The President's Pay Agent
Locality Pay Surveys
Under FEPCA, we must use salary surveys conducted by the Bureau of Labor Statistics (BLS) to set locality pay. Commencing with the 1996/97 surveys, BLS implemented a new survey design for its salary surveys. The new survey program, called the National Compensation Survey (NCS) program, was used in all BLS salary surveys started after September 1996. NCS uses probability sampling of occupations within survey establishments, rather than a fixed job list with detailed job descriptions, as had been used in the past.
The new survey process was not immediately accepted for use in the locality pay program. In fact, the Federal Salary Council recommended that the original NCS methods not be used to set Federal pay. After reviewing test data and several years of production surveys, the Pay Agent agreed with the Federal Salary Council's conclusion that the NCS program, as originally configured, should not be used for the locality pay program. However, the Pay Agent did not ask BLS to reinstate the previous methodology. The Pay Agent concluded that the NCS program has several advantages over the previous salary survey program, the Occupational Compensation Survey Program (OCSP). These include offering greater occupational coverage, being less costly, and being less burdensome on respondents.
The Pay Agent also concluded that certain major aspects of the NCS program would have to be improved before it would be prudent to use NCS data for making pay comparisons under the locality pay program. In 2002, Pay Agent and BLS staff implemented three of the five planned improvements in the NCS program, and the Federal Salary Council recommended that we begin to phase in the use of NCS data to set locality pay.
In 2002, the Council recommended and we agreed to begin using NCS data by averaging the OCSP and NCS results (on a 50-50 basis). In 2003, the Council recommended and we agreed to continue the phase-in by weighting NCS results 75 percent and OCSP results 25 percent. In 2004, the Council recommended that we continue to phase in NCS results by applying a 90 percent weight to NCS results and a 10 percent weight to OCSP results. This year, the Council recommended and we agree to use only NCS survey results for the locality pay program.
The same three improvements first introduced in 2002 are incorporated into surveys used this year:
1) The linkage of Federal and non-Federal jobs by developing an improved crosswalk between General Schedule occupations and the Standard Occupational Classification (SOC) System to permit weighting data by Federal employment.
2) The development of methods to identify and exclude survey jobs that would be graded above GS-15 in the Federal Government.
3) The development of an econometric model based on survey data to estimate salaries for jobs not found in the probability samples.
The remaining two improvements, which BLS is using in surveys now being conducted that will affect data delivered to the Pay Agent in future years, are the following:
1) The development and implementation of a four-factor job grading system with job family guides to improve grade leveling under the NCS program.
2) The development and implementation of better methods for grading supervisory jobs selected by probability sampling.
Industrial and Establishment Size Coverage
As required by FEPCA, BLS salary surveys used for the locality pay program include the collection of salary data from private industry and State and local governments, which have large numbers of workers, especially in certain occupations that are unique to government functions. Before 1991, BLS surveys for the pay comparability process covered only private sector goods-producing and service-producing industries.
BLS surveyed a total of 15,877 establishments for the data submitted for the locality pay program. In the 28 continuing separate metropolitan locality pay areas, BLS surveyed 8,453 establishments. The Rest of U.S. (RUS) locality pay survey covered 49 additional metropolitan areas and 70 non-metropolitan counties. A total of 7,424 establishments were surveyed in the RUS area, including establishments in Kansas City, St Louis, and Orlando, which the Pay Agent plans to merge with RUS in 2006, and establishments in Buffalo, Phoenix, and Raleigh, which the Pay Agent plans to make separate locality pay areas in 2006.
The industry scope of the surveys includes mining, construction, and manufacturing industries; service-producing industries, including transportation, communications, electric, gas, and sanitary services; wholesale trade; retail trade; finance, insurance, and real estate; services industries; and State and local governments. Households, agriculture, and the self-employed were excluded. The surveys covered establishments with 50 or more workers. In the future, BLS plans to extend the NCS program to cover all establishment sizes. The Pay Agent will review the data and consider the recommendations of the Federal Salary Council before expanding the scope of data used in the locality pay program.
