E8-12020. Nonforeign Area Cost-of-Living Allowance Rates; Puerto Rico and Hawaii County, HI

AGENCY:

Office of Personnel Management.

ACTION:

Final rule.

SUMMARY:

The Office of Personnel Management (OPM) is changing the cost-of-living allowance (COLA) rates received by certain white-collar Federal and U.S. Postal Service employees in Puerto Rico and Hawaii County, HI. The changes are the result of interim adjustments OPM calculated based on relative Consumer Price Index differences between the cost-of-living allowance areas and the Washington, DC, area. OPM is also making an additional one-time adjustment to the Puerto Rico COLA rate based on the impact of the new sales tax in Puerto Rico. This regulation increases the COLA rate for Puerto Rico to 13 percent and the COLA rate for Hawaii County, HI, to 18 percent.

DATES:

Effective date: June 30, 2008.

Implementation date: First day of the first pay period beginning on or after June 30, 2008.

FOR FURTHER INFORMATION CONTACT:

J. Stanley Austin, [phone]; fax: [phone]; or e-mail: [e-mail].

SUPPLEMENTARY INFORMATION:

Section 5941 of title 5, United States Code, authorizes Federal agencies to pay cost-of-living allowances (COLAs) to white-collar Federal and U.S. Postal Service employees stationed in Alaska, Hawaii, Guam and the Northern Mariana Islands, Puerto Rico, and the U.S. Virgin Islands (USVI). Executive Order 10000, as amended, delegates to the Office of Personnel Management (OPM) the authority to administer nonforeign area COLAs and prescribes certain operational features of the program. OPM conducts living-cost surveys in each allowance area and in the Washington, DC, area to determine whether, and to what degree, COLA area living costs are higher than those in the DC area.

As required by section 591.223 of title 5, Code of Federal Regulations, OPM conducts COLA surveys in the Alaska, Pacific, and Caribbean areas on a 3-year rotating basis, and in the Washington, DC, area on an annual basis. OPM sets the COLA rate for each area based on the results of these surveys. For areas not surveyed during a particular year, OPM computes interim adjustments to COLA rates based on the relative change in the Consumer Price Index (CPI) for the COLA area compared with the Washington, DC, area. (See 5 CFR 591.224-591.226.)

OPM adopted the COLA survey methodology pursuant to the stipulation for settlement in Caraballo et al. v. United States, No. 1997-0027 (D.V.I.), August 17, 2000. Caraballo was a class-action lawsuit in which the plaintiffs contested the prior methodology OPM used to determine COLA rates. In the Caraballo settlement, the parties agreed that if the Government adopted and maintained certain changes in the COLA program, the plaintiffs would be barred from bringing suit over these issues. The stipulation for settlement is available on OPM's Web site at http://www.opm.gov/​oca/​cola/​settlement.asp.

Before the settlement, the parties entered into a memorandum of understanding under which they engaged in a cooperative process to study living-cost and compensation issues. The research was exhaustive and covered essentially all aspects of the COLA program. A summary of that research is available at http://www.opm.gov/​oca/​cola/​research.asp.

Exhibit A of the Caraballo settlement agreement lists 26 “Safe Harbor Principles” that outline the changes to which the parties agreed. These principles formed the basis for a new COLA methodology, which OPM incorporated into its regulations. In developing these regulations, OPM consulted with the Survey Implementation Committee, which was established under the Caraballo settlement and is composed of representatives of the parties in Caraballo. The Survey Implementation Committee in turn consulted with the Technical Advisory Committee, which was also established under the Caraballo settlement and is composed of three economists with expertise in living-cost comparisons. OPM published proposed regulations incorporating the new methodology in the Federal Register for notice and comment on November 9, 2001, at 66 FR 56741, and a final rule on May 3, 2002, at 67 FR 22339. The Survey Implementation Committee and the Technical Advisory Committee worked closely with OPM in preparing for and implementing the 2002, 2003, and 2004 COLA surveys.

On September 12, 2007, at 72 FR 52169, OPM published a Federal Register notice conveying the results of the 2006 interim adjustments for the Pacific and Caribbean COLA areas. We did not compute interim adjustments for the Alaska COLA areas because we surveyed Alaska in 2006. The interim adjustments indicated that, except for Hawaii County and Puerto Rico, the COLA rates for the Pacific and Caribbean COLA areas were set at the appropriate levels. For Hawaii County, the adjustments indicated an increase in the COLA rate from 17 percent to 18 percent. For Puerto Rico, the adjustments indicated an increase in the COLA rate from 10.5 percent to 11 percent.

On September 6, 2007, at 72 FR 51200, OPM proposed to further increase the Puerto Rico COLA rate to 13 percent to account for the impact on prices of the new Puerto Rico sales tax. This increase supersedes the 1 percent reduction previously proposed by OPM on October 27, 2006, at 71 FR 63176, which was based on the 2005 Caribbean survey results.

Discussion of Comments

We address comments received in response to the 2007 proposed rule on the rate increases in Puerto Rico and Hawaii County, HI, in this section. We also received comments in response to the 2006 proposed rule to reduce the COLA rate in Puerto Rico. Although the rate reduction will not be implemented, we also respond to these comments in this section.

