The Catch.A permanent record of checked claims
SeriesThe job marketJuly jobs report · August 7, 2026

The jobs report · July 2026

The job market goes into reverse: payrolls fall by 23,000

Forecasters expected a gain of 83,000. Instead the economy lost jobs, the two prior months were revised down by a combined 103,000, and the unemployment rate fell for the wrong reason: fewer people working or looking.

Revised, September 4, 2026. The figures on this page are the July report as first published on August 7. In its September 4 release the agency revised July: "the change for July was revised up by 44,000, from -23,000 to +21,000," and June from 20,000 to 31,000. On today's series the twelve months to July hold four negative months, not five, and add 371,000 jobs; the seven-month revision total is minus 56,000, not minus 111,000. July gets a second revision on October 2. The August story carries the current figures.5

The storyJust the factsShow the work
-23,000payrolls in July
4.1 %unemployment rate
-103,000May + June, revised away
5 of 12months negative in the past year, as first reported

Story updated 2026-09-08.

This story rests on 3 records.

The report

What happened

The Bureau of Labor Statistics reported on August 7 that total nonfarm payroll employment fell by 23,000 in July and the unemployment rate was 4.1 percent.1

In the agency's own words, both “changed little.” The declines were concentrated: local government education lost 50,000 jobs “after showing little net change over the prior 12 months,” and retail trade lost 19,000. One caution before reading that 50,000 literally: these figures are seasonally adjusted. School employment always swings enormously in summer, and the adjustment is designed to remove that swing. The number does not mean 50,000 ordinary summer layoffs; it means school employment came in roughly 50,000 below the seasonal pattern the agency expected for July. Health care added 22,000, below its 12-month average of 36,000.1

  • In the agency's own words, both “changed little.” The declines were concentrated: local government education lost 50,000 jobs “after showing little net change over the prior 12 months,” and retail trade lost 19,000. One caution before reading that 50,000 literally: these figures are seasonally adjusted. School employment always swings enormously in summer, and the adjustment is designed to remove that swing. The number does not mean 50,000 ordinary summer layoffs; it means school employment came in roughly 50,000 below the seasonal pattern the agency expected for July. Health care added 22,000, below its 12-month average of 36,000.1
the record

Release: “Both nonfarm payroll employment (-23,000) and the unemployment rate (4.1 percent) changed little in July… Employment in local government education declined by 50,000 in July, after showing little net change over the prior 12 months. Retail trade lost 19,000 jobs in July… In July, employment in health care continued its upward trend (+22,000) but at a slower pace than the average monthly gain over the prior 12 months (+36,000).” The Employment Situation, July 2026.

May's gain was revised down by 66,000, from +129,000 to +63,000. June's was revised down by 37,000, from +57,000 to +20,000. Combined: 103,000 jobs previously reported are now gone from the record.1

  • 1The Employment Situation, July 2026The change in total nonfarm payroll employment for May was revised down by 66,000, from +129,000 to +63,000, and the change for June was revised down by 37,000, from +57,000 to +20,000.

Average hourly earnings rose 2 cents to $37.62; over the year they are up 3.2 percent.1

The composition of the loss matters: private employers added 30,000 jobs in July while government shed 53,000, so the headline decline is a public-sector story, led by the 50,000-job drop in local government education. The report landed nine days after the Federal Reserve voted 9–3 to hold interest rates, with three regional Fed presidents formally asking for a quarter-point hike, covered in our July FOMC story.1114

  • The composition of the loss matters: private employers added 30,000 jobs in July while government shed 53,000, so the headline decline is a public-sector story, led by the 50,000-job drop in local government education. The report landed nine days after the Federal Reserve voted 9–3 to hold interest rates, with three regional Fed presidents formally asking for a quarter-point hike, covered in our July FOMC story.1114

History

Where this sits

July was the fifth negative month in the past twelve. Over those twelve months the economy added a net 316,000 jobs, about 26,000 a month.6

That is a near-stall by the standard of any recent expansion, and choppy rather than steadily weak, with swings from minus 156,000 to plus 214,000 in a single spring.6

