate of approximately 33 percent.

Sensitivity Analyses. Unlike the Low- and Moderate-Income and Special Affordable Goals, the market estimates differ only slightly as one moves from a 13.5 percent MF mix to 16.0 percent MF mix. For example, reducing the assumed multifamily mix from 16.0 percent to 13.5 percent reduces the overall market projection for underserved areas by only 0.5-0.6 percentage points. This is because the underserved area differentials between owner and rental properties are not as large as the low- and moderate-income differentials reported earlier.

Similarly, the market estimates differ only slightly with changes in the investor mortgage share. Reducing the investor mix from 9.5 percent to 8.0 percent reduces the overall market projection for underserved areas by only 0.2-0.3 percentage points.

Case 2 (see Table D.9) considered slightly smaller underserved area percentages for rental properties (40 percent for SF rentals and 46 percent for MF rentals), as compared with the baseline Case 1, which assumed 42.5 percent and 46.0 percent, respectively. Incorporating these Case 2 assumptions reduces the underserved areas market estimate by only 0.6 percentage points. For example, if the SFO home purchase share is 28 percent, then the overall underserved area estimate is 32.3 percent under Case 2, as compared with 32.9 percent under Case 1 (see Table D.13).

Examples of Home Purchase and Refinance Environments. The above projection results for a home purchase environment can be compared with actual results for two home purchase years, 1999 and 2000 (see earlier description of these two years in the low-mod section, F.3.a). For 1999, the baseline model assumed a multifamily mix of 16.0 percent and a mortgage investor share of 8.2 percent. Under these assumptions, the projected 1999 market estimate (based on 1990-Census data) is 33.1 percent; if the 1999 MF mix was lower at 15.0 (14.0), then the estimate of the 1999 underserved areas market share would be only slightly lower at 32.9 (32.6) percent.^[54] For 2000, the baseline model assumed a multifamily mix of 17.2 percent and a mortgage investor share of 9.1 percent. Under these assumptions, the 2000 underserved areas market is estimated to be 34.9 percent. A lower MF mix of 16.0 (15.0) percent would reduce the estimated 2000 underserved areas market share slightly to 34.6 (34.4) percent.^[55]

The heavy refinance scenarios discussed for the low-mod market were also projected for the underserved areas market. Since the impact of a heavy refinancing period on the underserved areas market share will be covered in Section G.4, which incorporates 2000 Census data, there is no need for a detailed discussion in this section's analysis based on 1990 Census data. Still, it is useful to provide a quick review of the 1990-based underserved area estimates for three heavy refinancing environments (1998, 2001, 2002, and 2003). For 1998, the baseline model assumed a multifamily mix of 14.0 percent and a mortgage investor share of 6.8 percent. Under these assumptions, the 1998 market estimate is 29.9 percent. If the MF mix for 1998 had been 12.0 percent, instead of the baseline of 14.0 percent, then the estimated underserved area market share for 1998 would be 29.4 percent. For 2001, the baseline model assumed a multifamily mix of 13.5 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2001 market estimate is 32.1 percent, dropping to 31.7 percent if the MF mix was 12.0 percent. For 2002, the baseline model assumed a multifamily mix of slightly over 11.0 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2002 underserved areas market is estimated to be 31.6 percent, dropping to 31.1 percent if the MF mix is 9.5 percent. This analysis suggests that the underserved areas market based on 1990 Census data will be about 29-32 percent range during periods of heavy refinancing.^[56]

Additional sensitivity analyses were conducted to reflect the volatility of the economy and mortgage market. Recession and high interest rate scenarios assumed a significant drop in the underserved area percentage for single-family-owner mortgages. The single-family-owner percentage can go as low as 24 percent—which is 3 percentage points lower than the 1995-2003 average of 27 percent—and the estimated market share for underserved areas remains at almost 30 percent. In a more severe case, the overall underserved market share would be 27.5 percent if the single-family-owner share fell to 21 percent (its 1992 level), which is 7-9 percentage points lower than its 1999-2000 levels.

3. Adjustments: B&C Loans, the Rural Underserved Areas Market, and Manufactured Housing Loans

B&C Loans. The procedure for dropping B&C loans from the projections is the same as described in Section F.3.b for the Low- and Moderate-Income Goal. The underserved area percentage for B&C loans is 44.5 percent, which is much higher than the projected percentage for the overall market (which peaks at 35 percent as indicated in Table D.13). Thus, dropping B&C loans will reduce the overall market estimates. Consider the case of a single-family-owner percentage of 27 percent, which yields an overall market estimate for underserved areas of 33.1 percent, including B&C loans. When B&C loans are excluded from the projection model, the underserved areas market share falls by 0.9 percentage points to 32.2 percent, which is the figure reported in Table D.13.

Non-metropolitan Areas. Underserved rural areas are non-metropolitan counties with:

(a) County median income at or below 95 percent of the greater of statewide non-metropolitan median income or nationwide non-metropolitan income; or

(b) A minority composition equal to 30 percent or more and a county median income no more that 120 percent of statewide non-metropolitan median income.

HMDA's limited coverage of mortgage data in non-metropolitan counties makes it impossible to estimate the size of the mortgage market in rural areas. However, all indicators suggest that underserved counties in non-metropolitan areas comprise a larger share of the non-metropolitan mortgage market than the underserved census tracts in metropolitan areas comprise of the metropolitan mortgage market. For instance, underserved counties within rural areas include 54 percent of non-metropolitan homeowners; on the other hand, underserved census tracts in metropolitan areas account for only 34 percent of metropolitan homeowners.

During 1999-2003, 38.3 percent of the GSEs' single-family-owner (SFO) purchases in non-metropolitan areas were in underserved counties while 23.1 percent of their SFO purchases in metropolitan areas were in underserved census tracts. These figures suggest the market share for underserved counties in rural areas is higher than the market share for underserved census tracts in metropolitan areas. Thus, using a metropolitan estimate to proxy the overall market for this goal, including rural areas, is conservative.^[57]

The limited HMDA data available for non-metropolitan counties also suggest that the underserved areas market estimate would be higher if complete data for non-metropolitan counties were available. According to HMDA, underserved counties accounted for 41.6 percent of SFO mortgages originated in non-metropolitan areas between 1999 and 2003. By contrast, underserved census tracts accounted for approximately 24.9 percent of SFO mortgages originated in metropolitan areas between 1999 and 2003.^[58] Since non-metropolitan areas account for 13 percent of all single-family-owner mortgages ^[59] and estimating that the single-family-owner market accounts for 74.5 percent of newly-mortgaged dwelling units, then the non-metropolitan underserved area differential of 16.7 percent would raise the overall market estimate by 1.6 percentage point—16.7 percentage points times 0.13 (non-metropolitan area mortgage market share) times 0.745 (single-family owner mortgage market share). Based on this calculation, if the 16.7 point differential reflected actual market conditions, then the underserved areas market share estimated using metropolitan area data should be increased by 1.6 percentage points to account for the effects of underserved counties in non-metropolitan areas.^[60] A more conservative adjustment of 1.25 percentage points was made in Table D.13 for the 2005-2008 projection model. The non-metropolitan area issue will be discussed further in Section G.4 below, which incorporates the effects of the new 2000 Census data.

Small Loans and Manufactured Housing Loans. Excluding manufactured housing loans and small loans (less than $15,000) reduces the overall underserved area market estimates reported in Table D.13 by less than one percentage point. This is estimated as follows. First, excluding these loans reduces the unadjusted underserved areas percentage for single-family-owner mortgages in metropolitan areas by about 1.2 percentage points, based on analysis of recent home purchase environments (1995-97 and 1999 and 2000). Multiplying this 1.2 percentage point differential by the property share of single-family-owner units (74.5 percent) yields 0.9 percentage points, which serves as a proxy for the reduction in the overall underserved area market share due to dropping manufactured home loans from the market analysis. The actual reduction will be somewhat less because dropping manufactured home loans will increase the share of rental units, which increases the overall underserved areas market share, thus partially offsetting the 0.8 percent reduction. The net effect is probably a reduction of about three-quarters of a percentage point.

The small loan and manufactured housing effects can be considered separately. Dropping only manufactured housing loans would reduce the market estimates by approximately three-fourths of a percentage point. ICF argued that loans with less than $15,000 should be excluded. The impact of doing this on the market estimates would be about one-third of a percentage point. ICF also considered scenarios where one-half of manufactured loans would be dropped, as well as small loans less than $15,000. The impact of doing this on the market estimates would be three-fifths of a percentage point.

The next section discusses changes as a result of switching from 1990 to 2000 Census geography.

4. 2000-Based Underserved Area Market Shares

The above analysis has concluded that 29-34 percent would be a reasonable market range for the Geographically Targeted Goal based on past origination activity in underserved areas and on scenarios that cover a variety of economic and mortgage market conditions. That analysis, which included historical data going back to the early 1990's, necessarily used 1990 Census geography to define underserved census tracts. As explained in Appendix B, HUD will be defining underserved areas based on 2000 Census geography beginning in 2005, the first year covered by this final rule. Appendix B also explains that the number of census tracts in metropolitan areas covered by HUD's underserved area definition will increase from 21,587 tracts (based on 1990 Census) to 26,959 tracts (based on 2000 Census and OMB's respecification of metropolitan areas). This increase in the number of tracts defined as underserved means that the market estimate for the Geographically Targeted Goal will be about five percentage points higher than the 1990-based market estimate. Thus, this section provides a new range of market estimates for underserved areas defined in terms of 2000 Census data.

For the years 1999 to 2003, Table D.14a. reports the underserved areas share of the mortgage market for single-family-owner, investor (non-owner), and multifamily properties, with comparisons between 1990-based and 2000-based measures of underserved areas. HMDA data, which is the source of the mortgage data, were reported in terms of 1990 census tracts. For the years 1999 to 2002, HUD used various apportionment techniques to re-allocate 1990-based HMDA mortgage data into census tracts as defined by the 2000 Census; 2003 HMDA data were defined in terms of 2000 Census tracts, so no reallocation was required. The 1990-based underserved area market shares reported in Table D.14.a. are the same data reported earlier in Table D.12, while the 2000-based underserved area market shares result from re-allocating 1999-2002 HMDA data into 2000 Census geography. In addition, the data are defined in terms of the new OMB metropolitan area definitions.

Single-Family-Owner Loans. First, consider the market shares for single-family-owner properties in the top portion of Table D.14a. In 2002, the underserved area percentage for home purchase loans increases from 27.1 percent (1990-based) to 32.8 percent (2000-based), an increase of 5.7 percentage points; the corresponding percentages for refinance loans were 24.2 percent (1990-based) and 29.4 percent (2000-based), or an increase of 5.2 percentage points. Considering total owner loans (i.e., both home purchase and refinance owner loans), the weighted average of the “Differences” reported in Table D.14a. is 5.4 percentage points in 2002 for the conforming market. Between 1999 and 2003, 30.3 percent of mortgage originations were originated in underserved areas based on 2000 geography, compared with 25.2 percent based on 1990 geography—yielding the overall differential of 5.1 percentage points. (The unweighted 1999-2003 differential is 4.9 percent.)

The first column of Table D.14a. reports the 2000-based underserved areas share for home purchase loans for the years, 1999 to 2003. The share was about 31 percent in 1999 and 2001 and in the 32.6-33.7 percent range during 2000, 2002, and 2003. Notice that the peak share (33.7 percent) for home purchase loans occurred in the most recent year, 2003. It should be recalled that there was no need to re-apportion the 2003 data from 1990-based tracts to 2000-based tracts, as these 2003 data were already defined in terms of 2000 census geography. Whether this fact affects the various differentials between 2003 and earlier years is not clear. The years 1999 and 2000 exhibited higher underserved area shares for refinance loans than for home purchase loans; as discussed earlier, this pattern was largely, but not entirely, due to subprime refinance loans.

Single-Family Rental and Multifamily Loans. Next, consider the underserved area market shares reported for single-family rental (or non-owner) and multifamily properties in the middle and bottom portions of Table D.14a. In 2002, the underserved area percentage for home purchase investor loans increases from 42.0 percent (1990-based) to 47.7 percent (2000-based), an increase of 5.7 percentage points; the corresponding percentages for refinance loans were 45.6 percent (1990-based) and 50.8 percent (2000-based), or an increase of 5.3 percentage points. The multifamily differentials are somewhat higher at approximately 7-8 percentage points. Between 1999 and 2003, 60 percent (unweighted average) of multifamily originations were originated in underserved areas based on 2000 geography, compared with 52.6 percent based on 1990 geography.

In the 2004 proposed GSE Rule, HUD made the following 2000-based assumptions with respect to the underserved areas shares of single-family rental properties: 52.0% for Case 1 (baseline), 50.0% for Case 2, and 54.0% for Case 3. With respect to multifamily properties, the following assumptions were made with respect to underserved areas shares: 58.0% for Case 1 (baseline), 56.0% for Case 2, and 59.0% for Case 3. ICF criticized HUD's baseline assumptions (52 percent for SF investors and 58 percent for MF rentals) as being too high.^[61] ICF's best estimate was 50 percent for SF investors and 55 percent for MF rentals.^[62] Since SF rentals account for 10.6 percent of financed units, reducing the underserved area share by two percentage points from HUD's 52 percent to ICF's 50 percent would reduce the overall underserved areas goal by 0.21 percentage point. Since MF rentals account for 15.0 percent of financed units (in HUD's baseline model), reducing the underserved area share by three percentage points from HUD's 58 percent to ICF's 55 percent would reduce the overall underserved areas goal by an additional 0.45 percentage point. Thus, the combined effect of ICF's assumptions would be a 0.66 percentage point reduction in the underserved areas goal. Fannie Mae did not comment directly on this parameter other than to emphasize that HUD's Case 2 is the “most likely set of assumptions” for estimating the underserved areas share (Fannie Mae Appendix, p. I-38). HUD's Case 2 (see above) would drop the baseline underserved area share for both SF and MF by two percentage points; therefore, Fannie Mae's assumptions are similar to ICF's.

In this analysis supporting the Final Rule, HUD is retaining the same underserved areas shares for SF and MF rental properties that it used in the 2004 proposed GSE rule. HUD conducted several additional analyses that support its SF rental baseline of 52 percent and its MF rental baseline of 58 percent. These analyses are summarized below.

A report by Abt Associates ^[63] calculated 1990-based underserved areas shares using the 1995 AHS and POMS data, for (a) all SF rental properties, (b) all SF rental properties with a mortgage, (c) all SF rental properties with a conventional conforming mortgage, (d) all SF Rental properties with a new first mortgage, and (e) all SF rental properties with a new conventional conforming first mortgage. The underserved areas share for each of the groups of SF rental properties was approximately 50 percent. Adding a five percent adjustment to reflect 2000-based geography (see Table D.14a) would increase these estimates to 55 percent. While this information is dated, it is consistent with HUD's 52.0 percent baseline and its 54.0 percent assumption in Case 3. Abt Associates also reported similar data for MF rental categories (a)-(c). In this case the underserved areas share ranged from 51-54 percent; adding 7-8 percent adjustment to reflect 2000-based geography would increase these estimates to 55-62 percent, again providing support for HUD's baseline (58 percent) and Case 3 (59 percent) assumptions.^[64]

HUD had Census Bureau staff use the geocoded 2003 AHS file to calculate the distribution of the rental housing stock across served and underserved areas. This analysis, which was conducted in terms of 1990-Census geography, showed that 55.8 percent of the SF rental housing stock was located in underserved areas, as was 51.4 percent of the MF rental housing stock. Adding a five (7-8) percent adjustment to reflect 2000-based geography would increase these SF (MF) rental estimates to 60.8 (58.4-59.4) percent.

HUD also had Census Bureau staff use the geocoded, 2001 Residential Finance Survey (RFS) to calculate the distribution of rental mortgages and financed units across served and underserved areas. (See Table D.14b.) Unlike the AHS analysis mentioned above, this analysis was conducted in terms of 2000 Census geography. In 2001, 54.1 percent of newly-mortgaged SF rental units were located in underserved areas, as were 61.5 of newly mortgaged MF rental units. Similar underserved area percentages were obtained for SF investor and MF loans that were originated in 1999 and 2000 and still surviving at the time of the RFS survey in 2001. ^[65]

Finally, HUD examined the GSEs' own data. Between 1999 and 2003, 58 percent of the SF rental units financed by GSE purchases were located in underserved areas. Between 1999 and 2002, 57 percent of the multifamily units financed by GSE purchases were located in underserved areas.

Based on the above analyses, HUD retained the assumptions from the 2004 GSE proposed rule concerning underserved areas location of SF and MF rental properties. Specifically, the baseline underserved area share for SF rental units is 52 percent and that for MF rental units is 58 percent.

2000-Based Underserved Area Market Estimates. Table D.15 reports the results of the projection model assuming 2000 geography. Since Table D.15 has the same interpretation as Table D.13, there is no need for a detailed explanation of it. Considering a 15.0-percent MF mix and a 8.5-percent investor mortgage share, the market share estimate is 36.9 percent if the overall (both home purchase and refinance) single-family-owner percentage for underserved areas is 31 percent, which is the estimated 1994-2003 HMDA average as well as the recent 1999-2003 HMDA average.^[66]

The above results are based on averages across both home purchase and heavy refinance environments. The analysis can also be conducted in terms of home purchase environments, focusing on the underserved area percentages for home purchase loans reported in the first column of Table D.15. Again, considering a 15.0-percent MF mix and a 8.5-percent investor mortgage share, the underserved area market estimates reported in Table D.15 are: 37.8 percent if the SFO owner underserved area percentage is 32.3 percent (1999-2003 average home purchase share); ^[67] 37.6 if the SF owner percentage is 31.8 percent (estimated average home purchase share from 1994-2003); 36.9 percent if the owner percentage is 31 percent (approximate home purchase share in 1999 and 2001); 38.0 percent if the owner percentage is 32.5 percent (approximate home purchase percentage in 2000 and 2002); and 39.0 percent if the owner percentage is 33.7 percent (home purchase percentage in 2003). This analysis assumes that the underserved areas share of refinance loans is the same as those listed above for home purchase loans. But, as Table D.14a shows, in recent home purchase environments, the underserved areas share of refinance loans has been higher than that for home purchase loans, largely but not totally due to subprime refinance loans (see earlier discussion). In the year 2000, for example, the overall underserved areas share for SFO owner loans reached 34.2 percent; in this case, the market estimate is 39.4 percent in this case. However, the next highest overall (both home purchase and refinance loans) owner share is the 31.9 percent share in 1999, which yields at an overall market estimate of approximately 37.5 percent.

Fannie Mae reports its estimates of the 2000-Census-based underserved areas market in Table I.13 on page I-40. For SFO percentages of 30 percent and 32 percent (obtained by adding five percentage points to Fannie Mae's 1990-Census-based SFO percentages of 25 percent and 27 percent, respectively), Fannie Mae projects underserved area market shares of 35.1 percent and 36.8 percent, respectively. (It is interesting that these are the exact same market shares projected by HUD in Table D.15 for the “Fannie Mae assumptions” of 12.2-percent MF mix and an 8.0-percent investor mortgage share—suggesting that Fannie Mae's model produces the same results as HUD's model when the input assumptions are the same.) Fannie Mae concluded that the higher 36.8 percent market share was not appropriate because the SFO percentage of 32 percent was too high. However, as shown in Table D.14a, the 2000-based underserved area percentage for SFO home loans was greater than 32 percent in 2000, 2002, and 2003.

Multifamily Mix. As discussed earlier, compared with the low-mod and special affordable market estimates, the underserved area market estimates exhibit less variation as one moves from a 13.5 percent MF mix to 16.0 percent MF mix. For example, reducing the assumed multifamily mix from 16.0 percent to 13.5 percent reduces the overall market projection for underserved areas by only 0.6-0.7 percentage points. This smaller MF mix effect occurs because the underserved area differentials between owner and rental properties are not as large as the low- and moderate-income and special affordable differentials reported earlier. For example, the 1999-2003 average SF-owner underserved areas share (30.3 percent in Table D.14a) is only 22 percentage points less than the baseline SF-Rental underserved areas share (52.0); on the other hand, the 1999-2003 average SF-owner special affordable share (15.7 percent) is about 42 percentage points less than the baseline SF-Rental special affordable share (58.0 percent).

As shown in Table D.15, ICF's MF mix of 14.25 percent produces results intermediate between HUD's 13.5 percent and 15.0 percent. Estimates of the underserved areas based on a MF mix of 14.2 percent are only 0.2 percentage points less than those based on a MF mix of 15.0 percent.

Investor Mortgage Share. Similarly, the market estimates differ only slightly with changes the investor mortgage share. Reducing the investor mix from 9.5 percent to 8.0 percent reduces the overall market projection for underserved areas by only 0.2-0.4 percentage points. If the 10.0 percent baseline from the 2004 proposed GSE rule were used in this analysis, the market estimates would be approximately 0.3 (0.2) percentage points higher relative to the results reported in Table D.15 for a baseline of 8.5 (9.0) percent. Fannie Mae's model combined a MF mix of 12.3 percent with an investor mortgage share of 8.0 percent. If the underserved area share of home purchase loans is 32.3 percent (the average for 1999-2003), then the estimate for the overall underserved areas market is 37.0 percent based on Fannie Mae's assumptions. In contrast, HUD's estimates (with a MF mix of 15.0 percent and 8.5 percent investor share) are 37.8 percent—almost one percentage point higher. If the underserved areas share of home purchase loans is at its 2003 level (33.7 percent), then Fannie Mae's assumptions result in a market estimate of 38.3 percent while HUD's assumptions (see previous sentence) result in a market estimate of 39.0 percent. In its projection model, ICF assumed an underserved areas share of 31.5 percent for SF owner loans and produced an estimate of almost 37 percent for the overall underserved areas market during 2005-2008 (ICF Appendix, p.133).