Under the NCS program, BLS uses random sampling techniques to select occupations for survey within an establishment. The occupations are selected and weighted to represent all non-Federal occupations in the location and, based on the crosswalk published in Appendix VII of the 2002 Pay Agent's report, also represent virtually all GS employees. OPM provided the crosswalk between GS occupational series and the Standard Occupational Classification (SOC) system used by BLS to group non-Federal survey jobs. OPM also provided March 2004 GS employment counts for use in weighting survey job data to higher aggregates. (BLS completed delivery of the most recent NCS surveys in July 2005, before March 2005 GS employment counts became available.)
Matching Level of Work
In the NCS surveys, BLS field economists cannot use a set list of survey job descriptions because BLS uses a random sampling method and any non-Federal job can be selected in an establishment for leveling (i.e., grading). In addition, it is not feasible for BLS field economists to consult and use the entire GS position classification system to level survey jobs because it would simply take too long to gather all the information needed to level surveyed jobs. This would also place an undue burden on survey participants. Therefore, in its original NCS methodology, BLS adopted the primary standard of the GS Factor Evaluation System (FES) for use in leveling jobs that are selected randomly in the survey. The primary standard is a framework that guides OPM when developing detailed standards for occupations under the FES. However, when the FES was designed and tested in the 1970s, OPM's predecessor, the Civil Service Commission, found a high error rate when only the primary standard was used in leveling jobs. The Federal Salary Council and OPM staff concluded that tests of the NCS program methods revealed similar problems.
To improve grade leveling under the NCS program, OPM developed a simplified four-factor grade leveling system with 20 job family guides. These guides were designed to provide occupational-specific leveling instructions for the BLS field economists. The four factors were derived and validated by combining the nine factors under the existing FES. The factors were validated against a wide variety of GS positions and proved to replicate current grade levels.
The 20 job family guides cover the complete spectrum of white-collar work found in the Government. BLS and OPM have completed work on the guides, and BLS is now using the guides in its ongoing surveys. Fully implementing the new leveling system will take 5 years because of BLS' data collection cycle. See Appendix IV of the 2002 Pay Agent's report for a summary of the BLS data collection cycle. Appendix VI of the 2002 Pay Agent's report contains the 20 job family leveling guides.
Jobs above GS-15
For the NCS program, it was necessary to develop generic instructions for identifying white-collar jobs in the random surveys that would be graded above GS-15 if they existed in the Federal Government so that the data could be excluded from pay gap measurements. BLS developed and tested the guidance with assistance from OPM. Appendix V to the 2002 Pay Agent's report explains the process for identifying these jobs in the NCS program.
Grading Supervisory Positions
Grading supervisory jobs presented another problem for the NCS program because the Government does not use the FES approach to grade supervisory jobs. BLS' original NCS methodology included an experimental approach in which BLS first applied the FES to sampled supervisory positions and then added additional factor points for the level of supervision. OPM classifiers believed this experimental approach would not yield the correct grade level and suggested a new approach based on the highest level of work supervised. Under the new approach, BLS would grade the highest level of work supervised using the appropriate four-factor leveling guide, not the supervisory job itself, and then add one grade for a first-level supervisor, two grades for a second-level supervisor, and three grades for a third-level supervisor. BLS and OPM have completed work on developing this procedure, and BLS is now using the new procedures in its ongoing surveys. However, the data available for this report were not processed using the new approach.
As in the 2002 through 2004 reports, BLS excluded second- and third-level supervisors entirely from the NCS data this year. BLS graded first-level supervisors by using existing NCS grade leveling procedures. The Pay Agent issued these instructions to BLS because the grades of second- and third-level supervisors are more likely affected by their supervisory duties, while first-level supervisors are more likely graded based on other factors, such as technical expertise. This modification allowed us to use some of the data from supervisory positions.1 We anticipate that data from all levels of supervisory positions will be available and used next year.
While BLS surveys all white-collar jobs under the NCS program, it does not find all jobs at all work levels in each survey area. This is a serious problem with the NCS program because survey results and pay disparity measures can vary considerably based on which jobs are included. The Pay Agent asked BLS to develop an econometric model to provide estimates for jobs not found in NCS. The model is described later in this report and in Appendices II and III.
 Approximately 12 percent of the jobs sampled by BLS are supervisory, with 10 percent 1st level supervisors and 2 percent 2nd or 3rd level supervisors.