  • That is a near-stall by the standard of any recent expansion, and choppy rather than steadily weak, with swings from minus 156,000 to plus 214,000 in a single spring.6
Jobs gained and lost each monthJune 2025–July 2026 · Seasonally adjusted
-100k0k+100k+200kJune 2025: -20,000 jobsJun 25July 2025: +64,000 jobsAugust 2025: -70,000 jobsSeptember 2025: +76,000 jobsOctober 2025: -140,000 jobsNovember 2025: +41,000 jobsDecember 2025: -17,000 jobsJanuary 2026: +160,000 jobsJan 26February 2026: -156,000 jobsMarch 2026: +214,000 jobsApril 2026: +148,000 jobsMay 2026: +63,000 jobsJune 2026: +20,000 jobsJuly 2026: -23,000 jobsJul 26

July: -23,000 jobs. Inspect the source data

Bureau of Labor Statistics payroll counts, from the St. Louis Fed's FRED database as published August 31, 2026; later releases may revise the counts.

Read the values and calculation

Each bar subtracts the previous month's total jobs from that month's total. The database reports totals in thousands.

MonthJobs gained or lost
June 2025-20,000
July 2025+64,000
August 2025-70,000
September 2025+76,000
October 2025-140,000
November 2025+41,000
December 2025-17,000
January 2026+160,000
February 2026-156,000
March 2026+214,000
April 2026+148,000
May 2026+63,000
June 2026+20,000
July 2026-23,000

The unemployment rate dipped from 4.2 to 4.1 percent. The labor force shrank by 264,000 people while household employment fell by 87,000. The rate fell because the labor force contracted by three times more than employment did, not because more people found work.1

The jobs report is really two surveys: a survey of employers, which produced the payroll numbers above, and a survey of households, which produces the unemployment rate. Both moved in the same weak direction.1

  • The jobs report is really two surveys: a survey of employers, which produced the payroll numbers above, and a survey of households, which produces the unemployment rate. Both moved in the same weak direction.1

Participation among workers 25 to 54 rose to 83.4 percent in July from 83.3 in June. It remains below 84.0 in January.9

The exit is not uniform across ages. The July decline in overall participation came from outside the prime working ages; the spring slide touched prime-age workers too.19

  • The exit is not uniform across ages. The July decline in overall participation came from outside the prime working ages; the spring slide touched prime-age workers too.19

The people in the numbers

Who feels it

Average hourly earnings rose 2 cents to $37.62, up 3.2 percent over the year. Adjusted for inflation, real hourly earnings fell 0.1 percent in July and are down 0.2 percent over the year.3

Wages are rising about 3.2 percent a year while prices rise faster. Average hourly pay bought slightly less than a year earlier; average weekly pay, which includes hours worked, bought slightly more, up 0.1 percent. Neither average says what happened to the typical worker.13

  • Wages are rising about 3.2 percent a year while prices rise faster. Average hourly pay bought slightly less than a year earlier; average weekly pay, which includes hours worked, bought slightly more, up 0.1 percent. Neither average says what happened to the typical worker.13

People on temporary layoff rose by 153,000 to 921,000. The long-term unemployed (27 weeks or more) make up 25.5 percent of all unemployed people, one in four. People working part time because they could not find full-time work held at 4.8 million. People who want work but have stopped looking rose from 1.76 to 1.81 million.1

U-6, the broadest underutilization measure (adds those marginally attached workers and the involuntary part-timers to the officially unemployed), held at 7.9 percent, unchanged from June. The headline unemployment rate improved; the broadest measure did not.8

Put together, the shape of July is a hiring stall more than a burst of layoffs: employment growth has stopped and people are leaving the measured labor force, but neither the unemployment rate, the broad measure, nor the count of permanent job losers shows a sudden broad surge in joblessness. The 153,000 rise in temporary layoffs is the one caution in that picture.18

  • Put together, the shape of July is a hiring stall more than a burst of layoffs: employment growth has stopped and people are leaving the measured labor force, but neither the unemployment rate, the broad measure, nor the count of permanent job losers shows a sudden broad surge in joblessness. The 153,000 rise in temporary layoffs is the one caution in that picture.18