Different Underserved Area Shares for Rental Properties. Case 2 (see Table D.9) considered slightly smaller underserved area percentages for rental properties (50 percent for SF rentals and 56 percent for MF rentals), as compared with the baseline Case 1, which assumed 52 percent and 58 percent, respectively. Case 2 includes ICF's assumption (50 percent) for SF Rentals and is close to ICF's assumption (55 percent) for MF Rentals. Incorporating these Case 2 assumptions reduces the underserved areas market estimate by only 0.5 percentage points. For example, if the SFO home purchase share is 33 percent, then the overall underserved area estimate is 37.9 percent under Case 2, as compared with 38.4 percent under Case 1 (see Table D.15). As discussed earlier, the baseline Case 1 assumptions offer a reasonable approach for estimating the underserved area market shares.

Examples of Home Purchase Years. The above projection results for a home purchase environment can be compared with actual results for two home purchase years, 1999 and 2000 (see earlier description of these two years in the low-mod section, F.3.a). For 1999, the baseline model assumed a multifamily mix of 16.0 percent (see Section C) and a mortgage investor share of 8.2 percent (see Section D). Under these assumptions, the projected 1999 market estimate (based on 2000-Census data) is 37.6 percent; lowering the MF mix to 15.0 (14.0) percent instead of 16.0 percent reduces the estimate only slightly to 37.3 (36.9) percent. For 2000, the baseline model assumed a multifamily mix of 17.2 percent and a mortgage investor share of 9.1 percent. Under these assumptions, the 2000 underserved areas market is estimated to be 39.7 percent. A lower MF mix—for example, 16.0 (15.0) percent instead of 17.2 percent—would reduce the estimated 2000 underserved areas market share slightly to 39.4 (39.2) percent.^[68]

For 1999, the 2000-based underserved area estimate (37.6 percent) is 4.8 percentage points greater than the earlier-reported 1990 based estimate (32.8 percent); for the year 2000, the differential is 5.0 percentage points (39.7 versus 34.7). This approximately five percentage point differential can be used to obtain estimates of 2000-based underserved area shares for the earlier home purchase years, 1995 to 1997. Table D.9 of the proposed GSE rule reported 1990-based underserved area shares of 33.9 percent for 1995 and 1997 and 33.4 percent for 1996. These estimates, after adjustments for a lower HMDA-based mortgage investor share and a lower-than-baseline MF mix, would remain in the 32-33 percent range. Adding five percentage points would place these estimates in the 37-38 percent range in terms of 2000 Census geography.^[69] ICF's best estimates were approximately 37 percent for 1994-1997 and 39 percent for 1999 (ICF Appendix, p. 77); its lower bound estimates were approximately 34 percent during 1994-1997 and 1999, and 37 percent in 2000 (ICF Appendix, p.82). As noted earlier, ICF fills its report with numerous minimums that often appear unbelievable, such as the 32.8 percent projection for the overall underserved market in 2000 (ICF Appendix, p. 83), a time when the SF owner underserved areas percentage was 35.7 percent itself (see Table 14a)—in this case, the rental portion of the market was below the underserved share for owners, rather than the typical case where the rental portion is more “goals rich” than the owner portion.

Market Volatility. Additional sensitivity analyses were conducted to reflect the volatility of the economy and mortgage market. Recession and high interest rate scenarios assumed a significant drop in the underserved area percentage for single-family-owner mortgages. The single-family-owner home purchase percentage can go as low as 29 percent—which is almost 2.8 percentage points lower than the 1994-2003 average of 31.8 percent, 3.3 percentage points lower than the 1999-2003 average of 32.3 percent, and 4.7 percentage points lower than the underserved areas share of home purchase loans in 2003—and the estimated market share for underserved areas remains about 35 percent. In a more severe case, the overall underserved market share would be 33-34 percent if the single-family-owner home purchase share fell to 27 percent (its 1992 level), which is 5.3 percentage points lower than its 1999-2002 average.

Table D.11 shows the impact on the underserved areas market share under different assumptions about a refinancing environment. See the earlier discussion of the low-mod goal in Section F.2b for an explanation of the various model assumptions necessary to simulate a heavy refinance environment. The discussion focuses on the 65-percent refinance rate since that has characterized recent refinance waves. With respect to the underserved area characteristics of SF owner loans, two scenarios were considered: (A) Scenario A represents the average underserved area percentages for the last four refinance years (1998, 2001, 2002, and 2003)—32 percent for home purchase loans and 30 percent for refinance loans; and (B) Scenario B represents the average underserved percentages for the two most recent refinance years (2002, and 2003)—33 percent for home purchase loans and 29 percent for refinance loans. Thus, there is a 2-4 percentage point differential between home purchase loans and refinance loans in a heavy refinancing environment.

Under Scenario A, the underserved areas market shares varied by almost two percentage points (i.e., 1.6 percent), from 36.0 percent with a 12 percent MF mix to 34.4 percent with a 6 percent MF mix. These underserved area market shares are 3-5 percentage points lower than the underserved areas shares reported in Table D.15 for HUD's baseline home purchase environment. (The results were similar for Scenario B.) Notice that under Scenario A, the underserved areas share remains in the 34-35 percent range even if the MF mix falls to 6-8 percent. In addition to higher-income borrowers dominating the single-family market, the share of the “goals rich” rental market declines in a refinancing wave, which tends to further reduce the underserved areas share of market activity. The right-hand column of Table D.11 shows that the rental share falls to the 17-22 percent range, or 4-9 percentage points less than the almost 26-percent rental share in HUD's baseline model. This contributes to the underserved areas share of the market typically falling to 34-36 percent during a heavy refinancing period.

Model estimates were also made for the recent refinancing years of 2001, 2002, and 2003. For 2001, the baseline model assumed a multifamily mix of 13.5 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2001 market estimate is 36.9 percent.^[70] If the MF mix for 2001 had been 12.5 (12.0) percent, then the estimated underserved areas market share for 2001 would be 36.6 (36.4) percent. For 2002, the baseline model assumed a multifamily mix of slightly over 11.0 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2002 underserved areas market is estimated to be 36.2 percent.^[71] A lower MF mix—for example, 10.5 (9.5) percent instead of 11 percent—would reduce the estimated 2002 underserved areas market share to 36.0 (35.7) percent. ICF's best estimates for 1998, 2001, and 2002 were in the 34-35 percent range while its lower-bound estimates were in the 32-33 percent range.^[72]

As noted in Section F.3.b, HUD did not receive 2003 HMDA data until early August 2004 and therefore HUD has not been able to develop a complete projection model for 2003. Still, some rough projections for 2003 are provided here for different assumptions about the MF mix, recognizing that firm data on the 2003 multifamily market are not available. Combining an investor mortgage share of 8.2 from HMDA with different MF mixes produces the following estimates of the underserved areas market for 2003: 35.1 percent (MF mix of 8 percent); 34.7 percent (MF mix of 7 percent); and 34.4 percent (MF mix of 6.0 percent).

As shown by both the simulation results in Table D.10 and by the actual experience during 2001-2003, the underserved area share declines when refinances dominate the mortgage market. The above estimates suggest that the underserved areas share will not likely fall below 35 percent, although, as noted above, the estimates for 2003 (around 35 percent) are somewhat speculative.

Similar to 1999 and 2000, the 2001 and 2002 differences between the 1990-based and 2000-based underserved area market estimates are about five percentage points. For 2001, the 2000-based baseline estimate (36.9 percent) is 5.0 percentage points greater than the earlier-reported 1990 based estimate of 31.9 percent); for the year 2002, the differential is 4.9 percentage points (36.2 versus 31.3).^[73]

The analysis in this section suggests that a reasonable range for the overall market share for underserved areas based on 2000 geography might be 35-39 percent, which is consistent with the 30-34 percent range estimated earlier based on 1990-based geography.

Feasibility of Underserved Areas Goal in a Period of Heavy Refinancing. HUD received a number of public comments seeking a regulatory solution to the issue of the ability of the GSEs to meet the housing goals during a period when refinances of home mortgages constitute an unusually large share of the mortgage market. As explained in the Preamble, HUD is not addressing the refinance issue in this final rule. Elsewhere in the Federal Register, HUD is publishing an Advance Notice of Proposed Rulemaking that advises the public of HUD's intention to consider by separate rulemaking a provision that recognizes and takes into consideration the impact of high volumes of refinance transactions on the GSEs' ability to achieve the housing goals in certain years, and solicits proposals on how such a provision should be structured and implemented. HUD believes that it would benefit from further consideration and additional public input on this issue. HUD also notes that FHEFSSA provides a mechanism by which HUD can take into consideration market and economic conditions that may make the achievement of housing goals infeasible in a given year. (See 12 U.S.C. 1336(b)(e).)

B&C Loans. The procedure for dropping B&C loans from the projections is the same as described in Section F.3.c for the Low- and Moderate-Income Goal. The underserved areas percentage for B&C loans is 52.0 percent, which is larger than the projected percentages for the overall market given in Table D.15. Thus, dropping B&C loans (as well as all subprime loans) will appreciably reduce the overall market estimates. Consider the case of a single-family-owner percentage of 32 percent, which yields an overall market estimate for the underserved areas of 38.6 percent if B&C loans are included in the analysis. Dropping B&C loans from the projection model reduces the market share by one percentage point to 37.6 percent, as reported in Table D.15. Dropping all subprime loans (A-minus as well as B&C) would reduce the underserved areas market projection to 37.4 percent.

Non-metropolitan Areas. As explained in Section G.3, in order to account for the much larger coverage of underserved areas in non-metropolitan areas, 1.25 percent was added to the market share based on metropolitan area data, in order to arrive at a nationwide estimate of the market share for underserved areas. According to HMDA, underserved counties accounted for 42.7 percent of single-family-owner mortgages originated in non-metropolitan areas during the 1999-to-2002 period, based on 1990 geography. With 2000 geography and the new tract-based definition of underserved areas in non-metropolitan areas, the market share falls by 2.3 percentage points to 39.6 percent. This 2000-based underserved areas percentage of 39.6 percent for non-metropolitan areas is about eight percentage points less than the comparable percentage for metropolitan areas.^[74] This eight-point differential is lower than the 16-point differential used in the earlier 1990-based Census analysis. Assuming that non-metropolitan areas account for 13 percent of all single-family-owner mortgages and estimating that the single-family-owner market accounts for 74.5 percent of newly-mortgaged dwelling units, then the non-metropolitan underserved area differential of 8 percent would raise the overall market estimate by 0.78 percentage point—8 percentage points times 0.13 (non-metropolitan area mortgage market share) times 0.745 (single-family owner mortgage market share). Based on this calculation, if the 8 point differential reflected actual market conditions, then the underserved areas market share estimated using metropolitan area data should be increased by 0.78 percentage point to account for the effects of underserved counties in non-metropolitan areas, based on 2000 geography. A more conservative adjustment of 0.65 percentage points was made in Table D.15, which reports the results of the projection model.

Section G.3 reported that excluding manufactured housing loans (as well as small loans less than $15,000) reduced the overall underserved area market estimates based on 1990 geography by less than one percentage point (roughly three-quarters of a percentage point). Excluding manufactured housing loans leads to a similar reduction for the market estimates based on 2000 geography. As reported earlier, the small loan and manufactured housing effects can be considered separately. Dropping only manufactured housing loans would reduce the market estimates by approximately three-fourths of a percentage point. ICF argued that loans with less than $15,000 should be excluded. The impact of doing this on the market estimates would be about one-third of a percentage point. ICF also considered scenarios where one-half of manufactured loans would be dropped, as well as small loans less than $15,000. The impact of doing this on the market estimates would be three-fifths of a percentage point.

The above analyses of the effects of less affordable market conditions, different assumptions about the size of the rental market, and dropping different categories of loans from the market definition suggest that the 35-39 percent range described earlier is a reasonable range for the market estimate for underserved areas based on the projection model described earlier. This range incorporates market affordability conditions that are more adverse than have existed recently and it excludes B&C loans from the market estimates.

5. Conclusions

Based on the above findings as well as numerous sensitivity analyses, HUD concludes that 35-39 percent is a reasonable estimate of mortgage market originations that would qualify toward achievement of the Geographically Targeted Goal if purchased by a GSE. The 35-39 percent range is higher than the market range in the 2000 Rule mainly because it is based on 2000 Census geography which includes more underserved census tracts than 1990 Census geography. HUD recognizes that shifts in economic and housing market conditions could affect the size of this market; however, the market estimate allows for the possibility that adverse economic conditions can make housing less affordable than it has been in the last few years. In addition, the market estimate incorporates a range of assumptions about the size of the multifamily market and excludes B&C loans.

H. Size of the Conventional Conforming Market for the Special Affordable Housing Goal

This section presents estimates of the conventional conforming mortgage market for the Special Affordable Housing Goal. The special affordable market consists of owner and rental dwelling units which are occupied by, or affordable to: (a) Very-low-income families; or (b) low-income families in low-income census tracts; or (c) low-income families in multifamily projects that meet minimum income thresholds patterned on the low-income housing tax credit (LIHTC).^[75] HUD estimates that the special affordable market is 23-27 percent of the conventional conforming market.

HUD is proposing to establish each GSE's special affordable multifamily subgoal as 1.0 percent of its average annual dollar volume of total (single-family and multifamily) mortgage purchases over the 2000-2002 period. In dollar terms, the Department's proposal is $5.49 billion per year in special affordable multifamily purchases for Fannie Mae, and $3.92 billion for Freddie Mac. The multifamily special affordable goal, as well as the special affordable home purchase subgoal, are discussed further in Appendix C.

Section F described HUD's methodology for estimating the size of the low- and moderate-income market. Essentially the same methodology is employed here except that the focus is on the very-low-income market (0-60 percent of Area Median Income) and that portion of the low-income market (60-80 percent of Area Median Income) that is located in low-income census tracts. Data are not available to estimate the number of renters with incomes between 60 and 80 percent of Area Median Income who live in projects that meet the tax credit thresholds. Thus, this part of the Special Affordable Housing Goal is not included in the market estimate.

1. Special Affordable Shares by Property Type

The basic approach involves estimating for each property type the share of dwelling units financed by mortgages that are occupied by very-low-income families or by low-income families living in low-income areas. HUD combined mortgage information from HMDA, the American Housing Survey, the Property Owners and Managers Survey and the recently released 2001 Residential Finance Survey in order to estimate these special affordable shares.

a. Special Affordable Owner Percentages

HMDA data for the percentage of single-family-owners that qualify for the Special Affordable Goal are reported in Table D.16. That table also reports data for the two components of the Special Affordable Goal—very-low-income borrowers and low-income borrowers living in low-income census tracts. Focusing first on home purchase loans, HMDA data show that the special affordable share of the market has followed a pattern similar to that discussed earlier for the low- and moderate-income loans. The percentage of special affordable borrowers increased significantly between 1992 and 1994, from 10.4 percent of the conforming market in 1992 to 12.6 percent in 1993, and then to 14.1 percent in 1994. Between 1995 and 1998, the special affordable market was in the 14-16 percent range, averaging 15.1 percent. Over the past five years (1999-2003), the special affordable share of the home purchase loans has averaged 16.4 percent. It was about 17 percent during 1999 and 2000 and 16 percent during the most recent three years, 2001 to 2003.

Considering all (home purchase and refinance) loans during recent “home purchase” environments, the special affordable share averaged 18.7 percent during 1999-2000, over three percentage points more than the 15.4 percent average between 1995 and 1997. Excluding B&C (all subprime) loans from the analysis reduces this differential only slightly to 2.8 (2.4) percentage points. As mentioned earlier, lending patterns could change with sharp changes in the economy, but the fact that there have been several years of strong affordable lending suggests that the special affordable market has changed in fundamental ways from the mortgage market of the early 1990s.

Except for the four years of heavy refinancing (1998, 2001, 2002, and 2003), the special affordable share of the refinance market has recently been higher than the special affordable share of the home purchase market—a pattern discussed in Section F for low-mod and very-low-income loans. During 1999 (2000), for example, the special affordable share of the refinance market was 19.2 (22.6) percent, compared with 17.3 (16.9) percent for the home loan market. The higher special affordable percentages for refinance loans are reduced or even eliminated if subprime loans are excluded from the analysis. As shown in Table D.16, excluding B&C loans from the data practically eliminates the refinance-home-purchase differential for 1999 and reduces the differential for 2000 to 4.2 percentage points (from 5.7 percentage points). Going further and excluding A-minus loans from the year 2000 data would reduce the differential to 2.7 percentage points. HUD's projection model excludes B&C loans and sensitivity analyses will show the effects on the overall special affordable market of excluding all single-family subprime loans.

New 2000-Based Census Geography and New OMB Metropolitan Area Definitions. Going forward, HUD will be re-benchmarking its median incomes for metropolitan areas and non-metropolitan counties based on 2000 Census incomes, will be defining low-income census tracts (which are included in the definition of special affordable) in terms of the 2000 Census geography, and will be incorporating the effects of the new OMB metropolitan area definitions. As discussed earlier in Section F, HUD projected the effects of these three changes on the special affordable shares of the market for the years 1999-2003; the results for special affordable loans are reported in the top portion of Table D.8b. Under the historical MSA-based data, the (unweighted) average special affordable share of the conventional conforming market was 16.4 (16.3) percent for home purchase (total) loans (see Table D.16); the corresponding average with the CBSA-based projected data was 16.4 (16.4) percent, or practically the same. Given these small differences there is no need to adjust the overall market estimates reported below to account for the new data. However, it should be noted that the most recent year of 2003 does show a rather larger difference—the special affordable share of home purchase loans under the projected CBSA approach is 16.9 percent, which is a full percentage point higher than the special affordable share of 15.9 percent under historical data.^[76]

For the other two property types (single-family rental and multifamily), comparisons between projected and historical special affordable percentages were made using the GSEs' data. For single-family rental mortgages, the weighted average of Fannie Mae's (Freddie Mac's) special affordable percentage for the years 1999 to 2003 was 48.2 (48.7) percent using the historical data, compared with 49.6 (49.5) percent using the projected data. For multifamily mortgages, the weighted average of Fannie Mae's (Freddie Mac's) special affordable percentage for the years 1999 to 2003 was 50.9 (48.7) percent using historical data, compared with 51.6 (51.5) percent using the projected data. These comparisons suggest little difference between the historical and projected special affordable shares for rental properties. HUD also projected the overall special affordable percentage for each GSE. For the overall special affordable goal (considering all three property types), the unweighted average of Fannie Mae's (Freddie Mac's) special affordable percentage for the years 1999 to 2002 was 20.0 (18.9) percent using the projected data, compared with 20.0 (18.9) percent using the historical data. There is little difference in the GSEs' average special affordable performance between the projected and historical data.

b. Very-Low-Income Rental Percentages

Table D.14 in Appendix D of the 2000 Rule reported the percentages of the single-family rental and multifamily stock affordable to very-low-income families. According to the AHS, 59 percent of single-family units and 53 percent of multifamily units were affordable to very-low-income families in 1997. The corresponding average values for the AHS's six surveys between 1985 and 1997 were 58 percent and 47 percent, respectively. As discussed earlier in Section F, an important issue concerns whether rent data based on the existing rental stock from the AHS can be used to proxy rents of newly mortgaged rental units. HUD's analysis of POMS data during the 2000 rule-making process suggested that it could—estimates from POMS of the rent affordability of newly-mortgaged rental properties are quite consistent with the AHS data on the affordability of the rental stock. Fifty-six (56) percent of single-family rental properties with new mortgages between 1993 and 1995 were affordable to very-low-income families, as were 51 percent of newly-mortgaged multifamily properties. These percentages for newly-mortgaged properties from the POMS are similar to those reported above from the AHS for the rental stock. Based on this POMS analysis, HUD's baseline model in the 2004 proposed GSE rule assumed that 50 percent of newly-mortgaged, single-family rental units, and 47 percent of multifamily units, were affordable to very-low-income families. (See further discussion of this issue in Section H.1.d)

c. Low-Income Renters in Low-Income Areas

HMDA does not provide data on low-income renters living in low-income census tracts. As a substitute, HUD used the POMS and AHS data. As explained in the 2000 GSE Rule, the share of single-family and multifamily rental units affordable to low-income renters at 60-80 percent of area median income (AMI) and located in low-income tracts was calculated using the internal Census Bureau AHS and POMS data files.^[77] The POMS data showed that 8.3 percent of the 1995 single-family rental stock, and 9.3 percent of single-family rental units receiving financing between 1993 and 1995, were affordable at the 60-80 percent level and were located in low-income census tracts. The POMS data also showed that 12.4 percent of the 1995 multifamily stock, and 13.5 percent of the multifamily units receiving financing between 1993 and 1995, were affordable at the 60-80 percent level and located in low-income census tracts.^[78] The baseline analysis in HUD's proposed GSE rule assumed that 8 percent of the single-family rental units and 11.0 percent of multifamily units are affordable at 60-80 percent of AMI and located in low-income areas.

Combining the assumed very-low-income percentage of 50 percent (47 percent) for single-family rental (multifamily) units with the assumed low-income-in-low-income-area percentage of 8 percent (11 percent) for single-family rental (multifamily) units yields the special affordable percentage of 58 percent (58 percent) for single-family rental (multifamily) units. This was the baseline case in the 2004 proposed GSE rule.

d. Comments on the Special Affordable Rental Share and Additional Analysis

Both ICF and Fannie Mae commented that HUD overstated the special affordable share of the single-family rental and multifamily rental markets. They argued that updated 2001 AHS data showed that the affordability of the rental housing stock had declined since HUD had conducted its POMS and AHS analyses in 1995 and 1997, respectively. For both single-family (SF) and multifamily (MF) rentals, ICF used a special affordable range of 47-53 percent, with a baseline of 50 percent. ICF's special affordable range is much less than both HUD's 53-61 percent range (58 percent baseline) for single-family rentals and HUD's 54-62 percent range for multifamily rentals (also a 58 percent baseline). Since SF and MF rentals account for about 25 percent of financed units in HUD's model, reducing the SF and MF baselines from 58 percent (HUD's baseline) to 50 percent (ICF's baseline) would reduce the overall special affordable market estimate by two percentage points. Thus, this is an important issue.

