06-6568. Nonforeign Area Cost-of-Living Allowance; General Population Rental Equivalence Survey Report

AGENCY:

Office of Personnel Management.

ACTION:

Notice.

SUMMARY:

This notice publishes the “Nonforeign Area General Population Rental Equivalence Survey Report.” The General Population Rental Equivalence Survey (GPRES) was a special research project in which the Office of Personnel Management (OPM) collected data on homeowner estimates of the rental value of their homes and market rents in the nonforeign area cost-of-living allowance (COLA) areas and in the Washington, DC area. OPM conducted GPRES to determine whether rental survey data collected in the COLA surveys should be adjusted to account for homeowner shelter costs. Based on the GPRES results, OPM has determined that no adjustment is appropriate. OPM is publishing this report to inform interested parties of the research results and provide an opportunity for comment.

DATES:

Comments on this report must be received on or before September 29, 2006.

ADDRESSES:

Send or deliver comments to Jerome D. Mikowicz, Acting Deputy Associate Director for Pay and Performance Policy, Strategic Human Resources Policy Division, Office of Personnel Management, Room 7H31, 1900 E Street NW., Washington, DC 20415-8200; fax: [phone]; or e-mail: [e-mail].

FOR FURTHER INFORMATION CONTACT:

Donald L. Paquin, [phone]; fax: [phone]; or e-mail: [e-mail].

SUPPLEMENTARY INFORMATION:

The Office of Personnel Management (OPM) conducted the General Population Rental Equivalence Survey (GPRES) to determine whether OPM should adjust the rent indexes it computes from data collected in the nonforeign area cost-of-living allowance (COLA) surveys. The Federal Government pays COLAs to certain white collar Federal and U.S. Postal Service employees in Alaska, Hawaii, Guam and the Northern Mariana Islands, Puerto Rico, and the U.S. Virgin Islands. As provided by subpart B of title 5, Code of Federal Regulations, OPM conducts living-cost surveys to set COLA rates.

One of the items OPM surveys during the COLA surveys is market rents for detached houses, duplexes and triplexes, town and row houses, and apartments. We use rental data to estimate the relative price of shelter for both homeowners and renters between the COLA areas and the Washington, DC area. (For an example, see the 2004 Pacific COLA survey report published at 70 FR 44989-45023.) As applied to homeowners, this approach is called “rental equivalence” because it estimates the shelter value of owned homes rather than surveying homeowner costs directly.

OPM adopted the rental equivalence approach pursuant to the settlement in Caraballo, et al. v. United States, No. 1997-0027 (D.V.I), August 17, 2000. The settlement provides for several significant changes in the COLA methodology, including the use of rental equivalence. The settlement also established the Survey Implementation Committee (SIC), composed of seven plaintiffs' representatives and two OPM representatives, and the Technical Advisory Committee (TAC), composed of three economists with expertise in living-cost analysis. The TAC advises the SIC and OPM on living-cost issues. The SIC and the TAC agreed OPM could use, on an interim basis, market rents collected in the COLA surveys to estimate homeowner costs. The TAC noted, however, that the relative price of shelter for homeowners could differ compared with the relative price of market rents between the COLA areas and the DC area. If this were the case, it would be appropriate for OPM to adjust COLA survey market rent indexes before applying them to homeowners.

Therefore, OPM conducted a special research project, i.e., GPRES, to collect information on market rents and homeowner estimates of the rental value of their homes in the COLA areas and in the Washington, DC area. The SIC and the TAC were involved heavily in the design of the survey, and the TAC analyzed the survey results. The TAC also compared GPRES results with the results of the 1998 Federal Employee Housing and Living Patterns Survey (FEHLPS), which Joel Popkin and Company conducted as part of the research leading to the Caraballo settlement.

Using the GPRES results, the TAC found that no adjustment to the COLA survey market rents was appropriate because there were no statistically significant differences between homeowner estimated rents and market rents in the COLA areas compared with the DC area. The TAC found essentially the same results using FEHLPS. Therefore, the TAC recommended no rental equivalence adjustment be made. However, the TAC noted some differences between GPRES results and FEHLPS results and speculated these differences could reflect trends in relative rent prices/rental price estimates. Therefore, the TAC recommended OPM consider conducting additional GPRES-type surveys if OPM were to adopt a rental equivalence adjustment. Because OPM agrees that no rental equivalence adjustment is warranted, we do not plan to conduct additional GPRES-type surveys at this time.

Office of Personnel Management.

Linda M. Springer,

Director.

Nonforeign Area General Population Rental Equivalence Survey Report

TABLE OF CONTENTS

1. Introduction.

2. Purpose of GPRES.

2.1 Rental Equivalence and Rents.

2.2 Caraballo Settlement and Rental Equivalence.

3. Planning GPRES.

3.1 Consultation with the SIC and TAC.

3.2 Survey Instrument, Sampling Methodology, and Sample Size.

4. Conducting the Survey.

4.1 Survey Period.

4.2 Efforts To Ensure Quality Participation.

4.3 Survey Complications.

4.3.1 Home Size.

4.3.2 Prevalence of Subsidized Housing in Some Areas.

5. Survey Results and Response Rates.

5.1 GPRES Survey Results and Response Rates.

5.2. FEHLPS Survey Results and Response Rates.

6. Survey Analyses

6.1 Homeowner Factors: Comparison of Owner Rent Estimates and Market Rents.

6.2 Regional Comparisons.

6.3 COLA Survey Area Comparisons.

7. Summary and Conclusions.

List of Appendices

A. GPRES Survey Questionnaire.

B. GPRES Sample Size.

C. GPRES Data Collection Guidelines.

D. GPRES Number of Responses and Response Rates.

E. FEHLPS Samples Size, Responses, and Response Rates.

F. FEHLPS Survey Questionnaire—Housing Portion.

G. GPRES SAS Regression Results—Regional Analyses.

H. FEHLPS SAS Regression Results—Regional Analyses.

I. GPRES SAS Regression Results—Survey Area Analyses.

J. FEHLPS SAS Regression Results—Survey Area Analyses.

1. Introduction

This report provides the results of the General Population Rental Equivalence Survey (GPRES), which Westat, Incorporated, conducted for OPM in the winter of 2004/2005. In addition, the report provides for comparison purposes the results of the 1998 Federal Employee Housing and Living Patterns Survey (FEHLPS), which Joel Popkin and Company conducted for plaintiffs' representatives and Government representatives who were working collaboratively to resolve long-contested issues in the nonforeign area cost-of-living allowance (COLA) program. The collaborative work lead to the settlement of Caraballo, et al. v. United States, No. 1997-0027 (D.V.I.), August 17, 2000, and to major changes in the nonforeign area cost-of-living allowance (COLA) program. Therefore, although this report is principally about GPRES, it also covers the FEHLPS as it applies to rental equivalence analyses.

The report describes how OPM planned and prepared for the conduct of GPRES. In planning the survey, OPM consulted closely with the Survey Implementation Committee (SIC) and the Technical Advisory Committee (TAC), both established pursuant to the Caraballo settlement. The SIC has seven members—five plaintiffs' representatives from the COLA areas and two OPM representatives. The TAC has three members—economists who have expertise in living-cost measurement. The TAC performs research for and advises the members of the SIC.

