A preliminary benchmark check suggested payrolls were overstated by 79,000
The August 28 benchmark estimate lowered the March employment level by 79,000, a small percentage change that still matters for reading the hiring trend.
Published October 27, 2026. This retrospective examines BLS’s preliminary employment benchmark estimate, released August 28.
The Bureau of Labor Statistics said its preliminary annual benchmark comparison would reduce the estimated March 2026 payroll level by 79,000 jobs, or about 0.1 percent. The figure was preliminary. BLS said the final benchmark revision would be published in February 2027, so this August announcement did not yet rewrite the official monthly employment series.
Why a benchmark exists
Monthly payroll estimates rely on a survey of employers. BLS later compares the survey-based estimates with more complete administrative records, including unemployment insurance tax data covering nearly all jobs in the economy. The annual benchmark aligns the sample estimates with that fuller count and adjusts the level of employment.
A preliminary estimate is not a count of 79,000 people who suddenly lost jobs. It is a proposed adjustment to the statistical estimate of the number of payroll positions in the economy in March. It does not say that 79,000 layoffs happened in August or that the current month’s hiring was lower by that amount.
Small percentages can change the narrative
A 0.1 percent adjustment may sound tiny. But when monthly job gains are modest, a revision of tens of thousands can alter the average pace of hiring over several months. It can change whether the labor market looks like it is steadily expanding or barely holding level.
Readers should compare the adjustment with the size and timing of the original estimates. It does not invalidate every jobs report. The revisions themselves are part of the measurement system, and they provide a reason to avoid treating one monthly headline as perfectly precise.
What this did not tell us
The benchmark did not identify which employers or workers accounted for the difference in the public summary. It also did not tell whether the revision would be identical when finalized. Nor did it directly measure the unemployment rate, which comes from a separate household survey.
WallPost reports the August figure as preliminary and preserves its status. The final update is scheduled for February 2027. This matters because the year’s job reports were being read amid claims of a steady labor market and signs of slower hiring. Better measurement can clarify the trend, but it cannot replace the lived experience of workers navigating openings, wages and layoffs.
This kind of revision is why WallPost dates and labels its figures. An article published with the initial estimate should not quietly substitute a later figure without explanation. Readers deserve to know what the public record said at the time and how later evidence changed it. The benchmark process can improve accuracy while also reminding us that economic statistics are estimates, not a live census of every job.
For readers, the practical point is not to discard the employment survey. It is to keep the distinction between level and monthly change clear. A benchmark can lower the estimated total number of payroll jobs while the economy still adds positions from one month to the next. The preliminary -79,000 adjustment concerned the March level. It did not mean the economy lost 79,000 jobs in one month.
Sources
Spotted an error? Tell the editor. Corrections are logged on our corrections page.