Based on its analysis of the AHS (see Fannie Mae Appendix, I-31-I-32), Fannie Mae concluded that the very-low-income share for single-family rental properties had fallen from 58.3 percent in 1997 to 53.0 percent in 2001; similarly, the very-low-income (VLI) share of multifamily rental properties had fallen from 52.0 percent to 44.9 percent over this same period. (By comparison, ICF estimated that 47 percent of the SF rental stock and 42 percent of the MF rental stock were affordable to VLI families.) In its analysis, Fannie Mae provides a weight of 0.07 to the VLI share (25.7 percent) of recently-constructed single-family rental units in the AHS, and the residual 0.93 weight to the VLI share (53.6 percent) of the remaining existing units in the AHS. While Fannie Mae uses a VLI share of 46 percent for single-family rentals in its market sizing models, applying these weights to the 2001 AHS data (reported by Fannie Mae in Table I.7 on p. I-32) yields approximately 52 percent for the VLI share of single-family rental properties. Similarly, for multifamily properties, Fannie Mae provides a weight of 0.11 to the VLI share (22.2 percent) of recently-constructed multifamily rental units in the AHS, and the residual 0.89 weight to the VLI share (45.7 percent) of the remaining existing units in the AHS. In this case, applying the above weights to the 2001 AHS data yields 43 percent for the VLI share of multifamily rental properties—a figure similar to the 41-percent VLI share that Fannie Mae uses in its market sizing models. After computing a VLI share of 46 percent for SF rentals, Fannie Mae adds 8 percent to account for low-income renters living in low-income census tracts (the second component of the special affordable category); this yields 54 percent for the special affordable share of SF rentals. After computing a VLI share of 41 percent for MF rentals, Fannie Mae adds 11 percent to account for low-income renters living in low-income census tracts; this yields 52 percent for the special affordable share of MF rentals. Thus, Fannie Mae's estimates are intermediate between ICF's (50 percent) and HUD's (58 percent). Since SF rentals account for 10.6 percent of financed units in HUD's model, reducing the SF baseline from 58 percent (HUD's baseline) to 54 percent (Fannie Mae's baseline) would reduce the overall special affordable market estimate by 0.42 percentage points. Since MF rentals account for 15.0 percent of financed units in HUD's model, reducing the MF baseline from 58 percent (HUD's baseline) to 52 percent (Fannie Mae's baseline) would reduce the overall special affordable market estimate by 0.90 percentage points. Combining these two reductions yields a 1.32 percentage point reduction in the overall special affordable market.

HUD is retaining its baseline of 58 percent for the special affordable share of both SF and MF rentals. Several sets of analyses led to this decision.

HUD updated its analysis with 2001 and 2003 AHS data. Using ICF's assumptions for an AHS analysis (see ICF Appendix, p. 45), the 2003 AHS data showed that 57 percent (67 percent) of single-family (multifamily) rental units would qualify as being affordable to VLI families. This analysis of the 2003 AHS used a new geocoded file that identified the specific metropolitan area or county location for each observation in the AHS. This allowed HUD to link accurate area median incomes (used to determine affordability) to each AHS observation, which represents a substantial improvement over previous AHS analyses that did not have the specific household location and thus had to rely on estimates of area median income in order to compute affordability ratios. This more accurate approach appears to produce higher affordability estimates than earlier analyses based on the non-geocoded AHS.

To derive an overall special affordable percentage, one must add the second component of the special affordable category—low-income renters living in low-income areas—to the VLI share. HUD's analysis of POMS data and its analysis of 2003 AHS geocoded data suggest that low-income SF renters in low-income areas account for 22 percent of all SF low-income renters; GSE data for 2001 and 2002 suggest a slightly higher percentage.^[79] With respect to MF properties, HUD's analysis of POMS data and its analysis of 2003 AHS geocoded data suggest that low-income MF renters in low-income areas accounted for 24-25 percent of all MF low-income renters; GSE data for 2001 and 2002 suggest a slightly lower percentage (21 percent).These shares can be applied to the 2003 AHS results for low-income renters. For SF rentals, the 22 percent share for low-income renters living in low-income census tracts can be multiplied by the 20 percent figure that the 2003 AHS produces for low-income SF renters, yielding estimate of 4.4 percent. This 4.4 percent is added to the VLI percentage of 67 percent for SF rentals to arrive at a special affordable estimate of 71 percent, based on the 2003 AHS. For MF rentals, the 25 percent share for low-income renters living in low-income census tracts can be multiplied by the 27 percent figure that the 2003 AHS produces for low-income MF renters, yielding an estimate of 6.7 percent.^[80] This 6.7 percent is added to the VLI percentage of 57 percent for MF rentals to arrive at a special affordable estimate of 63 percent, based on the 2003 AHS. These 2003 AHS special affordable shares—67 percent for SF rental units and 63 percent for MF rental units—support HUD's use of a 58-percent baseline as the special affordable share of both SF and MF rental properties.

It is interesting to compare HUD's 58-percent baseline with the actual performance of Fannie Mae and Freddie Mac. For single-family rental mortgages, the weighted average of both Fannie Mae's and Freddie Mac's special affordable percentage for the years 1999 to 2003 was about 50 percent using projected CBSA data. For multifamily mortgages, the weighted average of Fannie Mae's special affordable percentage for the same years was 49 percent, while Freddie Mac's percentage was 52 percent. As ICF notes, the GSEs' below market performance may be due to their limited participation in the small multifamily market (ICF Appendix, p. 47).

2. Size of the Special Affordable Market

The size of the special affordable market depends in large part on the size of the single-family rental and multifamily markets and on the special affordable percentages of both owners and renters. Therefore, this section conducts several sensitivity analyses around these market parameters. As in the previous sections, this section initially assumes a refinance rate of 35 percent, which means that it initially focuses on a home purchase or low-refinancing environments. After presenting these results, market estimates reflecting a heavy refinance environment will be presented. In the 2000 GSE Rule, HUD assumed that the special affordable share of refinance loans was 1.4 percentage points lower than the special affordable share of borrowers purchasing a home. However, as discussed earlier, the special affordable share of refinance loans equaled or was greater than the special affordable share of home purchase loans during home purchase environments such as 1995-97 or 1999-2000; thus, the assumption of a lower special affordable share for refinance loans is initially dropped from the analysis but will be reintroduced during the sensitivity analysis and the discussion of heavy refinancing environments. If the special affordable share of refinance loans were assumed to be one percentage point less than that of home purchase loans, then the market shares in Table D.17 would be approximately one-quarter percentage point lower.^[81]

Considering a 15.0-percent MF mix and a 8.5-percent investor mortgage share, the special affordable market estimates reported in Table D.17 are: 27.3 percent if the owner percentage is 17 percent (home purchase share for 1999 and 2000); 26.8 if the owner percentage is 16.4 percent (average home purchase share from 1999-2003); 26.5 percent if the owner percentage is 16 percent (home purchase share for 1998, 2001, 2002, and 2003); and 25.7 percent if the owner percentage is 15 percent (home purchase average from 1995-97). Considering a range of 13.5-16.0 for the MF mix and a range of 8.5-9.0 for the investor mortgage share, the special affordable market estimates reported in Table D.17 are: 26.7-27.9 percent if the owner percentage is 17 percent; 26.2-27.4 percent if the owner percentage is 16.4 percent; 25.9-27.1 percent if the owner percentage is 16 percent; and 25.1-26.3 percent if the owner percentage is 15 percent.

If the special affordable percentage for home purchase loans fell to 13 percent—or by three percentage points below its 1995-2003 average level of approximately 16 percent—then the overall market estimate would be about 24 percent under the baseline assumptions. Thus, 24 percent is consistent with a rather significant decline in the special affordable share of the single-family home purchase market. A 24 percent market estimate allows for the possibility that adverse economic and housing affordability conditions could keep special affordable families out of the housing market. On the other hand, if the special affordable home purchase percentage stays at its recent levels (15-17 percent), the market estimate is in the 26-27 percent range.

Different Special Affordable Shares for Rental Properties. Case 2 (see Table D.9) considered smaller special affordable percentages for rental properties (53 percent for SF rentals and 54 percent for MF rentals), as compared with the baseline Case 1, which assumed 58 percent for both property types. Case 2 assumptions are close to Fannie Mae's assumptions—54 percent for SF Rentals and 52 for MF Rentals. Incorporating the Case 2 assumptions reduces the special affordable market estimate by 1.2 percentage points. For example, if the SFO home purchase share is 17 percent, then the overall special affordable estimate is 26.1 percent under Case 2, as compared with 27.3 percent under Case 1 (see Table D.17).

ICF's assumptions were even lower, 50 percent for both SF and MF rentals, a figure that is eight percentage points lower than HUD's baseline Case 1 assumption of 58 percent for each of these two property types. Given that these two property types account for 25 percent of all financed dwelling units, using ICF's 50-percent assumption (instead of HUD's 58-percent assumption) would reduce the overall special affordable market shares in Table D.17 by two percentage points. As discussed above, HUD's baseline Case 1 assumptions offer a reasonable approach for estimating the special affordable market shares.

Multifamily Mix. The volume of multifamily activity is also an important determinant of the size of the special market. While Section C explained the rationale for HUD's 15.0 percent range, it is useful, given the uncertainty surrounding the size of the multifamily market, to consider the effects of lower multifamily mix assumptions, even in a home purchase environment. Assuming a 13.5 percent MF mix reduces the overall special affordable market estimates by 0.4 percentage points compared with a 15 percent MF mix, and by 1.0 percentage point compared with a 16.0 percent mix. For example, when the special affordable share of the home purchase market is at 16.4 percent (its 1999-2003 average), the special affordable share of the overall market is 26.2 percent assuming a 13.5 percent multifamily mix, compared with 26.8 (27.4) percent assuming a 15 (16.0) percent multifamily mix.

As shown in Table D.17, the ICF's MF mix of 14.2 percent produces results intermediate between HUD's 13.5 percent and 15.0 percent. Estimates of the special affordable market based on a MF mix of 14.2 percent are only 0.3 percentage points less than those based on a MF mix of 15.0 percent. Fannie Mae's model combined an even lower MF mix of 12.3 percent with an investor mortgage share of 8.0 percent. If the special affordable share of home purchase loans is 16.4 percent (the 1999-2003 average), then the estimate for the overall special affordable market is 25.2 percent based on Fannie Mae's assumptions. In contrast, HUD's estimates (with a MF mix of 15.0 percent and 8.5-9.0 percent investor share) are 26.8-27.0 percent “ about one and a half percentage points higher. If the special affordable share of home purchase loans is 16 percent (its recent 2001-2003 level), then Fannie Mae's assumptions result in a market estimate of 25.2 percent while HUD's assumptions (see previous sentence) result in market estimates of 26.5-26.7 percent.

Investor Mortgage Share. As shown in Table D.17, increasing the investor mortgage share by one percentage point from 8.0 percent to 9.0 percent increases the special affordable market estimate by approximately 0.4-0.5 percentage point. If the 10.0 percent baseline from the 2004 proposed GSE rule were used in this analysis, the market estimates would be approximately 0.6 (0.4) percentage points higher relative to the results reported in Table D.15 for a baseline of 8.5 (9.0) percent.

Examples of Home Purchase Years. The above projection results for a home purchase environment can be compared with actual results for two home purchase years, 1999 and 2000, which were characterized by refinance rates of 34 percent and 29 percent, respectively. For 1999, the baseline model assumed a multifamily mix of 16.0 percent and a mortgage investor share of 8.2 percent. Under these assumptions, the 1999 market estimate is 27.9 percent; if the 1999 MF mix was lower—for example, 15.0 (14.0) percent instead of 16.0 percent—then the estimate of the 1999 special affordable market share would be 27.5 (27.2) percent.

The 2004 proposed rule (Table D.9 in Appendix D) reported a higher baseline market estimate for 1999 of 29.2 percent, as compared with the 27.9 percent reported in the previous paragraph—a differential of 1.3 percentage points. The difference is largely due to the treatment of single-family rental mortgages. For example, using the proposed rule's 10-percent assumption for the mortgage investor share (instead of the lower 8.2 percent HMDA-based mortgage investor shares reported in the text) would increase the 1999 estimate by 0.8 percentage points to 28.7 percent, only 0.5 percentage points lower than the 29.2 percent reported in the proposed rule. Other more minor changes that lower market estimate included: (a) Further reducing the SF mortgage investor share by excluding B&C investor loans from the HMDA data (see Section C); (b) using 1.6 percent (instead of 2.0 percent) for the mortgage share of single-family 2-4 property owners; and (c) using slightly lower dwelling-units-per-mortgage assumptions for SF 2-4 properties (2.20 instead of 2.25) and for SF investor mortgages (1.30 instead of 1.35). These changes, leading to this 1.3 percentage point differential, also affect the estimates reported in Table D.9 of Appendix D of the proposed rule for the three home purchase environments prior to 1999—28.9 percent for 1995, 28.7 for 1996, and 28.8 percent for 1997.^[82] Given (a)-(c) and the fact that the HMDA-reported mortgage investor share was approximately eight percent during these three years (instead of the assumed 10 percent in the earlier 1995-97 analysis), these estimates should probably be reduced by the above-mentioned 1.3 percentage points, which would place them at 27-28 percent assuming no adjustment in the baseline MF mix, and at 26-27 percent assuming a MF mix three percentage points lower than the baseline MF mix.^[83]

For 2000, the baseline model assumed a multifamily mix of 17.2 percent and a mortgage investor share of 9.1 percent. Under these assumptions, the 2000 special affordable market is estimated to be 29.1 percent. A lower MF mix—for example, 15.0 percent instead of 17.2 percent—would reduce the estimated 2000 low-mod market share to 28.2 percent.^[84]

ICF's best estimates for the special affordable market were 25-26 percent in 1995, 1997, 1999, and 2000, and a particularly low 23 percent for 1996 (ICF Appendix, p. 94). Its lower bound estimates were 22-23 percent for 1997 and 1999, 24 percent for 1995 and 2000, and 21 percent for 1996 (ICF Appendix, p. 99). As discussed earlier, two percentage points of the HUD-ICF differential involves ICF's lower assumptions about the special affordable characteristics of rental loans. Given that the SFO percentage was 18-19 percent during 1999 and 2000 (see Table D.16), ICF's 23-24 estimates for 1999 and 2000 are in need of further explanation.

Heavy Refinancing Environments. The special affordable share of the overall market declines when refinances dominate the market. Section F.3c, which presents the low-mod market estimates, explained the assumptions for incorporating a refinance environment into the basic projection model for 2005-08. Briefly, they are: the refinance share of single-family mortgages was increased to 65 percent (from 35 percent); the multifamily mix was allowed to vary from 6 to 12 percent; the market share for subprime loans was reduced to 8.5 percent (from 12 percent); and the mortgage investor share was set at 8.0 percent (its average during recent refinancing waves). With respect to MF mixes, it is likely that an 11-12 percent MF mix characterized 2001, 9-11 percent characterized 2002, and less than 7 percent characterized 2003, although there is some uncertainty with these estimates. In a refinancing wave, the special affordable percent is typically lower for refinance loans than home purchase loans, as middle- and high-income borrowers dominate the market. With respect to the special affordable characteristics of SF owner loans, the refinancing analysis assumed 16 percent for home purchase loans and 14 percent for refinance loans, which were the average special affordable percentage for the last four refinance years (1998, 2001, 2002, and 2003). There has been a two percentage point differential between home purchase loans and refinance loans during a heavy refinancing environment.

As shown in Table D.11, the special affordable shares varied by over two percentage points, from 24.1 percent with a 12 percent MF mix to 21.7 percent with a 6 percent MF mix. These special affordable market shares are 3-5 percentage points lower than the special affordable shares reported in Table D.17 for HUD's baseline home purchase environment. Notice that the special affordable share remains in the 22-23 percent range even if the MF mix falls to 6-8 percent. In addition to higher-income borrowers dominating the single-family market, the share of the “goals rich” rental market declines in a refinancing wave, which tends to further reduce the special affordable of market activity. The right-hand column of Table D.11 shows that the rental share falls to the 17-22 percent range, or 4-9 percentage points less that the almost 26-percent rental share in HUD's baseline model.

Model estimates were also made for the recent refinancing years of 1998, 2001, 2002, and 2003. For 1998, the baseline model assumed a multifamily mix of 14.0 percent and a mortgage investor share of 6.8 percent. Under these assumptions, the 1998 market estimate is 24.0 percent. If the MF mix for 1998 had been 13.0 (12.0) percent then the estimated special affordable market share for 1998 would be 23.5 (23.1) percent. For 2001, the baseline model assumed a multifamily mix of 13.5 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2001 market estimate for special affordable loans is 25.0 percent. If the MF mix for 2001 had been 12.0 percent, instead of the baseline of 13.5 percent, then the estimated special affordable market share for 2001 would be 24.4 percent. For 2002, the baseline model assumed a multifamily mix of slightly over 11.0 percent and a mortgage investor share of 7.8 percent. Under these assumptions, the 2002 special affordable market is estimated to be 24.3 percent.^[85] A lower MF mix—for example, 10.5 (9.5) percent instead of 11 percent—would reduce the estimated 2002 special affordable market share to 24.2 (23.7) percent.^{[86]  [87]}

As explained in Section F.3b, HUD has not yet completed its analysis of 2003 data. However, HUD developed some rough projections for different assumptions about the MF mix. Combining an investor mortgage share of 8.2 from HMDA with different MF mixes (ranging from 6 percent to 8 percent) produced estimates of 22.6 percent (MF mix of 6 percent) to 23.5 percent (MF of 8 percent).

As shown by both the simulation results in Table D.17 and the actual experience during 2001-2003, the special affordable share of the overall market declines when refinances dominate the market. The special affordable share was approximately 24 percent during 2001 and 2002 and 23 percent in 2003 (although there is some uncertainty with the 2003 estimate).

The various market estimates presented in Table D.17 for a home purchase environment and reported above for a refinance environment are not all equally likely. Most of them equal or exceed 23 percent. In the home purchase environment, estimates below 23 percent would require the special affordable share for home purchase loans to drop to 12 percent which would be 4 percentage points lower than the 1995-2003 average for the special affordable share of the home purchase market. As shown in Table D.11, dropping below 23 percent would be more likely in a heavy refinance environment, particularly those characterized by extremely low MF mixes of 7 percent or less.

As stated in Sections F and G above, HUD received a number of public comments seeking a regulatory solution to the issue of the ability of the GSEs to meet the housing goals during a period when refinances of home mortgages constitute an unusually large share of the mortgage market. As explained in the Preamble, HUD is not addressing the refinance issue in this final rule. Elsewhere in the Federal Register, HUD is publishing an Advance Notice of Proposed Rulemaking that advises the public of HUD's intention to consider by separate rulemaking a provision that recognizes and takes into consideration the impact of high volumes of refinance transactions on the GSEs' ability to achieve the housing goals in certain years, and solicits proposals on how such a provision should be structured and implemented. HUD believes that it would benefit from further consideration and additional public input on this issue. HUD also notes (see above) that FHEFSSA provides a mechanism by which HUD can take into consideration market and economic conditions that may make the achievement of housing goals infeasible in a given year. (See 12 U.S.C. 1336(b)(e).)

B&C Loans. The procedure for dropping B&C loans from the projections is the same as described in Section F.3.c for the Low- and Moderate-Income Goal. The special affordable percentage for B&C loans is 28.0 percent, which is similar to the projected percentages for the overall market given in Table D.17. Thus, dropping B&C loans (as well as all subprime loans) does not appreciably reduce the overall market estimates. Consider the case of a single-family-owner percentage of 16 percent, which yields an overall market estimate for Special Affordable Goal of 26.7 percent if B&C loans are included in the analysis. Dropping B&C loans from the projection model reduces the special affordable market share by 0.2 percentage points to 26.5, as reported in Table D.17. Dropping all subprime loans (A-minus as well as B&C) would reduce the special affordable market projection to 26.2 percent.

Manufactured Housing Loans and Small Loans. Excluding manufactured housing loans and small loans (loans less than $15,000) reduces the overall market estimates reported in Table D.17 by less than one percentage point. This is estimated as follows. First, excluding these loans reduces the special affordable percentage for single-family-owner mortgages in metropolitan areas by about 1.5 percentage points, based on analysis of recent home purchase environments (1995-97 and 1999 and 2000). Multiplying this 1.5 percentage point differential by the property share (0.745) of single-family-owner units yields 1.1 percentage points, which serves as a proxy for the reduction in the overall special affordable market share due to dropping manufactured home loans from the market analysis. The actual reduction will be somewhat less because dropping manufactured home loans will increase the share of rental units, which increases the overall special affordable share, thus partially offsetting the 1.1 percent reduction. The net effect is probably a reduction of three-quarters to one percentage point.

The effects can be considered separately. Dropping only manufactured housing loans would reduce the market estimates by approximately one-half of a percentage point. ICF argued that loans with less than $15,000 should be excluded. The impact of doing this on the market estimates would be about one-third to four-fifths of a percentage point. ICF also considered scenarios where one-half of manufactured loans would be dropped, as well as small loans less than $15,000. The impact of doing this on the market estimates would be three-fifths to three-quarters of a percentage point.

The above analyses of the effects of less affordable market conditions, different assumptions about the size of the rental market, and dropping different categories of loans from the market definition suggest that 23-27 percent is a reasonable range of estimates for the low- and moderate-income market. This range covers markets without B&C and allows for market environments that would be much less affordable than recent market conditions.

Tax Credit Definition. Data are not available to measure the increase in market share associated with including low-income units located in multifamily buildings that meet threshold standards for the low-income housing tax credit. Currently, the effect on GSE performance under the Special Affordable Housing Goal is rather small. For instance, adding the tax credit condition increased Fannie Mae's performance as follows: 0.42 percentage point in 1999 (from 17.20 to 17.62 percent); 0.59 percentage point in 2000 (from 18.64 to 19.23 percent); and 0.43 percent point in 2001 (from 19.29 to 19.72 percent). The increases for Freddie Mac have been lower (ranging from 0.24 to 0.38 percentage point during the same period).

3. Conclusions

Sensitivity analyses were conducted for the market shares of each property type, for the very-low-income shares of each property type, and for various assumptions in the market projection model. These analyses suggest that 23-27 percent is a reasonable estimate of the size of the conventional conforming market for the Special Affordable Housing Goal. This estimate excludes B&C loans and allows for the possibility that homeownership will not remain as affordable as it has over the past six years. In addition, the estimate covers a range of projections about the size of the multifamily market.

