Answer first
Summary
The July 2026 jobs report shows US nonfarm payroll employment fell by 23,000 and the unemployment rate at 4.1%, as the Bureau of Labor Statistics reported on 7 August 2026. Revisions cut May and June by a combined 103,000. Losses were concentrated in local government education, down 50,000, and retail trade, down 19,000, while health care continued to add jobs at a slowing pace. Average hourly earnings rose 3.2% over the year. The report describes a labour market that is close to flat rather than one that is contracting sharply, and its most useful lines are the revisions, the participation rate, and the split between temporary and permanent job losers.
The headline, and why it is only a headline
The Employment Situation release for July 2026 was published on 7 August 2026 under number USDL-26-1291. Its summary line is short: both nonfarm payroll employment, down 23,000, and the unemployment rate, at 4.1 percent, changed little in July.
That phrasing is doing careful work. In this release, changed little is a statistical statement, not a rhetorical one. Monthly payroll estimates carry a confidence interval of roughly plus or minus 136,000, so a 23,000 decline is not distinguishable from no change at all. The same applies to the unemployment rate, where a change of about 0.2 percentage point is the usual threshold for significance.
Two surveys produce these numbers, and they are separate instruments. The household survey samples households and produces the unemployment rate, participation, and demographic detail. The establishment survey samples employers and produces payrolls, hours, and earnings. They disagree regularly, for structural reasons, and that is not a defect in either one.
What the establishment survey reported
Payrolls fell by 23,000 against an average monthly gain of 34,000 over the prior twelve months. The industry detail matters more than the total, because the total is a net figure hiding movement in both directions.
| Sector | July change | Context in the release |
|---|---|---|
| Local government education | Down 50,000 | Little net change over the prior 12 months |
| Retail trade | Down 19,000 | Little net change over the prior 12 months |
| Financial activities | Down 14,000 | Down 121,000 since a peak in May 2025 |
| Health care | Up 22,000 | Below its 12-month average gain of 36,000 |
Within retail, general merchandise retailers including warehouse clubs and supercenters lost 21,000 and gasoline stations and fuel dealers lost 5,000, while sporting goods, hobby, musical instrument, book and miscellaneous retailers added 10,000. Within financial activities, credit intermediation lost 9,000 and insurance carriers lost 7,000.
Health care added 22,000, with ambulatory health care services contributing 18,000. The release is explicit that this is a continued upward trend but at a slower pace than the 36,000 monthly average of the prior year. Health care has been the main engine of payroll growth, and the engine is running slower.
Everything else was described as showing little change: mining, construction, manufacturing, wholesale trade, transportation and warehousing, information, professional and business services, social assistance, leisure and hospitality, and other services. That is a long list of no movement, and it is arguably the most important sentence in the establishment section.
The revisions are the real news
May was revised down by 66,000, from 129,000 to 63,000. June was revised down by 37,000, from 57,000 to 20,000. Together, employment in May and June is 103,000 lower than previously reported.
This deserves more attention than it usually receives. Two months that were reported as modest growth are now close to flat. A reader who followed the monthly headlines through the summer would have absorbed a materially more positive picture than the current data supports.
The release explains why revisions happen: additional reports arrive from businesses and government agencies after the first estimate is published, and seasonal factors are recalculated. This is the survey working as designed, not an error being corrected. The first print of any month is a partial count, and it is normal for later prints to move it.
The practical lesson is that the most recent month in any jobs report is the least reliable figure in it, and it is also the one that generates the headlines. The prior two months, having been revised, are firmer ground.
What the household survey reported
The unemployment rate was 4.1 percent, with 6.9 million people unemployed. Both were little changed over the month and over the year.
By group, the rates were 3.9 percent for adult men, 3.7 percent for adult women, 12.1 percent for teenagers, 3.6 percent for White workers, 6.3 percent for Black workers, 4.0 percent for Asian workers, and 4.6 percent for Hispanic workers. The teenage and Hispanic rates declined over the month; the others showed little or no change.
Those gaps are persistent features of the series rather than news in this particular month. The Black unemployment rate has run above the White rate throughout the published history of the series, and a single month's movement in any of these subgroup rates is heavily affected by sampling variability.
Temporary versus permanent job loss
One line in the household data changed meaningfully. The number of people on temporary layoff rose by 153,000 to 921,000. The number of permanent job losers was little changed at 1.7 million.
That distinction carries real information. Temporary layoffs are expected to reverse: the worker anticipates recall. Permanent separations are the ones that require a full search. A rise concentrated in temporary layoffs, with permanent losers flat, points toward short-term disruption rather than structural shedding.
It is worth holding that reading loosely. A temporary layoff is a survey classification based on expectation of recall, and expectations are not always met. If a share of those 921,000 do not get recalled, they will reappear later as permanent job losers.
Duration, and the shape of unemployment
The number of people unemployed less than five weeks edged down to 2.0 million and is down 344,000 over the year. The long-term unemployed, meaning those jobless 27 weeks or more, edged down to 1.8 million but changed little over the year and accounted for 25.5 percent of all unemployed people.