2007 Proposed Rate Increases

Rising Living Costs

We received 253 comments in response to the 2007 proposed rate increases in Puerto Rico and Hawaii County, HI. Most of the commenters in Puerto Rico said they support the proposed increase in the Puerto Rico COLA rate; however, many commenters believed the increase should be higher than proposed. A number of commenters cited a rise in the Consumer Price Index produced by the Puerto Rico Department of Labor and Human Resources and other indicators as a basis for a higher COLA rate.

As required by section 5941 of title 5, U.S. Code, we compare living costs in the COLA areas with living costs in the Washington, DC, area to determine COLA rates. We survey the prices of over 240 items to use in the cost comparisons. The comparisons result in indexes that reflect how COLA area prices measure against DC area prices over a given period of time. The comparisons result in indexes that do not necessarily correspond to rising (or falling) prices in the COLA areas. For instance, if living costs in a COLA area rise, but living costs in the DC area rise more sharply, the COLA rate for the area would decrease. Conversely, if COLA area living costs decrease, but DC area living costs decrease more sharply, the COLA rate for the area would increase.

This regulation increases the COLA rate in Puerto Rico from 10.5 percent to 13 percent based partially on the relative change in the 2006 CPI for Puerto Rico compared with the Washington, DC, area and partially to account for the new sales tax implemented after the 2005 survey. While this change provides a 2.5 percent increase in the COLA rate for Federal employees in Puerto Rico, the actual change from the 2005 survey index of 103.32 (3 percent) to the adjusted index of 112.94 (13 percent) correlates to an effective increase of 10 percent. (The 3-percent COLA rate indicated by the 2005 surveys was to be implemented in annual 1-percentage-point reductions.)

We used the CPI produced by the Puerto Rico Department of Labor and Human Resources for the 2006 interim adjustment for Puerto Rico. The Puerto Rico Department of Labor and Human Resources has since revised its methodology for producing the CPI. This change in producing the CPI does not affect the COLA index used for the Puerto Rico rate increase implemented in this regulation, but likely will affect future Puerto Rico interim adjustment comparisons.

A number of commenters noted that certain costs have increased since OPM conducted the survey. They cited the cost of gasoline, housing, utilities, airline tickets, grocery items, medical needs, automobile expenses, various fees and taxes, and other items. Several commenters believed we should survey more frequently. We recognize that prices for various items will increase in the COLA areas and/or the DC area between surveys. We collect prices in each survey area every 3 years on a rotating basis according to a schedule agreed upon by the parties in the Caraballo settlement. As noted previously, we adjust area price indexes in non-survey years based on the relative change in the CPI for the COLA area compared with the CPI for the Washington, DC, area.

One commenter said OPM should survey the cost of water, gas, electricity, and telephone utilities. We survey each of these items. We published a list of the items we surveyed in the Caribbean and DC areas in appendix 3 of the 2005 Caribbean Survey Report at 71 FR 63197.

Puerto Rico Sales Tax

Several commenters believed OPM did not fully account for coverage of the new Commonwealth and municipio sales tax, particularly with grocery items. We proposed a one-time adjustment to the Puerto Rico COLA index based on the sales tax, which had not yet been captured by the COLA surveys or reflected in the CPI adjustments. We obtained information on the applicability of the sales tax from the Puerto Rico Department of the Treasury (Hacienda). Using this information, we applied the sales tax to covered survey items to determine an aggregate indicator of the impact of the tax on the Puerto Rico COLA index. The index increased by 1.9 points, translating to a COLA rate increase of 2 percent. We did not attempt to account for variations in tax coverage by municipio as these variations would likely have an inconsequential effect on the index. Similarly, we did not account for price reductions for survey items no longer subject to the general excise tax because the effect would also likely be inconsequential. In the future, the sales tax will be added to the prices we survey and will be reflected in the Puerto Rico CPI used for the interim adjustments.

One association advocated making the Puerto Rico increase retroactive to November 5, 2007, the effective date of the new sales tax. Paragraph (d) of section 553 of title 5, U.S. Code, requires that regulations be issued with an effective date at least 30 days after publication. The Caraballo settlement agreement requires that we publish rate changes pursuant to section 553.

Recruitment and Retention

One commenter said the current economy in Puerto Rico is likely causing recruitment and retention problems. OPM is concerned about the Government's ability to recruit and retain a well-qualified workforce and notes that the Government has several pay authorities that are available to address recruitment and retention problems. Among these are special salary rates and recruitment, retention, and relocation incentives. OPM's Web site at http://www.opm.gov/​oca/​pay/​index.asp provides information on pay authorities to assist in agency recruitment and retention efforts.

Locality Pay

Several commenters noted their opposition to an Administration legislative proposal that would transition employees in the COLA areas to locality pay over a 7-year period. The commenters said a change to locality pay would lead to a decrease in net salaries in Puerto Rico because locality pay is subject to income tax. The proposed legislation would reduce COLAs by 85 percent of the added locality pay amount to help offset (by 15 percent) the tax liability of locality pay, but would not relieve employees of their total tax responsibility. Unlike with COLAs, all Federal employees, whether in the COLA areas or the 48 contiguous States and Washington, DC, must pay income tax on locality pay. Under the proposed legislation, a change to locality pay would also eliminate future COLA rate reductions and increases, confer retirement credit where the COLA did not, and provide higher pay potential not restricted by the 25-percent cap that applies to COLA rates.