Footnotes

1.  Fannie Mae and Freddie Mac have both announced their intention voluntarily to register their common stock with the Securities and Exchange Commission (SEC) under section 12(g) of the Securities Exchange Act of 1934. Fannie Mae's registration became effective March 31, 2003. Freddie Mac has stated that it will complete the process of voluntarily registering its common stock once it resumes timely reporting of its financial results.

2.  “Updated Estimates of the Subsidies to the Housing GSEs,” attachment to a letter from Douglas Holtz-Eakin, Director, Congressional Budget Office, to the Honorable Richard C. Shelby, Chairman, Committee on Banking, Houseing, and Urban Affairs, United States Senate, April 8, 2004. A related recent study is Wayne Passmore, “The GSE Implicit subsidy and Value of Government Ambiguity,” Board of Governors of the Federal Reserve System. Finance and Economics Discussion Series, FEDS Working Paper 2003-64, December 2003.

3.  Fiscal Year 2005 Budget of the U.S. Government: Mid-Session review (July 30, 2004). Office of Management and Budget, also posted at http://www.whitehouse.gov/​omb. The Budget and Economic Outlook: An Update, Washington, Congressional Budget Office, September 2004, also posted on http://www.cbo.gov.

4.  Mortgage Bankers Association of America, MBA Mortgage Finance Forecast, September 17, 2004.

5.  Margery Austin Turner, All Other Things Being Equal: A Paired Testing Study of Mortgage Lending Institutions, The Urban Institute Press, April 2002. Appendix A includes further discussion of this study.

6.  These studies are discussed in section B.1 of Appendix B.

7.  Unless otherwise noted, the conventional conforming market data reported in this section exclude an estimate of B&C loans; the less-risky A-minus portion of the subprime market is included in the market definition. See section d below and Appendix D for a discussion of primary market definitions and the uncertainty surrounding estimates of the number of B&C loans in HMDA data. As noted there, B&C loans are much more likely to be refinance loans rather than home purchase loans.

8.  These percentage shares are computed from Table A.30 in Appendix A. Note that B&C loans are excluded from these data. See also Table A.31b in Appendix A.

9.  As discussed in Appendix D, the GSEs questioned HUD's historical estimates of the multifamily market as too high. Section C of Appendix D discusses these comments and responds. As indicated in Table A.30, multifamily loans accounted for 14.8 percent of all financed units in the market, excluding B&C loans. As reported in Section G of Appendix A and Sections F-H of Appendix D, HUD also conducted sensitivity analyses that reduced its 1999-2002 multifamily shares for the market by approximately two percentage points. As a result, 1999-2002 multifamily units decreased from 7,018,044 units to 5,991,036 units (reducing the multifamily share from 14.8 percent to 12.6 percent). With these reduced multifamily market numbers, the GSEs' share of the multifamily market increased from 35 percent to 41 percent. The GSEs also accounted for higher shares of the goals-qualifying multifamily market: 42 percent for low-mod units, 34 percent for underserved area units, and 37 percent for special affordable units. In this case, the GSEs' shares of the overall goals-qualifying markets (including single-family-owner, single-family-rental, and multifamily mortgages) increased as follows: low-mod—from 48 percent (see right column of Table A.30 in Appendix A) to 50 percent (see right column of Table A.31b in Appendix A); underserved areas—from 48 percent to 50 percent; and special affordable—from 41 percent to 43 percent.

10.  By way of comparison, the refinance rate was 29 percent for both Fannie Mae and Freddie Mac in 2000.

11.  “Economic value” is the net present value of the fund's reserves plus expected future cash flows, and the “capital ratio” is economic value divided by insurance-in-force.

12.  A negative credit subsidy of 3.0 percent means that the net present value of FHA's revenues (premiums, fees, recoveries from claims paid, etc.) will exceed the net present value of FHA's program costs (claims and related expenses) by 3.0 percent of the total insured mortgage amount.

13.  This is suggested by recent experience of below-average multifamily mix in years where the volume of single-family refinancings has been high. Further support is provided by evidence of a relationship between interest rates and the multifamily share of the net change in residential mortgage debt between 1975 and 2002.

14.  The Freddie Mac 2002 figures in Table 4 differ from the corresponding figures in Table 3 in HUD's Proposed Rule. Subsequent to publication of the Proposed Rule, HUD discovered that HUD had credited some units toward Freddie Mac's Low- and Moderate-Income Housing Goal in 2002 that had been previously counted toward the goal in 2001. The units were associated with a large year-end Freddie Mac mortgage purchase transaction in 2002. Because HUD's regulations prohibit double counting, HUD has recalculated Freddie Mac's 2002 Low- and Moderate-Income Housing Goal performance. The recalculation also reflects correction of some coding errors discovered in HUD's recent review.

15.  The Freddie Mac 2002 figures in Table 6 differ from the corresponding figures in Table 4 in HUD's Proposed Rule. Subsequent to publication of the Proposed Rule, HUD discovered that HUD had credited some units toward Freddie Mac's Underserved Areas Housing Goal in 2002 that had been previously counted toward the goal in 2001. The units were associated with a large year-end Freddie Mac mortgage purchase transaction in 2002. Because HUD's regulations prohibit double counting, HUD has recalculated Freddie Mac's 2002 Underserved Areas Housing Goal performance. The recalculation also reflects correction of some coding errors discovered in HUD's recent review. With the recalculation, Freddie Mac fell slightly short of its 2002 Underserved Areas Housing Goal.

16.  HUD will begin defining underserved areas based on 2000 Census geography and new OMB definitions of metropolitan areas in 2005, the first year of the proposed rule. As explained in Appendix B of the proposed GSE Rule, the 2000-based definition of underserved areas includes 5,372 more census tracts in metropolitan areas than the 1990-based definition, which means the GSE-market comparisons had to be updated to incorporate tract designations from the 2000 Census. Therefore, for the years 1999, 2000, 2001, and 2002, HUD used various apportionment techniques to re-allocate 1990-based GSE and HMDA data into census tracts as defined by the 2000 Census. (Since 2003 HMDA and GSE data were gathered in terms of 2000 Census geography, no apportionment was required for that year.) Switching to the 2000-based tracts increases the underserved area share of market originations by about five percentage points. Between 1999 and 2002, 30.3 percent of mortgage originations (without B&C loans) were originated in underserved tracts based on 2000 geography, compared with 25.2 percent based on 1990 geography. As shown in Table B.8 of Appendix B of this Final Rule, the underserved areas share of each GSE's purchases also rises by approximately five percentage points. Thus, conclusions about the GSEs' performance relative to the market are similar whether the analysis is conducted in terms of 2000 Census geography or 1990 Census geography.

17.  The Freddie Mac 2002 figures in Table 8 differ from the corresponding figures in Table 5 in HUD's Proposed Rule. Subsequent to publication of the Proposed Rule, HUD discovered that HUD had credited some units toward Freddie Mac's Special Affordable Housing Goal in 2002 that had been previously counted toward the goal in 2001. The units were associated with a large year-end Freddie Mac mortgage purchase transaction in 2002. Because HUD's regulations prohibit double counting, HUD has recalculated Freddie Mac's 2002 Special Affordable Housing Goal performance. The recalculation also reflects correction of some coding errors discovered in HUD's recent review.

18.  For rental units, the 2000 Housing Goals Final Rule also established counting rules that allow the GSEs to estimate rents or exclude units from the denominator when rent data are missing. See 24 CFR 81.15(e)(6)(i) on the rules applicable to multifamily units and 24 CFR 81.15(e)(6)(ii) on the rules for single-family rental units.

1.  Mortgage denial rates are based on 2002 HMDA data for home purchase loans; manufactured housing lenders are excluded from these comparisons.

2.  The “affordable lending performance” of Fannie Mae and Freddie Mae refers to the performance of the GSEs in funding loans for low-income and underserved borrowers through their purchase (or guarantee) of loans originated by primary lenders. It does not, of course, imply that the GSEs themselves are lenders originating loans in the primary market.

3.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2004, p. 35.

4.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2003, p. 16.

5.  According to the National Association of Realtors, Housing Market Will Change in New Millennium as Population Shifts, November 7, 1998. Forty-five percent of U.S. household wealth was in the form of home equity in 1998. Since 1968, home prices have increased each year, on average, at the rate of inflation plus two percentage points.

6.  Todd Buchholz, “Safe At Home: The New Role of Housing in the U.S. Economy,” a paper commissioned by the Homeownership Alliance, 2002.

7.  Federal Reserve Board, “Recent Changes in U.S. Family Finances: Results from the 1998 and 2001 Survey of Consumer Finances,” January 2003, p. 16.

8.  Mark Zandi, “Housing's Rising Contribution,” June 2002, p. 5.

9.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 1998.

10.  Lawrence Yun, “The Forecast,” National Association of Realtors Real Estate Outlook, February 2004, p. 4.

11.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2004, p. 15.

12.  U.S. Department of Housing and Urban Development, “Economic Benefits of Increasing Minority Homeownership,” p. 7.

13.  http://www.whitehouse.gov/​infocus/​homeownership/​. Accessed July 28, 2004.

14.  Homeownership Alliance, “The Economic Contribution of the Mortgage Refinancing Boom,” December 2002, p. 2.

15.  Homeownership Alliance, “The Economic Contribution of the Mortgage Refinancing Boom,” December 2002, p. 4-5.

16.  Robert Dietz and Donald Haurin, “The Social and Private Consequences of Homeownership,” May 2001, p. 51.

17.  William M. Rohe, George McCarthy, and Shannon Van Zandt, “The Social Benefits and Costs of Homeownereship,” May 2000, p. 31.

18.  U.S. Deparmtent of Housing and Urban Development, “Economic Beneifts of Increasing Minority Homeownership,” p. 8-9.

19.  For a dicusssion of the causes of existing disparities in homeownership, see the various articles in Nicolas P. Retsinas and Eric S. Belsky (Eds), Low-Income Homeowernsip: Examining the Unexamined Goal, Washington, DC: Brookings Institution Press, 2002.

20.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2004, p. 15.

21.  U.S. Census Bureau, Current Population Survey, March 2000.

22.  Fannie Mae, Fannie Mae National Housing Survey, 2002, p. 11.

23.  Joseph Gyourko, Peter Linneman, and Susan Wachter. “Analyzing the Relationships among Race, Wealth, and Home Ownership in America,” Journal of Housing Economics 8 (2), p. 63-89, as discussed in Thomas P. Boehm and Alan M. Schlottmann. “Housing and Wealth Accumulation: Intergenerational Impacts,” in Low-Income Homeownership: Examining the Unexamined Goal, Brookings Institution Press (2002), p. 408.

24.  Fannie Mae, Fannie Mae National Housing Survey, 2002, p. 11.

25.  See Dan Immergluck, Stark Differences: The Explosion of the Subprime Industry and Racial Hypersegmentation in Home Equity Lending, Woodstock Institute, October 2000; and Daniel Immergluck and Marti Wiles, Two Steps Back: The Dual Mortgage Market, Predatory Lending, and the Undoing of Community Development, Woodstock Institute, Chicago, IL, November 1999. For a national analyses, see the HUD report Unequal Burden: Income and Racial Disparities in Subprime Lending in America, April 2000; and Randall M. Scheessele, Black and White Disparities in Subprime Mortgage Refinance Lending, Housing Finance Working Paper No. HF-114, Office of Policy Development and Research, U.S. Department of Housing and Urban Development, April 2002.

26.  Fannie Mae, Fannie Mae National Housing Survey, 2002, p. 9.

27.  See “Immigration Changes Won't Hurt Housing,” in National Mortgage News, January 27, 2003, p. 8.

28.  Donald S. Bradley and Peter Zorn, “Fear of Homebuying: Why Financially Able Households May Avoid Ownership,” Secondary Mortgage Markets, 1996.

29.  Margery Austin Turner, Stephen L. Ross, George Galster, and John Yinger, “Discrimination in Metropolitan Housing Markets,” The Urban Institute Press, November 2002.

30.  Martin D. Abravanel and Mary K. Cunningham, How Much Do We Know? Public Awareness of the Nation's Fair Housing Laws. A report prepared for HUD by the Urban Institute, Washington, DC, April 2002.

31.  Margery Austin Turner, John Yinger, Stephen Ross, Kenneth Temkin, Diane Levy, David Levine, Robin Ross Smith, and Michelle deLair, What We Know About Mortgage Lending Discrimination, The Urban Institute, contract report for the Department of Housing and Urban Development, December 1998.

32.  Margery Austin Turner, All Other Things Being Equal: A Paired Testing Study of Mortgage Lending Institutions, The Urban Institute Press, April 2002.

33.  Alicia H. Munnell, Geoffrey M.B. Tootell, Lynn E. Browne, and James McEneaney, “Mortgage Lending in Boston: Interpreting HMDA Data,” American Economic Review, 86, March 1996.

34.  See Charles W. Calomeris, Charles M. Kahn and Stanley D. Longhofer, “Housing Finance Intervention and Private Incentives: Helping Minorities and the Poor,” Journal of Money, Credit and Banking, 26, August 1994, pp. 634-74, for more discussion of this phenomenon, which is called “statistical discrimination.”

35.  Robert B. Avery, Patricia E. Beeson and Mark E. Sniderman, Understanding Mortgage Markets: Evidence from HMDA, Working Paper Series 94-21, Federal Reserve Bank of Cleveland, December 1994.

36.  HUD has published an update on “worst case housing needs,” which found that the number of such households rose from 4.86 million in 1999 to 5.07 million in 2001. However, detailed tables for 2001 have not been published.

37.  This does not constitute a significant difference from the 1999 figure of 4.9 million households. However, when the focus is narrowed to renters with incomes below 50 percent of AMI, a statistically significant change emerges; there were 4 percent fewer units affordable to this group in 2001 than there were in 1999.

38.  Very-low-income households are defined as those whose income, adjusted for household size, does not exceed 50 percent of HUD-adjusted area median income. This differs from the definition adopted by Congress in the GSE Act of 1992, which uses a cutoff of 60 percent and which does not adjust income for family size for owner-occupied dwelling units.

39.  Edward N. Wolff, “Recent Trends in the Size Distribution of Household Wealth,” The Journal of Economic Perspectives, 12(3), (Summer 1998), p. 137.

40.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing: 2000, p. 24.

41.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2004. p. 1-2, 4.

42.  Rent is measured in this report as gross rent, defined as contract rent plus the cost of any utilities that are not included in contract rent.

43.  US Housing Market Conditions, 4th Quarter, 2003. HUD Office of Policy Development and Research.

44.  Homeownership rates prior to 1993 are not strictly comparable with those beginning in 1993 because of a change in weights from the 1980 Census to the 1990 Census.

45.  National Association of Realtors, “Near Record Homes Sales Projected for 2003,” December 3, 2002.

46.  Fannie Mae, “Berson's Economic and Mortgage Market Development Outlook,” December 2003. http://www.fanniemae.com/​media/​pdf/​berson/​monthly/​2003/​121203.pdf.

47.  http://www.nahb.org.

48.  Mortgage Bankers Association of America, Mortgage Finance Forecast, December 17, 2003. http://www.mbaa.org/​marketdata/​forecasts/​mffore1103.pdf.

49.  Fannie Mae, “Berson's Economic and Mortgage Market Development Outlook,” December 2003.

50.  U.S. Census Bureau, Population Projections Table NP-T1.

51.  Martha Farnsworth Riche, “How Changes in the Nation's Age and Household Structure Will Reshape Housing Demand in the 21st Century,” in Issue Papers on Demographic Trends Important to Housing, Urban Institute Final Report to the Office of Policy Development and Research, U.S. Department of Housing and Urban Development, September 2002, p. 5.

52.  Joint Center for Housing Studies at Harvard University, State of the Nation's Housing 2004, p.10-11.

53.  Barry Chiswick, Paul Miller, George Masnick, Zhu Xiao Di, and Martha Farnsworth Riche, Issue Papers on Demographic Trends Important to Housing. Urban Institute Final Report to the Office of Policy Development and Research, U.S. Department of Housing and Urban Development, September 2002.

54.  Martha Farnsworth Riche, “How Changes in the Nation's Age and Household Structure Will Reshape Housing Demand in the 21st Century,” in Issue Papers on Demographic Trends Important to Housing. Urban Institute Final Report to the U.S. Department of Housing and Urban Development, September 2002, p. 4.

55.  Ibid. p. 6.

56.  The National Association of Homebuilders estimates base housing demand will average 1.84 million units but increases that estimate to 2.19 million units with high immigration.

57.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 1998, p. 14.

58.  Ibid. p. 15.

59.  Federation for American Immigration Reform, <http://www.fairus.org/​html/​042us604.htm#ins>, site visited December 13, 2002.

60.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2004, p. 11-12.

61.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2002, pp. 16-17.

62.  George S. Masnick and Zhu Xiao Di, “Projections of U.S. Households By Race/Hispanic Origin, Age, Family, Type, and Tenure to 2020: A Sensitivity Analysis,” in Issue Papers on Demographic Trends Important to Housing. Urban Institute Final Report to the U.S. Department of Housing and Urban Development, September 2002, p. 5.

63.  Fred Flick and Kate Anderson, “Future of Housing Demand: Special Markets,” Real Estate Outlook, 1998, p. 6.

64.  Riche, 2002, p. 1.

65.  Average new-home price: U.S. Census Bureau, <http://www.census.gov/​const/​uspriceann.pdf>.

66.  Riche, 2002, p. 17.

67.  All data in this paragraph are from the U.S. Census Bureau's Historical Income Table H2.

68.  Jennifer Cheeseman Day and Eric C. Newburger, The Big Payoff: Educational Attainment and Synthetic Estimates of Work-Life Earnings, U.S. Bureau of the Census, Current Population Reports P23-210, July 2002, p. 3.

69.  U.S. Census Bureau, Historical Income Table H13.

70.  “Mortgage Originations Hit Record-Busting $2.5 Trillion in 2002, IMF Numbers Reveal,” Inside Mortgage Finance, January 24, 2003, p. 3.

71.  Freddie Mac “Cash-Out Refi Report.”

72.  Interest rates in this section are effective rates paid on conventional home purchase mortgages on new homes, based on the Monthly Interest Rate Survey (MIRS) conducted by the Federal Housing Finance Board and published by the Council of Economic Advisers annually in the Economic Report of the President and monthly in Economic Indicators. These are average rates for all loan types, encompassing 30-year and 15-year fixed-rate mortgages and adjustable rate mortgages.

73.  U.S. Housing Market Conditions, 2nd Quarter 2002, August 2002, Table 14.

74.  U.S. Housing Market Conditions, 4th Quarter 2003, February 2004, pg.1.

75.  Mortgage Bankers Association website. MBA Weekly Survey of Mortgage Applications, Monthly Average Interest Rates On 30-Year Fixed-Rate Mortgages. http://www.mortgagebankers.org/​marketdata/​index.html.

76.  http://www.fhfb.gov/​mirs/​mirs_​t25.xls.

77.  http://www.fhfb.gov/​mirs/​mirstbl5.xls;​; data for 2003 is average of May through December data.

78.  This is discussed in more detail in Paul Bennett, Richard Peach, and Stavros Peristani, Structural Change in the Mortgage Market and the Propensity to Refinance, Staff Report Number 45, Federal Reserve Bank of New York, September 1998.

79.  http://www.fhfb.gov/​mirs/​mirs_​t1.xls.

80.  Other sources of data on loan-to-value ratios such as the American Housing Survey and the Chicago Title and Trust Company indicate that high-LTV mortgages are somewhat more common in the primary market than the Finance Board's survey. However, the Chicago Title survey does not separate FHA-insured loans from conventional mortgages. In addition, the statistics cited above pertain only to home purchase mortgages. Refinance mortgages generally have shorter terms and lower loan-to-value ratios than home purchase mortgages.

81.  The source for the refinance share and total mortgage originations is the Mortgage Bankers Association (http://www.mortgagebankers.org/​marketdata/​forecasts/​mffore1203.pdf,, http://www.mortgagebankers.org/​marketdata/​forecasts/​ffJUNE2004.pdf).

82.  Refinancing data is taken from Freddie Mac's monthly Primary Mortgage Market Survey.

83.  There is some evidence that lower-income borrowers did not participate in the 1993 refinance boom as much as higher-income borrowers—see Paul B. Manchester, Characteristics of Mortgages Purchased by Fannie Mae and Freddie Mac: 1996-97 Update, Housing Finance Working Paper No. HF-006, Office of Policy Development and Research, Department of Housing and Urban Development, August 1998, pp. 30-32.

84.  Economy.com, “The Economic Contribution of the Mortgage Refinancing Boom,” December 2002, p. 4.

85.  Fannie Mae, 2002 Fannie Mae National Housing Survey. <http://www.fanniemae.com/​global/​pdf/​media/​survey/​survey2002>,;, September 4, 2002, p. 2.

86.  Economy.com, “The Economic Contribution of the Mortgage Refinancing Boom,” December 2002, p. 4.

87.  Mark M. Zandi, “Refinancing Boom,” Regional Finance Review, December 2002, p. 11.

88.  Ibid. p. 14.

89.  Economy.com, “The Economic Contribution of the Mortgage Refinancing Boom,” December 2002, p. 9.

90.  Mortgage Bankers Association, “Mortgage Finance Forecast”, September 17, 2004. http://www.mortgagebankers.org/​marketdata/​forecasts/​mffore1203.pdf.

91.  Housing affordability varies markedly between regions, ranging in January 2004 from 194 in the Midwest to 107 in the West, with the South and Northeast falling in between.

92.  National Association of REALTORS. Housing Affordability Index, <http://www.realtor.org/​Research.nsf/​Pages/​HousingInx>, 2003.

93.  Fannie Mae, September 4, 2002, p.2.

94.  Ibid.

95.  U.S. Department of Commerce, Bureau of the Census, Money Income of Households, Families, and Persons in the United States: 1992, Special Studies Series P-60, No. 184, Table B-25, October 1993.

