Nonfarm Payrolls +162,000 vs 56K Est — and July Was Revised From -23,000 to +21,000

Fundamentals · 2026-09-04

Total nonfarm payrolls rose 162,000 in August against a consensus near 56,000, roughly triple the forecast, with unemployment unchanged at 4.1 percent; July was revised from -23,000 to +21,000 and June from +20,000 to +31,000, together 55,000 higher than previously reported; average hourly earnings +10 cents or 0.3 percent to $37.75, up 3.1 percent on the year; but the twelve-month average gain is only 31,000, and food services plus local government education contributed 101,000 of the 162,000; information -23,000, financial activities -11,000; the household survey added 683,000 to the labor force and cut involuntary part-time work by 414,000; the private diffusion index rose to 55.6.

What Is This?

• What it is: The BLS Employment Situation, released at 8:30am ET, drawn from two separate surveys. The establishment survey polls about 119,000 businesses covering 622,000 worksites — roughly 26 percent of all nonfarm payroll jobs — and produces the headline payroll figure, hours and earnings. The household survey polls about 60,000 households and produces the unemployment rate, participation and the demographic detail. They measure different universes and routinely disagree.
• Why it matters: It is the single most consequential monthly release, and this one arrives after a week in which every other labor indicator pointed down. It also lands seven days before August CPI and eleven days before an FOMC meeting at which three officials dissented in July in favor of a hike.
• How to read it: BLS publishes its own significance threshold and it is the most useful thing in the release. A monthly change of about 122,000 is statistically significant in the establishment survey; the household survey threshold is about 650,000. Most months fail that test. This one clears it. Read the revisions before the headline, and read the industry table before believing either.

Want the full explainer? Economic News Events, Explained breaks down this release and every other one we cover, in plain English.

Summary

Total nonfarm payrolls increased 162,000 in August against a consensus clustered near 56,000 — roughly triple the forecast, and the largest upside surprise of the year. The unemployment rate was unchanged at 4.1 percent. Average hourly earnings rose 10 cents, or 0.3 percent, to $37.75, up 3.1 percent over the year. On the face of it this is an emphatic repudiation of everything the data said this week.

Start with the honest part: every leading indicator we covered this week pointed the wrong way. ADP had private payrolls at +38,000 on Wednesday; the establishment survey put private payrolls at +127,000, an 89,000 gap on the same concept in the same month. JOLTS implied net employment of -18,000 for July. ISM manufacturing employment netted +0.3, ISM services employment printed 47.8 and contracting, and the claims four-week average had risen four consecutive weeks. Every one of those was consistent with a downside surprise. The print came in at three times consensus. That is worth stating plainly rather than explaining away.

The revisions made it worse for the bears. July was revised from -23,000 to +21,000 — a 44,000 upward revision that flipped the sign on what had been the month's headline job loss. June went from +20,000 to +31,000. Together the two months are 55,000 higher than previously reported. A good deal of the "labor market is cracking" case rested on a July figure that no longer exists.

Now the qualification, and it is substantial. BLS states in its own text that 162,000 is "higher than the average monthly gain of 31,000 over the prior 12 months." Thirty-one thousand. The three-month average is 71,000. So August did not reveal a strong labor market; it revealed one very strong month inside a trend running at about a fifth of that pace. One print at five times the trailing average is a data point, not a trend.

The composition narrows it further. Food services and drinking places added 59,000, well above their 12-month average of 12,000. Local government education added 42,000, which BLS describes as "largely offsetting a decrease in the prior month" and notes has shown "little net change since January 2025." Those two lines are 101,000 of the 162,000 — 62 percent of the entire print. Strip them out and payrolls grew about 61,000, which is almost exactly what consensus expected. The beat is two categories, one of them a seasonal give-back in school districts.

That said, the breadth measures argue against dismissing it. The private diffusion index — the share of 250 industries adding workers — rose to 55.6 from 52.8, and the manufacturing diffusion index jumped to 61.1 from 52.1. Employment gains were spread across more industries than in July even though the magnitude was concentrated. Both things are true and they sit together uncomfortably.

