Understanding the employment measures from the CPS and CES survey

Monthly Labor Review

Volumn 129, Issue 2, 2006, Pages 23-38

  Bowler, Mary ^a   Morisi, Teresa L ^a  

^a BUREAU OF LABOR STATISTICS   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 33644793318     PISSN: 00981818     EISSN: 00981818     Source Type: Journal    
DOI: None     Document Type: Article

References (40)

1. The National Bureau of Economic Research is generally recognized as the arbiter of business-cycle turning points. The organization determined that the latest recession began in March 2001 and ended in November of that year.
2. Since this article was prepared, data for recent years have been subject to routine revisions for seasonal adjustment, household survey population controls, and establishment survey benchmarking. These revisions do not significantly change the discrepancy between the surveys as it is described herein.
3. Detailed information on both the CES survey and the CPS is found in the BLS Handbook of Methods, on the Internet at http://www.bls.gov/opub/hom/homtoc.htm. (See chapter 1 for the CPS and chapter 2 for the CES survey.)
4. The household survey sample of approximately 60,000 addresses is selected from all countries and independent cities in the United States. The 3,141 countries and cities are formed into 2,007 sample areas (Primary Sampling Unites, or PSU's), which are then grouped into strata within each State. Then one sample area is chosen from within each stratum, with the probability of selection proportional to the population within the PSU; this totals 754 PSU'S. Sampling groups within these PSU'S are in the sample for 4 consecutive months, leave the sample for the next 8 months, and then return for the next 4 months. In each month, 1 of the 8 rotation groups is in its first month, while another rotation group is in its second months, and the group in its fifth month is returning after a break of 8 months. This rotation procedure allows the sampled households not to be overburdened. The CES sample includes about 160,000 businesses and government agencies in the United States, covering about 400,000 worksites. In June 2003, the CES program completed a redesign and moved from a quota-based sample to a probability-based sample. The program began to introduce the new sample in 2000 with one industry, and by June 2003, it was used for all industries under NAICS. The new sample design is a stratified, simple random sample of worksites clustered by Unemployment Insurance (UI) account number. The sampling frame for the CES survey is the list of employers in the Longitudinal Database (LDB) a universe file of business establishments that is derived from the UI tax system and collected by the BLS Quarterly Census of Employment and Wages (QCEW) program. The frame and the sample are updated twice a year. The sample is stratified by State, employment size, and industry. Each year, approximately 40,000 new units are added to the sample in order to rotate part of it, get representation from new firms, and realign the sample distribution with the universe distribution.
5. For more information on the 1994 CPS redesign, see "Revisions in the Current Population Survey Effective January 1994," Employment and Earnings (Bureau of Labor Statistics, February 1994), pp. 13-37.
6. For further details on changes in the CES survey in 2003, see Teresa L. Morisi, "Recent changes in the national Current Employment Statistics survey," Monthly Labor Review, June 2003, pp. 3-13; \ on the Internet at http://www.bls.gov/opub/mlr/2003/06/art1full.pdf.
7. Concurrent seasonal adjustment for the CES uses updated seasonal factors not only for the latest month, but also for the previous 2 months. With the release of each new estimate, these factors are routinely revised to incorporate additional sample responses for those months. Once a year, 5 years of historical monthly estimates are revised with the latest seasonal factors. For further information on concurrent seasonal adjustment for the CES, see Chris Manning, "Concurrent seasonal adjustment for national CES survey," Monthly Labor Review, October 2003, pp. 39-43; on the Internet at http://www.bls.gov/opub/mlr/2003/10/ressum2.pdf.
8. Some changes to the composite weighting also were implemented in the CPS in January 2003, and the survey questions on race and Hispanic ethnicity were modified to comply with new standards set by the Office of Management and Budget. Further details on the CPS changes in January 2003 are found in "Revisions to the Current Population Survey Effective in January 2003," on the Internet at http://www.bls.gov /cps/rvcps03.pdf,
9. and in Employment and Earnings (Bureau of Labor Statistics, February 2003), pp. 4-23.
10. In the CPS, only the latest month is seasonally adjusted with the latest seasonal factors. Once a year, 5 years of historical monthly estimates are revised with the latest seasonal factors.