The purpose of GPRES was two-fold. First, it was to determine whether there are statistically significant “homeowner factors” (HFs) that reflect the difference between homeowners' estimates of the rental value of their homes compared with market rents, holding rental unit characteristics constant. (The HF is the estimated rental value of owned homes divided by the market rent for homes of equivalent observed quality and quantity.) Second, GPRES was to determine whether HFs varied between the COLA areas and the Washington, DC area to a statistically significant degree. If so, OPM could use the results to adjust the market rents it collects during the COLA surveys to reflect homeowner shelter costs.

FEHLPS was used to look at the same two questions. The purpose of FEHLPS was to collect a wide range of information on Federal employees—much more than housing data. However, among the data FEHLPS collected were homeowner estimates of the rental value of their homes, so it was possible to use the survey to compute HFs and to examine whether these varied to a statistically significant degree between the COLA areas and the Washington, DC area. The scope of FEHLPS was more limited than GPRES. It had approximately a third fewer housing observations and was limited to Federal employees—a subset of the general population.

Comparing GPRES and FEHLPS results was very informative. This report describes those comparisons and why, based on the results and comparisons, no adjustment to rental indexes to account for homeowner shelter costs appears warranted at this time.

2. Purpose of GPRES

2.1 Rental Equivalence and Rents

There are two commonly accepted approaches for measuring the shelter value of owned homes. One is the user-cost approach. The other is rental equivalence. In simplistic terms, user costs are the costs of owning and maintaining a home minus the annual discounted expected capital gains that the owner will realize when he or she sells the home. Rental equivalence is what an owned home would rent for if it were available for rent in the rental market.

Rental equivalence is a well-known approach and is used by the Bureau of Labor Statistics (BLS) in the computation of the Consumer Price Index. Instead of measuring the change in owner user costs, which tend to be volatile, BLS attributes the change in market rents to homeowner shelter costs. This approach is supported by research that BLS conducted in the 1990's. Economists advising the plaintiffs' and Government representatives prior to the Caraballo settlement recommended that OPM adopt a similar approach for the COLA program, and the Caraballo settlement and OPM regulations adopted pursuant to the settlement prescribe that OPM use a rental equivalence approach to estimate the “price” of homeowner shelter.

Economic theory suggests that homeowners” estimates of the rental value of their homes will on average be higher than market rents for housing with equivalent observed characteristics (i.e., of equivalent observed quantity and quality). (See Akerlof, George A., 1970. “The Market for ‘Lemons': Quality Uncertainty and the Market Mechanism,” The Quarterly Journal of Economics, MIT Press, vol. 84(3), pages 488-500.) Imperfect market knowledge on the part of potential renters' and homeowners' awareness of unobserved amenities of their homes cause owner rent estimates to be higher than market rents. In other words, the HF should be greater than one. The size of the HF, however, could vary between one or more COLA areas and the Washington, DC area if owned homes in some areas have more unobserved amenities than owned homes in other areas.

Other factors could also affect owner rent estimates of the rental value of their homes, such as the owner's limited knowledge of local rental markets. Although some owners might have an excellent knowledge of rental markets and the rental value of their homes, most owners have little reason to pay much attention to the rental market, and their estimates might well be less accurate. In fact, GPRES results suggest that homeowners often relied on their mortgage payments to estimate the rental value of their homes, and mortgage payments are not necessarily correlated with market rents.

Although homeowner estimates may be somewhat inaccurate, the expectation is that the inaccurate estimates would be distributed normally in any area—some too high and some too low. Once again, it is possible that the effect might not be constant across all areas. Owners might overestimate in areas where home values are rising rapidly, even though market rents were trailing. On the other hand, owners might estimate more accurately in areas with a higher proportion of transient population because owners might have a greater opportunity to acquire rental market knowledge if homes near to them become available for rent. Variation in the accuracy of owner estimates among areas would make it difficult to compare differences between owner estimates and market rents from one area to the next.

Another factor that might lead to inaccurate homeowner estimates could be the pride of ownership. It is conceivable that home owners systematically might estimate high rental values because the owners take pride in their homes and think they should be worth more, regardless of any unobserved amenities. This could further contribute to the “noise” in the survey—i.e., undermine the survey's ability to reflect higher owner shelter values attributable to unobserved amenities. Whether the effect of this “pride factor” might vary among areas is speculative.

GPRES was designed to collect information that could be used to compare homeowner estimated rents with market rents. It also obtained information on many of the characteristics and amenities of the respondents' homes to allow the comparison of estimated rents and market rents while holding observed quality and quantity constant.

2.2 Caraballo Settlement and Rental Equivalence

As stated in the previous section, pursuant to the Caraballo settlement OPM adopted a rental equivalence approach to measure the shelter value of owner-occupied housing. Appendix A of the stipulation for settlement provides 26 “Safe Harbor Principles” (SHPs) concerning the operation of the COLA program. One of the key principles, SHP-18, describes how OPM will measure the relative cost of shelter:

18. Hedonic Housing Model and Rental Equivalence: Shelter price relatives will be estimated for owners and renters from the triennial regional sample. The sample for the region will be pooled with the comparison sample from the base area and price relatives for the COLA areas will be estimated using hedonic regression models to adjust for quality differences.

Discussion: OPM will adopt a rental-equivalence approach to estimate shelter costs and a hedonic regression approach to compare housing of similar quality. To identify the living communities to be surveyed, OPM will use the results of the 1992/93 employees survey, JPC's [Joel Popkin and Company] survey, and/or other appropriate information. How the housing data will be collected is not known or stipulated. OPM may survey Federal employees, collect the data on its own or through a contractor, enter into an interagency agreement with another Federal agency (e.g., the Department of Interior), or use some other appropriate approach.

OPM adopted this principle when it published final regulations at 67 FR 22339. Section 591.219 of title 5, Code of Federal Regulations, prescribes how OPM will compute shelter price indexes based on rental and rental equivalence prices and/or estimates. As noted in Section 2.1, rental equivalence compares the shelter value (rental value) of owned homes rather than total owner costs because the latter are influenced by capital gains (i.e., the investment value of a home). Most living-cost surveys do not compare how consumers invest their money.

In the COLA surveys, OPM surveys market rents in each of the COLA areas and in the Washington, DC area, obtaining over 80 characteristics of the rental units for use in the hedonic regression equations. (A hedonic regression is a statistical technique, specifically a form of multiple linear regression. For an explanation of how OPM applies these regressions, see “2004 Nonforeign Area Cost-of-Living Allowance Survey Report: Pacific and Washington, DC Areas,” published at 70 FR 44989.) The SIC and the TAC agreed that OPM could use market rents as an estimate for rental equivalence until the issue of rental equivalence could be explored more fully through a GPRES-type survey.

GPRES explored two questions. The first question was whether the rental value of owned homes in the COLA and DC areas differed to a statistically significant degree from market rents in the same area holding observed quality and quantity constant. To do this, the TAC computed homeowner factors, as described in Section 6.1. The second question was whether the COLA area homeowner factors differed to a statistically significant degree compared with the DC area homeowner factor. If the homeowner factors were significantly different, it might be appropriate for OPM to make a rental equivalence adjustment to account for homeowner shelter costs. As it turned out, no adjustment was appropriate because we did not find statistically significant differences between the COLA and DC areas.

