At 8:30 a.m. on August 7, the Bureau of Labor Statistics posted a July payroll number of minus 23,000. Unemployment held at 4.1%. Forecasters had penciled in a gain of roughly 83,000 to 95,000.
Two days before that, Challenger counted 33,429 announced job cuts in July, the fewest in two years. Read the two together and July stops looking contradictory. Fewer companies announced cuts. Fewer people were on payrolls. The subtraction did not come through the door marked layoffs.
What the report actually says
The 23,000 decline lands against a 34,000 average monthly gain over the prior 12 months. July did not slow the trend. It erased a year’s average pace and took a bite out of the far side.
Local government education dropped 50,000 in a single month after a year of essentially no net change. Retail trade lost 19,000, with warehouse clubs, supercenters and other general merchandise retailers accounting for 21,000 of it and gasoline stations another 5,000. Sporting goods, hobby, book and miscellaneous retailers added 10,000 back.
Health care kept growing, but slower. It added 22,000 against a prior-12-month average of 36,000, with ambulatory services contributing 18,000. Mining, construction, manufacturing, wholesale, transportation and warehousing, information, professional and business services, social assistance, and leisure and hospitality were all described with the same phrase: little change.
The revisions are the ugly part. May came down 66,000, from a reported 129,000 to 63,000. June came down 37,000, from 57,000 to 20,000. Together the two months are 103,000 lower than what the market traded on at the time.
Average hourly earnings finished at $37.62, up 2 cents on the month and 3.2% on the year.
The finance line has been falling for over a year
Financial activities lost 14,000 jobs in July. Credit intermediation and related activities took 9,000 of that. Insurance carriers and related activities took 7,000.
The single most important sentence in the establishment survey is the one BLS appends right after: financial activities employment is down by 121,000 since a recent peak in May 2025.
That line is worth pulling out because of where it falls, not how big it is. Credit intermediation and insurance carrier operations are the two deepest back offices in finance: reconciliation, document intake, first-pass review, first-touch claims. That is precisely the work the global banking industry has spent the past year saying out loud it intends to hand over. Standard Chartered named 7,000 roles, about 15%, on a 2030 clock, and the cuts sat in back office.
BLS does not attribute. The word “AI” appears nowhere in the release; the agency publishes net industry change and stops. So “financial activities is down 121,000” and “AI took 121,000 finance jobs” are not the same sentence, and between them sit the rate cycle, merger integration, branch consolidation and a stack of other explanations.
What the report does establish on its own terms is narrower and still useful. Over 14 months, financial activities has been one of the few industries running a sustained net decline, and running it in the back office. Most sectors this year have zigzagged. Finance has had one direction.
The cohort point is what makes 121,000 worth writing down. The largest single announced back-office program in banking this year is Standard Chartered’s, at 7,000 roles on a four-year clock. Financial activities in the U.S. alone has quietly shed more than 17 times that headcount in 14 months, and not one payroll of it arrived as a press release with a number attached. The announced programs are the visible portion of a subtraction that is mostly running through vacancies.
The exit is not the layoff door
Two more July numbers only mean something next to each other.
People on temporary layoff rose by 153,000 to 921,000. Permanent job losers were little changed at 1.7 million.
Nobody is being permanently cut in greater numbers. The growth is in people stood down temporarily, which is exactly what Challenger’s two-year low would predict.
So where did the 23,000 go? The labor force participation rate is 61.4%, down 0.7 percentage point since January. The employment-population ratio is 58.9%, down 0.5 point over the same stretch.
Alongside that: 2.0 million people jobless less than five weeks, down 344,000 year over year, and 1.8 million long-term unemployed, who now make up 25.5% of everyone out of work.
Two months ago the same agency printed a very different tape. We covered the May report when it came in at 172,000 against an 80,000 consensus, with unemployment steady at 4.3% and April revised up. That print has since been revised to 63,000. The strong month was not strong. It was a first estimate that took two rounds of revision to find its floor, which is the single best argument for reading the participation rate rather than the headline.
Put those together and the shape is a market that is slow in both directions. Fewer people entering unemployment. Fewer people leaving it. Not firing, not hiring. Payrolls fall anyway, because people retire, quit or stop looking, and the employer backfills half the seat and keeps the other half.
For an individual worker the two regimes carry completely different risk. In a high-layoff market the risk is getting named. In a low-hire, low-fire market the risk is that once you are off a payroll, for any reason, the door back is narrower than the one you came out of. July moved the risk decisively toward the second kind.
By occupation, the net decline concentrated in four places: bank and insurance back-office processing, underwriting and claims support, store and stockroom roles at general merchandise and warehouse-club retailers, and non-teaching positions in local government education. What links them is not that a model took the work. It is that when one of those seats empties, it is the easiest kind of seat to leave unfilled. An unfilled seat files no notice, enters no WARN list, and appears in nobody’s layoff count. We wrote two days ago that the announcement layer and the WARN layer describe two different countries. July’s payroll print adds a third layer, the one with no announcement at all, visible once a month as a net change.
The August report lands September 4 at 8:30 a.m. The number to watch is not the unemployment rate. It is the participation rate, and whether the finance line keeps going down.