The losers matter as much as the winners. Information shed 23,000 jobs, against losses averaging 8,000 a month over the prior year, and the internal detail is the most interesting line in the release: losses of 8,000 in computing infrastructure providers, data processing, web hosting and related services, 7,000 in publishing, and 5,000 in broadcasting and content. All week this feed has documented an AI capital-spending boom — capital goods imports up 46.9 percent year over year, GPUs added to ISM's short-supply list, electronic components scarce for 18 months. Here is the other side of it: the sector that actually runs the computing infrastructure is cutting staff, and cutting faster than its trailing average. That is what substituting capital for labor looks like in a payroll survey.

Financial activities lost 11,000 for a second consecutive month, which is now confirmed from three directions — JOLTS showed finance effectively frozen with a 1.4 percent hires rate against a 0.2 percent layoff rate, and ISM Services had Finance & Insurance contracting yesterday.

Two industries contradict this week's other surveys outright. Manufacturing added 16,000 and is up 58,000 since a December 2025 low, with machinery and fabricated metal products each adding 6,000 — against ADP's manufacturing -17,000 on Wednesday. And health care added only 13,000, explicitly "at a slower pace than the average monthly gain over the prior 12 months (+32,000)" — against ADP's education and health services +45,000, which was more than ADP's entire national total. On the two sectors where ADP was most emphatic, BLS says the opposite.

The household survey is where the report is unambiguously strong. The labor force rose 683,000, employment rose 569,000, and the number of people not in the labor force fell 551,000. Participation edged up to 61.6 percent, though it remains 0.5 points below January. People working part time for economic reasons fell 414,000 to 4.4 million, with those citing slack work down 204,000 and those who could only find part-time work down 165,000. And job leavers rose 121,000 — people quitting voluntarily, which is a confidence signal rather than a distress one. A labor market pulling half a million people back off the sidelines is not one that is deteriorating.

Two soft spots survive. Long-term unemployment rose 159,000 to 1.93 million, now 27.0 percent of all unemployed, and teenage unemployment jumped 2.0 points to 14.1 percent. So the labor market is absorbing returning workers while the people already stuck are staying stuck.

For the Fed the read is straightforward and it is hawkish. Warsh told Jackson Hole a week ago that labor markets were "consistent with full employment" and that low monthly job gains were arithmetic rather than distress. This print vindicates that view harder than he could have hoped, and it removes the employment mandate as an argument for anything. The one thing it does not do is give the hawks a wage story: average hourly earnings at 3.1 percent year over year sit well below headline PCE at 3.7 percent, so real pay is still falling. That is the same message as yesterday's productivity release, where unit labor costs rose 1.4 percent over four quarters and the labor share hit a 79-year low. Whatever is driving inflation, this report confirms again that it is not the price of labor.

Aggregate weekly payrolls — the product of employment, hours and earnings, and the best single proxy for labor income — rose 0.7 percent on the month, with aggregate hours up 0.3 percent and the average workweek up a tenth to 34.4 hours. That is a genuine income tailwind into the September quarter, and it sits awkwardly beside the soft consumer spending and record-low consumer sentiment this feed has documented over the past fortnight.

The Internals

The headline and its revisions:

Measure · August 2026 · July 2026 · June 2026 · Consensus

Total nonfarm payrolls · +162,000 · +21,000, revised from -23,000 · +31,000, revised from +20,000 · About +56,000

Total private payrolls · +127,000 · +71,000 · +26,000 · n/a

Government payrolls · +35,000 · -50,000 · +5,000 · n/a

Unemployment rate · 4.1 percent · 4.1 percent · 4.2 percent · 4.1 percent

Average hourly earnings, monthly · +0.3 percent · n/a · n/a · About +0.3 percent

Average hourly earnings, annual · +3.1 percent · n/a · n/a · n/a

June and July combined were revised 55,000 higher. July's revision of +44,000 flipped a reported job loss into a gain.

