Revisions, Retreats, and the Rewriting of the Recent Past. Also Known As: The July Jobs Report

Publicly, the July jobs report is likely going to get filed under "quiet month." Payrolls fell by 23,000. The unemployment rate ticked down to 4.1%. Wages rose 3.2% over the year. Nothing dramatic, nothing scary, nothing that moves a news cycle on a Friday in August.

And that framing is going to be wrong for a reason very few will report.

Because the most important thing that happened in the July jobs report did not happen in July. The same release that told us the economy lost 23,000 jobs last month also deleted 103,000 jobs from May and June. Not a rounding adjustment. Not a footnote. A revision that takes an entire quarter of what we were told was a recovering labor market and reduces it to static noise.

So no, this isn't a quiet month of steady resilience. It's a month of receipts coming in.

The Revisions Are the Report

Let me walk you through what happened to May, because it is genuinely remarkable and it is going to disappear from the conversation come Monday.

On June 5, the BLS reported that the economy added 172,000 jobs in May. That number more than doubled the consenus estimate of 80,000. It ran in every outlet as a blowout, a boon. It was, however briefly, the strongest month of the year.

On July 2, May was revised down to 129,000.

On August 7, today, May was revised down again, to 63,000.

That is a 63% reduction from the number that was celebrated. May did not add 172,000 jobs. May added 63,000 jobs, lower than the consensus estimate, and we spent two months believing otherwise.

June got the same treatment, if less dramatically: the +57,000 I wrote about last month is now +20,000.

Here is the ledger as it stands this morning:

Month	First reported	Now	Change
April 2026	+115,000	+148,000	+33,000
May 2026	+172,000	+63,000	-109,000
June 2026	+57,000	+20,000	-37,000
July 2026	-23,000	—	—

Now put the last three months side by side: +63,000, +20,000, -23,000. That is an average of 20,000 jobs per month across an entire quarter, in an economy of 162 million workers.

And here is the part that should genuinely unsettle you: the establishment survey's own margin of error on a one-month change is plus or minus 122,000 jobs. Which means every single one of those three months (May, June, and July) is statistically indistinguishable from zero.

Not weak. Not slowing.

Statistically indistinguishable from no job creation at all.

We celebrated a recovery, then got a flat line. We only realized it once the data caught up.

I flagged this exact dynamic in last month's report: "Keep that pattern in mind when June's +57,000 gets revised in August and September." It came in at 20,000. In May, I wrote that the World Cup hiring surge would swing back the other way once the tournament ended, and lo and behold, leisure and hospitality proceeded to shed 61,000 jobs in June and show no net change for the entire year.

I am not telling you this to take a victory lap. I'm telling you because the pattern is now predictable enough that I could call it in advance twice in a row, and it still gets reported as a surprise every single month. A number isn't a fact until it survives its revisions. Treat every first print accordingly.

The Bureau Finally Said It Out Loud

For two months I have been arguing in this space that the most ignored number in the entire report is labor force participation, and that the unemployment rate keeps "improving" because people are leaving the labor market rather than finding work in it.

This month, the Bureau of Labor Statistics put it in their own summary:

"Since January, the labor force participation rate declined by 0.7 percentage point, and the employment-population ratio decreased by 0.5 percentage point."

BLS does not editorialize. They report the month, note what changed, and move on. When they voluntarily zoom out to a seven-month trendline in the second paragraph of a household survey summary, you better believe it's because the trend has become impossible to characterize as just another line item.

Participation now sits at 61.4%. The employment-population ratio is at 58.9%. And July ran the same play as June, just on a slightly smaller scale:

• Employed Americans: down 87,000

• Unemployed Americans: down 178,000

• The Labor Force: down 264,000

• Not in the Labor Force: up 381,000, to 106.2 million

Once again, the unemployment rate fell while employment fell. Once again, the improvement is in the denominator. If July's 264,000 departures were still in the labor force and still looking, the rate would be roughly 4.2% instead of 4.1%.

I want to be straight with you about the scale here: that gap is much smaller than June's, when the same calculation produced 4.6% against a reported 4.2%. July's exodus was a fraction of June's. This is a milder month by that measure and I'm not going to inflate it.

But the direction has not changed once. Over the past twelve months, the labor force has shrunk by 1.3 million people while the population grew. The number of Americans outside the labor force entirely has grown by 2.8 million. Six million of them say they want a job right now. 1.8 million have looked in the past year. 476,000 have stopped looking because they've concluded there's nothing out there for them.

None of them are in the unemployment rate. Most of them aren't in the U-6 either, which held flat at 7.9% this month, because U-6 only reaches you if you've actively searched within the past twelve months. Walk way for thirteen and you're gone from both.

The gap between U-3 and U-6 is now 3.8 points, up from 3.6 a year ago. The margins keep widening while both headline numbers look calm.

A Correction, and What it Uncovered

Last month I wrote that June's labor force exits were not retirements; that employment among workers 55 and over had risen, while 25 to 34 year-olds lost 643,000 jobs in a single month. I called it an early-career exit and said it was a detail that mattered.