11. Two such situations are as follows: Faculty in schools. The CES survey has a special treatment for "regular" faculty members of elementary and secondary schools and colleges and universities. Regular faculty are those persons who are professional or certified and who have a contractual arrangement (written or otherwise) for 1 or more years. Regular faculty are counted as employed the entire year, regardless of whether they get paid year round or have 9-or 10-month payment plans. This means that regular faculty members who do not work during the summer break or other breaks are still counted as employed. The treatment of regular faculty members differs from that of workers in all other industries in the CES program, who must have been paid during the reference week in order to be counted. In the CES survey, if regular faculty take another job during the summer months, they are counted twice for those months. The household survey has no such special treatment for faculty. One of the categories for those "with a job, not at work" includes vacation; therefore, faculty with a written or verbal contract to return to teaching in the fall would still be counted as employed during the summer break. Because CPS employment is a count of those employed, faculty who take a second job during the summer are not counted twice. Census workers. Every 10 years, the Census Bureau hires a number of temporary workers to staff the decennial census. Although some hiring is done prior to the census year, the bulk of the increase in workers and of the subsequent decline occurs in approximately February through May of the census year. The influx of these workers causes spikes in Federal Government employment and in total nonagricultural employment in the CES survey as additional jobs are added to payrolls. In order to connt the new workers accurately the Census Bureau transmits to the CES program actual counts of temporary workers hired to staff the census, and the CES program incorporates the counts into the CES survey estimates. The household survey does not add in outside counts for temporary census workers. Because the CPS counts each person who is employed once and does not count jobs, persons who have second jobs as temporary census workers will not be counted twice, as they would be in the CES survey. Household survey employment does not show any discernible effects due to the hiring of temporary census workers.
12. These data on secondary jobs are available from outgoing rotation groups in the CPS sample.
13. "Measuring Employment and Unemployment," President's Committee to Appraise Employment and Unemployment Statistics, 1962, pp. 113-21.
14. Gloria P. Green, "Comparing employment estimates from house-hold and payroll surveys," Mmddy Labor Review, December 1969, pp. 9-20.
15. See note 10, earlier, for a description of the treatment of faculty.
16. Counting the Labor Force (National Commission on Employment and Unemployment Statistics, Labor Day 1979), pp. 193-99.
17. John F. Stinson, Jr., "Comparison of Nonagricultural Estimates from Two Surveys," Employment and Earnings, March 1994, pp. 6-9.
18. John F. Stinson, Jr., "Moonlighting: a key to differences in measuring employment growth," Monthly Labor Review, February 1987, pp. 30-31.
19. Paul O. Flaim, "How many new jobs since 1982? Data from two surveys differ," Monthly Labor Review, August 1989, pp. 10-15.
20. Thomas Nardone, Mary Bowler, Jurgen Kropf, Katie Kirkland, and Signe Wetrogan, Examining the Discrepancy in Employment Growth between the CPS and the CES, report to the Federal Economic Statistics Advisory Committee on October 17, 2003; on the Internet at http://www.bls.gov/bls/fesacp2101703.pdf. Note that the reconciliation methodology used at the time was somewhat different from the current procedure. As a result, the report to the committee shows the unexplained portion of the employment discrepancy for 1994-2000 to be 4.2 million, somewhat less than the 4.5 million derived with current methods for this article.
21. See Mary Bowler, Randy E. Ilg, Stephen Miller, Ed Robison, and Anne Polivka, "Revisions to the Current Population Survey Effective in January 2003," Employment and Earnings, February 2003,pp. 423; \ on the Internet at http://www.bis.gov/cps/rvcpsO3.pdf.
22. In the household survey, the questions designed to elicit household members' employment status are asked about all household members aged 15 years and older. The published estimates, however, reflect only those aged 16 years and older.
23. The idea that both job changing and the different survey reference periods contribute to the discrepancy between the surveys has been around for some time. Earlier research in this area-by Antos, Barkume, Mixon, and Triplett in 1976, for example (cited in Nardone, Bowler, Kropf, Kirkland, and Wetrogan, Examining the Discrepancy) - was hindered by even less information than is available now.
24. Research by Bruce C. Fallick and Charles A. Fleischman, economists at the Federal Reserve Board, came to a similar conclusion. (See The Importance of Employer-to-Employer Flows in the U.S. Labor Market, Finance and Economics Discussion Series working paper 2001-18 (Federal Reserve Board, Division of Research and Statistics \ and Monetary Affairs, April 2001).)
25. The establishment survey benchmark data are based primarily on administrative tax reports filed with State ui programs and provided to the Bureau of Labor Statistics for the agency's OCEW program. Prison workers were not formally excluded from coverage under Federal in law until 1997. Although the Bureau never identified any instances in which Prison workers were reported in the OCEW employment figures. the possibility that they were did exist before 1997.
