U.S. Payrolls Fall 23,000 in July 2026 as Unemployment Slips to 4.1%

Total nonfarm payroll employment fell by 23,000 in July 2026 and the unemployment rate edged down to 4.1 percent, the Bureau of Labor Statistics reported. Neither move is large enough to stand on its own: a payroll change this small sits well inside the establishment survey's sampling noise, which means July is statistically indistinguishable from no change at all in employment. The revisions are the part of this report that carries information.

Payrolls fell 23,000 in July, and with employment in May and June combined now 103,000 lower than previously reported, the last three months of hiring look materially thinner than they did a month ago. For context, the prior twelve months delivered an average monthly gain of 34,000 — itself a slow pace, and one this report pulls down further.

What the Revisions Did to the Prior Two Months

BLS revised May down by 66,000, from a gain of 129,000 to a gain of 63,000, and June down by 37,000, from a gain of 57,000 to a gain of 20,000. Revisions of this size are well past the threshold at which they move markets, and both went the same direction.

This is the pattern worth understanding rather than dismissing. The Current Employment Statistics model leans on historical trends to estimate business formations and closures, which makes it a reliable estimator in steady conditions and a lagging one at turning points. When the economy changes velocity, the initial prints tend to overstate the old trend and the revisions do the correcting. Three consecutive downward-revised months is not evidence of manipulation; it is what a decelerating labor market looks like inside a trend-based estimator.

Mark one date — BLS publishes the preliminary estimate of the annual benchmark revision on August 28, 2026, alongside first-quarter QCEW data. Benchmark revisions restate the establishment survey against near-universal unemployment insurance tax records, and in a period of soft prints the preliminary estimate is the most informative number of the month.

Where the Jobs Went: Sector Composition

The July decline was not broad. BLS describes employment as little changed over the month across mining, construction, manufacturing, wholesale trade, transportation and warehousing, information, professional and business services, social assistance, leisure and hospitality, and other services. The movement was concentrated in a handful of places.

Financial activities continued to trend down, falling 14,000 in July, with losses in credit intermediation and in insurance carriers. That supersector is now down by 121,000 since a recent peak in May 2025 — a sustained contraction rather than a single soft month.

The composition matters for reading the cycle. The cyclically informative supersectors — manufacturing, construction, professional and business services — did not decline; they simply did not grow. A month in which the cyclical core is flat, the one clearly contracting supersector is financial activities, and the headline is dragged negative by public education and retail is not a recession print. It is a print with no engine.

Industry Detail Below the Supersectors

Three sub-sector moves shaped the total:

• Retail trade lost 19,000 jobs, with the decline concentrated in general merchandise retailers and in gasoline stations and fuel dealers, partly offset by gains among sporting goods, hobby, musical instrument, book, and miscellaneous retailers. Retail employment had shown little net change over the prior twelve months.
• Local government payrolls fell, with the decline concentrated in education after that category showed little net change over the prior twelve months.
• Health care added 22,000, continuing its upward trend but below its average monthly gain of 36,000 over the prior twelve months, with ambulatory health care services leading the increase.

Health care decelerating is the quiet story here. It has been the most dependable source of payroll growth in this expansion, and a month at well under its trailing average removes the cushion that has been offsetting weakness elsewhere.

Wage Growth Keeps Cooling

Average hourly earnings for all private employees were $37.62 in July, up 2 cents on the month and 3.2 percent over the year. Production and nonsupervisory earnings were $32.40, up 4 cents.

At 3.2 percent, annual wage growth sits at a pace broadly consistent with the Federal Reserve's inflation target once ordinary productivity growth is accounted for, and the year-over-year rate has been decelerating. This is not a wage-driven inflation problem, and it has not been one for some time.

Hours were flat across the board. The average workweek held at 34.3 hours, manufacturing held at 40.4 hours with overtime down 0.1 hour to 3.1 hours, and the production and nonsupervisory workweek remained at 33.8 hours. Employers trimming hours before headcount is a standard early-weakening signal; that is not yet visible here.

The Household Survey Disagrees, and How It Disagrees Matters

The unemployment rate fell to 4.1 percent from 4.2 percent, with 6.9 million people unemployed. Over the year the rate is down 0.2 percentage points. Read alone, that is a tightening labor market.

It did not come from more people working. The labor force participation rate fell to 61.4 percent, down 0.1 percentage points on the month and 0.8 percentage points over the year, and the employment-population ratio was 58.9 percent. Since January, participation has declined 0.7 percentage points and the employment-population ratio 0.5 percentage points. An unemployment rate that falls while participation falls is arithmetic, not improvement: people leaving the labor force stop being counted as unemployed.

Two details cut against the benign reading. The number of people on temporary layoff increased by 153,000 to 921,000 — temporary layoffs are among the earliest household-survey signals of deteriorating demand. And while the number jobless less than 5 weeks edged down to 2.0 million and is down 344,000 over the year, the long-term unemployed held at 1.8 million and accounted for 25.5 percent of all unemployed people. Short-duration unemployment is falling; long-duration unemployment is not.

Among worker groups, unemployment rates declined for teenagers, at 12.1 percent, and for people who are Hispanic, at 4.6 percent. Rates showed little or no change for adult men (3.9 percent), adult women (3.7 percent), and people who are White (3.6 percent), Black (6.3 percent), or Asian (4.0 percent).

What to Watch

The August report publishes September 4, 2026, but the more consequential date is August 28, when the preliminary benchmark revision lands. If the benchmark estimate points down, the last year of payroll growth was slower than the monthly prints have shown, and a 34,000 average monthly gain becomes something thinner still.

In the September report itself, the number to isolate is participation. If the unemployment rate falls again while participation falls again, the rate is measuring exit rather than hiring, and the household survey stops being a useful check on the establishment survey. If participation stabilizes and the rate holds, July's 4.1 percent is real.