At the beginning of each month, the Bureau of Labor Statistics (BLS) reports the change in payroll employment for the previous month.1 This estimate of jobs gained or lost over the month is closely watched by policymakers and those who work in financial markets and the media. When the estimate is revised in subsequent months, however, data users sometimes perceive a very different picture of the job market than what was initially reported.2 Data users frequently ask why the number was revised. The short answer is, the revised estimate includes additional information that was not available at the time of the initial release—information that makes the revised estimate more accurate.
This Beyond the Numbers article explains the data collection process that BLS conducts every month to produce the estimate of U.S. employment change. The article will also clarify why BLS releases revisions to the initial release, so that users will better understand the change, if any, in the data.
The estimate of employment change is based on a monthly survey of about 560,000 worksites, selected to represent the millions of businesses throughout the country. (For simplicity, we will refer to worksites as businesses even though many individual businesses provide data for multiple worksites.) In the survey sample, businesses report the total number of people who worked or received pay during the pay period that includes the 12th of the month. Although BLS uses a variety of methods to gather these reports as quickly as possible, many businesses do not have their payroll data ready to report by the scheduled date that BLS initially releases the data. In 2012, for example, the average collection rate at the time of the initial release was 73.1 percent.
The initial estimate of job change for a month is based on the growth or loss of jobs at the businesses that have reported their data. Generally, BLS assumes that the employment situation at businesses that had reported is representative of the situation at those that had not yet reported. BLS continues to collect outstanding reports from the businesses in the sample as it prepares a second and then a third estimate for the month. With each subsequent estimate, more businesses have provided their information. In 2012, the average collection rate at the time of the third estimate for a month was 94.6 percent. (See chart 1.)
Differences between the initial and revised estimates generally indicate that the employment change that occurred at the businesses that had not initially reported was different than the change that occurred at the businesses that had initially reported. For example, if less employment growth occurred among those who had not reported at the time of the first estimate, the initial estimate would be revised down. If more growth occurred among the late responders, the initial estimate would be revised up. The estimation process is more complicated than this simple explanation, however; the key thing for users to remember is that the revised estimates represent a more complete and therefore more accurate picture of developments in the job market.
So why doesn’t BLS wait until it has all the reports to make the estimate and avoid revisions? Users of the data are intensely interested in the earliest possible read on labor market developments, and experience suggests that the initial estimate is generally very good. For example, in 2012, the average monthly employment change using the first estimate would have been +142,000, compared with a monthly average change of +165,000 using the third estimate. (See table 1.) Nevertheless, it is true that in some months, revisions are large enough that they change the users’ perspectives on the current state of the economy. In November 2012, for example, the initial estimate of over-the-month change was +146,000, while the third estimate was +247,000.
|Over-the-month change||Revision in over-the-month change|
|1st release||2nd release||3rd release||2nd - 1st||3rd - 2nd||3rd - 1st|
Mean absolute revision
Note: A full set of revision information is available at https://www.bls.gov/web/empsit/cesnaicsrev.htm.
In summary, data users should remember that the initial estimates of payroll employment are a preliminary look at what occurred in each month. It is the quick but lower-resolution snapshot of what went on in the job market for a particular month. Because the revised estimates are based on more complete data, they create a higher resolution picture—and occasionally the revised data produce a different picture altogether.
This Beyond the Numbers article was prepared by Thomas Nardone, Kenneth Robertson, and Julie Hatch Maxfield. Email: email@example.com. Telephone: (202) 691-6555.
Information in this article will be made available to sensory-impaired individuals upon request. Voice phone: (202) 691-5200. Federal Relay Service: 1 (800) 877-8339. This article is in the public domain and may be reproduced without permission.
Thomas Nardone, Kenneth Robertson, and Julie Hatch Maxfield, “Why are there revisions to the jobs numbers?” Beyond the Numbers: Employment & Unemployment, vol. 2, no. 17 (U.S. Bureau of Labor Statistics, July 2013), https://www.bls.gov/opub/btn/volume-2/revisions-to-jobs-numbers.htm
1 Payroll employment from the Current Employment Statistics (CES) program excludes proprietors, the unincorporated self-employed, unpaid volunteer or family employees, farm employees, and domestic employees. Salaried officers of corporations are included. Government employment covers only civilian employees; military personnel are excluded. Employees of the Central Intelligence Agency, the National Security Agency, the National Imagery and Mapping Agency, and the Defense Intelligence Agency also are excluded. CES is a monthly survey of about 145,000 businesses and government agencies, representing approximately 557,000 individual worksites. For more information on the program’s concepts and methodology, see “Technical notes to establishment survey data,”. To access CES data, see “Current Employment Statistics–CES (national),”.
2 Second preliminary estimates for a reference month are published the month following the initial release, and final sample-based estimates are published 2 months after the initial release.
Publish Date: Wednesday, July 10, 2013