96.  Chicago Title and Trust Family of Insurers, Who's Buying Homes in America, 1998.

97.  National Association of Realtors. “New NAR Survey of Home Buyers and Sellers Shows Growing Web Use in a Dynamic Housing Market.” http://www.realtor.org.

98.  U.S. Housing Market Conditions, 3rd Quarter 2001, November 2001, Table 4.

99.  National Association of Realtors. “New NAR Survey of Home Buyers and Sellers Shows Growing Web Use in a Dynamic Housing Market.” http://www.realtor.org.

100.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2002, p. 2.

101.  Office of Federal Housing Enterprise Oversight (OFHEO), Report to Congress, 2004, Tables 1 and 11.

102.  Office of Federal Housing Enterprise Oversight. “Mortgage Markets and The Enterprises in 2001,” August 2002, p. 13.

103.  http://www.financialservicesfacts.org/​financial2/​mortgage/​mortgages/​?table_​sort_​734796=​4.

104.  Mortgage market projections from the MBA's MBA Mortgage Finance Forecast, December 17, 2003. 2000 and 2001 numbers from the MBA's MBA Mortgage Finance Forecast, January 10, 2002.

105.  See Charles, K. K. and E. Hurst (2002). “The Transition to Home Ownership and the Black-White Wealth Gap.” The Review of Economics and Statistics, 84(2): 281-297; Mayer, C. and G. Engelhardt (1996). “Gift Down Payments and Housing Affordability.” Journal of Housing Research, 7(1): 59-77; and Quercia, R. G., G. W. McCarthy, et al. (2003). “The Impacts of Affordable Lending Efforts on Homeownership Rates.” Journal of Housing Economics, 12(1): 29-59.

106.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 8-9.

107.  Fannie Mae, “Fannie Mae's Comments on HUD's Proposed Housing Goals for Fannie Mae and Freddie Mac for the years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac,” July 16, 2004, p. I-58.

108.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, p. 6.

109.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 9-10.

110.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 62.

111.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 62.

112.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 62-64.

113.  Freddie Mac Public Comment Letter on HUD's Proposed Goals, July 2004, p. 2.

114.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 62-64.

115.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 22-24.

116.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 13-16.

117.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 17-22.

118.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 16.

119.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 17-18.

120.  Fannie Mae, “Fannie Mae's Comments on HUD's Proposed Housing Goals for Fannie Mae and Freddie Mac for the years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac,” July 16, 2004, p. I-60.

121.  Freddie Mac, News Release, January 15, 1999.

122.  Freddie Mac Public Comment Letter on HUD's Proposed Goals, July 2004, p. 3.

123.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 67.

124.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 66-67.

125.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 37-38.

126.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 38-39.

127.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 39-40.

128.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 42-43.

129.  Freddie Mac Public Comment Letter on HUD's Proposed Goals, July 2004, p. 4.

130.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 29-30.

131.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 30-34.

132.  Kenneth Temkin, Roberto Quercia, George Galster, and Sheila O'Leary, A Study of the GSEs' Single Family Underwriting Guidelines: Final Report. Washington DC: U.S. Department of Housing and Urban Development, April 1999.

133.  Temkin, et al. 1999, p. 28.

134.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, pp. 36-37.

135.  Freddie Mac Public Comment Letter on HUD's Proposed Goals, July 2004, p. 5.

136.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 19.

137.  Freddie Mac, 2002 Annual Housing Activities Report, 2003, p. 57.

138.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 11-12.

139.  Fannie Mae, “Fannie Mae's Comments on HUD's Proposed Housing Goals for Fannie Mae and Freddie Mac for the years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac,” July 16, 2004, p. I-57.

140.  Table A.3 also provides the same average (1999 to 2003) information as Tables A.1 and A.2 but for total (both home purchase and refinance) loans. Thus, it provides a complete picture of overall mortgage activity.

141.  The “Total Market” is defined as all loans (including both government and conventional) below the conforming loan limit of $240,000 in 1999, $252,700 in 2000, $275,000 in 2001, $300,700 in 2002 and $322,700 in 2003.

142.  The affordable market shares reported in Table A.1 for the “Conventional Conforming Market W/O B&C” were derived by excluding the estimated number of B&C loans from the market data reported by HMDA. Because B&C lenders operate mainly in the refinance sector, excluding these loans from the conforming market has little impact on the home purchase percentages reported in Table A.1. It should be recognized that there exists some uncertainty regarding the number of B&C loans in the HMDA data. The adjustment assumes that the B&C loans represent one-half of the subprime market. The adjustment for home purchase loans is small because supbrime (B&C) loans are mainly refinance loans. The method for excluding B&C loans is explained in Section E below and Appendix D.

143.  Almost two-thirds of the borrowers with an FHA-insured home purchase loan make a downpayment less than five percent, and over 80 percent are first-time home buyers. For discussions of the role of FHA in the mortgage market, see (a) Harold L. Bunce, Charles A. Capone, Sue G. Neal, William J. Reeder, Randall M. Scheessele, and Edward J. Szymanoski, An Analysis of FHA's Single-Family Insurance Program, Office of Policy Development and Research, U.S. Department of Housing and Urban Development, 1995; and (b) Office of Policy Development and Research, “FHA's Impact on Homeownership Opportunities for Low-Income and Minority Families During the 1990s” Issue Brief IV, U.S. Department of Housing and Urban Development, December 2000. For data on the credit characteristics of FHA borrowers, see Harold L. Bunce, William J. Reeder and Randall Scheessele, “Understanding Consumer Credit and Mortgage Scoring: A Work in Progress at HUD”, U.S. Department of Housing and Urban Development, Unpublished Paper, 1999.

144.  FHA, which focuses on low downpayment loans and also accepts borrowers with credit blemishes, experiences higher mortgage defaults than conventional lenders and the GSEs. Still, the FHA system is actuarially sound because it charges an insurance premium that covers the higher default costs. For the results of FHA's actuarial analysis, see Deloitte & Touche, Actuarial Review of MMI Fund as of FY 2000, report for the U.S. Department of Housing and Urban Development, January 2001.

145.  See Green and Associates, Fair Lending in Montgomery County: A Home Mortgage Lending Study, a report prepared for the Montgomery County Human Relations Commission, March 1998; and Calvin Bradford, Crisis in Déjà vu: A Profile of the Racial Patterns in Home Purchase Lending in the Baltimore Market. Report for The Public Justice Center, May 2000; and The Patterns of GSE Participation in Minority and Racially Changing Markets Reviewed from the Context of Levels of Distress Associated with High Levels of FHA Lending, GSE Study No. 11, U.S. Department of Housing and Urban Development, September 2000. For analysis suggesting some minorities receiving FHA loans could qualify for conventional loans, see Anthony Pennington-Cross, Anthony Yezer, and Joseph Nichols, Credit Risk and Mortgage Lending: Who Uses Subprime and Why? Working Paper No. 00-03. Research Institute for Housing America, 2000. Also see the series of recent studies concerning the lack of mainstream lenders in minority neighborhoods.

146.  For a comprehensive analysis of the GSEs' purchases of minority loans through 1999, see Harold L. Bunce, An Analysis of GSE Purchases of Mortgages for African-American Borrowers and their Neighborhoods, Housing Finance Working Paper No. 11, Office of Policy Development and Research, HUD, December 2000.

147.  Tables A.1, A.2, and A.3 include data for all home loans originated by depositories as well as for the subset of loans originated but not sold, the latter being a proxy for loans held in depository portfolios. (See the notes to Table A.1 for definitions of the depository data.)

148.  However, as shown in Table A.1 , depository institutions resemble other conventional lenders in their relatively low level of originating loans for African-American, Hispanic and minority borrowers. Within the conventional conforming market, Fannie Mae has done a better job than depositories in funding minority borrowers, particularly Hispanic borrowers and minority borrowers as a group. During the last three years, Fannie Mae has also funded African-American borrowers at a higher rate than have depository institutions.

149.  CRA loans are typically made to low-income borrowers earning less than 80 percent of area median income, and in moderate-income neighborhoods. For a comprehensive analysis of CRA and its impact on affordable lending, see Robert E. Litan, Nicolas P. Retsinas, Eric S. Belsky and Susan White Haag, The Community Reinvestment Act After Financial Modernization: A Baseline Report, U.S. Department of Treasury, 2000.

150.  In this case, the market includes all government and conventional loans, including jumbo loans.

151.  For a comprehensive analysis of CRA and its impact on affordable lending, see Robert E. Litan, Nicolas P. Retsinas, Eric S. Belsky and Susan White Haag, The Community Reinvestment Act After Financial Modernization: A Baseline Report, U.S. Department of Treasury, 2000.

152.  Board of Governors of the Federal Reserve System. The Performance and Profitability of CRA-Related Lending. Washington, DC, 2000.

153.  This discussion of urban lending draws from Jeff Siegel, “Urban Lending Helps Increase Volume and Meet CRA Requirements,” Secondary Marketing Executive, February 2003, pp. 21-23.

154.  Ibid.

155.  Fannie Mae, “Fannie Mae's Comments on HUD's Proposed Housing Goals for Fannie Mae and Freddie Mac for the years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac,” July 16, 2004, p. I-59.

156.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, pp. 8-9.

157.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 64.

158.  U.S. Department of HUD, Office of Policy Development and Research, U.S. Housing Market Conditions, May 2004, p. 81.

159.  Ibid.

160.  Fannie Mae, Fannie Mae National Housing Survey, 2002, p. 6.

161.  Ibid. p. 8.

162.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 2003, p. 15.

163.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing 1998, p. 20.

164.  John W. Straka, “A Shift in the Mortgage Landscape: The 1990s Move to Automated Credit Evaluations,” Journal of Housing Research, 2000, (11)2: p. 207.

165.  Peter M. Zorn, Susan Gates, and Vanessa Perry, “Automated Underwriting and Lending Outcomes: The Effect of Improved Mortgage Risk Assessment on Under-Served Populations. Program on Housing and Urban Policy,” Conference Paper Series, Fisher Center for Real Estate and Urban Economics. University of California Berkeley, 2001, p. 5.

166.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, p. 12.

167.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, p. 36.

168.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 55.

169.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 54.

170.  Freddie Mac Public Comment Letter on HUD's Proposed Goals, July 2004, p. 5.

171.  Ibid. pp. 208-217.

172.  Robert B. Avery, Raphael W. Bostic, Paul S. Calem, and Glenn B. Canner, Credit Scoring: Issues and Evidence from Credit Bureau Files, mimeo, 1998, p. 24.

173.  Fannie Mae, September 4, 2002, p. 33.

174.  Kenneth Temkin, Jennifer E.H. Johnson, and Diane Levy, Subprime Markets, The Role of GSEs, and Risk-Based Pricing, Washington: The Urban Institute. Report Prepared for the U.S. Department of Housing and Urban Development, 2002.

175.  Allen J. Fishbein, “Is Credit Scoring a Winner for Everyone?” Stone Soup, 2000, 14(3): pp. 14-15. See also Fitch IBCA, Inc., Residential Mortgage Credit Scoring, New York, 1995 and Jim Kunkel, “The Risk of Mortgage Automation,” in Mortgage Banking, 1995, 57(8): pp. 69-76.

176.  Zorn et al., 2001, pp. 19-20.

177.  Kenneth Temkin, Jennifer E.H. Johnson, and Diane Levy, Subprime Markets, The Role of GSEs, and Risk-Based Pricing, Washington: The Urban Institute. Report Prepared for the U.S. Department of Housing and Urban Development, 2002.

178.  Zorn, et al., 2001, pp. 14-15.

179.  Ibid. p. 5.

180.  Ibid. pp. 18-19.

181.  Subprime origination data are from Inside Mortgage Finance. For the 2002 estimates, see “Subprime Origination Market Shows Strong Growth in 2002,” Inside B&C Lending, published by Inside Mortgage Finance, February 3, 2003, page 1.

182.  Temkin et. al., 2002, p. 1.

183.  Kenneth Temkin, Jennifer E.H. Johnson, Diane Levy, Subprime Markets, The Role of GSEs, and Risk Based Pricing, Washington: The Urban Institute. Report Prepared for the Department of Housing and Urban Development, 2002, p. 4.

184.  U.S. Department of Housing and Urban Development/U.S. Department of the Treasury, Curbing Predatory Lending Report, 2000, p.31.

185.  “Wholesale Dominates Subprime Market Through 3rd Quarter '02,” Inside B&C Lending, published by Inside Mortgage Finance, December 16, 2002, pp. 1-2.

186.  Inside B&C Lending, November 16, 2002, p.2.

187.  Mortgage Information Corporation, The Market Pulse, Winter 2001, pp. 4-6.

188.  Inside B&C Lending, published by Inside Mortgage Finance, February 17, 2003, page 13.

189.  Daniel Immergluck, The Predatory Lending Crisis in Chicago: The Dual Mortgage Market and Local Policy, testimony before the Chicago City Council, April 5, 2000. Immergluck found that subprime lenders received 74 percent of refinance applications in predominantly black tracts compared to 21 percent in predominantly white tracts in 1998. According to Immergluck, these racial disparities provide evidence that the residential finance market in Chicago is hypersegmented, resulting in the increased likelihood that minorities receive mortgage credit from a subprime, rather than a prime, lender in Chicago. Also see Daniel Immergluck, Stark Differences: The Explosion of the Subprime Industry and Racial Hypersegmentation in Home Equity Lending, Woodstock Institute, October 2000

190.  See Randall M. Scheessele, Black and White Disparities in Subprime Mortgage Refinance Lending, Housing Finance Working Paper HF-014, Office of Policy Development and Research, U.S. Department of Housing and Urban Development, April 2002.

191.  For an update to 2001, see The Association of Community Organizers for Reform Now (ACORN), Separate and Unequal Predatory Lending in America, 2002. In 2001, subprime lenders originated 27.8 percent of all conventional refinance loans for African-Americans, 13.6 percent for Hispanic homeowners, and just 6.3 percent for white homeowners. Overall, African-Americans were 4.4 times more likely to use a subprime lender than whites, and Hispanics were 2.2 times more likely to do so.

192.  Howard Lax, Michael Manti, Paul Raca, and Peter Zorn, “Subprime Lending: An Investigation of Economic Efficiency,” February 25, 2000.

193.  It should also be noted that higher interest rates are only one component of the higher cost of subprime loans since borrowers also often face higher origination points. The Freddie Mac study did not find a large differential between prime and subprime loans in points paid, but the study notes that subprime loans often have points rolled into the loan principal, which cannot be identified with their data.

194.  Freddie Mac, We Open Doors for America's Families, Freddie Mac's Annual Housing Activities Report for 1997, March 16, 1998, p. 23.

195.  Rommy Fernandez, “Fannie Mae Eyes Half of the Subprime Market,” in The American Banker, March 1, 2002. Also see “Fannie Mae Vows More Minority Lending,” Washington Post, March 16, 2000, p. EO1.

196.  For an overview of these studies, see Harold L. Bunce, Debbie Gruenstein, Christopher E. Herbert, Randall M. Scheessele, Subprime Foreclosures: The Smoking Gun of Predatory Lending, 2000. Also see Abt Associates Inc., Analyzing Trends in Subprime Originations and Foreclosures: A Case Study of the Atlanta Metro Area, February 2000 and Analyzing Trends in Subprime Originations and Foreclosures: A Case Study of the Boston Metro Area, September 2000; National Training and Information Center, Preying on Neighborhoods: Subprime Mortgage Lenders and Chicagoland Foreclosures, 2000; and the HUD study, Unequal Burden in Baltimore: Income and Racial Disparities in Subprime Lending, May 2000.

197.  “OCC Cites Fannie, Freddie Predatory Lending Rules As Model,” Dow Jones Business News, February 25, 2003.

198.  Temkin et al., 2002, p. 1.

199.  David A. Andrukonis, “Entering the Subprime Arena,” Mortgage Banking, May 2000, pp. 57-60.

200.  Subprime Lenders Mixed on Issue of GSE Mission Creep,” Inside B and C Lending, March 19, 2001.

201.  Fannie Mae, “Fannie Mae's Comments on HUD's Proposed Housing Goals for Fannie Mae and Freddie Mac for the years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac,” July 16, 2004, p. I-59.

202.  Freddie Mac Public Comment Letter on HUD's Proposed Goals, July 2004, p. 6.

203.  See Lederman, et al., Op cit.

204.  Kenneth Temkin, Jennifer E. H. Johnson, and Diane K. Levy, “Subprime Markets, the Role of GSEs, and Risk-Based Pricing,” Urban Institute, August 2001, p. 1.

205.  Inside Mortgage Finance's, “Inside MBS & ABS,” December 15, 2000 and March 8, 2002.

206.  Statement by Mercy Jimenez of Fannie Mae in “Fannie Mae: Forges Ahead in Subprime,” Secondary Marketing Executive, February 2003, p.15.

207.  Temkin et al., 2002, p. 1.

208.  See Lax et al., 2000.

209.  Zorn, et al., 2001, p. 5.

210.  Fannie Mae, Remarks Prepared for Delivery by Franklin Raines, Chairman and CEO of Fannie Mae to the National Community Reinvestment Coalition. Washington, D.C. March 20, 2000.

211.  Annual Percentage Rate takes into account points, fees, and the periodic interest rate.

212.  Temkin et al., 2002, p.29.

213.  For example, see Radian's product offerings at http://www.radiangroupinc.com.

214.  Vanessa Bush, “Risk-Based Pricing Trend Could Make Mortgage Lending More Efficient,” America's Community Banker, October 1, 1998.

215.  “Improving Credit Quality, Maturing Business Stoke Confidence in Subprime MBS Market,” Inside MBS & ABS, published by Inside Mortgage Finance, February 21, 2003.

216.  Ibid.

217.  U.S. Department of HUD, Office of Policy Development and Research, U.S. Housing Market Conditions: 4th Quarter 2003, February 2004, p. 3.

218.  U.S. Department of HUD, Office of Policy Development and Research, U.S. Housing Market Conditions: 4th Quarter 2003, February 2004, p. 84.

219.  See, for example, Marcus & Millichap Research Services, National Apartment Report, January 2003.

220.  Marcus & Millichap Research Services, National Apartment Report, January 2004.

221.  U.S. Department of Labor, Bureau of Labor Statistics, “Bureau of Labor Statistics Data,” Accessed July 31, 2004, http://data.bls.gov/​servlet/​SurveyOutputServlet?​data_​tool=​latest_​numbers&​series&​lowbar;​id=​LNS14000000.

222.  “Apartment Landlords Gather to Dreary Outlook for Sector,” Wall Street Journal, January 15, 2003, Section B.

223.  Mortgage Bankers Association of America, “MBA News Link: Rental Market Demographics “Favorable,” Report Says,” January 2003.

224.  U.S. Department of HUD, Office of Policy Development and Research, U.S. Housing Market Conditions: 4th Quarter 2003, February 2004, p. 70.

225.  Center for Housing Policy/National Housing Conference, “Housing America's Working Families: A Further Exploration,” New Century Housing, Vol. 3, No. 1, March 2002; Mark Obrinsky and Jill Meron, “Housing Affordability: The Apartment Universe,” National Multi Housing Council, 2002.

226.  “Housing Affordability in the United States: Trends, Interpretations, and Outlook,” a report prepared for the Millennial Housing Commission by J. Goodman, November, 2001.

227.  Joint Center for Housing Studies of Harvard University, State of the Nation's Housing, 2002.

228.  Center for Housing Policy/National Housing Conference, “America's Working Families and the Housing Landscape 1997-2001,” New Century Housing, Vol. 3, No. 2, November 2002

229.  Urban Land Institute, The ULI Forecast, 2002; Lendlease and PriceWaterhouseCoopers, Emerging Trends in Real Estate, 2003.

230.  Merrill Lynch, A New Look at FHA Prepayments and Defaults, September 2002.

231.  Eight percent inflation adjusted.

232.  Freddie Mac Public Comment Letter on HUD's Proposed Goals, July 2004, p.3.

233.  “No Mistaking GSEs for Twins in Multifamily,” American Banker, October 2, 2002.

234.  This change was a percentage decrease but a volume increase.

235.  For background information on the Freddie Mac TAF, see pages 65054 and 65067-65068 of the 2000 Rule.

236.  Fannie Mae's 2002 Annual Housing Activities Report, pages 24-27; Freddie Mac's Annual Housing Activities Report for 2002, pages 41-47.

237.  Abt Associates Inc., An Assessment of the Availability and Cost of Financing for Small Multifamily Properties, a report prepared for the U.S. Department of Housing and Urban Development, Office of Policy Development and Research, August 2001.

238.  Fannie Mae, 2003 Annual Housing Activities Report, March 15, 2004, p. 26.

239.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 44.

240.  Fannie Mae, 2003 Annual Housing Activities Report, March 15, 2004, p. 27.

241.  Fannie Mae, 2003 Annual Housing Activities Report, March 15, 2004, p. 28.

242.  Fannie Mae, 2003 Annual Housing Activities Report, March 15, 2004, p. 29.

243.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 47.

244.  “Fannie Courting Multifamily Sellers; Small Banks Balking,” American Banker, January 13, 2003.

245.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Repoert for 2003, March 15, 2004, p. 47 & 49.

246.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 50-52.

247.  Fannie Mae, 2002 Annual Housing Activities Report, 2003, p. 27.

248.  Abt Associates, “Study of Multifamily Underwriting and the GSEs” Role in the Multifamily Market,” Final Report to the U.S. Department of Housing and Urban Development, Office of Policy Development and Research, August 2001.

249.  Federal Reserve, Survey of Professional Forecasters, November 2003.

250.  Board of Governors of the Federal Reserve System, Monetary Policy Report to the Congress, February 11, 2003, page 4.

251.  Jack Goodman, “The Changing Demography of Multifamily Rental Housing,” Housing Policy Debate, Winter 1999.

252.  Remarks by Franklin D. Raines, Chairman and CEO, Fannie Mae, to the Executive Committee of the National Association of Home Builders, January 18, 2003. See also Edward Glaeser and Joseph Gyourko, “The Impact of Zoning on Housing Affordability,” Working Paper 8835, National Bureau of Economic Research, March 2002.