Context that reframes the headline:

Measure · Reading

August payroll gain · +162,000

Average monthly gain, prior 12 months · +31,000

Three-month average, total nonfarm · +71,000

Three-month average, total private · +75,000

BLS threshold for a statistically significant monthly change · About 122,000

Does this print clear that threshold · Yes

Employment change by industry, over the month, in thousands:

Industry · Aug 2026 · Jul 2026 · Jun 2026 · Aug 2025

Total nonfarm · 162 · 21 · 31 · -70

Total private · 127 · 71 · 26 · -20

Government · 35 · -50 · 5 · -50

Goods-producing · 41 · 29 · 14 · -35

Construction · 22 · 18 · 3 · -22

Manufacturing · 16 · 14 · 13 · -10

Durable goods · 15 · 24 · 13 · -15

Motor vehicles and parts · -4.5 · 10.7 · -0.4 · -12.2

Mining and logging · 3 · -3 · -2 · -3

Private service-providing · 86 · 42 · 12 · 15

Leisure and hospitality · 62 · -21 · -54 · 22

Private education and health services · 29 · 12 · 50 · 23

Health care and social assistance · 28.4 · 13.1 · 38.6 · 22.8

Professional and business services · 10 · 15 · 36 · -28

Temporary help services · 6.8 · 5.2 · 16.9 · -10.9

Wholesale trade · 7.8 · 5.3 · 4.6 · -8.6

Transportation and warehousing · 5.0 · 13.8 · -10.3 · -1.6

Utilities · 2.5 · 1.4 · 0.2 · -1.2

Retail trade · 1.4 · 13.2 · 9.6 · 10.1

Other services · 3 · 8 · -7 · 19

Financial activities · -11 · -11 · 2 · -19

Information · -23 · 5 · -19 · -1

Hours and earnings:

Measure · August 2026 · July 2026 · August 2025

Average hourly earnings, all employees · $37.75 · $37.65 · $36.62

Average hourly earnings, production and nonsupervisory · $32.53 · Up 11 cents · n/a

Average weekly hours · 34.4 · 34.3 · 34.2

Average weekly earnings · $1,298.60 · $1,291.40 · $1,252.40

Manufacturing workweek · 40.5 hours · 40.4 hours · n/a

Manufacturing overtime · 3.1 hours · 3.1 hours · n/a

Index of aggregate weekly hours, monthly change · +0.3 percent · +0.1 percent · -0.1 percent

Index of aggregate weekly payrolls, monthly change · +0.7 percent · +0.2 percent · +0.4 percent

Diffusion indexes, the share of industries adding workers, where 50 is an even balance:

Index · Aug 2026 · Jul 2026 · Jun 2026 · Aug 2025

Total private, 250 industries · 55.6 · 52.8 · 53.6 · 45.4

Manufacturing, 72 industries · 61.1 · 52.1 · 57.6 · 41.7

Household survey detail, seasonally adjusted, in thousands:

Measure · Aug 2026 · Jul 2026 · Change · Aug 2025

Civilian labor force · 169,777 · 169,094 · +683 · 170,750

Employed · 162,746 · 162,177 · +569 · 163,370

Unemployed · 7,031 · 6,916 · +115 · 7,380

Not in labor force · 105,638 · 106,189 · -551 · 103,251

Participation rate · 61.6 percent · 61.4 percent · +0.2 · 62.3 percent

Employment-population ratio · 59.1 percent · 58.9 percent · +0.2 · 59.6 percent

Part time for economic reasons · 4,390 · 4,804 · -414 · 4,755

Slack work or business conditions · 2,815 · 3,019 · -204 · 3,069

Could only find part-time work · 1,263 · 1,428 · -165 · 1,332

Unemployed 27 weeks and over · 1,930 · 1,771 · +159 · 1,924

Job leavers · 914 · 793 · +121 · 785

Job losers and completed temporary jobs · 3,245 · 3,309 · -64 · 3,447

Marginally attached · 1,704 · 1,806 · -102 · 1,836

Discouraged workers · 441 · 476 · -35 · 519

Unemployment rates by group:

Group · Aug 2026 · Jul 2026 · Change

Total, 16 and over · 4.1 · 4.1 · 0.0

Adult men · 4.0 · 3.9 · +0.1

Adult women · 3.5 · 3.7 · -0.2

Teenagers · 14.1 · 12.1 · +2.0

White · 3.7 · 3.6 · +0.1

Black or African American · 6.0 · 6.3 · -0.3

Asian · 3.2 · 4.0 · -0.8

Hispanic or Latino · 4.8 · 4.6 · +0.2

Less than a high school diploma · 4.7 · 5.4 · -0.7

High school graduates, no college · 4.4 · 4.0 · +0.4

Bachelor's degree and higher · 2.7 · 2.7 · 0.0

Where The Jobs Came From

Two industries produced almost two-thirds of the headline:

Component · Contribution · Share of the 162,000

Food services and drinking places · +59,000 · 36 percent

Local government education · +42,000 · 26 percent

Those two combined · +101,000 · 62 percent

Everything else · +61,000 · 38 percent

BLS notes food services and drinking places averaged +12,000 a month over the prior year, so August ran nearly five times that pace. Local government education is described as "largely offsetting a decrease in the prior month," with "little net change since January 2025" — a timing artifact rather than new hiring. Strip both and payrolls grew about 61,000, close to what forecasters expected.