Taxes

A number of commenters noted that the Puerto Rico income tax is higher than the Federal income tax paid by employees in other areas. By law, we must compare living costs in the COLA areas with living costs in the Washington, DC, area to determine the COLA rates for the areas. In the DC area, employees pay Federal income tax, State income tax (Virginia and Maryland), city income tax (DC), local income tax (Maryland counties), and personal property tax (Virginia counties). Employees in all areas have varying tax obligations depending on income, dependents, deductions, and other factors. Because of the complexity involved, we do not attempt to determine the aggregate income tax liability for employees in the COLA areas and the DC area for comparison purposes. The extent to which the total tax burden may be higher in a COLA area than in the DC area is covered by the adjustment factor we add to the price index for each COLA area pursuant to the Caraballo settlement agreement.

Rate Variations

A number of commenters said Hawaii and the U.S. Virgin Islands have a similar economic situation to Puerto Rico, but have higher COLA rates. One commenter thought that costs in Puerto Rico justified the same 25 percent COLA received by employees in Alaska and Hawaii. Two commenters said that prices are higher in Puerto Rico because items must be imported. There are innumerable economic influences that affect prices in an area, including poverty rate, housing vacancy rate, availability of goods and services, competition, and importation costs. We survey using the same methodology and essentially the same marketbasket in all areas. We survey the final cost to the consumer of items and services in each area. The final cost includes any overhead, transportation and shipping costs, taxes, competition, and other price influences. Additionally, we survey catalog prices for a number of items and include in the price the costs for shipping, sales tax, and excise tax, which are often higher in the COLA areas relative to the Washington, DC, area.

We use this data to compare living costs in the COLA areas with living costs in the DC area. The surveys and subsequent interim adjustments have indicated a 25-percent COLA rate for the U.S. Virgin Islands and 3 of the 4 allowance areas in Hawaii, but a 10-percent rate for Anchorage, AK, a 19-percent rate for Fairbanks, AK, and a 20-percent rate for Juneau, AK. The Anchorage COLA index is below the index for Puerto Rico. Actual COLA rates are currently higher in Alaska because the Caraballo settlement established rates based on historical levels in the areas. The COLA rates in Alaska remain higher than indicated by OPM's surveys because we may reduce rates by no more than 1 percent in a 12-month period. We have published at 73 FR 772 a proposed second rate reduction, from 24 to 23 percent, for Anchorage, Fairbanks, and Juneau.

One commenter described how his living costs increased on moving to Puerto Rico from Texas. As noted earlier, section 5941 of title 5, U.S. Code, requires that we compare living costs in Puerto Rico with living costs in the Washington, DC, area to set the Puerto Rico COLA rate. We do not conduct cost-of-living surveys in other areas of the continental United States.

Housing Costs

Two commenters noted the high cost of housing in safe neighborhoods and high mortgage rates in Puerto Rico. As stipulated by the Caraballo settlement, we use rental equivalence to determine shelter costs in the COLA areas. We discuss the rental survey, including neighborhood selection, later in this section.

Hawaii County

Two commenters said there should be separate COLA rates for the east (Hilo) and west (Kona) sides of the island of Hawaii because prices in these areas are not equal. There are communities in each of the nonforeign COLA areas (and in the DC area) that are more expensive than other communities within the same COLA area. It is not feasible or practical to segment each of these communities, many of which share numerous economic characteristics, into independent survey areas with separate COLA rates. For this reason, we do not plan to split Hawaii County into two separate COLA areas. However, we remain open to a mutual recommendation on this issue from the COLA Advisory Committees in Hilo and Kona.

2006 Proposed Reduction

We received 204 comments in response to the 2006 proposed reduction in the Puerto Rico COLA rate published at 71 FR 63176. Although the increase implemented by these regulations supersedes the 2006 proposed reduction, we respond to the comments we received in the discussion that follows.

Increasing Costs

Many of the commenters said OPM should not reduce COLA rates because Puerto Rico living costs were increasing. As noted previously, section 5941 of title 5, U.S. Code, requires that we measure costs in the COLA areas against costs in the Washington, DC, area to determine COLA rates. We increase the COLA rate if the difference in living costs between the COLA area and the DC area increases and reduce the rate if the difference in living costs between the COLA area and the DC area decreases. As provided by 5 CFR 591.228(c), we reduce COLA rates by no more than 1 percentage point in a 12-month period.

A number of commenters referred to publications or other surveys showing high or rising costs in Puerto Rico, indicating the COLA rate should be set higher. We measure costs using the methodology stipulated in the Caraballo settlement and cannot comment on the methodology used by other publications and surveys. We conduct on-site surveys in each survey area and collect more than 4,600 prices on over 240 items representing typical consumer purchases. We collect prices at over 900 outlets, including grocery, hardware, electronics, and department stores, as well as automobile dealers, doctors, dentists, insurance companies, and many other providers of goods and services. We collect these prices in both the COLA and DC areas to use in the price comparisons that determine each area's COLA rate.

