The 258k Employment Revision in the BLS Jobs Report: Why More Large Revisions are Expected
Introduction
This Edgeworth Insight takes a closer look at the substantial revisions to payroll employment estimates in the BLS jobs report released on August 1st.[1] There are several important lessons from this report.
- The revision to total employment, across both the private and government sectors, was larger than any in the past two decades, other than one revision in the spring of 2020 during COVID lockdowns.[2]
- While the revision to private sector employment is also relatively large, it is consistent with the lack of precision in estimated employment changes in the establishment survey.
- Government employment revisions for educational institutions accounted for a substantial share of the revision to estimated total employment.
- The revision to local government education (in percentage terms) was more than seven times larger than the revision to private sector employment.
- The revision to state government education (in percentage terms) was more than 17 times larger than the revision to private sector employment.
- Future substantial revisions should be expected based on recent history and the fact that the establishment survey response rate has fallen below 43%.[3]
- On September 9th the projected benchmark revision to payroll employment will be announced. While this is a different type of revision, it may result in another change in payroll employment estimates of several hundred thousand employees.
The CES Survey
The payroll employment figures in the jobs report come from the Current Employment Statistics (CES) establishment survey.[4] While there are over 12 million establishments in the U.S.[5] the CES surveys about 650,000 establishments each month.
The jobs report includes preliminary payroll employment totals for the previous month and revisions are made in the next two reports because some establishments submit their survey responses late. The BLS explains that monthly revisions “result from additional reports received from businesses and government agencies since the last published estimates and from the recalculation of seasonal factors.”[6] Seasonal factors need to be recalculated when survey responses are updated because the BLS relies on dynamic moving-average seasonal factors to seasonally adjust payroll employment.
The BLS reports that the standard error of an estimated month-to-month change in total non-farm payroll employment is about 83,000 employees.[7] The total non-farm payroll employment gain of 73,000 between June and July that was reported on August 1 (and that is subject to two future revisions) has a 95% confidence interval between a decline of 89,500 jobs and an increase of 235,500 jobs (plus or minus 162,500 jobs).[8] Although the unusually large downward revisions of private sector job gains from April to May and May to June were 68,000 and 71,000 respectively, the differences between the revised and preliminary estimates of these month-to-month job changes are not statistically significant.
Revisions in the August 1 Report
An important headline from the August 1st jobs report was the downward revision of employment estimates for June 2025. The revision decreased the estimate of seasonally adjusted (SA) non-farm employment by 258,000. The previous report released on July 3rd included a preliminary estimate of SA non-farm employment for June 2025 of 159,724,000. The August 1st report estimated non-farm employment to be 159,466,000 for June.[9]
The downward revision to SA private sector payroll employment for June was substantially smaller than the non-farm employment revision which covers both the private and government sectors. Total private sector employment was revised by -139,000, or a decrease of one-tenth of one percent (10 basis points). While this revision was one of the largest in the past 20 years it was not unprecedented.[10] There was an almost identical downward revision to private sector employment for March 2023 (in the jobs report released in May 2023). Even larger downward revisions to private sector payroll employment occurred in the spring of 2020, during COVID lockdowns, and in the fall of 2008 at the beginning of the Great Recession.
The following charts show the revisions to SA employment for June (from comparing the reports released in July and August) for dozens of major industries as well as the private sector and government, overall. Figure 1 shows that the revisions in SA estimates for June were -16 basis points for total nonfarm employment, and -10 basis points for total private sector employment. The largest revisions for June estimates were for government educational institutions. For SA estimates the revisions for June were -181 basis points for state government education and -74 basis points for local government education. The largest revisions in the private sector, in percentage terms, were for the Arts, Entertainment and Recreation and the Administrative and Support Services industries. The revisions for June were -44 basis points for Arts, Entertainment and Recreation and -31 basis points for Administrative and Support Services.
Figure 2 reports SA employment revisions in terms of changes measured in thousands of jobs. It is noteworthy that while total non-farm payroll employment was revised downward by 258,000 jobs, state and local government educational employment was revised down by over 109,000 jobs.
Figure 3 shows that the revisions in “not seasonally adjusted” (NSA) estimates for June (from comparing the reports released in July and August) were -11 basis points for total non-farm employment, and -8 basis points for total private sector employment. For NSA estimates the revisions for June were -205 basis points for state government education and -12 basis points for local government education. The substantial difference between the SA and NSA revisions for local government education (-74 basis points compared to -12 basis points) indicates how dynamic estimates of seasonal factors matter for local government employment estimates.