Where jobs were lost, and why it matters:

Industry · Change · Detail from the release

Information · -23,000 · Against losses averaging 8,000 a month over the prior 12 months

Computing infrastructure, data processing, web hosting · -8,000 · The sector that operates AI infrastructure, shedding staff

Publishing industries · -7,000 · Continued contraction

Broadcasting and content providers · -5,000 · Continued contraction

Financial activities · -11,000 · A second consecutive monthly loss

Motor vehicles and parts · -4,500 · Reversing a 10,700 gain in July

The information detail is the line worth sitting with. This feed has spent the fortnight documenting an AI capital-spending boom from every angle — capital goods imports up 46.9 percent year over year while domestic core capital goods orders managed 0.2 percent, GPUs and memory added to short-supply lists, electronic components scarce for 18 consecutive months, and Warsh putting equipment and intangibles investment near 9 percent with over half AI-related. The payroll data now supplies the other half of that trade. Computing infrastructure providers, data processing and web hosting shed 8,000 jobs in a single month. The capital is being bought; the headcount that would once have accompanied it is not.

Where this report contradicts the week's other surveys:

Sector · This report · Other source · Source

Manufacturing · +16,000, up 58,000 since December 2025 · -17,000 · ADP, Wednesday

Health care · +13,000, below a +32,000 12-month average · Education and health services +45,000 · ADP, Wednesday

Total private · +127,000 · +38,000 · ADP, Wednesday

Services employment · Private services +86,000 · Employment index 47.8, contracting · ISM Services, Thursday

Net employment change · +162,000 · Implied -18,000 for July · JOLTS, Tuesday

Five contradictions in one table. ADP's private figure missed the establishment survey's by 89,000 on the same concept in the same month, and got the sign wrong on manufacturing. This is a useful reminder of how loose the ADP-to-BLS relationship actually is, and it is worth carrying into next month.

Where this report agrees with the week:

Finding · This report · Corroboration

Financial sector frozen or shrinking · -11,000, second straight loss · JOLTS finance hires rate 1.4 percent, layoff rate 0.2 percent; ISM Finance & Insurance contracting

Wages are not driving inflation · AHE +3.1 percent against 3.7 percent PCE · Unit labor costs +1.4 percent over four quarters; ADP base pay 3.2 percent

Real pay is falling · +3.1 percent nominal, below inflation · Real hourly compensation -3.3 percent in Q2, labor share at a 1947-series low

Leisure and hospitality strength · +62,000 · ISM Services had Accommodation & Food Services among the fastest-growing industries

That last row is the one genuine cross-survey confirmation of the headline. ISM Services named Accommodation & Food Services among its five fastest-growing industries yesterday, and food services and drinking places produced 59,000 payroll jobs today. On the single largest contributor to the beat, two independent surveys agree.

Impact on USD

• Bullish — a print at roughly three times consensus with upward revisions removes the employment mandate as an argument for anything dovish.
• The change clears BLS's own 122,000 significance threshold, so this is a real signal rather than the statistical noise most monthly prints amount to.
• July's revision from -23,000 to +21,000 retires the single data point the softening narrative leaned on hardest.
• Warsh said a week ago that labor markets were "consistent with full employment." This vindicates that and hands the three July dissenters a much easier argument into September 15-16.
• The offset is wages: average hourly earnings at 3.1 percent year over year sit below 3.7 percent headline PCE, so there is still no wage-push inflation case here.