July's data says I read too much into one month, and I want to correct that here rather than let it sit.

The 25 to 34 cohort gained back 283,000 in July. Prime-age employment overall rose 273,000, while workers 55 and over fell 288,000—the precise inverse of June. That is what statistical anomalies look like in the household survey's age data, and June's number was substantially an anomaly. I should have hedged harder; consider it hedged now.

But here's why I'm not just deleting the section and moving on: when you stop looking at one month and look at the full year, the story relocates rather than just disappears.

Year over year, prime-age employment (25 to 54) is down 779,000. And the concentration isn't in the 25 to 34 bracket at all; that group is essentially flat over twelve months. It's here:

Age group	Change, July 2025 → July 2026
25 to 34	+17,000
35 to 44	-226,000
45 to 54	-570,000
55 and over	-350,000

Forty-five to fifty-four. Peak earning years. Two decades or more into a career. The people carrying mortgages, tuition payments, and aging parents simultaneously.

This is the population I wrote about back in February, when the Professional and Business Services sector shed 30,000 jobs and I described "highly skilled, highly trained professionals, decades into their careers, being met mid-life with erasure and a lack of options." I was looking at the right people; I just spent last month pointing one age bracket too young.

The mid-career erasure is not a one-month blip. It's a 570,000-person trend line, and it has been running underneath every headline all year.

The Job That Isn't a Whole Job

Here is the thread I think matters most in this report, and I haven't seen anyone else pull it.

Over the past twelve months, the number of Americans working full time fell by 1.29 million. Over the same twelve months, the number working part time rose by 311,000.

Sit with that.

The economy did not simply stop creating jobs. It converted them. Full-time positions, the kind that come with health insurance thresholds and retirement matching and predictable income, are being replaced at the margins by part time ones that don't.

And people are responding exactly how you'd expect.

The number of Americans working more than one job has risen every single month since March:

Month	Multiple jobholders	Share of employed
March 2026	8.36M	5.1%
April 2026	8.43M	5.2%
May 2026	8.43M	5.2%
June 2026	8.55M	5.3%
July 2026	8.69M	5.4%

That's up 335,000 over the year, and the share is the highest in this cycle.

In my May report I argued that the falling multiple-jobholder number was misleading, because the decline was entirely concentrated in stable "moonlighting" arrangements while precarious multi part-time work was rising underneath it. That argument is now obsolete for the simplest possible reason: the number isn't falling anymore. It's climbing, and it's climbing in a labor market that is simultaneously shedding full-time positions.

One more from the same table: self-employment (unincorporated) jumped 304,000 in a single month. That category is where displaced workers land when they start driving, contracting, consulting, and freelancing because nothing else is available. One month doesn't make a trend (I've learned that lesson publicly this month) but combined with everything above, it's pointing the same direction.

Meanwhile, 4.8 million Americans are working part time who don't want to be. And within that group, "could only find part-time work" now sits at 1.43 million, up from 1.21 million in May.

The jobs report counts jobs. It does not count whether a job is enough to liv eon. That distinction is now doing an enormous amount of hidden work in these numbers.

Where the Losses Landed

July's -23,000 wasn't spread evenly. It was concentrated, and the concentrations are telling:

Local government education (-50,000). The single biggest drag in the report, in a category that had shown essentially no net change over the prior twelve months. This is teachers, aides, support staff, and administrators. Public education payrolls don't usually move like this outside of budget shocks.

Retail trade (-19,000). Warehouse clubs, supercenters, and general merchandise retailers shed 21,000 alone. Gas stations and fuel dealers lost another 5,000.

Financial activities (-14,000). Credit intermediation lost 9,000, insurance carriers 7,000. This sector is now down 121,000 jobs from its May 2025 peak, and it pays roughly $49 an hour. The high-wage contraction I have documented in every one of these reports since February has not paused for a single month.

Health care (+22,000). Still the most durable engine in this labor market, and still decelerating; the 12-month average is +36,000, and this is the second consecutive month below trend. When the one reliably growing sector starts slowing, that's worth more attention than a 23,000-job headline.

Leisure and hospitality: little change. And with that, the World Cup arc closes exactly as forecast in May. The sector surged on pre-tournament hiring, gave it all back in June, and has now shown no meaningful net change across the entire year. Four months of distorted data that told us nothing about the underlying economy, reported at the time as strength.

Average hourly earnings came in at $37.62, up two cents on the month and 3.2% over the year, down from 3.5% last month. Wage growth is decelerating alongside everything else.

921,000 People Are on Temporary Layoff

This is the number I would watch most closely over the next two reports.

The count of Americans on temporary layoff jumped by 153,000 in July to 921,000. Permanent job loss, by contrast, barely moved: steady at roughly 1.7 million.

Temporary layoff is a distinct category from a permanent separation. It means your employer has sent you home with an expectation of a recall. It's the category taht fills up when businesses are hedging: cutting labor costs immediately while hoping conditions turn before they have to make it final.