Numerous commenters noted that certain costs increased after OPM conducted the 2005 survey and that the survey data were outdated. They cited the cost of gasoline, housing, utilities, grocery items, medical needs, various fees and taxes, and other items. Many commenters requested that OPM survey again. As noted previously, we recognize that prices for items may increase in the COLA areas and/or the DC area between surveys. We collect prices in each survey area every 3 years on a rotating basis according to a schedule agreed upon by the parties in the Caraballo settlement. As stipulated in the settlement, we adjust COLA rates annually between surveys based on the relative change in the CPI for the COLA area as compared with the Washington, DC, area. We discuss this adjustment in the notice on the 2006 interim adjustments published at 72 FR 52169. These adjustments are designed to account for price fluctuations between surveys. The 2006 interim adjustment calculation raised the Puerto Rico index, making the proposed COLA rate reduction no longer necessary.

New Sales Tax

We received a number of comments on the new sales tax in Puerto Rico. As we discussed previously, we are implementing an adjustment to account for the impact of the sales tax as part of the rate increase to 13 percent.

Rate Change Delay

One agency commented on the delay in implementing COLA rate adjustments. As set out in the Caraballo settlement, we survey each COLA area on a triennial basis and make interim adjustments based on CPI changes in the years between surveys. We also may make adjustments based on special circumstances, such as with the Puerto Rico sales tax adjustment. While we make efforts to implement COLA rate adjustments in a timely fashion, we must follow the rulemaking procedures mandated by the Administrative Procedure Act (section 553 of title 5, U.S. Code) and various other statutory and regulatory requirements before implementing any rate change. These requirements largely determine the interval for making a rate change effective.

Comparison With DC

The same agency also commented on the use of Washington, DC, as the basis for COLA living-cost comparisons. While the agency conceded that this requirement is in statute (section 5941 of title 5, U.S. Code), it observed that the DC area has become more expensive over time, resulting in less variance between the DC and COLA areas. Because the requirement to use Washington, DC, as the basis for comparison is mandated in the statute that authorizes COLAs, we do not have authority to address this issue by regulation.

COLA/Locality Pay

The agency also raised the issue of replacing the nonforeign area COLA with locality pay. The Federal Employees Pay Comparability Act of 1990 authorizes locality pay only for Federal employees in the contiguous 48 States and Washington, DC. We do not have authority to address this issue by regulation. However, as noted earlier in this section, the Administration has submitted proposed legislation for consideration by Congress that would convert employees from COLAs to locality pay over time.

Recruitment, Relocation, and Retention Incentives

The agency noted that reductions in COLA rates may require greater use of discretionary authorities, such as recruitment, relocation, and retention incentives. As noted previously, OPM's Web site at www.opm.gov/​oca/​pay/​index.asp provides information on pay authorities to assist in agency recruitment and retention efforts.

Employee Involvement

One commenter believed OPM did not conduct the survey in Puerto Rico with local Federal employees. The commenter indicated that OPM should have surveyed a sample of Puerto Rico Federal employees. We conduct on-site price surveys of a marketbasket of goods and services representing typical consumer purchases as prescribed by the Caraballo settlement. Observers from the Puerto Rico COLA Advisory Committee, which is composed of current Federal employees who live in Puerto Rico, accompanied the OPM data collectors during the non-rental price survey. Before the 2005 Caribbean survey, we established a COLA Advisory Committee in each of the survey areas. As described in 5 CFR 591.243, each Committee is composed of approximately 12 agency and employee representatives from the survey area and two representatives from OPM. We held 3-day meetings with the COLA Advisory Committees in each area to be surveyed to plan the COLA surveys. During the 2005 survey, the Committee members assisted OPM staff in collecting non-rental data, and after the survey the Committee members had the opportunity to review all of the survey results, including the results of the rental survey. Although COLA Advisory Committee members helped plan the rental survey and had the opportunity to review the rental survey results in detail, Committee members did not participate in the rental data collection as observers.

Rental Surveys

One local union in Puerto Rico offered extensive comments on the Puerto Rico rental survey. The union disputed the overall veracity, reliability, and adequacy of the rental data collected in Puerto Rico. The union claimed OPM knowingly and willfully harmed Puerto Rico employees through the fashion in which it collected, evaluated, analyzed, and utilized the rental data in Puerto Rico. The union and many other commenters asserted that OPM's actions did not conform to the Caraballo settlement or Safe Harbor Principles 5 (regarding quality and quantity comparisons), 18 (regarding the hedonic housing model and rental equivalence), and 22A (regarding survey plans and methodology).

As noted previously, the Caraballo settlement prescribed the methodology we use to conduct COLA surveys and set COLA rates. The settlement stipulates that OPM use a rental equivalence approach to estimate shelter costs and a hedonic regression approach to compare housing of similar quality. The Technical Advisory Committee economists worked with OPM and the Survey Implementation Committee to develop methodologies for the rental equivalence and hedonic regression processes. The settlement agreement did not require OPM to use a particular method to collect rental data; however, OPM provided its draft rental data collection specifications and procedures to the Survey Implementation Committee and Technical Advisory Committee for review and comment.

We contracted for the services of a company with an outstanding depth of experience in rental data collection to survey rental properties in the COLA areas for the 2004 through 2007 surveys. The contractor collected the data in essentially the same manner in all areas. Using parameters defined by OPM, the contractor collected rental data on-site in Puerto Rico from March 13 through May 5, 2005. Following its survey of the Caribbean areas, the contractor surveyed the Washington, DC, area. The contractor delivered the rental survey data to OPM in June 2005. We manually reviewed the rental data and performed various computer-based quality assurance checks on the rental database. We believe the 2005 rental survey was in full conformance with the settlement agreement and was stringently conducted under the rental equivalence and hedonic regression methodology mutually developed by the Survey Implementation Committee, Technical Advisory Committee, and OPM.