Figure 4 reports NSA employment revisions in terms of changes measured in thousands of jobs. It is noteworthy that while total non-farm payroll employment was revised downward by 182,000 jobs, government employment was revised down by 68,000 jobs including a decline in state government education of over 48,000 jobs.
Why Payroll Employment Revisions Matter
Employment revisions in the monthly jobs report matter because the Federal Reserve and other policymakers rely on job growth data to set government policies, and private investors and entrepreneurs may also rely on high frequency changes in job growth. While a downward revision of estimated private sector employment of one-tenth of one percent may appear quite small, over the past 20 years the average monthly growth rate in private sector employment has been about one-twelfth of one percent. In other words, the revisions reported on August 1st erased more private sector jobs than are added in a typical month.
Are More Large Revisions Expected?
Looking forward, we should not be surprised if there are future downward revisions to private sector employment for at least two reasons.
Figure 5 shows that the response rate for the CES has declined from just over 60% during the period 2016 to 2019 to less than 43% over the past two years making payroll employment estimates less reliable and less accurate than in earlier years.[11] Figure 5 also shows that the initiation rate, or the rate at which new establishments are successfully brought into the CES survey, has declined from over 70% in 2015 to less than 40% over the past year.
Since 2022 three out of four revisions to private sector employment have reduced employment and one of three revisions decreased the number of private sector jobs by at least 50,000.[12] This suggests that more monthly downward revisions should be expected if CES response rates continue to be relatively low.
Another reason for future revisions is that in each year CES estimates are benchmarked to comprehensive counts of employment from the Quarterly Census of Employment and Wages (QCEW). The QCEW is based on state unemployment insurance tax records that nearly all employers are required to file and therefore constitutes a virtual census of payroll employment, making it an accurate alternative to the CES for payroll employment figures.[13] However, the QCEW is released with an approximately six-month lag. This year, the QCEW figures from March 2025 will be issued on September 9, the same day that the BLS will release the “preliminary estimate” of its annual benchmark revision to the CES survey data.[14]
Last year when the BLS published its preliminary revision due to the QCEW benchmark, it projected a reduction of 818,000 jobs. In February 2025, when these benchmark adjustments to the QCEW were officially adopted for calendar year 2024, the average reduction in SA total nonfarm employment was about 626,000 employees or a reduction of about 40 basis points. Similarly, in February 2023, when QCEW benchmark adjustments were officially made for calendar year 2022, the average increase in seasonally adjusted total nonfarm employment was about 586,000 employees or an increase of about 38 basis points.[15]
Therefore, the recent history of QCEW benchmark revisions suggests that the BLS announcement on September 9 may include projected payroll employment revisions of at least several thousand jobs (either added or subtracted).
CITATIONS
[1] https://www.bls.gov/news.release/pdf/empsit.pdf
[2] These comparisons do not consider the monthly reports released in February and March each year because these reports included both the typical monthly revisions as well as the benchmark revisions discussed below.
[3] https://www.bls.gov/news.release/pdf/empsit.pdf
[4] An establishment is generally an individual workplace, or a “single physical location” that produces goods or services.
[5] https://www.bls.gov/cew/data.htm
[6] https://www.bls.gov/news.release/pdf/empsit.pdf
[7] When a sample is surveyed instead of the entire population, there is a chance that the estimates calculated from the sample will vary from the true population values. The standard error provides a statistical measure of this variation. A smaller standard error indicates a more precise estimate.
[8] Standard errors can be used to create confidence intervals, which provide a range of values within which the true population value is likely to fall. See, https://www.bls.gov/news.release/pdf/empsit.pdf
[9] The figures in this section are obtained by comparing the estimates of June employment disclosed in the July 3rd and August 1st jobs reports.
[10] These comparisons do not consider the monthly reports released in February and March each year because these reports included both the typical monthly revisions as well as the benchmark revisions discussed below.
[11] https://www.bls.gov/osmr/response-rates/
[12] These comparisons do not consider the monthly reports released in February and March each year because these reports included both the typical monthly revisions as well as the benchmark revisions discussed below.
[13] One limitation of the CES survey is that the sampling framework makes it difficult to account for employment changes due to establishment births and deaths within a calendar year. This is exactly why the QCEW benchmark revisions are so important.
[14] https://www.bls.gov/news.release/pdf/empsit.pdf
[15] In February 2024 the calendar year benchmark revisions for 2023 were 105,000, on average, or a reduction of about 7 basis points.
Experts
Partner