August payrolls rise 162,000 and July's decline is revised away

Total nonfarm payroll employment rose by 162,000 in August 2026 and the unemployment rate was unchanged at 4.1 percent, the Bureau of Labor Statistics reported on 4 September.¹ The increase was “higher than the average monthly gain of 31,000 over the prior 12 months,” and above the 55,000 consensus recorded before the release by Forex Factory’s economic calendar, an aggregator whose consensus figures differ from other providers’.¹² The same release revised the two preceding months upward: July, first reported as a fall of 23,000, became a gain of 21,000, and June rose from 20,000 to 31,000. “With these revisions, employment in June and July combined is 55,000 higher than previously reported.”¹ Employment rose in food services and drinking places and in local government education; the information industry lost jobs.¹ Average hourly earnings rose 0.3 percent on the month to $37.75 and 3.1 percent over the year, and the labour force participation rate edged up to 61.6 percent.¹

1. The upward revisions do not undo the earlier markdown. This corpus reported the July FOMC minutes against the 7 August release, setting the Committee’s description of payroll gains as having “strengthened this year” against three figures from that release: a fall of 23,000 in July, a combined markdown of 103,000 across May and June, and a prior-twelve-month average of 34,000 (see July FOMC minutes record the Chairman’s case for a six-meeting calendar).³⁴ The 4 September release reverses the first: July is now a gain of 21,000.¹ The second stands substantially intact. May is unrevised at 63,000 against an original 129,000 and June is 31,000 against an original 57,000, leaving those two months 92,000 below their first prints, against the 103,000 first reported.¹⁴ The third is not a like-for-like comparison: the 31,000 average published with the August figure and the 34,000 published with the July figure cover twelve-month windows offset by one month, and every revision between the two releases was upward, so the lower average reflects the window moving rather than the record being marked down.¹⁴ What the revisions leave standing is a markdown of 92,000 across May and June and a twelve-month average of hiring near 30,000.¹⁴

2. The revisions were smaller than either forecast error. August’s release came in 107,000 above the recorded consensus; the July release a month earlier came in 108,000 below it.¹² The revisions themselves were 44,000 for July and 11,000 for June.¹ The preliminary benchmark revision published on 28 August put the total error in the establishment survey between March 2025 and March 2026 at −79,000, or −0.1 percent, against a ten-year absolute average of 0.2 percent — below the historical norm rather than above it.⁵

3. The rebound is concentrated in two categories. Food services and drinking places added 59,000 against a prior twelve-month average of 12,000, and local government education added 42,000, “largely offsetting a decrease in the prior month.”¹ As first published, that category had fallen 50,000 in July — more than the 23,000 then reported for the month as a whole; the August release does not restate the July figure for the category.¹⁴ Manufacturing rose 16,000 and is “up by 58,000 since a recent low in December 2025.”¹ Against that, the information industry lost 23,000 jobs, “following losses that had averaged 8,000 per month over the prior 12 months,” including declines of 8,000 in computing infrastructure providers, data processing and web hosting, 7,000 in publishing and 5,000 in broadcasting and content providers.¹ Only the first of those three sits near what the July minutes discussed: the minutes address “the financing of the rapid buildout of AI-related infrastructure” and “capital spending in the AI sector,” and use neither the term “information industry” nor “information sector.”³ On employment specifically the minutes recorded that “a few participants assessed that AI-related developments appeared to have had a limited net effect on employment so far” and that “several participants observed that fears about AI leading to widespread layoffs had not materialized to date.”³

4. The measures of long-duration joblessness did not move with the headline. The number of long-term unemployed — those jobless 27 weeks or more — “changed little at 1.9 million,” and the long-term unemployed accounted for 27.0 percent of all unemployed people, against 25.5 percent in July; the release makes no over-the-month comparison of that share.¹⁴ Participation edged up to 61.6 percent but remains “down by 0.5 percentage point since January,” and the employment-population ratio at 59.1 percent “changed little over the month and since January.”¹ The number working part time for economic reasons fell by 414,000 to 4.4 million.¹ The July minutes had recorded that “a few participants noted some lingering signs of softness in the labor market, including the low job-finding rate and the persistently elevated long-term unemployment rate.”³