The coverage

What the coverage got right, and what it got wrong

Three outlets' coverage of this report is checked here: CNBC's and CBS's stories from release day, and NBC's morning-of preview, published before the number came out. Checked against the release:111213

CNBC reported minus 23,000, the 4.1 percent rate, the May and June revisions, retail's loss of 19,000, health care's below-trend 22,000, and the 2-cent earnings gain. All match the release exactly.11

Its explanation of the falling unemployment rate (fewer people working or looking) matches the release's participation numbers. CNBC's sector figures (40,000 lost in leisure and hospitality, 53,000 lost in government, 30,000 gain in private payrolls) all match the release. What separates the two documents is emphasis, not arithmetic: the release classes leisure and hospitality among industries whose employment “showed little change,” while CNBC headlined the sector's drop and attached a cause, “a possible consequence of the World Cup tournament ending,” that appears nowhere in the release. The number is the agency's; the explanation is the outlet's speculation.111

  • Its explanation of the falling unemployment rate (fewer people working or looking) matches the release's participation numbers. CNBC's sector figures (40,000 lost in leisure and hospitality, 53,000 lost in government, 30,000 gain in private payrolls) all match the release. What separates the two documents is emphasis, not arithmetic: the release classes leisure and hospitality among industries whose employment “showed little change,” while CNBC headlined the sector's drop and attached a cause, “a possible consequence of the World Cup tournament ending,” that appears nowhere in the release. The number is the agency's; the explanation is the outlet's speculation.111

CBS reported the 23,000 loss, the 50,000 and 19,000 sector losses, health care's 22,000, the combined 103,000 revision, and the participation explanation for the falling unemployment rate. All match the release. It adds a second forecast: economists polled by FactSet expected a gain of 95,000.12

NBC's morning-of preview expected hiring to pick up to 83,000 and unemployment to hold at 4.2 percent. Both were wrong by mid-morning.13

The preview also fixes the backdrop the report landed in: gasoline at $4.06 a gallon, up 36 percent since late February amid the war with Iran, and inflation at 3.5 percent against the Federal Reserve's 2 percent target.13

  • The preview also fixes the backdrop the report landed in: gasoline at $4.06 a gallon, up 36 percent since late February amid the war with Iran, and inflation at 3.5 percent against the Federal Reserve's 2 percent target.13

One claim, examined

Claim checked against the record

The claim: “The 12-month average [payroll gain is] down to just 34,000” (CNBC, August 7). The record today: recomputing from the current published series gives about 26,000 a month over the same window.6

Neither number is an error: CNBC's was correct on the data available August 7, and later revisions pulled the average lower still.1110

  • Neither number is an error: CNBC's was correct on the data available August 7, and later revisions pulled the average lower still.1110

An earlier version of this page read that aging as proof that “this year's first estimates have repeatedly overstated the job market.” The full first-print record says something more specific. Three months were revised down, three up, one is unchanged so far. The overstatement is real but concentrated: May and June were cut by 109,000 and 37,000. The spring months were revised up.10

  • An earlier version of this page read that aging as proof that “this year's first estimates have repeatedly overstated the job market.” The full first-print record says something more specific. Three months were revised down, three up, one is unchanged so far. The overstatement is real but concentrated: May and June were cut by 109,000 and 37,000. The spring months were revised up.10
2026 payroll changes: first estimate vs. today's estimate (thousands)
MonthFirst estimateCurrentRevision
January+130+160+30
February-92-156-64
March+178+214+36
April+115+148+33
May+172+63-109
June+57+20-37
July-23-230

Across the seven months: average revision −16,000; cumulative −111,000; average miss in either direction 44,000. First estimates from the ALFRED vintage archive; current estimates from the July release of August 7. The revision note at the top of the page carries the September 4 figures.