3. Planning GPRES

3.1 Consultation With the SIC and TAC

OPM worked closely with the SIC and TAC to plan and develop GPRES. In August 2001, OPM provided the SIC and TAC with a rough draft of a survey questionnaire that could be used with homeowners and renters to obtain and compare information about estimated rental values and market rents. The SIC and TAC subsequently met on several occasions to refine the questionnaire and begin planning GPRES. The goal was to design a survey that was sufficiently brief as to encourage renters and owners to participate but sufficiently detailed so that OPM could compare market rents and rental equivalence estimates for comparable housing. By early 2002, the SIC and TAC had developed such a questionnaire. Later that year, at the request of the SIC and TAC, the Caraballo trustee entered into a contract with Joel Popkin and Company (JPC) to review draft plans for GPRES, review current literature regarding rental equivalence, and to make recommendations to the SIC and TAC concerning GPRES. JPC's research emphasized the importance of conducting GPRES. The SIC and TAC reviewed JPC's findings, incorporated them as appropriate in the survey, and recommended that OPM proceed with the conduct of GPRES. This OPM did.

OPM continued to consult with the SIC and TAC as it finalized plans for GPRES and kept them apprised during the conduct of GPRES. The TAC analyzed GPRES results, and OPM and the TAC discussed those results with the SIC.

3.2 Survey Instrument, Sampling Methodology, and Sample Size

In the fall of 2002, OPM contracted with Westat, Inc., a statistical research firm, to review JPC's research, propose a survey methodology, develop a survey instrument, and recommend sample sizes and sampling strategies for GPRES. In terms of a survey methodology, Westat recommended the use of Computer Assisted Telephone Interviews (CATIs). This approach appeared to offer the probability of greater response rates at reasonable cost compared with other approaches, such as mail-out questionnaires. Appendix A shows the GPRES questionnaire that Westat developed as modified by OPM.

To develop sample sizes, Westat used the results of FEHLPS and OPM's 2002 Caribbean and DC area COLA rental survey, applying standard sample size calculations. (See Cochran, W.G., Sampling Techniques: third edition, New York: John Wiley & Sons, Inc., 1977) Westat used FEHLPS to estimate the standard deviation of homeowner estimated rents for each COLA area and the Washington, DC area. Westat also used the results of the survey to estimate the standard deviation of market rents by area, except for the Caribbean and DC areas. For these areas, Westat used the results of the 2002 COLA survey because that survey had more observations and covered the general population, not just Federal employees. From the surveys, Westat developed sample sizes for owner and renters for the COLA areas and the Washington, DC area. Westat developed two sets each for owners and renters. One set was the sample size necessary for estimating rent or rental equivalence within a margin of error of +/− $500 in annual rent with 90 percent confidence level, and the other was the sample size for estimating rent or rental equivalence at the same margin of error at the 95 percent confidence level. Subsequent to the 2003 Alaska COLA survey, OPM modified the renter sample sizes for the Alaska and DC areas based on the additional rental data that OPM had collected in these areas. Appendix B shows the sample sizes Westat recommended, as modified by OPM.

Within each area, OPM limited the geographic scope of GPRES to the zip code areas in which OPM collected rental data in the annual COLA surveys. In the Washington, DC area, OPM further allocated the sample among the District of Columbia and the Counties of Montgomery, MD; Prince Georges, MD; Arlington, VA; Fairfax, VA; and Prince William, VA; and the independent cities therein, based on the relative numbers of owners and renters within these areas as reflected by the 2000 Census.

OPM obtained approval for GPRES from the Office of Management and Budget (OMB) as required by 5 CFR Part 1320, and OMB assigned GPRES an information collection number. Federal surveys and other information collections that Federal agencies conduct are covered by the Paperwork Reduction Act (44 U.S.C. 3501 et seq.). Participation in GPRES was voluntary, and any identifying information regarding the respondents is protected under the Privacy Act (5 U.S.C. 552a) and the Freedom of Information Act (5 U.S.C. 552).

4. Conducting the Survey

4.1 Survey Period

In the fall of 2004, OPM awarded a second contract to Westat to conduct GPRES. Using CATI, Westat began collecting data in October 2004 and finished in March 2005. Although Westat started data collection in some areas before others, Westat essentially collected data in all of the areas throughout this entire time period. Westat provided OPM with interim deliverables throughout the survey so that OPM and the TAC could begin testing analyses prior to receiving the final deliverable. Westat provided the final deliverable in early April 2005.

4.2 Efforts to Ensure Quality Participation

Westat used commercially available lists of phone numbers and addresses of owners and renters for the Washington, DC area and all of the COLA areas, except Guam, Puerto Rico, and the U.S. Virgin Islands for which such lists were unavailable. Using the sampling strategy described in Section 3.2, Westat drew the sample using commercial data bases where available. Westat then mailed letters to the prospective respondents informing them of the survey and asking for their cooperation. The letter was prepared by OPM on OPM letterhead and signed by Donald J. Winstead, who at that time was OPM's Deputy Associate Director for Pay and Performance Policy, Strategic Human Resources Policy Division. For those areas where commercial mailing/phone lists were unavailable, Westat was unable to mail advance letters; and Westat used simple random sampling to select potential participants.

At the beginning of each telephone interview, Westat surveyors explained the purpose of the survey, that the survey was voluntary, and provided the respondent the OMB-provided information collection number. Westat made certain that the respondent was a knowledgeable adult who could answer questions relating to the housing unit. If the adult was not available, Westat made arrangements to call back at a more convenient time to conduct the interview. The complete interview took approximately 8 minutes.

It was critically important that GPRES collect accurate information from persons who either owned their own homes or rented homes at current market rents. To this end, some GPRES questions were designed to eliminate respondents who did not meet these criteria. For example, Westat discontinued the survey if the respondent lived in rent-subsidized or rent-controlled housing, occupied military housing, or rented from relatives or other persons at rates other than market rates. Likewise, Westat discontinued the survey if the respondent was renting a room in a home or was living in a mobile home or similar lodging.

In addition, OPM identified for Westat several “threshold” questions that were critical to the survey and instructed Westat to discontinue the survey if the respondent could not or would not answer these questions. For example, if the respondent did not know or refused to answer how many bathrooms or bedrooms were in the home, Westat was instructed to discontinue the survey. The questionnaire in Appendix A shows the threshold questions. They are identified by the interview instruction “GO TO END.” Similarly, OPM provided Westat with guidelines to help ensure that respondents did not provide frivolous responses or occupied housing so atypical as to be outside the scope of the survey. Appendix C shows the Guidelines that Westat used to help identify frivolous and highly atypical responses.

4.3 Survey Complications

Westat encountered two unexpected complications in conducting GPRES. One involved the respondent's lack of knowledge concerning home size. The other involved an unexpectedly high proportion of the population in certain areas residing in subsidized or rent-controlled housing.

4.3.1 Home Size

One problem that Westat encountered was that respondents often did not know and could not estimate or guess the number of square feet in their home. As shown in Appendix A, OPM had identified this as a critical threshold question; and as shown in Appendix C, OPM provided guidelines concerning acceptable data. Westat noted that invalidating these responses was increasing the non-response rate and the cost of the survey. Westat suggested that OPM reconsider whether home size should be a threshold question and/or subject to the guidelines.

OPM discussed the issue with the TAC. The TAC was not surprised and noted that BLS, the Bureau of the Census, and other housing surveys encountered the same problem and dropped home size as a question in their surveys. The TAC suggested that OPM use room count and a limited number of other characteristics to impute home size for respondents who were unknowledgeable or provided atypical responses. OPM tested this approach using the rental data it had collected in the COLA surveys and found it feasible. Therefore, OPM informed Westat to continue survey interviews even when respondents did not know and could not estimate home size and instructed Westat not to apply guidelines to flag atypical responses. OPM and the TAC later tested whether to use imputed home sizes but decided against it because the imputation process had a systematic error in estimating the size of relatively small and relatively large homes.

