Why Are U.S. Employment Statistics Revised Every Month? How July Changed from -23 to +21
The fact that U.S. employment figures changed from negative to positive within a month does not necessarily mean an error was discovered and corrected after the fact. To release data quickly, the U.S. Bureau of Labor Statistics (BLS) publishes preliminary estimates based on employer responses available at the time, then recalculates the figures for the same month over the following two months as additional survey responses arrive.
July 2026 nonfarm payroll employment was initially reported as a decline of 23, but it was revised to an increase of 21 in the August employment report released on September 4. This illustrates why revisions to the previous two months deserve as much attention as the latest month's headline figure.
How Did July Employment Change from -23 to +21?
The July figure was revised upward by 44.
Initial release: -23
Revised figure on September 4: +21
Size of revision: +44
In the same report, June employment was also revised upward by 11, from the previously reported +2 to +31. Together, the revisions to June and July showed that employment was 55 higher than previously reported.
Because the figure even changed sign, the revision may appear especially dramatic. However, the BLS monthly employment report is a sample-based survey whose estimates are subject to revision from the outset.
Why Not Publish Completely Accurate Figures from the Start?
The key distinction is speed versus completeness.
U.S. nonfarm payroll employment is measured through the BLS establishment survey, known as Current Employment Statistics (CES). When the initial report is published, responses have not yet arrived from every establishment included in the survey.
Financial markets and policymakers, however, need timely information about current labor market conditions rather than waiting several months for more complete statistics. The BLS therefore publishes an initial estimate using the responses already available and subsequently revises it as late survey submissions are incorporated.
The BLS explains that monthly CES estimates may be revised for up to two months after their initial release as additional sample responses are incorporated.
This means that the August employment increase of +162 should not be treated as a completely final figure as of September 4. It may be revised in subsequent employment reports as additional responses become available.
What Is the Difference Between Monthly Revisions and Annual Benchmark Revisions?
Both are called revisions, but their underlying reasons and the way their significance should be assessed differ.
| Category | Monthly revisions | Annual benchmark revisions |
|---|---|---|
| Main reason | Incorporating late survey responses | Comparing estimates against more comprehensive employment records |
| Timing | During the two months following the initial release | 1 time per year |
| Primary data | CES establishment sample survey | Unemployment insurance tax records, including QCEW data |
| Significance | Refines estimates for recent months | Recalibrates previously estimated employment levels |
Annual benchmarking uses data from the Quarterly Census of Employment and Wages (QCEW), which covers a much broader range of establishments than the sample survey. QCEW is based on state unemployment insurance tax records, although those records take longer to become available.
In other words, monthly employment reports provide timely estimates, while benchmark revisions arrive later and use a much broader set of actual employment records to recalibrate earlier estimates.
How Large Was the 2026 Benchmark Revision?
The preliminary benchmark revision for March 2026, announced by the BLS on 2026-08-28, indicated that total nonfarm employment was 79 lower than previously estimated (-0.1%).
The figure 79 may look substantial in isolation, but it represents 0.1% of total nonfarm employment. According to the BLS, the average absolute annual benchmark revision over the past 10 years was 0.2% of total nonfarm employment.
However, when monthly job gains themselves are relatively small, as they have been recently, even a revision that is small relative to total employment can carry greater significance for investors when compared with monthly employment growth.
Does a Large Revision Mean the Initial Report Cannot Be Trusted?
Not necessarily.
The purpose of the initial release is to estimate current conditions as quickly as possible. Subsequent revisions improve the accuracy of those estimates as more information becomes available. The BLS also explains that employment statistics are revised to reflect economic conditions more accurately as additional data are collected.
What investors should avoid is treating a single initial figure as final.
For example, someone looking only at the initial July figure of -23 might have concluded that U.S. employment had entered a period of contraction. But when that figure was subsequently revised to +21, the starting point for interpreting labor market conditions changed.
Three Employment Figures Worth Examining Together in Every U.S. Jobs Report
Rather than focusing exclusively on the latest month's nonfarm payroll figure, it is useful to examine three components together whenever an employment report is released.
First, new jobs added during the latest month.
Nonfarm payroll employment increased by 162 in August.
Second, revisions to the previous two months.
This time, the combined figures for June and July were revised upward by 55. Notably, the July figure changed direction from a decline to an increase.
Third, the trend across several months.
The BLS also explains that examining trends over a longer period can be more useful than placing excessive emphasis on a single month's figure.
Stock markets may react quickly to the difference between forecasts and actual results immediately after a report is released. However, when assessing the economy's underlying direction, considering both revisions and broader trends can help investors avoid overreacting to fluctuations in preliminary figures.
Could the August Increase of 162 Also Change Later?
Yes. The August increase of +162 reported on September 4 is an initial estimate and is therefore subject to subsequent revision.
Just as the July figure changed from -23 to +21, the August figure could be revised upward or downward when additional survey responses arrive. The direction of any future revision cannot be determined in advance.
Therefore, when examining future employment reports, the first question should not simply be, "How many jobs were added this month?"
Looking at new jobs added this month → revisions to the previous two months → unemployment rate and labor force participation rate → wage growth makes it much easier to assess whether the labor market is genuinely strengthening or whether the initial report reflects temporary fluctuations.
In particular, assessments of the August 2026 report suggest that strong job creation has once again increased the relative importance of inflation indicators, rather than labor market conditions, in the Federal Reserve's interest rate decisions. Nevertheless, a single employment report cannot determine future policy rate decisions with certainty.
※ This article is provided for informational purposes to help readers understand market trends and does not constitute an investment recommendation concerning any particular security or asset. Readers are responsible for their own final decisions and investment outcomes.
Related Articles
U.S. August Employment Rises by 162 as July Figure Changes from Negative to Positive