Nonfarm payrolls fell by 23,000 in July against a consensus looking for somewhere between 80,000 and 83,000. That is the number the wires led with, and it is the least interesting number in the release. The Bureau of Labor Statistics also revised May down from 129,000 to 63,000 and June down from 57,000 to 20,000, removing 103,000 jobs from a two-month window that had already been read, priced, and used to justify a policy stance. The three-month average is now approximately zero. Hiring did not stop in July. It stopped some time in the spring, and the data caught up.
That distinction matters because the hawkish case inside the Fed was built on labor market resilience. Several officials spent the past two weeks arguing openly for tightening, and the argument rested on a specific premise: the jobs market is stable enough to absorb higher rates while the Fed deals with inflation that has refused to cooperate. Core PCE ran 3.7% year over year in June. The premise, as of Friday morning, no longer holds in the form it was stated. The labor market was not stable through the second quarter. It was reported as stable and then quietly restated.
The unemployment rate improved to 4.1% from 4.2%, and the improvement is worse than a deterioration would have been. Participation fell to 61.4%, the lowest reading in more than five years. The employment level has declined by roughly 833,000 across 2026. A jobless rate falls for two reasons, and only one of them is good. This was the other one. Anyone using the headline rate as evidence of labor market health is measuring the size of the denominator and calling it demand.
Composition offers some cover and then takes it back. Local government education shed 50,000, a category that swings hard on seasonal adjustment and school-calendar timing, and one that had been broadly flat for a year before this. Leisure and hospitality lost 40,000, plausibly a hangover from the World Cup tournament ending. Retail lost 19,000, concentrated in warehouse clubs and supercenters. Financial activities gave back 14,000. Private payrolls actually rose by 30,000 while government fell 53,000, which is the version of the report a bull would prefer to quote. The problem is healthcare. It added 22,000 against a twelve-month average of 36,000. Healthcare has been carrying this report almost single-handedly for two years, and it just decelerated by nearly 40% against its own trend. Education seasonality washes out next month. A healthcare slowdown does not.
Wages went quiet at the same time. Average hourly earnings are running 3.2% year over year, and production and nonsupervisory pay rose four cents to $32.40. The average workweek held at 34.3 hours, manufacturing at 40.4 with overtime down a tenth to 3.1. There is no wage-push inflation signal anywhere in this release, which removes the last labor-side justification for a hike and leaves the inflation problem sitting entirely on the goods and energy side, where monetary policy is a slower and blunter instrument.
Markets took it as unambiguously good news, which requires examining. The S&P 500 closed at a record 7,757.64, up 0.62%. The Nasdaq added 1.3% to 26,690.62 and the Dow gained 151.83 points to 54,036.93. On the week the S&P rose 3.6%, the Nasdaq 5.2%, and the iShares Semiconductor ETF more than 7%, the strongest stretch for the major indexes since mid-April. Ten-year yields fell to 4.64%, the dollar index slipped 0.3% to 99.60 with the euro at a seven-week high of $1.1567, and gold posted one of its better sessions in weeks.
The mechanism behind that rally is not the usual one. In the standard soft-print response, equities rally because weak employment pulls forward rate cuts and cuts are stimulus. That is not what happened here. Rate futures moved from a 57% probability of a September hike to 43.9%, with the odds of a hold rising from 43.2% to 60.4%. Money markets still price tightening before the end of 2026, just not before December. Nothing in Friday’s price action reflects expected easing. It reflects the removal of an expected tightening. Equities did not get a tailwind. They got a headwind withdrawn, and they were repriced as though a tailwind had arrived.
That is thinner fuel than a 3.6% week suggests, and it leaves the index vulnerable to a specific sequence. Strong earnings have been doing the real work, with 87% of the 440 S&P 500 companies reported so far beating estimates and average year-over-year earnings growth near 25%. Cooling oil on signs of progress toward an Iran settlement has been doing the rest. Both of those are independent of Friday’s data. Strip them out and what remains is a market that rallied because the Fed might not hurt it, on a report showing the economy is producing no net jobs.
The Fed now sits in a position with no comfortable exit. Cutting to support employment feeds an inflation rate still running well above target with energy risk unresolved. Hiking into a labor market that just printed negative and has shed 833,000 employed persons this year invites the recession it spent two years avoiding. Holding is the only defensible choice, and holding is what the market just priced. That much is coherent. What is not coherent is an equity market at a record high and a rate market pricing a hike in December on the same set of facts. One of them is going to be repriced.
Three dates settle it. July CPI arrives on 12 August and remains the actual decider, because a hot print means a cooling labor market will not be enough to quiet the hike camp. August payrolls land on 4 September, immediately ahead of the FOMC. Between them, on 28 August, the BLS publishes its preliminary annual benchmark revision, which rechecks the past year of establishment data against state unemployment insurance tax records. Given that Friday’s release just moved 103,000 jobs out of a two-month window, the benchmark exercise is the one to watch. It has the capacity to restate the whole of 2026 hiring, and every position currently justified by labor market resilience is sized against numbers that have already proven revisable.