4.3.2 Prevalence of Subsidized Housing in Some Areas

Westat also discovered difficulties obtaining the desired sample of renters in certain areas because an unexpectedly large portion of the renter population appear to occupy subsidized or rent-controlled housing. This was most noticeable in Guam, Puerto Rico, and the U.S. Virgin Islands (USVI), as well as in the District of Columbia. Under the contract, OPM paid Westat on a price-per-completed-survey-response basis. When Westat began encountering unexpectedly high respondent invalidation rates, Westat informed OPM that it would not be able to provide the desired sample sizes in certain areas because the company had reached the breakeven point at which further data collection would not be profitable.

Therefore, OPM modified the price schedule in the contract to ensure that Westat could obtain at least the “minimum” sample size shown in Appendix B in all areas. As shown in Appendix D, Westat exceeded this level in several areas, but it was unable to obtain the minimum number of renter samples in Guam and Puerto Rico.

5. Survey Results and Response Rates

5.1 GPRES Survey Results and Response Rates

Appendix D shows the number of renter and owner observations that Westat obtained by area. Except in Guam and Puerto Rico, Westat obtained a sample that equaled or exceeded the sample size necessary for estimating rent or rental equivalence within a margin of error of +/-$500 in annual rent with a 90 percent confidence level. In all, Westat obtained 6,170 observations.

To do this, Westat made more than 152,000 phone calls. Therefore, one simplistic measure of the response rate might be 4 percent (i.e., 6,170 divided by 152,000). Many of those calls, however, particularly in the areas for which commercial phone list data as described in Section 4.2 were unavailable, were screening calls to businesses, facsimile machines, and other non-residential phone numbers. Also, many of the residential respondents (e.g., those occupying rent-controlled or subsidized housing) were not eligible to be part of the survey universe. Therefore, another and perhaps more meaningful way to look at the response rate is to compare the number of respondents with the total number of those who were determined, after the screening questions, to be part of the survey universe. According to Westat, a total of 23,662 respondents passed the screening questions. Using this as a basis, the response rate was 26.1 percent (i.e., 6,170 divided by 23,662). This does not, however, include respondents who become ineligible in the “extended interview,” i.e., the main part of the interview that followed the screening questions. Taking this into consideration, the overall GPRES response rate according to Westat was 28 percent. Appendix D shows this type of response rate for each COLA area and the for Washington, DC area.

5.2 FEHLPS Survey Results and Response Rates

JPC conducted FEHLPS in cooperation with OPM in 1998. It was a survey of a sample of non-U.S. Postal Service Federal employees in the COLA areas and in the Washington, DC area. JPC selected a sample size of approximately 15,800, of which 11,478 were to be drawn from the COLA areas and 4,324 were to come from the Washington, DC area. The sample was drawn from OPM's Central Personnel Data File (CPDF), which is essentially a census of non-Postal Federal employees. According to the CPDF, there were approximately 44,027 non-Postal Federal employees in 1998 in the COLA areas and 258,304 in the DC area.

JPC collected 5,662 responses from the COLA areas, which makes the average response rate for those areas 49.3 percent. JPC collected 1,081 responses from the Washington, DC area, which makes the DC area response rate 25 percent. Appendix E shows the FEHLPS sample sizes, responses, and response rates by COLA area and for the Washington, DC area. Not all of the respondents provided usable housing data. Therefore, the TAC could use only 4,275 FEHLPS observations in its analyses.

The survey was a “mail out” survey, delivered to employees at their worksite. Agencies were encouraged to grant employees time at work to complete the survey. FEHLPS covered numerous topics, including transportation and travel, K-12 private education, college education, medical costs, and housing. Appendix F shows the housing related portion of the survey.

6. Survey Analyses

The TAC performed most of the analyses of the GPRES results, with OPM's support and oversight. OPM also contracted with JPC to review the GPRES results and analyses. JPC concurred with the TAC's analyses, findings, and recommendations.

6.1 Homeowner Factors: Comparison of Owner Rent Estimates and Market Rents

As discussed in Section 2, one purpose for conducting GPRES was to compare owner estimates of the rental value of their homes with market rents for comparable housing in terms of quality and quantity. The goal was to express mathematically the relationship of rents and rent estimates within each COLA area and the Washington, DC area. The second purpose was to examine whether those relationships varied significantly between the COLA areas and the Washington, DC area.

The TAC computed homeowner factors (HFs) to express the relationship of homeowner rent estimates and market rents in and among the COLA areas and the Washington, DC area. The HF is the estimated rental value of owned homes divided by the market rent for homes of equivalent observed quality and quantity. To compute the HF, the TAC used hedonic regressions to hold quality and quantity constant.

The TAC used two distinctly different approaches to analyze HFs. One approach involved comparing HFs by COLA region with the DC area HF. The other involved estimating HFs for each COLA survey area and comparing these with the DC area HF. The results of the two approaches were quite different but lead to the same conclusion.

6.2. Regional Comparisons

The COLA areas are divided into three regions—the Alaska, Pacific, and Caribbean regions. The Alaska region is composed of the Anchorage, Fairbanks, and Juneau COLA survey areas. The Pacific region is composed of the Honolulu County; Hilo and Kailua Kona, Hawaii County; Kauai County; Maui County; and Guam COLA survey areas. The Caribbean region is composed of the Puerto Rico; St. Croix, USVI, and St. Thomas/St. John, USVI, COLA survey areas.

The TAC noted that there were virtually no previous studies to serve as a guide on how to analyze HFs by area and compare them between areas. The TAC believed if there were systematic differences in HFs across areas, the TAC would need as many observations as possible to identify these relationships. Pooling the data by region allowed the use of all of the survey observations (GPRES or FEHLPS) at one time.

The TAC applied semi-logarithmic hedonic regressions to compute rental equivalence indexes and market rent indexes for the COLA regions relative to the Washington, DC area, holding quantity and quality of housing constant. The dependent variable of the regression was the logarithm of rent. Appendix G shows the SAS GPRES regression results that the TAC used. (SAS is a proprietary statistical analysis computer software package.) The independent variables for the GPRES regression are listed below:

Type of dwelling (e.g., detached house, townhouse, apartment),

Whether the unit had central air conditioning,

Number of baths,

Number of bedrooms,

Number of baths crossed with type of dwelling, and

Tenure (i.e., owned or rented) by the COLA region or DC area in which unit is located.

The parameter of interest in this regression was tenure by COLA region and the results are shown in the table below. The HF is shown in column (1). (The logarithmic form of the HFs and standard errors and t values are shown in columns (2) through (4).) An HF of 1.223 for Alaska means that homeowner estimates of the rental value of their homes are on average 22.3 percent higher than market rents holding observed quality and quantity of the housing unit characteristics constant. The critical values of “t” at the 5 percent and 1 percent levels are 1.96 and 2.58 respectively. In other words, HFs with t-values equal to or greater than 2.58 are significant at a 99 percent confidence level or higher.