The reaction

What the markets said

Right after the Fed's July 29 decision, NBC News quoted CME FedWatch odds of a September rate hike at 59 percent. After this report, CNBC quoted them at 44 percent. Stock futures rose; Treasury yields fell.1511

The move, not the level, is the story. On July 29 the Fed had voted 9–3 to hold, with three members formally asking for a quarter-point hike. Between the Fed's decision and this report, nine days apart, a rate increase went from more likely than not to less; this report is the main event in that window, not the only one, and markets treated the move as good news.41511

  • The move, not the level, is the story. On July 29 the Fed had voted 9–3 to hold, with three members formally asking for a quarter-point hike. Between the Fed's decision and this report, nine days apart, a rate increase went from more likely than not to less; this report is the main event in that window, not the only one, and markets treated the move as good news.41511

Since this page was written

The August 28 benchmark check

Once a year the agency checks its survey-based payroll count against near-complete counts from state unemployment-insurance records. For March 2026: the survey was 79,000 jobs too high (0.1 percent), with private payrolls 178,000 too high. The average size of this correction over the past ten years is 0.2 percent.2

This correction does not change the figures on this page yet: it is applied to the published series in February 2027. Private payrolls 178,000 too high and the total only 79,000 too high means government payrolls were counted about 99,000 too low as of March. The check covers one month, March, and does not say which months the error sits in, so it neither confirms nor complicates July's school figures.2

  • This correction does not change the figures on this page yet: it is applied to the published series in February 2027. Private payrolls 178,000 too high and the total only 79,000 too high means government payrolls were counted about 99,000 too low as of March. The check covers one month, March, and does not say which months the error sits in, so it neither confirms nor complicates July's school figures.2
the record

Release: “The preliminary estimate of the CES national benchmark revision to total nonfarm employment for March 2026 was -79,000 (-0.1 percent)… The preliminary benchmark revision for total private employment was -178,000 (-0.1 percent).” Preliminary Benchmark (National), March 2026, August 28, 2026.

Open questions

What we do not know yet

  • Why local government education fell by 50,000 in one month.
  • Which months created the benchmark gap. The month-by-month distribution arrives with the final revision in February 2027.
  • Whether the labor-force exit continues. One month of a shrinking labor force alongside flat broad underutilization reads as a hiring stall; a second month would start to read as something worse.

Sources

The records

Everything this page rests on, numbered where the story cites it. We keep a dated copy of each one.

Primary documents

  1. 1The Employment Situation, July 2026 · released August 7 · every figure the story marks as record
  2. 2Preliminary Benchmark (National), March 2026 · released August 28 · the benchmark check
  3. 3Real Earnings, July 2026 · released August 12 · the inflation-adjusted wage figures (copy captured via the Internet Archive; disclosed in the source ledger)
  4. 4FOMC statement, July 29, 2026 · released July 29 · the 9 to 3 vote and the three requests for a quarter-point increase
  5. 5The Employment Situation, August 2026 (September 4 revision of July) · the revision note at the top of the page

Official data

  1. 6Total nonfarm payrolls, monthly · the chart, the five-in-twelve count, the 26,000 average
  2. 7Unemployment rate, monthly · the 4.2 to 4.1 move
  3. 8U-6 underemployment rate, monthly · the 7.9 percent broad measure
  4. 9Prime-age (25–54) participation, monthly · the 83.4 percent counterweight
  5. 10Payroll estimates as first published (ALFRED vintages) · the revisions table

Coverage checked

  1. 11CNBC, release day · the headline figures, the World Cup explanation, the 34,000 claim, the market reaction
  2. 12CBS News, release day · the FactSet forecast of 95,000
  3. 13NBC News, morning-of preview · the 83,000 expectation and the backdrop
  4. 14CNBC on long-term unemployment · context for the one-in-four figure
  5. 15NBC News · after the July 29 decision · the 59 percent September-hike odds before the report

How we check.

Figures in the source documents 4

Unemployment rate

4.1 percentU.S. Bureau of Labor Statistics
Recorded passage
Both nonfarm payroll employment (-23,000) and the unemployment rate (4.1 percent) changed little in July, the U.S. Bureau of Labor Statistics reported today.

Payroll revisions for May and June

-103,000 jobsU.S. Bureau of Labor Statistics
Recorded passage
The change in total nonfarm payroll employment for May was revised down by 66,000, from +129,000 to +63,000, and the change for June was revised down by 37,000, from +57,000 to +20,000. With these revisions, employment in May and June combined is 103,000 lower than previously reported.
Source history

Sources

Source reader

Source