September Jobs Miss: Payrolls Plunge To 29k

I refreshed the screen twice before I trusted the number. Twenty-nine thousand. Not a rounding error, not a delayed seasonal quirk sitting in the fine print, just a September hiring print that landed under every serious estimate I had scribbled down the night before. Yields were already twitchy. A few minutes later they were sliding, the kind of slide you feel in your shoulders if you have spent the week listening to people swear the labor market could not possibly crack this fast. Maybe that is the point. The headline missed. The revisions made the miss worse. And yet another survey, the one households fill out, told a louder and more confusing story. If you care about rates, paychecks, or whether the soft-landing story still has legs, this is the report worth sitting with.

What the September Payrolls Miss Actually Showed

The establishment side of the report, the one built from company payroll records, came in at 29,000 jobs. August, which had first looked sturdy at 162,000, was cut to 133,000. July did not merely cool. It flipped. A print that had been nudged positive last month is now negative again, down 10,000. Put the two revisions together and employment in July and August is 60,000 lower than the figures markets were trading on a month ago. I have watched enough of these releases to know revisions are not a footnote. They are the report, arriving late.

There is a bitter little irony in the sequence. August originally cleared the high end of forecasts. After the pencil came out, it sits roughly in line with what people expected the first time. September, by contrast, missed the entire range. That pattern, a strong first print that gets walked back, then a weak follow-through, is how confidence in the series erodes. Traders do not need a conspiracy theory. They need a reason to stop treating the first number as gospel.

A payroll figure is a snapshot taken with a shaky camera. The revisions are the second look, and sometimes the second look changes the whole room.

– Market strategist, after a revision-heavy release

Unemployment did not cooperate with the soft headline either. The rate rose to 4.2 percent from 4.1 percent, above calls for an unchanged print. Unemployed workers increased by about 78,000, to 7.109 million. The labor force jumped far harder, up 485,000, to 170.262 million. Among major groups, the jobless rate for Black workers jumped to 7.0 percent. Rates for adult men, adult women, teenagers, and for White, Asian, and Hispanic workers showed little change on the month. One group moving while others hold still is not a rounding story. It is a distribution story, and distribution is where labor-market stress usually shows up first.

Why Two Surveys Can Disagree So Sharply

Here is the split that makes this release hard to summarize in a single sentence. The household survey, which asks people whether they worked, showed a 406,000 surge in employed workers. That is the second-largest gain since January, behind only August’s 569,000. Total employment on that measure reached 163.152 million, the highest since the start of the year. So payrolls stumbled, and the count of people saying they have a job jumped. Both can be true in the same month. They rarely feel true at the same time.

I have found that readers get lost when analysts pretend the two surveys are interchangeable. They are not. Payrolls come from businesses and miss some self-employment, some new firms, and some informal work. The household survey catches people, including those who just entered the labor force, and it is noisier month to month. A 406,000 pop can be real activity. It can also be sampling noise that fades. Treating either number as the whole economy is how people talk themselves into a trade they regret by Thursday.

Participation helps explain the unemployment uptick. The labor force grew faster than employment in a way that pushed the jobless rate higher even as the household count of workers rose. Participation has jumped over the past two months after sinking to a five-year low in July. In September the participation rate sat at 61.8 percent and the employment-population ratio at 59.2 percent, little changed on the month and little changed on net since January. “Little changed” is the official phrase. It hides a sharper two-month rebound that matters if you are trying to judge whether people are coming back because jobs exist or because savings ran out.

Wages Gave the Inflation Crowd a Quiet Win

Average hourly earnings rose 0.1 percent, under the 0.3 percent that had been penciled in. That pulled the yearly pace to 3.0 percent, from 3.1 percent, and below an unchanged forecast. For production and nonsupervisory workers, pay rose 7 cents, or 0.2 percent, to $32.60. The workweek held at 34.4 hours for all private employees. Manufacturing stayed at 40.6 hours, with overtime stuck at 3.0. Production workers held a 33.8-hour week. Hours not falling is a small mercy. Hours collapsing would have made the weak payroll print look like demand destruction rather than a hiring pause.

Perhaps the most interesting aspect is how calm the wage number looks next to the chaos in the job count. A labor market that is adding almost no payroll jobs and still paying people only a tenth of a percent more is not an overheating machine. It is a machine idling. Whether that idle is the soft landing people have been promising, or the first cough before something rougher, is the argument this report refuses to settle.

A Quick Map of the Headline Figures

Numbers land cleaner in a grid than in a paragraph, so here is the spine of the release without the commentary piled on top.

If you only remember one row, remember the revision row. A 29,000 print is a miss. A 29,000 print after 60,000 jobs were erased from the prior two months is a different animal. Policy that was set on the old July figure was set on a number that no longer exists.

The People Behind the Rate

Long-term unemployment, defined as 27 weeks or more without work, was essentially unchanged at 1.9 million. That group accounted for 27.1 percent of all unemployed people. A stable long-term share next to a rising headline rate usually means the new unemployment is fresh, not a pile-up of people who have already been shut out for half a year. Fresh unemployment can reverse. Entrenched unemployment is stickier, and more expensive for households.