Table 1.—GPRES Homeowner Factors by Region

COLA region	HF	Logarithmic HF	Standard error	t-value
	(1)	(2)	(3)	(4)
Alaska	1.223	0.201	0.027	7.50
Pacific	1.171	0.158	0.018	8.74
Caribbean	1.117	0.111	0.023	4.94
Washington, DC Area	1.153	0.142	0.031	4.62

The TAC also computed homeowner factors on a regional basis using the results of FEHLPS. Again, the dependent variable was the log of rent, but the independent variables were somewhat different than those used in the GPRES analyses. Appendix H shows the TAC's regression results using the FEHLPS data. The homeowner factors are shown in Table 2, below:

Table 2.—FEHLPS Homeowner Factors by Region

COLA region	HF	Logarithmic HF	Standard error	t-value
	(1)	(2)	(3)	(4)
Alaska	1.274	0.242	0.0301	8.03
Pacific	1.092	0.088	0.0195	4.49
Caribbean	1.168	0.155	0.0326	4.75
Washington, DC Area	1.254	0.226	0.0479	4.71

The HFs from both surveys are statistically significant and greater than 1 when the results are analyzed on a regional basis. HFs greater than one is what economic theory would predict. The key question is whether there are statistically significant differences between the HFs for the COLA regions compared with the DC area HF. To do this, the TAC again used a t-test where the standard error is the difference between HFs calculated from a covariance matrix of the regression coefficients on owners and renters. Tables 3 and 4 below show the results for GPRES and FEHLPS respectively.

Table 3.—GPRES Test of Difference Between Regional HFs and DC Area HF

COLA region	COLA region HF divided by DC	Logarithmic COLA region HF	Standard error	t-value
	(1)	(2)	(3)	(4)
Alaska	1.061	0.0595	0.0375	1.58
Pacific	1.016	0.0161	0.0328	0.49
Caribbean	0.970	−0.0301	0.0353	−0.85

Table 4.—FEHLPS Test of Difference Between Regional HFs and DC Area HF

COLA region	COLA region HF divided by DC Area HF	Logarithmic COLA region HF −DC Area HF	Standard error	t-value
Alaska	1.016	0.0161	0.0548	0.29
Pacific	0.871	−0.1379	0.0500	−2.76
Caribbean	0.932	−0.0705	0.0560	−1.26

As shown in Table 3, the TAC found, based on the GPRES results, the differences between the COLA region HFs and the DC area HF were not statistically significant. Similarly, as shown in Table 4, the TAC found, based on the FEHLPS results, there was no statistically significant difference between the COLA region HFs and the DC area HF. Therefore, no adjustment to the COLA survey rental index was appropriate to account for homeowner shelter values (rental equivalence).

Although analyses of both surveys found no statistically significant differences between the COLA and DC area HFs, the TAC also noted the significant differences between the GPRES results compared with the FEHLPS results. For example, GPRES showed the Pacific region HF was slightly higher than the DC area HF, but FEHLPS show the Pacific region HF to be somewhat lower than the DC area HF. Unless Federal employees were atypical of the general population with regard to market rents and homeowner estimates, it appeared that the HFs changed substantially over the 6-year interval between FEHLPS and GPRES. The TAC found the apparent lack of stability over time troubling.

6.3 COLA Survey Area Comparisons

The second approach the TAC used to analyze GPRES and FEHLPS results was to compute HFs by COLA survey area and compare these with the DC HF. The advantage of this approach was more consistency with the COLA program, which sets COLA rates by COLA area, not COLA region. It also allowed the HFs to be computed separately for each area, using different equations as appropriate. The disadvantage was that each regression used far less data than in the regional analyses.

To compute HFs for each of the COLA survey areas, the TAC pooled the survey data by region and computed HFs for each of the COLA survey areas within the region. Appendix I has an example of the SAS regression results for one of the survey areas—the Pacific region—using GPRES. Appendix J has an example of the SAS regression results for one of the survey areas—the Caribbean region—using FEHLPS. Table 5 shows the HFs by area and their relationship to the DC HF using GPRES. Table 6 shows the same results using FEHLPS.

Table 5.—GPRES HFs by COLA Survey Area

Survey area	HF	Logarithmic HF	Standard error	t ratio
	(1)	(2)	(3)	(4)
Anchorage	1.025	0.0250	0.0354	0.70
Fairbanks	0.958	−0.0434	0.0416	−1.04
Juneau	0.935	−0.0667	0.0392	−1.70
Honolulu	1.061	0.0588	0.0321	1.81
Hilo	0.986	−0.0141	0.0499	−0.28
Kailua Kona	0.957	−0.0440	0.0546	−0.81
Kauai	0.930	−0.0728	0.0396	−1.84
Maui	1.013	0.0134	0.0355	0.38
Guam	0.997	−0.0030	0.0351	−0.09
Puerto Rico	1.002	0.0018	0.0495	0.04
St. Croix	1.141	0.1321	0.0395	3.35
St. Thomas/St. John	1.124	0.1166	0.0442	2.64
DC Area	1.110	0.1040	0.0415	2.51

Unlike the COLA region analyses, the GPRES results in Table 5 show that the HFs are less than 1 in half of the COLA survey areas. This is contrary to what economic theory would predict. In addition, 10 of the 13 COLA survey area HFs are not statistically significant at a 95 percent confidence level. By comparison, the results using FEHLPS are quite different. (See Table 6.) All of the HFs are greater than 1, which conforms with economic theory, and only four of the HFs are not significant at a 95 percent confidence level.

Table 6.—FEHLPS HFs by COLA Survey Area

Survey area	HF	Logarithmic HF	Standard error	t ratio
	(1)	(2)	(3)	(4)
Anchorage	1.278	0.2451	0.0397	6.17
Fairbanks	1.011	0.0106	0.0623	0.17
Juneau	1.222	0.2006	0.0707	2.84
Honolulu	1.120	0.1130	0.0240	4.71
Hawaii County	1.011	0.0108	0.0424	0.25
Kauai	1.083	0.0798	0.0587	1.36
Maui	1.176	0.1618	0.0495	3.27
Guam	1.168	0.1549	0.0488	3.17
Puerto Rico	1.208	0.1888	0.0497	3.80
St. Croix	1.045	0.0440	0.0784	0.56
St. Thomas/St. John	1.468	0.3842	0.0839	4.58
DC Area	1.279	0.2461	0.0450	5.46

As with the regional analysis, the key question is whether the COLA survey area HFs are statistically significantly different from the DC area HF. The TAC used the same approach it used to produce Tables 3 and 4 in the region analyses. As shown in Table 7, the GPRES results indicate that the HFs in the COLA survey areas are lower than the DC area HF except in the USVI. The t-ratios, however, show that these results are not significant at the 95 percent confidence level in 8 out of 12 cases. (Keep in mind that 10 of the 13 HFs were not statistically significant at that level, which further weakens the statistical validity of the comparison.). Table 8, which shows the FEHLPS results, also shows that the COLA survey area HFs are lower than the DC area HF, except in St. Thomas/St. John, USVI. (Note: Unlike GPRES, it was not possible using FEHLPS data to split Hawaii County into the Hilo and Kailua Kona survey areas.) In addition, the FEHLPS differences are not statistically significant at a 95 percent confidence level in 7 out of 13 areas.