253.  “Capital Markets Outlook 2003,” Apartment Finance Today, Vol. 7, No. 1 (January/February 2003).

254.  Performance for the 1993-95 period was discussed in the October 2000 rule.

255.  To separate out the effects of changes in counting rules that took effect in 2001, this section also compares performance in 2001 to estimated performance in 2000 if the 2001 counting rules had been in effect in that year.

256.  See Congressional Record, December 15, 2000, pp. H12295-96.

257.  Prior to the October 2000 rule, purchases of these government-backed mortgages were only eligible for credit under the special affordable goal.

258.  Exclusion of loans with missing information had a greater impact on Fannie Mae's goal performance than on Freddie Mac's goal performance.

259.  Federal Register, October 31, 2000, Footnote 145, p. 65141.

260.  “Fannie Courting Multifamily Sellers; Small Banks Balking,” American Banker, January 13, 2003, p. 1.

261.  In New England, MSAs were defined through mid-2003 in terms of Towns rather than Counties, and the portion of a New England county outside of any MSA was regarded as equivalent to a county in establishing the metropolitan or non-metropolitan location of a property. The MSA definitions established by the Office of Management and Budget (OMB) in June, 2003 defined MSAs in New England in terms of counties.

262.  The procedure is explained in detail in annual releases entitled “HUD Methodology for Estimating FY [year] Median Family Incomes” for years 1993 through 2002, issued by the Economic and Market Analysis Division, Office of Economic Affairs, PD&R, U.S. Department of Housing and Urban Development.

263.  The procedure applicable to the decennial census data used to generate estimated rents is explained in connection with data used to define Underserved Areas in Appendix B.

264.  Transition from the 2002 methodology to the 2005 methodology is occurring in stages in 2003 and 2004. To generate the area median income estimates used to score GSE loans in 2003, data from the 2000 census on 1999 area median incomes were adjusted to 2001 using Bureau of Labor Statistics survey data on rates of change in average incomes for MSAs and counties between 1999 and 2000, data on rates of change in median incomes for the United States and individual States between 1999 and 2001 from Census Bureau's Current Population Survey and American Communities Survey, and an assumed 3.5 percent per year inflation factor between 2001 and 2003. (See “HUD Methodology for Estimating FY 2003 Median Family Incomes,” issued by the Economic and Market Analysis Division, op cit.) A similar procedure has been used to generate area median income estimates for scoring GSE loans in 2004.

265.  HUD has deferred application of the 2003 MSA specification to 2005, pending completion of the present rulemaking process.

266.  The “affordable lending performance” of Fannie Mae and Freddie Mac refers to the performance of the GSEs in funding loans for low-income and underserved borrowers through their purchase (or guarantee) of loans originated by primary lenders. It does not, of course, imply that the GSEs themselves are lenders originating loans in the primary market.

267.  Throughout this analysis, the terms “home loan” and “home mortgage” will refer to a “home purchase loan,” as opposed to a “refinance loan.” As noted earlier, the mortgage data reported in this paper are for metropolitan areas, unless stated otherwise. Restricting the GSE data to metropolitan areas is necessary to make it comparable with the HMDA-reported conventional primary market data, which is more reliable for metropolitan areas. The analysis of first-time homebuyers in Sections E.9 and E.12 cover both metropolitan and non-metropolitan areas.

268.  Unless otherwise noted, the conventional conforming market data reported in this section exclude an estimate of B&C loans; the less-risky A-minus portion of the subprime market is included in the market definition. See Section E.7 and Appendix D for a discussion of primary market definitions and the uncertainty surrounding estimates of the number of B&C loans in HMDA data. As noted there, B&C loans are much more likely to be refinance loans rather than home purchase loans.

269.  Fannie Mae had a particularly poor year during 1999. Therefore, the text also reports averages for 2000-2003, dropping the year 1999 (see Table A.13 in Section E.9).

270.  As explained in Section E.9, deducting B&C loans from the market totals has more impact on the market percentages for total (both home purchase and refinance) loans than for only home purchase loans. The effects of excluding B&C loans from the total market can be seen by comparing the third and sixth columns of data in Table A.19 in Section E.10.

271.  See Glenn B. Canner, Wayne Passmore, and Brian J. Surette, “Distribution of Credit Risk Among Providers of Mortgages to Lower-Income and Minority Homebuyers' in Federal Reserve Bulletin, 82(12): 1077-1102, December, 1996.

272.  In this comparison, a higher special affordable percentage for HMDA-reported mortgage originations that lenders report as also being sold to the GSEs—as compared with the special affordable percentage for newly-originated mortgages that the GSEs report as being actually purchased by them—would suggest that HMDA market data are biased; that is, in this situation, the special affordable percentage for all mortgage originations reported in HMDA would likely be larger than the special affordable percentage for all new mortgage originations, including those not reported in HMDA as well as those reported in HMDA.

273.  The market definition in this section is narrower than the “Total Market” data presented earlier in Tables A.1 and A.2, which included all home loans below the conforming loan limit, that is, government loans as well as conventional conforming loans. The market share analysis reported in Section E.12 also examine the GSEs' role in the overall market.

274.  And there is some evidence that many subprime loans are not even reported to HMDA, although there is nothing conclusive on this issue. See Fair Lending/CRA Compass, June 1999, p. 3.

275.  The list of subprime lenders as well as Scheessele's list of manufactured housing lenders are available at http://www.huduser.org/​publications/​hsgfin.html.

276.  The one-half estimate is conservative as some observers estimate that B&C loans account for only 30-40 percent of the subprime market. However, varying the B&C share from 50 percent to 30 percent does not significantly change the following analysis of home purchase loans because subprime loans are mainly for refinance purposes. Overstating the share of B&C loans in this manner also allows for any differences in HMDA reporting of different types of loans—for example, if B&C loans account for 35 percent of all subprime loans, then assuming that they account for 50 percent is equivalent to assuming that B&C loans are reported in HMDA at 70 percent of the rate of other loans.

277.  The reductions in the market shares are more significant for total loans, which include refinance as well as home purchase loans; for data on total loans, see Table A.19 in Section 10. Subprime lenders have been focusing more on home purchase loans recently. The home purchase share of loans originated by the subprime lenders in Scheessele's list increased from 26 percent in 1999 to 36 percent in 2000 before dropping to about 30 percent during the heavy refinancing years of 2001 and 2002.

278.  In 2001 (2002), lenders reported in HMDA that they purchased 851,735 (906,684) conventional conforming, home purchase loans in metropolitan areas; this compares with 2,763,230 (2,929,197) loans that these same lenders reported that they originated in metropolitan areas.

279.  See Randall M. Scheessele, HMDA Coverage of the Mortgage Market, Housing Finance Working Paper No. HF-007. Office of Policy Development and Research, U.S. Department of Housing and Urban Development, July, 1998.

280.  In this example, HMDA-reported purchased loans insured by FHA have been reduced from 411,930 to 100,251 by a procedure that accounts for missing data and overlapping purchased and originated loans. See Harold L. Bunce, The GSEs' Funding of Affordable Loans: A 2000 Update, Working Paper HF-013, Office of Policy Development and Research, HUD, April 2002, for an alternative analysis showing that a market estimate based on adding HMDA-reported purchased loans to HMDA-reported originations would substantially overstate the volume of FHA mortgage originations in metropolitan areas.

281.  See Chapter III, “Reporting of Brokered and Correspondent Loans under HMDA”, in Exploratory Study of the Accuracy of HMDA Data, by Abt Associates Inc. under contract for the Office of Policy Development and Research, HUD, February 12, 1999, page 18.

282.  The percentage shares for purchased loans are obtained after eliminating purchased loans without data and purchased loans that overlap with originated loans. The calculations included 138,536 purchased loans for 2001 and 182,290 purchased loans for 2002.

283.  Readers not interested in these technical issues may want to proceed to Section E.9, which compares GSE performance to the primary market.

284.  See Jim Berkovec and Peter Zorn, “How Complete is HMDA? HMDA Coverage of Freddie Mac Purchases,” The Journal of Real Estate Research, Vol. II, No. 1, Nov. 1, 1996.

285.  For another discussion of this issue, see Randall M. Scheessele, HMDA Coverage of the Mortgage Market, Housing Finance Working Paper HF-007, Office of Policy Development and Research, Department of Housing and Urban Development, July 1998. Scheessele reports that HMDA data covered 81.6 percent of the loans acquired by Fannie Mae and Freddie Mac in 1996. The main reason for the under-reporting of GSE acquisitions is a few large lenders failed to report the sale of a significant portion of their loan originations to the GSEs. Also see the analysis of HMDA coverage by Jim Berkovec and Peter Zorn. “Measuring the Market: Easier Said than Done,” Secondary Mortgage Markets. McLean VA: Freddie Mac, Winter 1996, pp. 18-21; as well as the Berkovec and Zorn study cited in the above footnote.

286.  Between 1993 and 1996, the GSEs' purchases of prior-year loans were not as targeted as they were after 1996; thus, during this period, HMDA provided reasonable estimates of the goals-qualifying percentages of the GSEs' purchases of all (both current-year and prior-year) loans, with a few exceptions (see Table A.11).

287.  The “prior-year” share dropped to 16 percent during the heavy refinancing year of 2003. During the 1990s, the GSEs increased their purchases of seasoned loans; see Paul B. Manchester, Goal Performance and Characteristics of Mortgages Purchased by Fannie Mae and Freddie Mac, 1998-2000, Housing Finance Working Paper No. HF-015, Office of Policy Development and Research, HUD, May 2001.

288.  Freddie Mac's underserved area figures for 2002 and 2003 showed particularly large discrepancies. As shown in Table A.11, Freddie Mac reported that 25.0 (23.4) percent of the current-year loans it purchased during 2002 (2003) financed properties in underserved areas, a figure much higher than the 21.4 (20.3) percent that HMDA reported as underserved area loans sold to Freddie Mac during 2002. These discrepancies are the largest in Table A.11, and it is not clear what explains them. This downward bias for HMDA data, is the opposite of that suggested by Berkovec and Zorn, who argued that affordability percentages from HMDA data are biased upward.

289.  The data in Table A.12 that support Berkovec and Zorn are the 1993-95 special affordable and low-mod data (particularly for Freddie Mac) that show HMDA over reporting percentages by more than a half percentage point. Otherwise, the data in Table A.12, as well as Table A.11, do not present a picture of HMDA's having an upward bias in reporting targeted loans. In fact, the recent years' data suggest a downward bias in HMDA's reporting of targeted loans.

290.  Of course, on an individual year basis, the GSEs' current-year data can differ significantly from the HMDA-reported data on GSE purchases. The other annual data reported in Table A.11 show a mixture of results—in some cases the HMDA percentage is larger than the GSE “current year” percentage (e.g., Fannie Mae's special affordable purchases in 2000) while in other cases the HMDA percentage is smaller than the GSE current year percentage (e.g., Freddie Mac's special affordable purchases in recent years). As noted in the text, the differential is typically in the opposite direction to that predicted by Berkovec and Zorn, particularly on the underserved areas category.

291.  Table A.12 also includes aggregates for the more recent period, 1999-2003. The ratios of HMDA-reported-to-GSE-reported averages for this sub-period are similar to those reported for 1996-2003.

292.  Under the origination-year approach, GSE performance for any specific origination year (say year 2000) at the end of a particular GSE purchase year (say year 2003) is subject to change in the future years. Table A.16 (in Section E.9 below) reports that 13.7 percent of year-2000 mortgage originations that Fannie Mae purchased through year 2002 qualify as special affordable; the special affordable share for the market was 16.6 percent in 2000, which indicates that, to date, Fannie Mae has lagged the primary market in funding special affordable mortgages originated during 2000. However, Fannie Mae's special affordable performance could change in the future as Fannie Mae continues to purchase year-2000 originations during 2004 and the following years. Of course, whether Fannie Mae's future purchases result in it ever leading the 2000-year market is not known at this time.

293.  As shown in Table A.13, the depository percentage is higher (16.8 percent) if the analysis is restricted to those newly-originated loans that depositories do not sell (the latter being a proxy for loans held in depositories' portfolios). Note that during the recent, 1999-to-2003 period (also reported in Table A.13), there is less difference between the two depository figures.

294.  Unless stated otherwise, the market in this section is defined as the conventional conforming market without estimated B&C loans.

295.  Table A.14 reports annual market percentages that exclude the effects of manufactured housing, small loans, and subprime loans. Freddie Mac's performance is closer to the market average under the alternative market definitions, particularly during 2001 and 2002.

296.  Prior to 2002, Freddie Mac's performance on the underserved areas category had not approached the market even under the alternative market definitions reported in Table A.14.

297.  Freddie Mac, on the other hand, fell further behind the market during this period. In 1992, Freddie Mac had a slightly higher underserved areas percentage (18.6 percent) than Fannie Mae (18.3 percent). However, Freddie Mac's underserved areas percentage had only increased to 19.8 percent by 1998 (versus 22.7 percent for Fannie Mae). Thus, the “Freddie Mac-to-market” ratio fell from 0.84 in 1992 to 0.82 in 1998.

298.  These figures include estimates of first-time homebuyer loans for those home purchase loans with a missing first-time homebuyer indicator; the estimates were obtained by multiplying the GSE's first-time homebuyer share (based only on data with a first-time homebuyer indicator) by the number of loans with a missing first-time homebuyer indicator.

299.  The first-time homebuyer share for Fannie Mae was almost 35 percent between 1996 and 1998; it then dropped to 30 percent in 1998 and to 26 percent in 1999. The first-time homebuyer share for Freddie Mac was approximately 29 percent in 1996 and 1997 before dropping to about 25 percent in 1998 and 1999.

300.  See Harold L. Bunce and John L. Gardner, “First-time Homebuyers in the Conventional Conforming Market: The Role of the GSEs” (unpublished paper), January 2004. An update of this work to include data for 2002 and 2003 shows similar patterns as those reported in the text for 1999-2001. See Harold L. Bunce and John L. Gardner, “First-time Homebuyers in the Conventional Conforming Market: The Role of GSEs: An Update” (October, 2004).

301.  The GSE total (home purchase and refinance) data in Tables A.18-A.20 are presented on a purchase-year basis; Table A.21 presents similar data on an origination-year basis.

302.  Following the purchase-year approach used in Sections E.9 and E.10, the GSE purchase data include their acquisitions of “prior-year” as well as “current-year” mortgages, while the market data include only newly-originated (or “current year”) mortgages.

303.  As explained in Section E.7, the GSEs' affordable lending performance is evaluated relative to the conventional conforming market, as required by Congress in the 1992 GSE Act that established the housing goals. However, it is insightful to examine their overall role in the mortgage market and to contrast them with other major sectors of the market such as FHA. There is no intention here to imply that the GSEs should purchase the same types of loans that FHA insures.

304.  As explained in the notes to Table A.25, HMDA data are the source of the market figures. It is assumed that HMDA data cover 85 percent of all mortgage originations in metropolitan areas. If HMDA data covered higher (lower) percentages of market loans, then the market shares for both the GSEs and FHA would be lower (higher).

305.  See Harold L. Bunce, The GSEs' Funding of Affordable Loans: A 2000 Update, Housing Finance Working Paper No. HF-013, Office of Policy Development and Research, HUD, April 2002.

306.  Bunce explains numerous assumptions and caveats related to combining American Housing Survey data on homebuyers with FHA and GSE data on mortgages. For example, the American Housing Survey (AHS) data used by Bunce included both financed home purchases and homes purchased with cash. If only financed home purchases were used, the market shares of both FHA and the GSEs would have been slightly higher (although the various patterns would have remained the same). The AHS defines first-time homebuyers as buyers who have never owned a home, while FHA and the GSEs define a first-time homebuyer more expansively as buyers who have not owned a home in the past three years. If it were possible to re-define the FHA and GSE data to be consistent with the AHS data, the FHA and GSE first-time homebuyer shares would be lower (to an unknown degree). For additional caveats with the AHS data, also see David A. Vandenbroucke, Sue G. Neal, and Harold L. Bunce, “First-Time Homebuyers: Trends from the American Housing Survey”, November 2001, U.S. Housing Market Condition, a quarterly publication of the Office of Policy Development and Research at HUD. In some years, home purchases as measured by the AHS declined while home purchases as measured by other data sources (e.g., HMDA) increased. In addition, the AHS home purchase data for separate minority groups (e.g., African-Americans, Hispanics) sometimes exhibited shifts inconsistent with other sources.

307.  BNV's methodology for estimating first-time borrowers consists of three steps: (1) estimate the total number of home purchase loans originated during a particular year using a mortgage market model that they develop; (2) disaggregate the home purchase loans in step (1) into racial and ethnic groups using HMDA data for metropolitan areas; and (3) for each racial and ethnic group in step (2), estimate the number of first-time homebuyers using mortgage and first-time homebuyer information from the American Housing Survey.

308.  See Bunce, Neal, and Vandenbroucke, op. cit., for comparisons of various estimates of the market shares for FHA and the GSEs using different data bases and estimation methods. One can compare (a) the 1999-2001 market shares for FHA and the conventional conforming market in metropolitan areas calculated using the same methodology as Table A.25 with (b) the 1999-2001 market share estimates reported in Table A.25 for the entire mortgage market (including jumbo loans and covering non-metropolitan areas as well as metropolitan areas). The results are strikingly consistent. For the 1999-to-2001 period, the FHA share of the overall (African American and Hispanic) home loan market is estimated to be 19.0 percent (35.8 percent) under (a) versus 16.4 percent (31.2 percent) under (b). Lower percentage shares are expected for (b) because (b) includes jumbo loans. For the same period, the GSE share of the overall (African American and Hispanic) home loan market is estimated to be 46.0 percent (25-28 percent) under (a) versus 41.5 percent (24.3 percent) under (b).

309.  For other analyses of the GSEs' market role, see the following study by economists at the Federal Reserve Board: Glenn B. Canner, Wayne Passmore, and Brian J. Surette, “Distribution of Credit Risk among Providers of Mortgages to Lower-Income and Minority Homebuyers” in Federal Reserve Bulletin, 82(12): 1077-1102, December, 1996. This study considered several characteristics of the GSEs' loan purchases (such as amount of downpayment) and concluded that the GSEs have played a minimal role in providing credit support for underserved borrowers.

310.  Canner, et al., op. cit.

311.  The Impact of Secondary Mortgage Market and GSE Purchases on Underserved Neighborhood Housing Markets: Final Report to HUD. July 2002.

312.  GSE Service to Rural Areas, 2002.

313.  An Analysis of the Effects of the GSE Affordable Goals on Low- and Moderate-Income Families, 2001.

314.  Van Order, Robert. 1996. “Discrimination and the Secondary Mortgage Market.” In John Goering and Ronald Wienk, eds. Mortgage Discrimination, Race, and Federal Policy. The Urban Institute Press, Washington, DC: 335-363.

315.  Are the GSEs Leading, and if So Do They Have Any Followers? An Analysis of the GSEs' Impact on Home Purchase Lending to Underserved Markets During the 1990s. University of Notre Dame Working Paper and Technical Series Number 2003-2. 2002.

316.  A detailed discussion of the GSEs' activities in this area is contained in Theresa R. Diventi, The GSEs' Purchases of Single-Family Rental Property Mortgages, Housing Finance Working Paper No. HF-004, Office of Policy Development and Research, Department of Housing and Urban Development, (March 1998).

317.  Senate Report 102-282, May 15, 1992, p. 35.

318.  Tables A.30 and A.31 examine GSE purchases on a “going forward basis by origination year”. Specifically, it considers GSE purchases of: (a) 2000 mortgage originations during 2000, 2001, 2002 and 2003; (b) 2001 originations during 2001, 2002 and 2003; and (c) 2002 originations during 2002 and 2003. In other words, this analysis looks at the GSEs' purchases of a particular origination year cohort through 2003. This approach contrasts with the approach that examines GSE purchases on a “backward looking basis by purchase year”, for example, GSE purchases during 2000 of both new 2000 originations and originations during previous years (the latter called “prior-year” or seasoned loans). Either approach is a valid method for examining GSE purchases; in fact, when analyzing aggregated data such as the combined 1999-2002 data in Table A.30, the two approaches yield somewhat similar results. HUD's methodology for deriving the market estimates is explained in Appendix D. B&C loans have been excluded from the market estimates in Tables A.30 and A.31.

319.  Based on Table A.30, multifamily properties represented 14.8 percent of total units financed between 1999 and 2002 (obtained by dividing 7,018,044 multifamily units by 47,551,039 “Total Market” units). Increasing the single-family-owner number in Table A.30 by 2,648,757 to account for excluded B&C mortgages increases the “Total Market” number to 50,199,796, which produces a multifamily share of 14.0 percent. See Appendix D for discussion of the B&C market.

320.  Abt Associates, op. cit. (August 2002).

321.  The problem of secondary market “adverse selection” is described in James R. Follain and Edward J. Szymanoski. “A Framework for Evaluating Government's Evolving Role in Multifamily Mortgage Markets,” Cityscape: A Journal of Policy Development and Research 1(2), 1995.

322.  This section is based heavily on “DU and LP Usage Continues to Rise,” in Inside Mortgage Technology published by Inside Mortgage Finance, January 27, 2003, page 1-2.

323.  Fannie Mae, 2002 Annual Housing Activities Report, 2003, pp. 10-11.

324.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 14.

325.  Freddie Mac, 2002 Annual Housing Activities Report, 2003, p. 52.

326.  Inside Mortgage Finance, “Online Volume Comprises One-Fourth of Total Originations in First Half '02,” September 20, 2002, p. 8.

327.  Freddie Mac, Opening Doors for America's Families: Freddie Mac's Annual Housing Activities Report for 2003, March 15, 2004, p. 38.

328.  Fannie Mae, 2002 Annual Housing Activities Report, 2003, p. 12.

329.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, p. 13.