The union further maintained the rental data did not accurately reflect the areas or types of housing units where Federal employees live. The union said that only a small percentage of the rental observations in Puerto Rico were in areas where the median income level equals or exceeds the local average Federal salary. The union also said OPM and the Technical Advisory Committee invented new categories of housing units, “apartments in home” and “other,” almost exclusively for Puerto Rico. The union said these two categories of housing units were substandard and not representative of where Federal employees live.

We used data from the 2000 census that show the number of Federal employees and the number of housing units by municipio to determine which locations to survey and how many samples the rental survey contractor should attempt to collect in each location. We allocated more samples to locations that have a large number of Federal employees and a large number of housing units and fewer samples to locations that have a small number of Federal employees and housing units. If the location had no Federal employees, we excluded the location from the survey.

We held a 3-day meeting with the Puerto Rico COLA Advisory Committee on January 18-20, 2005, to plan the 2005 Puerto Rico rental and non-rental surveys. At the meeting, we shared with the Committee a map that showed the rental survey locations and the requested number of samples from each location. At the Committee's request, we agreed to further refine the survey locations using zip codes where practical. We did this for the San Juan, Carolina, and Bayamon municipios.

We also collect information that reflects the quality of neighborhoods and use additional information from the Bureau of the Census to introduce supplementary variables to the hedonic regressions that indicate neighborhood quality. To do this, we identify the census tract in which each rental observation is found and then add variables, such as median income, percent of school-age persons, and percent of people in the area with B.A. degrees or higher, to the hedonic regressions. Those variables that prove to be statistically significant and increase the precision of the rent index are used in the final hedonic regression equation.

In the 2005 hedonic regression analysis, we tested whether median income or median income paired with median income squared should be included in the equation. We found that median income was not a statistically significant variable at the 99.9 confidence level and dropped it from the hedonic regression.

The variable “Type of Unit” has eight subcategories: (1) Detached house, (2) duplex, (3) triplex, (4) townhouse/row house, (5) in-home apartment, (6) walk-up apartment, (7) high rise apartment, and (8) other. An “in-home apartment” is usually in a structure one to three stories tall with generally four or five units within the structure. Sometimes the original structure is a large, older home that has been converted to apartments. In other cases, the original structure may have been a triplex or quadplex. These units were found only in Puerto Rico and the U.S. Virgin Islands—none were found in the DC area. We tested the effect of dropping all in-home apartments in Puerto Rico in the final hedonic regression equation. The net result was a slight increase in the Puerto Rico rent index from 63.49 to 63.95, which had an inconsequential effect on the final survey living-cost index.

Units classified as “Other” are apartments in larger buildings that are not duplexes, triplexes, high rise apartments, typical walk-up apartment complexes, or in-home apartments. These were found mainly in Puerto Rico. In consultation with the Technical Advisory Committee, we collapsed “Type of Unit” into three subcategories: (1) Apartments of any kind, (2) townhouse/row house/duplex/triplex, and (3) detached house. “Collapsing” means combining two or more variables or subcategories within a variable. We generally do this when the variable or subcategory parameter estimates are similar and doing so improves the accuracy of the survey area parameter estimates. We assigned units classified as “in-home apartment” and “other” to the “apartments of any kind” subcategory. We then used hedonic regressions to compare the COLA area rents with DC area rents, while holding quality and quantity constant.

The union claimed OPM did not exercise any supervision over the contractor's data collection and accepted all data submitted by the contractor. We engaged a number of controls on the rental data furnished by the contractor. We established the specifications and locations for the rental survey in the contract and provided that payment would be made only for properties meeting the specifications. We required progress reports, shortfall reports, and other documentation during the course of the rental survey. As noted previously, we performed quality assurance checks on the data delivered by the contractor, including manually comparing property data against the property photograph(s) and sketch. We mapped the properties using longitude and latitude coordinates to verify geographic locations. Additionally, we provided the rental data to Puerto Rico agency and union representatives on the Puerto Rico COLA Advisory Committee for evaluation and comment.

Disparate Treatment

The union stated that OPM treated Puerto Rico COLA employees in a disparate fashion because of national origin and without regard to unique linguistic and cultural differences. The union cited misspellings in the rental data as evidence that the data collectors encountered a serious language barrier.

OPM and the rental contractor respect linguistic and cultural differences in Puerto Rico. Both OPM and the contractor assigned Spanish-speaking data collectors, some of whom were former residents of Puerto Rico, to the price and rental surveys. In addition, OPM arranged for observers from the Puerto Rico COLA Advisory Committee, which is composed of current Federal employees who live in Puerto Rico, to accompany the data collectors surveying non-rental prices. The rental data contained some misspellings in business names and street addresses, but the overall rental data were high-quality and fulfilled the COLA survey specifications for rental prices in Puerto Rico. Misspellings in names and addresses did not affect the rental prices used to determine the rent index.