Table 7.—GPRES Test of Difference Between Survey Area HFs and DC Area HF

Survey area	COLA area HF divided by DC area HF	Logarithmic COLA area HF− DC area HF	t ratio
	(1)	(2)	(3)
Anchorage	0.924	−0.0790	−1.45
Fairbanks	0.863	−0.1474	−2.51
Juneau	0.843	−0.1707	−2.99
Honolulu	0.956	−0.0452	−0.86
Hilo	0.889	−0.1181	−1.82
Kailua Kona	0.862	−0.1480	−2.16
Kauai	0.838	−0.1768	−3.09
Maui	0.913	−0.0906	−1.66
Guam	0.899	−0.1070	−1.97
Puerto Rico	0.903	−0.1022	−1.58
St. Croix	1.028	0.0281	0.49
St. Thomas/St. John	1.013	0.0126	0.21
DC Area	1.000	0.0

Table 8.—FEHLPS Test of Difference Between Survey Area HFs and DC Area HF

Survey area	COLA area HF divided by DC area HF	Logarithmic COLA area HF−DC Area HF	t ratio
	(1)	(2)	(5)
Anchorage	0.999	−0.0010	−0.02
Fairbanks	0.790	−0.2355	−3.06
Juneau	0.956	−0.0455	−0.54
Honolulu	0.875	−0.1331	−2.61
Hawaii County	0.790	−0.2353	−3.80
Kauai	0.847	−0.1663	−2.25
Maui	0.919	−0.0843	−1.26
Guam	0.913	−0.0912	−1.37
Puerto Rico	0.944	−0.0573	−0.85
St. Croix	0.817	−0.2021	−2.23
St. Thomas/St. John	1.148	0.1381	1.45
DC Area	1.000	0.0

As with the regional analyses, the TAC found troubling the significant differences between the GPRES and FEHLPS results. Once again, the question was whether there were trends over the 6-year period between the surveys that could explain these differences or the differences were simply inherent in the populations surveyed and/or survey techniques used. The TAC recommended that OPM not implement any adjustments to the rental data based on the COLA survey area analyses without first conducting additional GPRES-like surveys.

7. Summary and Conclusions

OPM conducted GPRES to determine whether OPM should adjust rental data that it collects during its annual COLA surveys. In these annual surveys, OPM collects prices on market rents on various types of housing units. OPM uses rental data to estimate the relative price of shelter for both homeowners and renters between the COLA areas and the Washington, DC area.

The TAC analyzed the GPRES results and compared them with similar analyses using rental data and estimates from an earlier survey of Federal employees—FEHLPS. Using regression analyses, the TAC computed homeowner estimated rent and market rent indexes and from these computed homeowner factors (HFs), which were homeowner indexes divided by the market rent indexes for units of equivalent observed quality and quantity. Economic theory suggests that HFs will be greater than 1.

The TAC conducted two significantly different analyses—one pooled the COLA region and DC area data and the other treated each COLA area separately. The TAC conducted these analyses using GPRES results and then using FEHLPS results for comparison. For both surveys, the regional analyses showed that the HF were greater than 1 for all areas, which means that homeowner rent estimates are higher than market rents, holding observed housing characteristics constant. This is as economic theory would predict. But the TAC also found that for both surveys, the COLA area HFs did not differ to a statistically significant degree compared with the DC area HF. Therefore, no adjustments to the COLA survey rent index to account for rental equivalence are appropriate. In addition, the differences between the results using GPRES and those using FEHLPS raised questions of whether HFs are changing over time.

The TAC also analyzed the results of both surveys on a COLA survey area basis. These analyses showed that the COLA area HFs were generally less than 1, which is the opposite of the findings from the regional analyses and what economic theory would predict. Most of these HFs were not statistically significant using GPRES, and many were not significant using FEHLPS. For both surveys, the COLA area HFs were lower than the DC area HF, with the exception of the USVI HFs, but several of the COLA area HFs did not differ to a statistically significant degree from the DC area HF. As with the regional analyses, the COLA survey area analyses indicates that no adjustments to the COLA survey rent index are appropriate. In addition, the differences between the results using GPRES and those using FEHLPS were even more extreme and raised more questions of whether HFs are changing over time.

Based on these analyses, the TAC recommended that no adjustments be made in the COLA survey rent index to account for homeowner shelter costs. The TAC further recommended that OPM conduct additional GPRES-like surveys before considering any such adjustment. OPM hired JPC to review the TAC's analyses. JPC found the TAC's analyses to be appropriate and comprehensive and concurred with the TAC's recommendations. Therefore, OPM will not adjust COLA survey rent indexes to account for homeowner shelter costs. OPM does not see a need to conduct additional GPRES surveys at this time.

Appendix A—GPRES Survey Questionnaire

The interviewer must provide the following information to each respondent: My name is {INTERVIEWER'S NAME} and I am calling on behalf of the U.S. Office of Personnel Management. We are conducting a study to determine housing costs in your area. Although the results of the study may be public, we will not divulge any information that would allow someone to identify you or your home.

Your participation is voluntary and very important to the success of this study. This study should take approximately 8 minutes. You may send any comments concerning this study to the Office of Personnel Management. [IF NEEDED: The address is office of Personnel Management, Forms Officer, Washington, DC 20415-8900]. We invite comments about how long the study takes and how this time could be reduced.

The Office of Management and Budget has approved this study and assigned it a collection number of 3206-0247. We would not be able to conduct this study without this approval. The approval expires 5/31/2007.

1. Do you own or rent your home?

OWN—1 GO TO Q8a

RENT—2 GO TO 2

OTHER (SPECIFY ______)—91 GO TO END

REFUSED—−7 GO TO END

DON'T KNOW—8 GO TO END

RENTERS ONLY

2. Which of the following best describes your rental agreement? Would you say . . .

You live in subsidized or rent controlled housing—1 GO TO END

You live in military housing—2 GO TO END

You rent from a family member or friend who does not charge you market rate for your home—3 GO TO END

You pay the market rate for renting your home—4

REFUSED—−7 GO TO END

DON'T KNOW—−8 GO TO END

3. What is the length of your lease?

YEAR—1

6 MONTHS—2

NO LEASE (e.g., month-to-month)—3

OTHER—91

(SPECIFY)—

REFUSED—−7

DON'T KNOW—−8

4a. What is your monthly rent?

$__,___ MONTHLY RENTAL AMOUNT

REFUSED—−7 GO TO END

DON'T KNOW—−8 GO TO END

4b. Are any utilities included in the rent?

YES—1

NO—2 GO TO Q5

REFUSED—−7 GO TO Q5

DON'T KNOW—−8 GO TO Q5

4c. Which of the following utilities are included in the rent? Does it include . . .

		YES	NO	REF	Don't know
4ca	Water?	1	2	−7	−8
4cb	Electric?	1	2	−7	−8
4cc	Gas?	1	2	−7	−8
4cd	Heat?	1	2	−7	−8

5. Are any of the following included in the rent? How about . . .

		YES	NO	REF	Don't know
5a	Maintenance, e.g. faucet/appliance repair?	1	2	−7	−8
5b	Lawn care?	1	2	−7	−8
5c	Snow removal?	1	2	−7	−8
5d	Trash removal?	1	2	−7	−8
5e	Parking in covered public style garage?	1	2	−7	−8
5f	Furnishings?	1	2	−7	−8

6a. Are pets allowed at your rental unit?

YES—1

NO—2 GO TO 7a

REFUSED—−7 GO TO 7a

DON'T KNOW—−8 GO TO 7a

6b. Is there an additional fee for pets?

YES—1

NO—2 GO TO 7a

REFUSED—−7 GO TO 7a

DON'T KNOW—−8 GO TO 7a

6c How much is the additional fee?

$______ AMOUNT OF PET FEE

MONTHLY—1

ANNUALLY—2

ONE-TIME DEPOSIT—3

OTHER (SPECIFY) ______—91

REFUSED—−7

DON'T KNOW—−8

7a. Approximately how long have you rented at this location?

