The revision that shrank

Eleven times smaller

A year ago, the preliminary benchmark revision released September 9, 2025 erased 911,000 jobs from the official count, 0.6 percent of payrolls, and fed a season of argument about whether the numbers could be trusted at all. This year’s jobs report revision, released by the Bureau of Labor Statistics on August 28, cut March 2026 nonfarm payrolls by 79,000. That is 0.1 percent, roughly one eleventh of last year’s correction.

The count held. That is the headline. It is also the least useful part of the release.

The total hid a trade

Private-sector employment was revised down 178,000. Government was revised up 99,000. The two moves nearly cancel, which is how a 79,000 correction can sit on top of a much larger redistribution. The economy the headline series described had more private jobs and fewer public ones than the economy that existed in March.

A map of where the picture was wrong

Table 1 of the revision reads like a correction of geography. Transportation and warehousing had 135,100 more jobs than previously reported. Information had 87,000 more, financial activities 85,000 more, construction 62,000 more.

The other column is heavier. Retail trade had 154,600 fewer jobs than the published series showed. Wholesale trade was down 86,200, private education and health services down 96,000, professional and business services down 76,000, manufacturing down 67,000.

Someone who spent the past year applying in retail was searching a market weaker than the one the data described. Someone in warehousing was searching a stronger one. Neither of them was told.

The average is not your market

The corrected picture lands in a market that is slow by every fresh measure. Job openings held at 7.3 million in July, little changed, in the JOLTS data BLS released September 1. ADP’s National Employment Report, out September 2, counted 38,000 private-sector jobs added in August, the smallest gain since January.

Headline payroll numbers are an average across every sector at once, and an average can be right about the country while wrong about the specific market a person is searching. Last September’s question was whether the data could be trusted. This release answers a better one: not whether the count was wrong, but where. CoBlack matches people against openings sourced from employer career pages and ATS feeds, sector by sector, because a job search happens in one market, not in the average of all of them.

The 79,000 is small. The 178,000 sitting inside it is the number a job seeker was living in.