26. See 2000 CIA Directory (Baltimore, Correctional Industries Association, 2000).
27. On a related note, the CES employment measure includes a relatively small number of American workers employed by U.S. companies overseas. These workers are not in the scope of the CPS. In the mid- to late 1990s, there were fewer than 10,000 such workers, representing only a tiny portion of the discrepancy between the surveys.
28. See Patterns of Reporting Errors and Irregularities by 100 Employers with the Most Suspended Wage Items, Audit Report A-0398-31009 (Office of the Inspector General, Social Security Administration, September 1999);
29. and The Social Security Administration's Earnings Suspense File Tactical Plan and Efforts to Reduce the File's Growth and Size, Evaluation Report A-03-97-31003 (Office of the Inspector General, Social Security Administration, February 2000).
30. See, for example, Manuel de la Puente, Using Ethnography to Explain Why People Are Missed or Erroneously Included by the Census: Evidence from Small Area Ethnographic Shoes (U.S. Census Bureau, Center for Survey Methods Research, 1995); on the Internet at http://www.census.gov/srd/papers/pdf/mdp9501.pdf.
31. Simon Potter and Chinhui Juhn of the Federal Reserve Bank of New York present an excellent discussion of undercoverage in their article "Explaining the Recent Divergence in Payroll and Household Employment Growth," Current Issues in Economics and Finance (Federal Reserve Bank of New York, December 1999); on the Internet at http://www.ny.frb.org/research/current_issues/ci5-16.html.
32. The response error that occurred in the CES benchmark data - monthly employment figures reported by employers in their State UI tax reports and provided to the Bureau of Labor Statistics for the QCEW program - was discovered when CES employment estimates and the benchmark data diverged significantly in 1991.
33. Although many focused on the absolute difference in employment growth between the surveys, others conducted a more thoughtful analysis that compared the trend in the employment-population ratio from the CPS - a statistically stronger measure than the CPS employment level - with CES employment and found that they both indicated a similar, less-than-robust labor market. (See Mark Schweitzer and Guhan Venkatu, "Employment Surveys are Telling the Same (Sad) Story," Economic Commentary (Federal Reserve Bank of Cleveland. Research Department. May 15, 2004)); on the Internet at http://www.clevelandfed.org/Research/Com2004/05-15.pdf.
34. More detailed information is arailable on the bls Internet site at http:/ /www.bls.gov/ces/cesjobch.pdf.
35. In its seasonal adjustment process, the Bureau of Labor Statistics uses x-12 ARIMA software developed by the Bureau of the Census. For further information on both the net birth/death model and benchmark revisions, see Kirk Mueller, "bls Establishment Estimates Revised to Incorporate March 2005 Benchmarks," Employment and Earnings, February 2006; on the Internet at http://www.bls.gov/web/cesbmart.htm.
36. On a related note, some who criticized the establishment survey for failing to capture new firms have used the terms "new firms" and "startups" interchangeably with "self-employed." These are distinct employment concepts and such usage is misleading. As previously mentioned, the CES survey does not include the unincorpoated self-employed; the household survey provides this information. The CES birth/death model is used to estimate employment associated with new firms that have payroll employees.
37. For additional information, see "Assessing the Timeliness of Business Births in BLS Establishmett Statistics" (Bureau of Labor Statistics, Apr 7, 2004); on the Internet at http://www.bls.gov/eew /eta581study.pdf
38. The IRS Statistics of Income Division tabulated for the Bureau the number of taxpayers with nonemployee income reported on Form 1099-misc from Individual Statistics of Income samples matched to Forms 1099-misc from the Information Returns Master File. These samples excluded instances in which a corporation issued a 1099-misc to another corporation. Even this refined tabulation, however, would include a number of types of payments that were not necessarily for independent contractor services, so it should not be construed as an estimate of the number of independent contractors.
39. Statistics of Income Division, Internal Revenue Service, Table 11, Partnership Returns: Selected Balance Shed and Income Statement Items for Specific Income Years. These types of companies are under-stated because some busmesses failed to answer the question about type of partnership on their tax returns.
40. The BLS analysis categorized paid and unpaid work as follows: paid work includes unsubsidized employment, subsidized private employment and subsidized public employment; unpaid work includes work experience programs on-the-job training, job searches, community service, vocational education, job skills training, education, providing childcare, and other TANF-approved activities.