NOTE: LESS THAN 1 MONTH = 1 MONTH

______ TIME RENTED AT THIS ADDRESS MONTHS

______ TIME RENTED AT THIS ADDRESS YEARS

REFUSED—−7

DON'T KNOW—−8

7b. Would you consider the place that you're renting a permanent rental property, that is, the property is consistently rented out, or is it a temporary rental, for example the owner is abroad and intends to return?.

PERMANENT—1 GO TO 11a

TEMPORARY—2 GO TO 11a

REFUSED—−7 GO TO 11a

DON'T KNOW—−8 GO TO 11a

OWNERS ONLY

8a. If you were to rent your home on a long term basis, not as a vacation rental, what do you think your home would rent for per month? We are not asking you whether you want to rent it, only to estimate what it might rent for if it were for rent.

$______ MONTHLY RENTAL AMOUNT—SKIP TO 8c

REFUSED—−7 GO TO 8b

DON'T KNOW—−8 GO TO 8b

8b. Would you estimate that your home would rent for . . .

Less than $200 per month—1 GO TO END

$201 to $500 per month—2 GO TO 8c

$501 to $1,000 per month—3 GO TO 8c

$1,001 to $1,500 per month—4 GO TO 8c

$1,501 to $2,000 per month—5 GO TO 8c

$2,001 to $2,500 per month—6 GO TO 8c

$2,501 to $3,000 per month—7 GO TO 8c

$3,001 to $6000 per month—or 8 GO TO 8c

Over $6000 per month?—9 GO TO END

REFUSED—−7 GO TO END

DON'T KNOW—−8 GO TO END

8c. How did you arrive at the rental amount? Was it based on . . .

	NOTE: ALL RESPONDENTS WILL BE ASKED ABOUT EACH REASON	YES	NO	REF	Don't know
8ca	Other neighborhood rentals?	1	2	−7	−8	GO TO 10a
8cb	Rental ads in newspapers, etc?	1	2	−7	−8	GO TO 10a
8cc	Realtor or property manager advide?	1	2	−7	−8	GO TO 10a
8cd	Previous experience renting this home?	1	2	−7	−8	GO TO 9a
8ce	Cost incurred, for example, receiving enough to cover your mortgage?	1	2	−7	−8	GO TO 10a
8cf	Something else? (Specify):______	1	2	−7	−8	GO TO 10a

9a. How long ago did you rent it?

______ TIME SINCE RENTED MONTHS

______ TIME SINCE RENTED YEARS

REFUSED—−7

DON'T KNOW—−8

9b. How much rent did you charge?

$______ PER

9b.1 MONTH—1

WEEK—2

YEAR—3

REFUSED—−7

DON'T KNOW—−8

10a. What is the approximate monthly mortgage payment on your home?

$______ MORTGAGE PAYMENT

REFUSED—−7

DON'T KNOW—−8

10b. Given current market conditions in your area, at what price would your home sell?

$________

REFUSED—−7

DON'T KNOW—−8

OWNERS AND RENTERS

11a. Which one of the following best describes where you currently live? Do you live in a . . .

One-family detached house—1 GO TO Q12a

Duplex or triplex—2 GO TO Q12a

Townhouse or rowhouse—3 GO TO Q12a

Apartment—4 GO TO Q11b

Rented room in a house—5 GO TO END

Trailer, or—6 GO TO END

Somewhere else?—91 GO TO END

REFUSED—−7 GO TO END

DON'T KNOW—−8 GO TO END

11b. Would you say that your home is . . .

An apartment in a home—1

An apartment in a building without an elevator or—2

An apartment in a building with an elevator—3

REFUSED—−7

DON'T KNOW—−8

12a. Approximately how many square feet of living space do you have?

_,_____ LIVING SPACE IN SQUARE FEET GO TO NOTE 1

REFUSED—−7 GO TO 12b

DON'T KNOW—−8 GO TO 12b

12b. Would you estimate that your living space is

Less than 250 square feet,—1 GO TO END

250 to less than 500 square feet,—2 SEE PROGRAMMER NOTE, ABOVE

500 to 1,000 square feet,—3 GO TO NOTE 1

1,001 to 1,500 square feet,—4 GO TO NOTE 1

1,501 to 2,000 square feet,—5 GO TO NOTE 1

2,001 to 2,500 square feet,—6 GO TO NOTE 1

2,501 to 3,000 square feet,—7 GO TO NOTE 1

3,001 to less than 6,000 square feet, or—8 GO TO NOTE 1

Over 6,000 square feet,—9 GO TO NOTE 1

REFUSED—−7 GO TO END

DON'T KNOW—−8 GO TO END

13. What is the lot size of your property?

__,____.__ PROPERTY LOT SIZE

13.1 ACRES—1

SQUARE FEET—2

REFUSED—−7

DON'T KNOW—−8

14. Does your home have an exceptional view, for example, overlooking a body of water or a city skyline?

YES—1

NO—2

REFUSED—−7

DON'T KNOW—−8

15a. How old is your home?

LESS THAN 1 YEAR = 1 YEAR

____ TIME IN YEARS

REFUSED—−7

DON'T KNOW—−8

15b. How many years has it been since it was remodeled/renovated?

LESS THAN 1 YEAR = 1 YEAR

____ TIME IN YEARS

NOT REMODELED/RENOVATED—N

REFUSED—−7

DON'T KNOW—−8

16a. Do you live in a studio or efficiency apartment?

YES—1 GO TO 17A.

NO—2 GO TO 16

REFUSED—−7 GO TO 16

DON'T KNOW—−8 GO TO 16

16. Please tell us how many bedrooms you have?

_ NUMBER OF BEDROOMS

REFUSED—−7 GO TO END

DON'T KNOW—−8 GO TO END

17a. How many full bathrooms are in your home?

__ NUMBER OF FULL BATHS

REFUSED—−7 } GO TO END

DON'T KNOW—−8 } GO TO END

17b. How many 1/2 bathrooms are in your home?

__ NUMBER OF HALF BATHS

REFUSED—−7 } GO TO END

DON'T KNOW—−8 } GO TO END

18. Excluding the bedrooms and bathrooms you just mentioned, how many other rooms are there? (Note: Closets and hallways are not rooms.)

__ NUMBER OF OTHER ROOMS

REFUSED—−7

DON'T KNOW—−8

19. Do you have a security system or live in a gated or guarded community?

YES—1

NO—2

REFUSED—−7

DON'T KNOW—−8

20a. Do you have air conditioning?

YES—1

NO—2 GO TO Q21a

REFUSED—−7 GO TO Q21a

DON'T KNOW—−8 GO TO Q21a

20b. Is it central air or individual room units?

CENTRAL AIR—1

ROOM UNIT—2

BOTH—3

REFUSED—−7

DON'T KNOW—−8

21a. How do you mainly heat your home?

SPACE HEATERS [electric or kerosene]—1

WALL UNIT [gas, electric]—2

BASEBOARD [electric, hot water]—3

CENTRAL HEAT [forced air]—4

NONE—5 GO TO Q22

OTHER—91

(SPECIFY)—

REFUSED—−7 GO TO Q22

DON'T KNOW—−8 GO TO Q22

21b. What type of fuel does it use?