330.  Fannie Mae, 2003 Annual Housing Activities Report, 2004, p. 13.

331.  Fannie Mae, 2002 Annual Housing Activities Report, 2003, p. 10.

332.  The 22% decrease in Fannie Mae's 2002 net income resulted primarily from a $4.508 billion increase in purchased options expense, which occurred due to an increase in the notional amount of purchased options outstanding and the declining interest rate environment. Recorded purchased options expense for 2001 was only $37 million by comparison. Fannie Mae 2002 Annual Report, 2003, p. 23.

333.  Fannie Mae, 2003 Annual Report, “Financial Highlights.”

334.  Fannie Mae, 2003 Annual Report, “Financial Highlights.”

335.  Fannie Mae, 2003 Annual Report, “Financial Highlights” and United States Securities and Exchange Commission form 10-K, p. 108.

336.  Fannie Mae, 2003 Annual Report to Shareholders, Financial Highlights and Financial Information.

337.  Freddie Mac, Consolidated Statements of Income 2003 and Freddie Mac Core Tables 2003.

338.  “The Standard and Poor's Five Hundred: Performance Ranking S&P 500”, Business Week, April 5, 2004, p. 127.

339.  “Fortune 500 Largest U.S. Corporations,” Fortune, April 5, 2004, p. F-1.

340.  “Fortune 500 Largest U.S. Corporations,” Fortune, July 26, 2004, p. 159.

341.  National Association of Home Builder, 2004 Spring Construction Forecast Conference, April 21, 2004.

342.  As shown in Table A.31b, the GSEs' share of the rental market increases to 41 percent when a lower multifamily share is assumed in the market analyses.

343.  Senate Report 102-282, May 15, 1992, p. 36.

1.  This analysis excludes Puerto Rico. In addition, tracts are excluded if median income is suppressed in the underlying census data. There are 379 such tracts. When reporting analysis of mortgage loan denial, origination, and application rates later in this appendix, tracts are excluded if there are no purchase or refinance applications. Tracts are also excluded if: (1) Group quarters constitute more than 50 percent of housing units or (2) there are less than 15 home purchase applications in the tract and the tract denial rates equal 0 or 100 percent. Excluded tracts account for a small percentage of mortgage loan applications (1.4 percent). These tracts are not excluded from HUD's underserved areas if they meet the income and minority thresholds. Rather, the tracts are excluded to remove the effects of outliers from the analysis.

2.  Kalawao County, Hawaii, which has a very small population, is excluded from the analysis for 1990 but included for 2000.

3.  In this appendix, the term “central city” is used to mean “OMB-designated central city.”

4.  The actual denial rates were as follows: 23.6 percent for low-income (80% AMI or less) African Americans, 15.5 percent for upper-income (120% AMI or more) African Americans. 11.4 percent for low-income Whites, and 5.6 percent for upper-income Whites. The overall denial rate in the conventional conforming home purchase market was 9.7 percent in 2002. The data exclude applications to lenders that specialize in manufactured home lending.

5.  Alicia H. Munnell, Lynn E. Browne, James McEneaney, and Geoffrey M.B. Tootell, “Mortgage Lending in Boston: interpreting HMDA Data,” American Economic Review, March 1996.

6.  William C. Hunter, “The Cultural Affinity Hypothesis and Mortgage Lending Decisions,” WP-95-8, Federal Reserve Bank of Chicago, 1995. Hunter confirmed that race was a factor in denial rates of marginal applicants. While denial rates were comparable for borrowers of all races with “good” credit ratings, among those with “bad” credit ratings or high debt ratios, minorities were significantly more likely to be denied than similarly-situated whites. The study concluded that the racial differences in denial rates were consistent with a cultural gap between white loan officers and minority applicants, and conversely, a cultural affinity with white applicants.

7.  For a reassessment of the Boston Fed study, see Stephen Ross and John Yinger, The Color of Credit, MIT Press 2002, and other studies cited there.

8.  Since upfront loan fees are frequently determined as a percentage of the loan amount, lenders are discouraged from making smaller loans in older neighborhoods, because such loans generate lower revenue and are less profitable to lenders.

9.  Traditional underwriting practices may have excluded some lower income families that are, in fact, creditworthy. Such families tend to pay cash, leaving them without a credit history. In addition, the usual front-end and back-end ratios applied to applicants' housing expenditures and other on-going costs may be too stringent for lower income households, who typically pay larger shares of their income for housing (including rent and utilities) than higher income households.

10.  Margery A. Turner and Felicity Skidmore, eds., Mortgage Lending Discrimination: A Review of Existing Evidence. The Urban Institute: Washington, DC, June 1999.

11.  Margery Austin Turner, All Other Things Being Equal: A Paired Testing Study of Mortgage Lending Institutions, The Urban Institute Press, April 2002.

12.  Margery Austin Turner, Stephen L. Ross, George Galster, and John Yinger, Discrimination in Metropolitan Housing Markets, The Urban Institute Press, November 2002.

13.  How Much Do We Know? Public Awareness of the Nation's Fair Housing Laws, prepared for HUD by Martin D. Abravanel and Mary K. Cunningham of the Urban Institute, April 2002.

14.  U.S. Bureau of the Census, August 2002. The co-authors of the study were John Iceland and Daniel H. Wienberg. For a summary of the study, see “Residential Segregation Still Prevalent”, National Mortgage News, January 6, 2003, page 1.

15.  See Randall M. Scheessele, Black and White Disparities in Subprime Mortgage Refinance Lending, Housing Finance Working Paper No. HF-14, Office of Policy Development and Research, U.S. Department of Housing and Urban Development, April 2002.

16.  These studies, which were conducted at the census tract level, typically involved regressing the number of mortgage originations (relative to the number of properties in the census tract) on characteristics of the census tract including its minority composition. A negative coefficient estimate for the minority composition variable was often interpreted as suggesting redlining. For a discussion of these models, see Eugene Perle, Kathryn Lynch, and Jeffrey Horner, “Model Specification and Local Mortgage Market Behavior,” Journal of Housing Research, Volume 4, Issue 2, 1993, pp. 225-243.

17.  For critiques of the early HMDA studies, see Andrew Holmes and Paul Horvitz, “Mortgage Redlining: Race, Risk, and Demand,” The Journal of Finance, Volume 49, No. 1, March 1994, pp. 81-99; and Michael H. Schill and Susan M. Wachter, “A Tale of Two cities: Racial and Ethnic Geographic Disparities in Home Mortgage Lending in Boston and Philadelphia,” Journal of Housing Research, Volume 4, Issue 2, 1993, pp. 245-276.

18.  Like early HMDA studies, an analysis of deed transfer data in Boston found lower rates of mortgage activity in minority neihborhoods. The discrepancies held even after controlling for income, house values and other economic and non-racial factors that might explain differences differences in demand and housing market activity. The study concluded that “the housing market and the credit market together are functioning in a way that has hurt African American neighborhoods in the city of Boston.” Katherine L. Bradbury, Karl E. Case, and Constance R. Dunham, “Geographic Patterns of Mortgage Lending in Boston, 1982-1987,” New England Economic Review, September/October 1989, pp. 3-30.

19.  Using an analytical approach similar to that of Bradbury, Case, and Dunham, Anne Shlay found evidence of fewer mortgage loans originated in black census tracts in Chicago and Baltimore. See Anne Shlay, “Not in That Neighborhood: The Effects of Population and Housing on the Distribution of Mortgage Finance within the Chicago SMSA,” Social Sciene Research, Volume 17, No. 2, 1988, pp. 137-163; and “Financing Community: Methods for Assessing Residential Credit Disparities, Market Barriers, and Institutional Reinvestment Performance in the Metropolis,” Journal of Urban Affiars, Volume 11, No. 3, 1989, pp. 201-223.

20.  Holmes and Horvitz, op. cit.

21.  Schill and Wachter, op. cit.

22.  Schill and Wachter, page 271. Munnell, et al. reached similar conclusions in their study of Boston. They found that the race of the individual mattered, but that once individual characteristics were controlled, racial composition of the neighborhood was insignificant.

23.  Fred J. Phillips-Patrick and Clifford V. Rossi, “Statistical Evidence of Mortgage Redlining? A Cautionary Tale”, The Journal of Real Estate Research, Volume 11, Number 1, 1996, pp.13-23.

24.  Samuel L. Myers, Jr. and Tsze Chan, “Racial Discrimination in Housing Markets: Accounting for Credit Risk”, Social Science Quarterly, Volume 76, Number 3, September 1995, pp. 543-561.

25.  For another study that uses HMDA data on reasons for denial to construct a proxy for bad credit, see Steven R. Holloway, “Exploring the Neighborhood Contingency of Race Discrimination in Mortgage Lending in Columbus, Ohio”, Annals of the Association of American Geographers, Volume 88, Number 2, 1998, pp. 252-276. Holloway finds that mortgage denial rates are higher for black applicants (particularly those who are making large loan requests) in all-white neighborhoods than in minority neighborhoods, while the reverse is true for white applicants making small loan requests.

26.  See Geoffrey M. B. Tootell, “Redlining in Boston: Do Mortgage Lenders Discriminate Against Neighborhoods?”, Quarterly Journal of Economics, 111, November, 1996, pp. 1049-1079; and “Discrimination, Redlining, and Private Mortgage Insurance”, unpublished manuscript, October 1995.

27.  Tootell notes that both omitted variables and the strong correlation between borrower race and neighborhood racial composition in segregated cities have made it difficult for previous studies to distinguish the impacts of geographic redlining from the effects of individual borrower discrimination. He can unravel these effects because he includes a direct measure of credit history and because over half of minority applicants in the Boston Fed data base applied for mortgages in predominately white areas.

28.  Stephen L. Ross and Geoffrey M. B. Tootell, “Redlining, the Community Reinvestment Act, and Private Mortgage Insurance”, unpublished manuscript, March 1999.

29.  William W. Lang and Leonard I. Nakamura, “A Model of Redlining,” Journal of Urban Economics, Volume 33, 1993, pp. 223-234.

30.  Paul S. Calem, “Mortgage Credit Availability in Low- and Moderate-Income Minority Neighborhoods: Are Information Externalities Critical?” Journal of Real Estate Finance and Economics, Volume 13, 1996, pp. 71-89.

31.  David C. Ling and Susan M. Wachter, “Information Externalities and Home Mortgage Underwriting,” Journal of Urban Economics, Volume 44, 1998, pp. 317-332.

32.  Robert B. Avery, Patricia E. Beeson, and Mark S. Sniderman, “Neighborhood Information and Home Mortgage Lending,” Journal of Urban Economics, Volume 45, 1999, pp. 287-310.

33.  William Shear, James Berkovec, Ann Dougherty, and Frank Nothaft, “Unmet Housing Needs: The Role of Mortgage Markets,” Journal of Housing Economics, Volume 4 , 1996, pp. 291-306. These researchers regressed the number of mortgage originations per 100 properties in the census tract on several independent variables that were intended to account for some of the demand and supply (i.e., credit risk) influences at the census tract level. See also Susan Wharton Gates, “Defining the Underserved,” Secondary Mortgage Markets, 1994 Mortgage Market Review Issue, 1995, pp. 34-48.

34.  See Avery, et al.

35.  Methodological and econometric challenges that researchers will have to deal with are discussed in Mitchell Rachlis and Anthony Yezer, “Serious Flaws in Statistical Tests for Discrimination in Mortgage Markets,” Journal of Housing Research, Volume 4, 1993, pp. 315-336.

36.  The purchase affordability index assesses the extent to which a family with the median income of a given area would be able to afford a housing unit that carries the median purchase price of that area. For example, a purchase affordability index number less than 100 means that a family with the median income would not qualify for a mortgage on a unit with the median value; a purchase affordability index equal to 100 means that a family with the median income has exactly the level of income needed to qualify for a mortgage on a unit with the median value; and an index number greater than 100 means that a family with the median income has 20 percent more than the level of income needed to qualify for a mortgage on a unit with the median value. The rental affordability index is similarly constructed.

37.  J.J. Mikesell, “Housing Problems across Types of Rural Households”, Rural Conditions and Trends, Volume 9, Number 2, pp. 97-101, 1999.

38.  Performance for the 1993-95 period was discussed in the October 2000 rule.

39.  To separate out the effects of changes in counting rules that took effect in 2001, this section also compares performance in 2001 to estimated performance in 2000 if the 2001 counting rules had been in effect in that year.

40.  Unlike the low- and moderate-income and special affordable goals, there is no exclusion of units from the denominator for units with missing information about the area in which a property is located. That is, such units are counted in the denominator, but not in the numerator, in determining underserved areas goal performance.

41.  See Congressional Record, December 15, 2000, pp. H12295-96.

42.  65 FR 65141 & n. 145 (2000).

43.  In New England, MSAs were defined through mid-2003 in terms of Towns rather than Counties, and the portion of a New England county outside of any MSA is regarded as equivalent to a county in establishing metropolitan or non-metropolitan location of a property. The MSA definitions established by the Office of Management and Budget (OMB) in June, 2003 defined MSAs in New England in terms of counties.

44.  The procedure used to generate estimated rents in connection with the Low- and Moderate Income and Special Affordable Housing Goals, as mentioned in Appendixes A and C, uses similar data series.

45.  HUD has deferred application of the 2000 census data and 2003 MSA designations to 2005, pending completion of the present rulemaking process.

46.  8,717 tracts included both served and underserved area, out of a total of 61,493 tracts that could be classified as served or underserved or assigned an underservice factor.

47.  Heather MacDonald, “Fannie Mae and Freddie Mac in Nonmetropolitan Housing Markets: Does Space Matter?” Cityscape: A Journal of Policy Development and Research, Volume 5, 2001, pp. 219-264.

48.  Jeanette Bradley, Noah Sawyer and Kenneth Temkin, Factors Influencing GSE Service to Rural Areas, The Urban Institute, prepared for U.S. Department of Housing and Urban Development, 2002.

49.  Affordable loans are defined as borrowers earning less than 80 percent the Area Median Income.

50.  Underserved areas make up about 56 percent of the census tracts in nonmetropolitan areas and 47 percent of the census tracts in metropolitan areas. This is one reason why underserved areas comprise a larger portion of the GSEs' single-family mortgages in nonmetropolitan areas (39 percent) than in metropolitan areas (23 percent).

51.  60 FR 61,925-58 (1995) (Appendix B).

52.  In areas with 30 percent or greater minority population, all families with income in excess of 120 percent of the greater of State or national median income are counted as qualifying as “in need” for these computations. Similarly, in areas with less than 30 percent minority, those minority (headed) families with income between 95 and 120 percent of the applicable median income are not classified as “in need.”

53.  A more comprehensive presentation of this analysis may be found in Economic Systems, Inc., Indicators of Mortgage Market Underservice in Non-Metropolitan Areas, Interim Report to HUD, March 2003, Chapter 6.

54.  Note that, unlike the other panels in tables 6.3 and 6.8, “underserved population” is defined according to the applicable definition. Thus, eliminating the national median income test, narrows the defined cohort of underserved families. Despite this, coverage falls.

55.  Denial rates are computed for mortgage applications without manufactured housing loans. Origination rates equal home purchase and refinance mortgages (without subprime loans) per 100 owner occupants in a census tract.

56.  The differentials in denial rates are due, in part, to differing risk characteristics of the prospective borrowers in different areas. However, use of denial rates is supported by the findings in the Boston Fed study which found that denial rate differentials persist, even after controlling for risk of the borrower. See Section B for a review of that study.

57.  See Dan Immergluck, Star Differences: The Explosion of the Subprime Industry and Racial Hypersegmentation in Home Equity Lending, Woodstock Institute, October 2000; and Daniel Immergluck and Marti Wiles, Two Steps Back: The Dual Mortgage Market, Predatory Lending, and the Undoing of Community Development, Woodstock Institute, Chicago, IL, November 1999. For a national analyses, see the HUD report Unequal Burden: Income and Racial Disparities in Subprime Lending in America, April 2000; and Randall M. Scheessele, Black and White Disparities in Subprime Mortgage Refinance Lending, Housing Finance Working Paper No. HF-114, Office of Policy Development and Research, U.S. Department of Housing and Urban Development, April 2002.

1.  Performance for the 1993-95 period was discussed in HUD's Housing Goals 2000 Finale Rule.

2.  To separate out the effects of changes in counting rules that took effect in 2001, this section also compares performance in 2001 to estimated performance in 2000 if the 2001 counting rules had been in effect in that year. Freddie Mac's goal performance in 2002 has been revised due to coding errors that were discovered in HUD's review of 2002 data, as discussed in HUD's press release No. 04-105, October 15, 2004.

3.  During 1996-2000 Freddie Mac took steps to acquire representations and warranties from lenders to attest that they were “recycling” the proceeds from the sales of qualifying loans. Fannie Mae did not take such steps; rather, Fannie Mae excluded such loans from the denominator in making its own calculations of its special affordable goal performance. In 1996-2000 HUD counted all eligible loans in the denominator, and, in the absence of measures to verify “recycling” by Fannie Mae, did not award credit in the numerator of the special affordable goal for most of Fannie Mae's seasoned mortgage purchases.

4.  See Congressional Record, December 15, 2000, pp. H12295-96.

5.  The revised requirements are codified at 24 CFR 81.14(e)(4). The changes are discussed in detail in the rule preamble, 68 FR 65074-76 (October 31, 2000).

6.  Exclusion of loans with missing information had a greater impact on Fannie Mae's goal performance than on Freddie Mac's goal performance.

7.  “Fannie Courting Multifamily Sellers; Small Banks Balking.” American Banker, January 13, 2003, p.1.

8.  In New England, MSAs were defined through mid-2003 in terms of Towns rather than Counties, and the portion of a New England county outside of any MSA was regarded as equivalent to a county in establishing the metropolitan or non-metropolitan location of a property. The MSA definitions established by the Office of Management and Budget (OMB) in June 2003 defined MSAs in New England in terms of counties.

9.  HUD has deferred application of the 2003 MSA specification to 2005, pending completion of the present rulemaking process.

10.  Tabulations of the 2001 American Housing Survey by HUD's Office of Policy Development and Research. The results in the table categorize renters reporting housing assistance as having no housing problems.

1.  Dixie M. Blackley and James R. Follain, “A Critique of the Methodology Used to Determine Affordable Housing Goals for the Government Sponsored Housing Enterprises,” unpublished report prepared for Office of Policy Development and Research, Department of Housing and Urban Development, October 1995; and “HUD's Market Share Methodology and its Housing Goals for the Government Sponsored Enterprises,” unpublished paper, March 1996.

2.  See Freddie Mac, “Comments on Estimating the Size of the Conventional Conforming Market for Each Housing Goal: Appendix III to the Comments of the Federal Home Loan Mortgage Corporation on HUD's Regulation of the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac)”, May 8, 2000, page 1.

3.  See Fannie Mae, “Fannie Mae's Comments on HUD's Regulation of the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac)”, May 8, 2000, page 53.

4.  See Freddie Mac, “Comments of the Federal Home Loan Mortgage Corporation on HUD's Proposed Housing Goals for the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac) for the Years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac,” July 16, 2004; and Fannie Mae, “Fannie Mae's Comments on HUD's Proposed Housing Goals for the Federal National Mortgage Association (Fannie Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac) for the Years 2005-2008 and Amendments to HUD's Regulation of Fannie Mae and Freddie Mac,” July 16, 2004.

5.  Readers not interested in this overview may want to proceed to Section C, which begins the market analysis by examining the size of the multifamily market.

6.  Sections 1332(b)(4), 1333(a)(2), and 1334(b)(4).

7.  So-called “jumbo” mortgages, greater than $333,700 in 2004 for 1-unit properties, are excluded in defining the conforming market. There is some overlap of loans eligible for purchase by the GSEs with loans insured by the FHA and guaranteed by the Veterans Administration.

8.  The owner of the SF 2-4 property is counted in (a).

9.  Property types (b), (c), and (d) consist of rental units. Property types (b) and (c) must sometimes be combined due to data limitations; in this case, they are referred to as “single-family rental units” (SF-R units).

10.  The property shares and low-mod percentages reported here are based on one set of model assumptions; other sets of assumptions are discussed in Section E.

11.  This goal will be referred to as the “Underserved Areas Goal”.

12.  The example in Table D.1 is based on 1990 Census tract geography. As explained in Section G, switching to 2000 Census tract geography (scheduled for 2005) increases the underserved areas market share by approximately five percentage points.

13.  This section is based on analysis by Jack Goodman under contract with the Urban Institute.

14.  1990 is excluded from this calculation because of the unusually high multifamily mix that year. Also, the estimated multifamily mix from the HUD New Method is also provided for 2002 since it was greater than the estimate from the Flow of Funds method.

15.  The projection model for 2002 showed the following multifamily mixes for 2002: 11.1 percent for the HUD New multifamily estimate ($67.4 billion); 10.5 percent for the top end ($64 billion) of the Flow of Funds multifamily range ($60-64 billion), 10.3 percent for the mid-point ($62 billion), and 9.9 percent for the low end ($60 billion). In Sections F-H, HUD will consider multifamily mixes as low as 9.5 percent for 2002.

16.  For purposes of sensitivity analysis, the lower MF mixes were derived as follows: three percentage points were subtracted from the 1995-1997 mid-point MF mixes, which were in the high 18-to-21-percent range; two percentage points were subtracted from the 1998-2000 mid-point MF mixes, which were in the 14-to-17-percent range; and 1.5 percentage points were subtracted from the 2001-2002 mid-point MF mixes, which were less than 13 percent.

17.  HUD estimated ICF's MF mixes by including subprime loans in the data that ICF reported on pages 58-60 of its Appendix (for the Best Estimate) and on pages 63-65 of its Appendix (for the Lower Bound Estimate). To the extent that ICF also excluded other single-family loans (in addition to subprime SF loans), the estimates reported in the text overstate ICF's initial MF mixes.

18.  The multifamily origination data in this paragraph reflect a recent release of the RFS; other single-family and multifamily data in this section draw from an earlier version of the RFS. HUD will continue its analysis of the RFS data as new versions are released by the Census Bureau.

19.  Estimates of new MF units were created by comparing the historical estimates of numbers of units added by HUD and REIS, creating a ratio, and then applying that ratio to the REIS' future projections.