We conduct COLA surveys the same in all areas using the methodology prescribed by the Caraballo settlement. The rental survey contractor similarly does not vary its approach for collecting rental data in the COLA areas. To the extent cultural differences in Puerto Rico affect prices, the survey accounts for such differences. Additionally, we add 7 points to the Puerto Rico COLA index to account for other costs that may be influenced in part by local or cultural differences. For the rental surveys, we note that cultural differences likely explain variations in advertising methods (e.g., more rent-by-owner signs in Puerto Rico) and the quantities of certain housing types (such as in-home apartments) between Puerto Rico and the DC area. We discuss these variations elsewhere in this section.

The union said that OPM made significant changes to the Alaska and Pacific rental surveys based on the union's comments, but did not employ the changes in the 2005 Caribbean rental survey. We made refinements in the hedonic regression analysis, including adding listing source and self identification refusal as variables based on the union's comments, but applied all changes uniformly to the 2005, 2006, and 2007 surveys. We also initiated trial observation of the rental survey in the 2006 Alaska survey based on the union's comments; however, as we note in the discussion that follows, we plan to extend the trial period through the 2008 Caribbean survey.

Rental Survey Observers

The union believed OPM should have allowed observers from the Puerto Rico COLA Advisory Committee to accompany the contractor on the 2005 rental surveys in Puerto Rico. We permitted observers from the COLA Advisory Committees to accompany OPM data collectors conducting the non-rental price surveys beginning with the 2002 Caribbean surveys, but did not similarly arrange for observers to accompany the contractor conducting the rental surveys. The union originally requested that we permit observers for the 2005 rental surveys during a pre-survey meeting of the Puerto Rico COLA Advisory Committee on January 18, 2005. At that time, the contract for the rental surveys did not provide for observers. We determined there was not sufficient time to consider and resolve various issues (e.g., higher contract costs, logistical problems, and possible conflict of interest), establish ground rules for observers, and issue a contract modification before the scheduled 2005 rental data collection. Although we could not provide for rental survey observers in 2005, we did arrange for the contractor to meet with the Puerto Rico COLA Advisory Committee and answer questions regarding the data collection process.

Following the 2005 rental surveys, we negotiated with the contractor to permit rental survey observers on a trial basis. We have extended the trial observation period through the 2008 Caribbean surveys so that all COLA area committees will have an opportunity to observe, but not otherwise participate in, the rental data collection process.

Rental survey observations enable COLA Advisory Committee members to see how the contractor collects rental data in the field. To maintain survey integrity, we instruct the observers not to attempt to advise, direct, or influence the data collectors. Committee members have an opportunity to participate in setting the rental survey parameters in the pre-survey meeting. Regardless of whether committee members observe the collection, we provide the collected rental data to the committee for review and comment.

The union said that OPM did not provide the Puerto Rico COLA Advisory Committee truthful and/or accurate information regarding the rental contractor's work hours. OPM had noted the contractor's late work hours as one of the impediments to permitting Committee members to observe the rental survey. The union said that because the photographs of the rental units were taken during daylight hours, the contractor could not have worked evenings and/or nights. We would not have found photographs taken in the dark acceptable, so are not surprised that the contractor arranged to photograph the units in the daylight. We note that dawn to dusk in Puerto Rico is approximately 6 a.m. to 7 p.m. in April. We have since negotiated to permit Committee members to observe the rental survey during normal work hours.

Manual Data Review

The union said OPM's difficulties in providing the Puerto Rico COLA Advisory Committee with a printed copy of the rental data meant OPM could not have conducted a manual review, because this would have required a printed copy. We received a printed copy of the rental data from the rental survey contractor and used this for our manual review. We did not provide the contractor copy to the Puerto Rico COLA Advisory Committee, but instead elected to print a new copy from our electronic database. We did this because the contractor's copy did not reflect the changes we made following our manual and automated reviews; the pictures and sketches on the copy we produced were larger, which we believed made it easier to see details; copying the contractor's two-sided forms on a copier was less reliable than printing from the rental database; and we added census tract information to the rental database, which was not on the contractor's copy. We encountered initial difficulties in printing the copy from our database, but we resolved the problem and were able to provide a copy to the COLA Advisory Committee.

Data Quality

The union stated that OPM did not follow the established protocol for developing a reliable hedonic regression model. The union said OPM should have reviewed the rental data, verified the accuracy of the data, eliminated unverifiable data, and determined that the remaining data were not sufficient to support a reliable hedonic model. The union further said OPM and the Technical Advisory Committee knowingly ran statistical programs over the deficient data and that OPM and the Technical Advisory Committee should have known that the rental data were not of comparable quality and therefore not fit to support a reliable hedonic regression model. The union said a process must be developed whereby an adequate sample of accurate, verifiable, and comparable rental data is utilized before any hedonic regression to adjust for quality differences is made.

The current process provides ample accurate, verifiable, and comparable rental data to determine rental equivalence. In the 2005 Caribbean survey, the sample consisted of over 400 rental observations in Puerto Rico and over 900 observations in the Washington, DC, area. To assure the data were accurate, we conducted various automated and manual reviews as described earlier in this section. To enable data to be verified, we obtained housing unit addresses, geographic coordinates, and photographs, which we reviewed and provided to the Puerto Rico COLA Advisory Committee. To assure comparability, we employed hedonic regression analysis as described in the 2005 Caribbean survey report at 71 FR 63184. The Technical Advisory Committee economists developed the hedonic regression model in consultation with OPM and the Survey Implementation Committee in accordance with the Caraballo settlement and Safe Harbor Principle 18.