Historically, temporary layoffs lead permanent ones. Employers don't jump straight into the irreversible decision; they pause first, then decide. A 153,000-person surge in the "we'll call you" category, in an economy where hiring is already statistically flat, is a leading indicator I don't love.

Pair it with the other side of the ledger: job leavers (people who quit voluntarily because they believe something better is out there) sits at 793,000. Slightly up from June's 776,000, but still near the floor. Nobody is quitting. Nobody is hiring. And now employers are starting to send people home "temporarily."

An entire market holding it's breath.

The Middle Is Filling Up

The long-term unemployment share fell this month, from 27.3% to 25.5%. I'd love to tell you that's recovery. I don't think it is, and the rest of the duration table explains why.

Look at where the unemployment pool moved in July:

• Jobless less than 5 weeks: down 222,000 (and down 344,000 over the year)

• Jobless 5 to 14 weeks: up 113,000

• Jobless 15 to 26 weeks: up 105,000

• Jobless 27 weeks or more: down 166,000

  
  

The front door may be closing: fewer people are entering unemployment, which is consistent with a low-firing market. But the middle is swelling by 218,000 combined. People aren't exiting unemployment. They're aging through it, from the short-term pool into the medium-term pool, on their way to the long-term one.

And the long-term decline carries the same asterisk I flagged last month: when someone who has been unemployed for eight months finally gives up and stops looking, they don't show up as re-employed. They vanish from the unemployment count entirely and land in that "not in the labor force" bucket that grew by 381,000 this month.

Roughly 42% of all unemployed Americans have now been out of work for 15 weeks or longer. The ratio of newly unemployed to long-term unemployed sits at 1.11 to 1; effectively unchanged from June, and still far below the 1.6-to-1 that characterized the pre-pandemic labor market in 2019.

The composition of unemployment in this economy is still tilted toward people the market has already passed over.

The Number That Matters Going Forward

On August 28, three weeks from today, BLS publishes its preliminary annual benchmark revision.

Here's why that date should be circled on your calendar. Every month, the establishment survey estimates payroll employment from a sample of about 119,000 businesses. Once a year, BLS checks that estimate against something far more reliable: actual unemployment insurance tax records, which nearly every employer in the country is legally required to file. The benchmark revision is the moment the survey gets re-anchored to reality.

Now consider what we learned this morning. May lost 63% of itself to ordinary monthly revisions: the routine kind, driven by late-arriving survey responses. If the ordinary revision process is finding errors of that magnitude, the annual re-anchoring to tax records is worth watching very closely.

I've flagged this date in two consecutive reports. I'm flagging it a third time. If August 28 delivers a significant downward re-anchor, then the story of 2026 will not be "a labor market that slowed in the second half." It will be "a labor market that was weaker than reported all year, and we found out in arrears."

Which brings me back to where we started.

The unemployment rate is 4.1%, and it is the lowest it has been in over a year. Payrolls fell by 23,000, which is small enough to shrug at. Nothing in this report is going to frighten anyone into paying attention.

But underneath: an entire quarter of job growth just evaporated into the margin of error. Full-time work is down more than a million positions in a year. Nearly 8.7 million people are working more than one job. 921,000 are sitting at home waiting for a call that may not come. 106 million Americans are outside the labor force entirely, and 2.8 million of them left in the last twelve months.

The labor market is not crashing. It is doing something quieter and, I'd argue, harder to reverse: it's getting smaller, getting worse-paid, and getting less honest about both; one revised number at a time.

Watch August 28.

Key Data At-A-Glance

Indicator	July 2026	July 2025	Direction
Nonfarm payrolls (monthly)	-23k	+73k	↓↓
3-month average (May–July)	+20k/mo	+90k/mo	↓↓
Unemployment rate (U-3)	4.1%	4.3%	↓
Broad underutilization (U-6)	7.9%	7.9%	→
Labor force participation	61.4%	62.2%	↓↓
Employment-population ratio	58.9%	59.6%	↓
Not in labor force	106.2M	103.4M	↑↑
Labor force (monthly change)	-264k	—	↓
Full-time employed	133.6M	134.8M	↓↓ (-1.29M)
Part-time employed	28.8M	28.5M	↑ (+311k)
Multiple jobholders	8.69M / 5.4%	8.36M / 5.1%	↑↑
On temporary layoff	921k	—	↑↑ (+153k m/m)
Long-term unemployed (27+ wks)	1.77M	1.82M	↓
LT unemployed (% of total)	25.5%	—	↓
Jobless less than 5 weeks	1.96M	2.30M	↓↓
Job leavers (voluntary quits)	793k	785k	→
Part-time for economic reasons	4.80M	4.69M	↑
Avg hourly earnings (all private)	$37.62	—	+3.2% YoY
Prior-month revisions	May -66k, June -37k	—	↓↓

Sources: BLS Employment Situation, July 2026 (USDL-26-1291), released August 7, 2026; Summary Tables A and B; Tables A-9, A-11, A-12, A-15, B-1, B-3. Prior-vintage figures from the June 2026 (USDL-26-1125) and May 2026 (USDL-26-0786) releases.

https://www.bls.gov/news.release/pdf/empsit.pdf