20.  The HMDA data reported in this section ignore HMDA loans with “non-applicable” for owner type.

21.  These data without subprime loans are presented merely to provide a sense of the likely changes if one excludes subprime investor loans. Three comments should be made about them. First, HUD's procedure is to drop one-half of subprime loans as a proxy for B&C loans, which one reduce the one-half percentage point differential mentioned in the text to a one-quarter point percentage differential. Second, the comparisions in Table D.6a do not deduct single-family owner subrpime loans; doing that would raise the investor shares from those in middle portion of the table. Third, HUD's model starts with investor and owner property shares that include subprime loans (such as those in the top portion of Table D.6a) and then excludes the subprime loans as part of the derivations within the model. See Section F for an explanation of this procedure.

22.  Dixie M. Blackley and James R. Follain, “A Critique of the Methodology Used to Determine Affordable Housing Goals for the Government Sponsored Housing Enterprises,” report prepared for Office of Policy Development and Research, Department of Housing and Urban Development, October 1995; and “HUD's Market Share Methodology and its Housing Goals for the Government Sponsored Enterprises,” unpublished paper, March 1996.

23.  Blackley and Follain (1996), p. 20.

24.  It should be mentioned that ICF's 12.0 percent assumption for the SF rental share seems at odds with ICF's Exhibit 6.4, which suggests that ICF's 1994-2002 average SF rental share is 14.9 percent. A 14.9 percent SF rental share would be consistent with a 12 percent investor mortgage share.

25.  The property distribution reported in Table D.1 is an example of the output of the market share model. Thus, this section completes Step 1 of the three-step procedure outlined above in Section B.

26.  According to estimates by the Mortgage Bankers Association of America (MBAA), the conventional share of the 1-4 family market was between 86 and 88 percent of the market from 1993 to 1999, with a one-time low of 81 percent in 1994. Calculated from “1-4 Family Mortgage Originations” tables (Table 1—Industry and Table 2—Conventional Loans) from “MBAA Mortgage and Market Data,” at www.MBAAa.org/​marketkdata/​ as of July 13, 2000. More recent unpublished estimates by MBAA are slightly higher. As discussed in the text, the market sizing shares are affected by parameters other than this one, such as the multifamily share of newly-mortgaged dwelling units.

27.  In its August 17, 2004 forecast, Fannie Mae projected approximately $1.6 billion for 2005 and 2006 while the MBAA projected $1.8 billion for 2005 in its August 13, 2004 forecast. As discussed later, single-family originations could differ from $1,700 billion during the 2005-2008 period that the goals will be in effect. As recent experience shows, market projections often change. For example, in January 2003, the MBAA projected $1,246 billion for 2003; of course, actual 2003 mortgage originations were triple the latter amount. (See http://www.MBAAa.org/​marketdata/​forecasts for January 2003 Mortgage Finance Forecasts.) While Sections F-H will report the effects on the market estimates of alternative estimates of single-family mortgage originations, it should be emphasized that the important parameter for the market sizing estimates is the share of single-family-owner units relative to the share of single-family and multifamily rental units, not the absolute level of single-family originations.

28.  The model requires an estimated refinance rate because purchase and refinance loans can have different shares of goals-qualifying units. In 2003, the refinance rate was almost 70 percent. In its August 13, 2004 forecast, the MBAA projects 25 percent for 2005, as did Fannie Mae in its August 17, 2004 forecast. The baseline model uses a higher refinance rate of 35 percent because conforming conventional loans tend to refinance at a higher rate than the overall market. Sensitivity analyses for alternative refinance rates are presented in Sections F-H.

29.  The average 2002 purchase loan amount is estimated at $135,060 for owner occupied units using 2002 HMDA average loan amounts for single-family home purchase loans in metropolitan areas. A small adjustment is made to this figure to account for a small number of two-to-four and investor properties (see Section D above). This produces an average purchase loan size of $133,458 for 2002 which is then inflated 3 percent a year for three years and then rounded to arrive at an estimated $146,000 average loan size for home purchase loans in 2005.

30.  The average refinance loan amount is estimated by averaging the relationship between HMDA average purchase and refinance loan amounts for 1999 and 2000, which were non-refinance environments. Applying this average of 90 percent (refinance loan amount/purchase loan amount) to the $146,000 average loan amount for purchase loans gives a rounded estimate of $131,000 for average refinance loan amounts. When refinance environments are used, $146,000 average loan amounts are used for both purchase and refinance loans. This relationship is consistent with the observed relationship in past refinance years such as 1998, 2001, and 2002.

31.  Based on the 2001 RFS, there is an average of 2.2 housing units per mortgage for 2-4 properties and 1.3 units per mortgage for single-family investor properties. See earlier discussion.

32.  The share of the mortgage market accounted for by owner occupants is (SF-O)/TOTAL; the share of the market accounted for by all single-family rental units is SF-RENTAL/TOTAL; and so on.

33.  Because of rounding, the two rental component shares do not add to the overall rental share.

34.  Because of rounding, the two rental component shares do not add to the overall rental share.

35.  HMDA data are expressed in terms of number of loans rather than number of units. In addition, HMDA data do not distinguish between owner-occupied one-unit properties and owner-occupied 2-4 properties. This is not a particular problem for this section's analysis of owner incomes.

36.  Sensitivity analyses will focus on how the results change during a heavy refinancing environment.

37.  The annual averages of the goals-qualifying mortgages reported in this appendix are unweighted averages; for analyses using weighted averages see Appendix A.

38.  Between 1999 and 2002, the average single-family-owner differential between the historical and projected low-mod percentages was 1.1 percentage point for Fannie Mae and 1.3 percentage point for Freddie Mac.

39.  See Randall M. Scheessele, 1998 HMDA Highlights, op. cit. and “HUD Subprime and Manufactured Home Lender List” at http://www.huduseer.org/​datasets/​manu.html.

40.  Since most HMDA data are for loans in metropolitan areas and a substantial share of manufactured homes are located outside metropolitan areas, HMDA data may not accurately state the goals-qualifying shares for loans on manufactured homes in all areas.

41.  While many fewer manufactured home loans were identified in the 2002 and 2003 HMDA data, the loans showed similar goals-qualifying shares: low-mod (77.6 percent and 75.4 percent, respectively), special affordable (45.0 percent and 47.1 percent, respectively), and underserved areas (46.9 percent and 45.2 percent, respectively).

42.  In 2002, 75 percent of GSE purchases of single-family rental units and 89 percent of their purchases of multifamily units qualified under the Low- and Moderate-Income Goal, excluding the effects of missing data.

43.  Applying Fannie Mae's weights to data from the 2003 AHS produces low-mod shares of slightly over 90 percent for both single-family and multifamily rental properties.

44.  These three estimates were initially reported in HUD's 2000 Final Rule, and repeated in Table D.9 of Appendix D of the 2004 proposed GSE rule.

45.  Given that the midpoints of the multifamily mixes for 1995-1997 are in the high 18-20 percent range (see Table D.5b), three percentage points were dropped in the sensitivity analysis.

46.  To provide some confirmation for these 1995-1997 estimates, HUD went back and re-estimated the model for 1997. As shown in Table D.9 of the 2004 GSE Proposed Rule (as well as in Table D.15 of the 2000 GSE Rule), HUD had earlier estimated a low-mod share of 57.5 percent for 1997 (which was about the same as the 57.3-percent low-mod share estimated for 1995 and 1996). With a lower investor share (8.4 percent instead of 10.0 percent) and other changes mentioned in the text, the new estimate for the 1997 low-mod market was 56.4 assuming a multifamily mix of 19.3 percent. If the multifamily mix is reduced to 17.3 (16.3) percent, the low-mod share of the 1997 market is 55.5 (55.0) percent. The 55.0-56.4 percent range for 1997 is the same as the range reported in the text for 1995-1997.

47.  The baseline estimates for 1998 (51.9 percent), 2001 (53.4 percent) and 2002 (53.2 percent) are lower than those (53.8 percent, 54.9 percent and 54.1 percent, respectively) reported in Table D.9 of Appendix D of the proposed rule. As explained earlier, the differences between the results in the proposed rule and this Final Rule are mainly due to the treatment of single-family rental mortgages. (In addition, the SF owner percentages for 2002 were also lowered by approximately 0.5 percentage point in the Final Rule.) Notice that in 1998, the investor mortgage share dropped to 6.8 percent, or 3.2 percentage points lower than that assumed in the proposed rule; this differential accounts for the reduction of 1.9 percentage points (53.8 percent to 51.9 percent) in the low-mod market estimate for 1998.

48.  Section 1336(b)(3)(A).

49.  Estimates of the subprime market for all years since 1995 are as follows (dollar and market share): 1995 ($65 billion, 10 percent); 1996 ($96.5 billion, 12.3 percent); 1997 ($125 billion, 15 percent); 1998 ($150 billion, 10 percent; 1999 ($160 billion, 12.5 percent); 2000 ($138 billion, 12.1 percent); 2001 ($174 billion, 8.5 percent); 2002 ($213 billion, 8.6 percent), and 2003 ($332 billion, 8.7 percent). The uncertainty about what these various estimates include should be emphasized; for example, they may include second mortgages and home equity loans as well as first mortgages, which are the focus of this analysis. The source for these estimates is Inside Mortgage Finance (various years).

50.  The one-half assumption for A-minus loans is conservative because it probably underestimates (overestimates) the share of A-minus (B&C) loans. According to data obtained by the Mortgage Information Corporation (see next footnote), 57 percent of all subprime loans were labeled A-minus (as of September 30, 2000). According to Inside B&C Lending, which is published by Inside Mortgage Finance, the A-minus share of the subprime market was 61.6 percent in 2000, 70.7 percent in 2001 (see March 11, 2002 issue), 75 percent in 2002 (see the September 15, 2003 issue), and 82 percent during the first nine months of 2003 (see the December 8, 2003 issue). A more recent analysis by Inside Mortgage Finance found that 81.4 percent of subprime loans originated during the first quarter of 2002 were A-minus or better (see Inside B&C Lending, Vol. 9, Issue 12, June 14, 2004).

51.  The Mortgage Information Corporation (MIC) reports the following serious delinquency rates (either 90 days past due or in foreclosure) by type of subprime loan: 3.36 percent for A-minus; 6.67 percent for B; 9.22 percent for C; and 21.03 percent for D. The D category accounted for only 2 percent of subprime loans and of course, is included in the “B&C” category referred to in this appendix. By comparison, MIC reports a seriously delinquent rate of 3.63 percent for FHA loans. See MIC, The Market Pulse, Winter 2001, page 6.

52.  The goals-qualifying percentages for subprime lenders are much higher than the percentages for the overall single-family conventional conforming market; for example, the 1999-2003 average low-mod percentage for all single-family owner loans was 44 percent. For further analysis of subprime lenders, see Randall M. Scheessele, 1998 HMDA Highlights, Housing Finance Working Paper No. HF-009. Office of Policy Development and Research, U.S. Department of Housing and Urban Development, October 1999.

53.  Dropping B&C loans in the manner described in the text results in the goals-qualifying percentages for the non-B&C market being underestimated since HMDA coverage of B&C loans is less than that of non-B&C loans and since B&C loans have higher goals-qualifying shares than non-B&C loans. For instance, the low-mod shares of the market reported in the text underestimate (to an unknown extent) the low-mod shares of the market inclusive of B&C loans; so reducing the low-mod owner shares by dropping B&C loans in the manner described in the text would provide an underestimate of the low-mod share of the non-B&C owner market. A study of 1997 HMDA data in Durham County, North Carolina by the Coalition for Responsible Lending (CRL) found that loans by mortgage and finance companies are often not reported to HMDA. For a summary of this study, see “Renewed Attack on Predatory Subprime Lenders” in Fair Lending/CRA Compass, June 9, 1999.

54.  Table D.15 of the 2000 GSE Rule also reported underserved area shares of 33.9 percent for 1995 and 1997 and 33.4 percent for 1996. These estimates, after adjustments for a lower HMDA-based mortgage investor share and a lower-than-baseline MF mix, would still remain in the 32-33 percent range. To provide some confirmation for this, HUD went back and re-estimated the model for 1997. As shown in Table D.15 of the 2000 GSE Rule, HUD had earlier estimated an underserved areas share of 33.9 percent for 1997 (which was the same as the 33.9-percent underserved areas estimate for 1995 and similar to the 33.4-percent estimate for 1996). With a lower investor share (8.4 percent instead of 10.0 percent) and other changes mentioned in the text, the new estimate for the 1997 underserved areas market was 32.7 assuming a multifamily mix of 19.3 percent. If the multifamily mix is reduced to 17.3 (16.3) percent, the underserved areas share of the 1997 market is 32.3 (32.0) percent. Thus, this 32.0-32.7 percent range for 1997 is consistent with a 32-33 percent range for 1995-1997.

55.  The baseline 34.9 percent estimate for 2000 is 0.4 percentage points lower than the 35.3 percent share reported in Table D.9 of the proposed rule. The difference is mostly explained by the different treatment of single-family rental mortgages.

56.  For the years 1999 to 2002, Fannie Mae estimated an underserved areas share of 32-33 percent. (See their Table I.9, page I-34.) This compares with HUD's estimate of 32.5 percent to 32.9 percent for the same period.

57.  Between 1999 and 2001, the non-metropolitan portion of the Underserved Areas Goal contributed 1.1 to 1.4 (0.7 to 1.3) percentage points to Freddie Mac's (Fannie Mae's) overall performance (i.e., including both metro and non-metro loans), compared with a goals-counting system that only included metropolitan areas.

58.  These data do not include loans originated by lenders that specialize in manufactured housing loans, as well as estimated B&C loans.

59.  Federal Housing Finance Board data.

60.  Mortgage Interest Rate Survey (MIRS) data reported by the Federal Housing Finance Board separate conventional home purchase loans by their metropolitan and non-metropolitan location. The average non-metropolitan share between 1999 and 2002 was about 13 percent.

61.  ICF incorrectly said HUD's baseline underserved areas share for MF rentals was 60 percent, rather than 58 percent (ICF Appendix, p. 47).

62.  Freddie Mac says “ICF estimates the multifamily underserved share to be just 56 percent and the single-family renter underserved area share to be just 50 percent” (at Appendix IV-24). However, ICF uses a 50 percent share in its projection model (ICF Appendix, p. 133); therefore, 55 percent is used here as the ICF number. Also, ICF's lower (upper bound) projection was 47 (53) percent for SF rental properties and 56 (58) percent for multifamily properties.

63.  “Affordability and Geographic Distribution of the Housing Stock and the Use of Mortgage Finance,” Abt Associates, October 22, 2001.

64.  As shown in Table D.12, excluding B&C investor loans reduces the market's underserved area share for SF investor loans. An adjustment for B&C investor loans is made within HUD's model, along the same lines as that B&C owner loans are excluded from the analysis. See Section F.3.c for further explanation.

65.  It is encouraging that the RFS underserved area percentage (31.7 percent) for SF-owner mortgages originated in metropolitan areas during 2001 was similar to the corresponding percentage (31.0 percent) reported by HMDA.

66.  In this case, the 2000-based underserved area percentages for years prior to 1999 (i.e. 1994 to 1998 in this example) are estimated by adding 4.9 percent to the corresponding 1990-based underserved area percentages reported in Table D.12. The 4.9 percent is the unweighted difference of the 2000-based and 1990-based underserved area shares for total (home purchase and refinance) SFO owner loans reported in Table D.14. This procedure will be used throughout this section.

67.  The market share estimates are interpolated from Table D.15. For example, the overall market estimate for a SFO percentage of 32.3 percent is obtained by adding [.3*(38.4 minus 37.6)] to 37.6, to obtain the 37.6 figure reported in the text.

68.  The baseline 39.7 percent estimate for 2000 is 0.7 percentage points lower than the 40.4 percent share reported in Section G.4 of Appendix D of the proposed rule, mainly for the reasons discussed in the previous footnote. The difference is mostly explained (a) by the different treatment of single-family rental mortgages and (b) by a 0.4 percentage point decline in HUD's projections (in terms of the 2000 Census data) of the 2000 underserved areas percentage for SF owners.

69.  As explained earlier in Section G.2, HUD re-estimated the underserved areas share for 1997 under the new assumptions (e.g., a lower, HMDA-based mortgage share for investor loans), obtaining a range of 32.0 percent (with a 16.3 MF mix) to 32.7 percent (with a 19.3 percent MF mix). These estimates assume 1990 Census geography. Adding five percentage points to reflect 2000 Census geography yields estimates of 37.0 percent to 37.7 percent for the 1997 underserved areas market.

70.  The baseline 36.9 percent estimate for 2001 is 0.8 percentage point lower than the 37.7 percent share reported in Section G.4 of Appendix D of the proposed rule. The difference is mostly explained (a) by the different treatment in this Final Rule of single-family rental mortgages and (b) by a 0.2 percentage point decline in HUD's projections (in terms of the 2000 Census data) of the 2001 underserved areas percentage for SF owners.

71.  The baseline 36.2 percent estimate for 2002 is one percentage point lower than the 37.2 percent share reported in Section G.4 of Appendix D of the proposed rule. The difference is mostly explained (a) by the different treatment in this Final Rule of single-family rental mortgages and (b) by a 0.4 percentage point decline in HUD's projections (in terms of the 2000 Census data) of the 2002 underserved areas percentage for SF owners.

72.  For the years 1999 to 2002, Fannie Mae estimated a 2000-Census-based underserved areas share of 37-38 percent, obtained by adding five percentage points to Fannie Mae's 32-33 percent estimate for the underserved areas market based on 1990 Census data. (See their Table I.9, page I-34.) This compares with HUD's estimate of 37.1 percent to 37.6 percent for the same period.

73.  The differentials reported in Table D.14 for the three individual property types tend to be greater than five percentage points, which raises the question of why the overall differential is only five percentage points. As explained later, the upward adjustment to account for underserved areas in non-metropolitan areas is about 0.65 percentage point less using the 2000-based Census data than it was using the 1990-based Census data.

74.  Between 1999 and 2002, 2000-based underserved census tracts accounted for 31.4 percent (unweighted annual average) of all mortgages in metropolitan areas. This 1999-02 average percentage for single-family owners in metropolitan area is lower than the underserved area percentage reported in previous paragraphs. To be comparable with the non-metropolitan data, these metropolitan area data do not include loans originated by lenders that specialize in manufactured housing loans and B&C loans, excluding these loans lowers the underserved areas share.

75.  There are two LIHTC thresholds: at least 20 percent of the units are affordable at 50 percent of AMI or at least 40 percent of the units are affordable at 60 percent of AMI.

76.  As noted earlier, this discrepancy could be due to mis-measurement from the technique for apportioning 2003 data, which is defined in 2000-census geography, to a 1990-based geography.

77.  Affordability was calculated as discussed earlier in Section F, using AHS monthly housing cost, monthly rent, number of bedrooms, and MSA location fields. Low-income tracts were identified using the income characteristics of census tracts from the 1990 Census of Population, and the census tract field on the AHS file was used to assign units in the AHS survey to low-income tracts and other tracts. POMS data on year of mortgage origination were utilized to restrict the sample to properties mortgaged during 1993-1995.

78.  During the 1995 rule-making process, HUD examined the rental housing stock located in low-income zones of 41 metropolitan areas surveyed as part of the AHS between 1989 and 1993. While the low-income zones did not exactly coincide with low-income tracts, they were the only proxy readily available to HUD at that time. Slightly over 13 percent of single-family rental units were both affordable at the 60-80 percent of AMI level and located in low-income zones; almost 16 percent of multifamily units fell into this category.

79.  Fannie Mae's data exhibited some variation, standing at 33 percent in 2001 and 19 percent in 2001. Freddie Mac's percentage was 29 percent in both years.

80.  These adjustments for low-income renters living in low-income areas may be conservative. For SF (MF) rentals, the 2001 and 2002 figures for the GSEs were in the nine (eight) percent range.

81.  This is obtained by multiplying (a) 1.0 percentage point by (b) the refinance rate of 0.35 by (c) the 0.745 property share for SF owner loans.

82.  These three estimates were initially reported in HUD's 2000 Final Rule, and repeated in Table D.9 of Appendix D of the 2004 proposed GSE rule.

83.  To provide some confirmation for these 1995-1997 estimates, HUD went back and re-estimated the model for 1997. As shown in Table D.9 of the 2004 GSE Proposed Rule (as well as in Table D.15 of the 2000 GSE Rule), HUD had earlier estimated a special affordable share of 28.8 percent for 1997 (which was practically the same as the 28.9-percent share estimated for 1995 and the 28.7-percent share estimated for 1996). With a lower investor share (8.4 percent instead of 10.0 percent) and other changes mentioned in the text, the new estimate for the 1997 special affordable market was 28.0 assuming a multifamily mix of 19.3 percent. If the multifamily mix is reduced to 17.3 (16.3) percent, the special affordable share of the 1997 market is 27.1 (26.7) percent. The 26.7-28.0 percent range for 1997 is consistent with the 1995-1997 ranges reported in the text.

84.  Using the projected CBSA data (instead of the historical 1990-based MSA data) did not change the special affordable market estimate in either 1999 or 2000.

85.  The baseline estimates for 2001 (25.0 percent) and 2002 (24.3 percent) are lower than those (26.5 percent and 25.8 percent, respectively) reported in Table D.9 of Appendix D of the proposed rule. As explained earlier, the differences between the results in the proposed rule and this Final Rule are mainly due to the treatment of single-family rental mortgages. In addition, the SF0 percentage for home purchase loans originated during 2002 was lowered by approximately 0.2 percentage point in the Final Rule.

86.  Using the projected CBSA data (instead of the historical 1990-based MSA data) resulted in only small changes in the special affordable market estimates for 2001 (a 0.1 percentage point decline) and 2002 (a 0.5 percentage point decline).

87.  For the years 1999 to 2002, Fannie Mae estimated a special affordable market share of 23-25 percent. (This is their estimate assuming no missing data; see their Table I.9, page I-34.) This compares with HUD's estimate of 25.9 percent to 26.6 percent. As discussed in Section C.6, Fannie Mae assumes a rather low MF mix (approximately 10 percent) in the model that generates its historical estimates.