A quarter of the unemployed being long-term is the detail most relevant to anyone currently searching. Fewer people are entering unemployment than a year ago, but the population that has been out for six months or more has not meaningfully shrunk. That is consistent with a market where separations are low and re-entry is slow, and it lines up with what the weekly claims data has been showing.
Participation, and the people outside the count
The labour force participation rate was 61.4 percent and the employment-population ratio was 58.9 percent, both little changed on the month. Since January, however, participation has fallen by 0.7 percentage point and the employment-population ratio by 0.5 point.
That drift matters for reading the unemployment rate, because the rate only counts people who are actively looking. When participation falls, the unemployment rate can hold steady or improve without any improvement in the number of people working.
The release also counts people the headline rate excludes. 4.8 million were working part time for economic reasons, meaning they wanted full-time work but had their hours cut or could not find a full-time job. 5.9 million were outside the labour force but wanted a job. Of those, 1.8 million were marginally attached, having looked in the past year but not the past four weeks, and 476,000 were discouraged workers who believed no jobs were available for them.
None of those people appear in the 4.1 percent. That is not a criticism of the measure, which is defined precisely and consistently. It is a reason not to treat one rate as a complete description of labour market slack.
Earnings and hours
Average hourly earnings for all private nonfarm employees were 37.62 dollars, up 2 cents on the month and 3.2 percent over the year. For production and nonsupervisory employees, average hourly earnings were 32.40 dollars, up 4 cents.
The average workweek was unchanged at 34.3 hours. In manufacturing it was unchanged at 40.4 hours, with overtime down 0.1 hour to 3.1 hours.
Hours are worth watching as a leading indicator. Employers commonly cut hours and overtime before cutting headcount, so a falling workweek can precede falling payrolls. In this release the workweek did not move, and the manufacturing overtime decline was small.
On earnings, the release reports the nominal figure only. Whether 3.2 percent represents a real gain depends on inflation over the same period, which this release does not measure and this article will not infer.
Why the two surveys can disagree
A recurring source of confusion is that payrolls can fall while the unemployment rate holds steady, as happened here. That is not a contradiction, and understanding why makes every future release easier to read.
The establishment survey counts jobs, not people. Someone holding two part-time jobs is counted twice. The household survey counts people, so that same person counts once. The establishment survey also excludes the self-employed and unpaid family workers entirely, while the household survey includes them. Agricultural workers sit outside the payroll count as well.
The two also differ in precision. The establishment survey draws on a very large sample of employers, which makes its monthly change comparatively stable but subject to later revision as more reports arrive. The household survey is a smaller sample of about 60,000 households, which makes its monthly movements noisier, particularly for demographic subgroups.
So when a month shows a small payroll decline and an unchanged unemployment rate, the honest reading is that neither survey detected a meaningful change. Reaching for a story that reconciles them usually means inventing one.
What this release does not establish
- It does not establish a downturn. A 23,000 decline is within the margin of error for the payroll survey, and the release describes it as little changed.
- It does not establish a cause. The report counts jobs and does not attribute movements to policy, interest rates, technology, or any other factor.
- It is not a forecast. These are estimates for July 2026, and the release makes no claim about the months ahead.
- The July figure is provisional and will be revised, most likely twice, as more employer reports arrive.
- National totals describe no state, metropolitan area, occupation, or employer, and monthly subgroup rates carry substantial sampling error.
- The two surveys measure different things, so a divergence between the payroll change and the unemployment rate is expected and is not evidence that either is wrong.
- Nothing in it indicates how long any individual search will take.
How to read this if you are searching
The composite picture across this release is a labour market with low separations, slow hiring, and a long-term unemployed share that is not falling. That combination produces a specific experience: it is not especially likely that you lose your job, and it is unusually slow to find a new one.
Sector detail is more actionable than the total. Health care is still adding jobs, though more slowly. Local government education and retail shed jobs this month. Financial activities has been declining for over a year and is down 121,000 from its May 2025 peak, which is a trend rather than a monthly wobble.
- Read the revisions before the headline. They tell you whether the recent past was as reported, and they are the most reliably informative part of the release.
- Check your own sector's line rather than the national total, since the total is a net figure that can hide both growth and decline.
- Treat the long-term unemployed share as your planning input. At a quarter of the unemployed, it argues for a longer expected search.
- Watch the average workweek and overtime in your industry, because hours usually move before headcount does.
- Keep a specific, dated record of your work so that when an opening does appear you can move on it immediately rather than reconstructing your history.
- Do not adjust a career decision on one month of data. Wait for the revision, then decide.
The next release
The Employment Situation for August 2026 is scheduled for 8:30 a.m. Eastern on Friday 4 September 2026. It will contain a first estimate for August and revised figures for July and June.
When it arrives, the July number in this article will change. That is not a flaw in this reading; it is how the series works. The most useful habit is to treat every first print as provisional, and to give the revisions the attention that the headline usually takes.