Data Verification

The union said OPM knew most of the data were not verifiable because names and addresses of information sources were not provided. Approximately 17 percent of the 2005 rental observations in Puerto Rico were from sources who refused to provide self-identifying information, and no observations in the DC area were from such sources. COLA rental surveys are voluntary; therefore, OPM cannot require the source to provide self-identifying information. In reviewing the rental data, we found no indication that the information was misrepresented or collected in an unacceptable manner. The contractor provided the address, geographic coordinates (longitude and latitude), and a photograph for each unit, among other information. We believe this information is sufficient for verification.

We also analyzed whether source refusals to provide self-identifying information had a statistically significant influence on rental rates in Puerto Rico. We added self-identification refusal as a variable in the hedonic regression analysis. The hypothesis was that properties belonging to or managed by individuals who refused to provide self-identifying information would rent for less than equivalent properties where the source provided self-identifying information. We found that self-identification refusal was not a statistically significant variable. This means that whether or not the source provided self-identifying information did not appear to have an influence on rental rates. Therefore, we see no reason to exclude observations where self-identifying information is withheld.

The union said the data also were not verifiable because a high percentage of the Puerto Rico rental samples were obtained by “drive-by” observations and supplemented later from the contractor's non-local headquarters office. A “drive-by” property is one that is advertised by a sign posted on the property. The contractor collects information from five types of sources: local newspaper/publication, Internet, agent/broker, drive-by/sign posted, and other. The contractor collected data from all types except “other” in both Puerto Rico and the DC area, but the distribution of observations by listing source type varied by area.

The contractor often finds properties with “For Rent” signs while driving through rental survey neighborhoods. If the property appears to meet contract specifications, the contractor takes photographs of the unit, records measurable and visual observations, and notes the telephone number and other contact information provided on the “For Rent” sign. The contact can be a private individual, but in the DC area, the “For Rent” signs often provide the name of a property management company. The contractor then calls the point of contact, either locally or from its non-local headquarters, and obtains the rest of the required information about the rental unit. We do not require the contractor to document these calls separately because the provided survey data fully documents the information collected.

To determine whether listing source influenced rental rates, we added listing source as a variable in the hedonic regression analysis. We found that the variable was statistically significant, but that it raised the standard error of the survey area parameter estimates. Therefore, we did not use listing source as a variable in the final hedonic regression equation.

Addressing Union Concerns

The union stated it consistently notified OPM of its concerns with the rental survey and data collection, but OPM did not make a serious attempt to acknowledge, recognize, and address the many valid issues the union raised. We received several letters from the union regarding rental survey issues in response to two proposed COLA rate reductions in Puerto Rico. We replied to each of the union's concerns in detailed letters and also addressed its concerns in final regulations published in the Federal Register on August 2, 2006 (71 FR 43897). We again address the union's concerns in this discussion.

Rental Survey Support

The union criticized OPM for consistently supporting the contractor's work with respect to the 2005 rental survey. We support the 2005 COLA rental and non-rental surveys because the surveys were conducted in accordance with the methodology prescribed by the Caraballo settlement and developed in full collaboration with the Survey Implementation Committee and the Technical Advisory Committee economists. The contractor supplied rental data that fulfilled the contract specifications set by OPM for acquiring sufficient data to determine rental equivalency under the settlement methodology.

Substitutions

The union claimed OPM accepted endless substitutions of housing units in Puerto Rico from the contractor, allowing the introduction of bias to the housing sample. This is not correct. The contract prescribed the order in which the contractor would attempt to collect the data and specified the steps the contractor would take if it were unable to collect the requested number of observations within a class in a listed location. The contract allowed the contractor to do this without our direct supervision or involvement so that the rental survey could be conducted within a relatively short timeframe and because we did not have superior knowledge about what was available for rent in the local rental market. Although we were not involved in the substitution process, we received required reports that showed how the contractor allocated the shortfalls. The following is a brief description of how the contract addressed substitutions.

We determined the Puerto Rico sample size mainly based on the number of observations a contractor could reasonably be expected to collect within the survey time period. Next, we used information from the 2000 Census to distribute the samples in Puerto Rico by zip code among the locations where Federal employees live. We used the same approach in the DC area and in the U.S. Virgin Islands.

Within each location, we asked the contractor to collect information on the following six classes of housing: Class A—four bedroom, single family unit not to exceed 3200 square feet; Class B—three bedroom, single family unit not to exceed 2600 square feet; Class C—two bedroom, single family unit not to exceed 2200 square feet; Class D—three bedroom apartment unit not to exceed 2000 square feet; Class E—two bedroom apartment unit not to exceed 1800 square feet; Class F—one bedroom apartment unit not to exceed 1400 square feet. In most cases, we distributed the location target sample among the class on an equal basis, although sometimes we varied the class distribution based on the projected distribution within the location.

In designing the rental specifications, we recognized it was unlikely that a contractor would be able to find observations that exactly corresponded to the target distribution in the contract. Therefore, we established a process in the contract that enables the contractor to adjust the distribution throughout the survey by successively redistributing the shortfall according to a series of rules. The shortfall was the difference between the target amount and what was actually found.