Impact on US Indices (ES / NQ / YM)

• Mixed, lean bearish — good news is bad news when the Fed has explicitly deprioritized employment and three voters already want a hike.
• Aggregate weekly payrolls rose 0.7 percent on the month, which is a genuine income tailwind for consumption and the clearest bullish read in the report.
• Technology takes the specific hit: information shed 23,000 jobs, including 8,000 in computing infrastructure, data processing and web hosting.
• Restaurants and leisure are the winners, with food services adding 59,000 — five times their trailing average — and ISM corroborating that sector's strength yesterday.
• Position for the composition rather than the headline: strip food services and local government education and the print was 61,000, right at consensus.

Impact on Gold

• Bearish — a hawkish labor surprise lifts real yields, and there is no growth scare left in this data to support the metal.
• Unemployment steady at 4.1 percent with 683,000 people rejoining the labor force is the opposite of the distress case gold has been trading.
• Offsetting: wages at 3.1 percent against 3.7 percent inflation mean real pay is still falling, which keeps the distributional stress narrative intact.
• Conditional: August CPI on September 11 now carries almost all the remaining weight, since employment has been settled for this cycle.

What To Watch

• August CPI — Friday, September 11, 8:30am ET. With employment resolved, inflation is the only variable left before the FOMC. ISM services prices at a four-year high argue one way, S&P Global's easing-cost read the other.
• FOMC decision and dot plot — Tuesday and Wednesday, September 15-16. Three officials dissented for a hike in July. This report removes their biggest obstacle.
• Whether 162,000 survives revision. July was revised 44,000 higher and June 11,000 higher this month; the same machinery can run the other way.
• The 12-month average, not the headline. At 31,000 a month the trend is weak. Two more prints near 160,000 would change that; one does not.
• Information sector payrolls. Down 23,000 against an 8,000 monthly average, with computing infrastructure and data processing leading. If the AI capex boom keeps coinciding with information-sector job losses, that is the story of 2027.
• Next Employment Situation — Friday, October 2, 8:30am ET.

TLDR

The Employment Situation, August 2026 (released September 4):

• Total nonfarm payrolls: +162,000 (vs about +56,000 est) — roughly triple the forecast and the year's largest upside surprise
• Unemployment rate: 4.1 percent, unchanged; average hourly earnings +0.3 percent to $37.75, up 3.1 percent on the year
• July was revised from -23,000 to +21,000 and June from +20,000 to +31,000 — together 55,000 higher than previously reported
• But the average monthly gain over the prior 12 months is just 31,000, and the three-month average is 71,000
• Food services and drinking places +59,000 and local government education +42,000 are 101,000 of the 162,000 — 62 percent. Strip both and payrolls grew about 61,000, right at consensus
• Local government education is described by BLS as "largely offsetting a decrease in the prior month," with little net change since January 2025
• Information -23,000, against an 8,000 monthly average loss, including -8,000 in computing infrastructure, data processing and web hosting; financial activities -11,000 for a second month
• Manufacturing +16,000 and health care +13,000 — both directly contradicting ADP, which had manufacturing at -17,000 and education and health services at +45,000
• Private payrolls +127,000 against ADP's +38,000: an 89,000 gap on the same concept in the same month
• Household survey strong: labor force +683,000, not in labor force -551,000, participation up to 61.6 percent, involuntary part-time work -414,000, job leavers +121,000
• Soft spots: long-term unemployed +159,000 to 1.93 million, 27.0 percent of all unemployed; teenage unemployment +2.0 points to 14.1 percent
• Breadth improved: private diffusion index 55.6 from 52.8, manufacturing diffusion 61.1 from 52.1
• Aggregate weekly payrolls +0.7 percent on the month, the best labor-income reading in a year

Every leading indicator this week pointed down and every one of them was wrong. ADP said +38,000 private; the establishment survey said +127,000. JOLTS implied net job losses. ISM services employment was contracting. The print came in at roughly three times consensus, it clears BLS's own 122,000 significance threshold, and July's reported 23,000 job loss was revised into a 21,000 gain. Two qualifications keep it honest: the twelve-month average is still only 31,000 a month, so this is one very strong month inside a weak trend, and food services plus local government education account for 62 percent of the beat — strip them and payrolls grew 61,000, right at consensus. For the Fed it settles the employment question in the hawks' favor a week after Warsh called labor markets consistent with full employment, while giving them nothing on wages, which at 3.1 percent still trail 3.7 percent inflation. CPI on September 11 now carries the entire remaining weight, with the FOMC on September 15-16.

_For informational purposes only. Not investment advice._