GAS [Includes LP/ Propane]—1

ELECTRIC—2

OIL—3

OTHER—91

(SPECIFY)—

REFUSED—−7

DON'T KNOW—−8

22. What type of water system do you have? Is your water provided via* * *

Municipal water system,—1

Well,—2

Cistern, or—3

Something else?—91

(SPECIFY)—

REFUSED—−7

DON'T KNOW—−8

23. Do you have a garage? By this I mean your own garage, not a large public style parking garage.

YES—1

NO—2

REFUSED—−7

DON'T KNOW—−8

24. Do you have a carport?

YES—1

NO—2

REFUSED—−7

DON'T KNOW—−8

25a. Do you work outside of the home either full or part time?

YES—1

NO—2 GO TO 26

REFUSED—−7 GO TO 26

DON'T KNOW—−8 GO TO 26

25b. What is the one-way distance, in miles, from your home to your work?

LESS THAN ONE MILE—1

1-5 MILES—2

6-10 MILES—3

11-15 MILES—4

16-20 MILES—5

21-25 MILES—6

26-30 MILES—7

MORE THAN 30 MILES—8

REFUSED—−7

DON'T KNOW—−8

26. Do you or a member of your household work for the Federal Government?

YES—1

NO—2

REFUSED—−7

DON'T KNOW—−8

27. What is your zip code?

□□□□□ZIP CODE—

REFUSED—−7

DON'T KNOW—−8

END.

Appendix B—GPRES Sample Sizes

Geographic area	“Minimum” Quantity	“Target” Quantity
		Renter quantity	Owner quantity	Total quantity	Renter quantity	Owner quantity
Area A: District of Columbia	105	43	148	151	61	212
Area B: Montgomery Co., MD	72	88	160	103	126	229
Area C: Prince Geo. Co., MD	78	75	153	112	107	219
Area D: Arlington Co., VA	35	16	51	50	23	73
Area E: Fairfax Co., VA	82	108	190	116	155	271
Area F: Prince William Co., VA	20	7	27	28	9	37
Area G: Anchorage, AK	239	182	421	342	260	602
Area H: Fairbanks, AK	122	126	248	174	179	353
Area I: Juneau, AK	174	114	288	249	162	411
Area J: Honolulu County, HI	412	279	691	587	398	985
Area K: Hilo Area, HI	112	107	219	159	153	312
Area L: Kailua Kona Area, HI	85	69	154	121	98	219
Area M: Kauai County, HI	187	155	342	268	221	489
Area N: Maui County, HI	237	246	483	337	352	689
Area O: Guam	278	246	524	396	351	747
Area P: Puerto Rico	256	361	617	365	515	880
Area Q: St. Croix, USVI	185	295	480	264	422	686
Area R: St. Thomas, USVI	219	234	462	312	346	658
Area T: St. John, USVI	17	25	42	25	35	61
Totals	2,915	2,776	5,700	4,159	3,973	8,133

Note:

The “Minimum” set was the sample size necessary for estimating rent or rental equivalence within a margin of error of +/-$500 in annual rent with 90 percent confidence level, and the “Target” set was the sample size for estimating rent or rental equivalence at the same margin of error at the 95 percent confidence level.

Appendix C—Guidelines for Possible Flags to Identify Potentially Spurious or Highly Atypical Responses

Responses outside the range are assumed to be spurious and/or highly atypical and are not acceptable.

Question	Flag
4a. Monthly Rent	$200 to $5000.
6c. Typical Pet fees	Suggest this field not be flagged.
7a. Time at this address	Suggest this field not be flagged.
8a. Rental equivalence	$200 to $6000.
9a. Time since last rented	Suggest this field not be flagged.
9b. Rent charged	$200 to $5000.
10a. Mortgage payment	$0 to $6000.
10b. Market Price	$10,000 to $1,000,000.
12a. Home square footage	Apartments: 250 to 3000; Houses: 500 to 6000.
13. Lot size	One-Family Detached House: Greater than home square footage and 5 acres or less. All other homes: Suggest this field not be flagged.
15a. Age in years	0 to 200.
15b. Years since remodeled	0 to 50.
16. Bedrooms	Apartments: 0 to 4; All others: 1 to 8.
17a. Full baths	1 to the number of bedrooms plus 1.
17b. Half baths	0 to the number of bedrooms minus 1.
18. Other rooms	Apartments: 1 to 5; All other: 2 to 8.

Appendix D—General Population Rental Equivalence Survey—Final Response Rates

Area	#Wrkd	Hsehlds	Screener				Extended			Respondents
			Refusals	Ineligible	Eligible	Response (percent)	Refusals	Response (percent)	Ineligible	Total	Owners	Renters	Combined* (percent)
A—DC	820	594	231	3	360	61	72	80	122	166	61	105	49
B—Mont Co	858	676	302	3	371	55	77	79	95	199	126	73	44
C—PG Co	795	581	247	3	331	57	63	81	80	188	109	79	47
D—Arlington	320	252	148	3	101	41	25	75	20	56	21	35	31
E—Fairfax	1,016	798	365	2	431	54	77	82	109	245	155	90	45
F—PW	178	123	69	4	50	44	5	90	15	30	10	20	40
G—Anchorage	4,054	2,869	1,640	70	1,159	43	208	82	454	497	248	249	35
H—Fairbanks	1,436	1,135	444	47	644	61	90	86	282	272	150	122	52
I—Juneau	12,878	5,225	2,638	351	2,236	50	315	86	1,597	324	163	161	43
J—Honolulu	10,563	7,908	5,313	16	2,579	33	427	83	1,445	707	288	419	27
K—Hilo	2,953	2,339	1,382	57	900	41	167	81	505	228	123	105	33
L—Kona	9,454	4,009	2,857	19	1,133	29	238	79	723	172	87	85	23
M—Kauai	14,862	9,261	7,064	263	1,934	24	310	84	1,243	381	210	171	20
N—Maui	11,239	6,489	4,660	23	1,806	28	429	76	894	483	246	237	21
O—Guam	20,791	2,638	1,249	2	1,387	53	136	90	781	470	247	223	47
Puerto Rico	39,613	18,788	14,127	1	4,660	25	477	90	3,718	465	363	102	22
Q-St. Croix	10,004	3,178	1,405	1	1,772	56	392	78	662	718	533	185	43
R-St. Thomas	7,020	1,797	611	5	1,181	66	418	65	238	525	278	247	43
T-St. John	3,672	1,931	1,304	0	627	32	352	44	231	44	27	17	14
	152,526	70,591	46,056	873	23,662	35	4,278	82	13,214	6,170	3,445	2,725	28
* Combined response rate.

Appendix E—1998 Federal Employee Housing and Living Patterns Survey Sample Size, Responses, and Response Rates

Survey area	Number of non-postal federal employees	Sample size	Responses	Response rate (percent)
Anchorage	7,549	1,379	748	54.2
Fairbanks	1,625	519	320	61.7
Juneau	814	412	248	60.2
Rest of AK	2,413	524	336	64.1
Honolulu County	16,073	3,768	1,923	51.0
Hawaii County	728	577	378	65.5
Kauai County	332	332	182	54.8
Maui County	471	471	216	45.9
Guam	2,026	820	338	41.2
Puerto Rico	11,195	1,875	629	33.5
U.S. Virgin Islands	801	801	344	42.9
St. Croix			155
St. Thomas/St. John			184
COLA Areas Subtotal	44,027	11,478	5,662	49.3
Washington DC Area	258,304	4,324	1,081	25.0
Total	302,331	15,802	6,743	42.7