At the lowest level, the contract distribution specified the target amount for a housing class within a location in a survey area. If the contractor could not find that amount, the contractor allocated the shortfall to the next most similar housing class within the location. For example, if we asked the contractor to collect six Class A observations in a location but the contractor could only find four, the contractor assigned the shortfall to the next most similar housing class within that location, and repeated the process. By the time the contractor had completed surveying the location, if it still had a shortfall, the contract required the contractor to allocate that shortfall among the observations in the next location. For example, if the target amount for a particular location was 36 but the contractor could only collect 30, the contract required the contractor to distribute the shortfall among the housing classes in the next location.

In the last step, if the contractor was unable to collect the number of samples requested for the survey area, the contractor was required to distribute the shortfall to the next survey area listed in the contract. In the case of the 2005 survey, the contractor obtained 445 of the requested maximum 480 samples in Puerto Rico, so it redistributed 35 samples to the Washington, DC, area.

Hedonic Regressions

The union also claimed the hedonic regressions performed by OPM and the Technical Advisory Committee to arrive at the 2005 housing index for Puerto Rico were inaccurate and invalid. The methodology we used to produce the rent indexes was an objective, multi-step process by which we eliminated variables that were not statistically significant. As required by Safe Harbor Principle 18 of the Caraballo settlement, we use hedonic regressions to analyze the rental data. We do not use a “matched-model” approach; i.e., we do not compare the price of a 1,000 square foot, 3-bedroom apartment in a COLA area with the same size 3-bedroom apartment in the DC area. Hedonic regressions are a type of multiple regression, which is a commonly used mathematical process that describes how one or more things—the independent variables—affect something else—the dependent variable. The regression results show the influence, on average, of each independent variable on the dependent variable while holding all of the other independent variables constant.

We use the logarithm of rent as the dependent variable. This is a commonly used approach and was recommended by the Technical Advisory Committee economists. The independent variables we use are various rental unit characteristics. Variables may be continuous—like square footage, number of bedrooms, or number of bathrooms—or class variables, like external condition (good, fair, etc.), availability of air conditioning (yes, no), or the particular COLA survey area in which the rental unit is located. The resulting hedonic regression allows OPM to hold rental unit characteristics constant between the COLA area and the Washington, DC, area while comparing rents. In other words, we use hedonic regressions to compare rents for non-identical but comparable rental units by holding quality and quantity constant, to the extent practical. It is not practical to survey every characteristic of a rental unit. For example, we do not collect information on floor coverings, size and types of windows, color of bathroom fixtures, and size of closets. Instead, working with the Survey Implementation Committee, Technical Advisory Committee, and COLA Advisory Committees, we identified over 80 characteristics that seem likely to have an influence on rental prices. Similarly, it is not desirable from a statistical standpoint to use all 80-plus characteristics in the hedonic regressions. Therefore, OPM and the Technical Advisory Committee, in consultation with the Survey Implementation Committee, developed objective procedures to determine which rental unit characteristics to include in the regression equation.

Home Purchase Costs

One commenter believed OPM should survey home purchase costs instead of rental value. Under the Caraballo settlement, the parties agreed to adopt a rental equivalence approach similar to the one the Bureau of Labor Statistics uses for the Consumer Price Index. Rental equivalence compares the shelter value (rental value) of owned homes, rather than total owner costs, because the latter are influenced by the investment value of the home (i.e., what homeowners hope to realize as a profit when they sell their homes). As a rule, living-cost surveys do not compare how consumers invest their money.

Executive Order 12866, Regulatory Review

This rule has been reviewed by the Office of Management and Budget in accordance with Executive Order 12866.

Regulatory Flexibility Act

I certify that this regulation will not have a significant economic impact on a substantial number of small entities because the regulation will affect only Federal agencies and employees.

List of Subjects in 5 CFR Part 591

• Government employees
• Travel and transportation expenses
• Wages

Office of Personnel Management.

Linda M. Springer,

Director.

Accordingly, OPM amends subpart B of 5 CFR part 591 as follows:

PART 591—ALLOWANCES AND DIFFERENTIALS

Subpart B—Cost-of-Living Allowance and Post Differential—Nonforeign Areas

1. The authority citation for subpart B of 5 CFR part 591 continues to read as follows:

Authority: 5 U.S.C. 5941; E.O. 10000, 3 CFR, 1943-1948 Comp., p. 792; and E.O. 12510, 3 CFR, 1985 Comp., p. 338.

2. Revise appendix A of subpart B to read as follows:

Appendix A to Subpart B of Part 591—Places and Rates at Which Allowances Are Paid

This appendix lists the places approved for a cost-of-living allowance and shows the authorized allowance rate for each area. The allowance rate shown is paid as a percentage of an employee's rate of basic pay. The rates are subject to change based on the results of future surveys.

Geographic coverage	Allowance rate (percent)
State of Alaska:
City of Anchorage and 80-kilometer (50-mile) radius by road	24
City of Fairbanks and 80-kilometer (50-mile) radius by road	24
City of Juneau and 80-kilometer (50-mile) radius by road	24
Rest of the State	25
State of Hawaii:
City and County of Honolulu	25
Hawaii County, Hawaii	18
County of Kauai	25
County of Maui and County of Kalawao	25
Territory of Guam and Commonwealth of the Northern Mariana Islands	25
Commonwealth of Puerto Rico	13
U.S. Virgin Islands	25