Part-time work for economic reasons held near 4.5 million. These are people who wanted full-time hours and did not get them, either because hours were cut or because a full-time role never appeared. Little change here is mildly reassuring. A spike would have said employers are hoarding heads and cutting schedules, which is often the step before layoffs. The number of people outside the labor force who still want a job was little changed at 5.8 million. They are not counted as unemployed because they did not look in the prior four weeks, or could not take a job. Marginally attached workers, a subset who looked sometime in the past year, fell by 236,000 to 1.5 million. Discouraged workers, those who think nothing is out there for them, were little changed at 414,000.

• Long-term unemployed held at 1.9 million, about 27 percent of the jobless total.
• Involuntary part-time work stayed near 4.5 million, no fresh surge.
• People wanting a job but not actively looking sat at 5.8 million.
• Marginally attached workers dropped by 236,000, to 1.5 million.
• Discouraged workers were roughly flat at 414,000.

That list is less dramatic than the payroll miss, and that is useful. Drama in the headline with calm in the margins is a different regime from drama everywhere. I would rather see discouraged workers falling than rising. A drop in the marginally attached count can mean people either found something or stopped even the occasional search. The report does not split that hair cleanly. You have to watch the next two prints before you decide which story won.

Where the Jobs Actually Landed

Health care kept hiring, but the pace slowed. The sector added 17,000 jobs, against an average monthly gain of 33,000 over the prior year. Ambulatory care added 13,000 and hospitals added 12,000, while nursing and residential care facilities lost 9,000. Health care has been the reliable engine for so long that a half-speed month feels like a warning light. It may only be a pause. Engines that run hot for years do not stall in a single print, but they do cough.

Construction was little changed, up 11,000, in line with a 10,000 average over the past year. Nonresidential specialty trade contractors continued to trend higher, adding 12,000. Manufacturing was also little changed, up 9,000, yet it is up 72,000 since a recent low in December. Plastics and rubber products and machinery each added about 5,000. A factory sector that is no longer bleeding jobs is not the same as a factory sector that is booming. Still, “not bleeding” is an improvement plenty of industrial towns will take.

Financial activities slipped by 7,000 and are down 129,000 since a peak in May, with insurance carriers and related work accounting for most of that, about 90,000. That is not a one-month blip. That is a sector that has been quietly shrinking for months while headlines stayed glued to the aggregate. Other major industries showed little change: mining and energy extraction, wholesale, retail, transportation and warehousing, information, professional and business services, social assistance, leisure and hospitality, other services, and government. A report where most sectors are flat and one former engine is slower is a cooling report, not a collapse report. The difference matters for anyone pricing a recession versus a long muddle.

Sector snapshot, September:
  Health care        +17,000   (slower than the +33,000 trend)
  Construction       +11,000   (near the yearly average)
  Manufacturing       +9,000   (up 72,000 since December low)
  Financial activities -7,000  (down 129,000 since May peak)
  Most other majors   flat

Full-Time, Part-Time, and Who Got Hired

Under the surface, part-time jobs rose by 205,000 to 28.746 million, while full-time jobs rose by 88,000 to 134.376 million. Both up is better than a pure swap from full-time into part-time. The mix still leans toward the smaller category growing faster, which is what you see when employers are cautious about committing to benefits and hours. I do not read that as panic. I read it as hesitation, and hesitation is how hiring slowdowns usually start.

The native-born versus foreign-born split did what it has often done. US-born employment rose by 298,000. Foreign-born employment rose by 473,000. Back, in other words, toward the pattern that dominated earlier stretches of this cycle. Whatever you think that mix means for wages, housing, or politics, the labor-market arithmetic is plain. A large share of the household-survey strength came from foreign-born workers. Ignoring that and talking only about the aggregate is how commentary stays tidy and wrong.

How Markets Took the Miss

Bonds did not wait for the nuance. Yields slipped from around 5.22 percent to 5.16 percent, a fresh weekly low, as a heavy short base in Treasuries got squeezed. The setup had been whispered for days: August seasonals were flipping, positioning was crowded, and a soft payroll number was the match. The match arrived. When the print misses every estimate, the people leaning the other way do not get a debate. They get a gap.

Markets, and by extension policy watchers, still lean on the payroll side of this report more than the household side. That is a habit, not a law. It is also why a mixed release produced a clean rates reaction. Payrolls missed. Revisions cut the past. Wages cooled. The unemployment rate rose. The household survey was strong enough to complicate the story, not strong enough to cancel the squeeze. If you were short duration into this print, the morning was unpleasant. If you had been arguing that the first August number was too good to last, the morning felt like vindication, at least until the next revision.

Positioning is the hidden payroll number. A miss only hurts if the crowd is standing on the wrong side of it.

Would the last rate increase have happened if July had already been known as negative? That is the counterfactual everyone reaches for, and it is not answerable with a spreadsheet. Policymakers hike on the data they have, plus forecasts, plus a fear of looking late. A later revision does not rewind the meeting. It does change the burden of proof for the next one. A labor market that just printed 29,000, with the prior two months 60,000 lighter, is a harder backdrop for another restrictive step than the backdrop they thought they had.

What a Cooling Print Does to the Soft-Landing Story

The soft-landing pitch needs hiring to slow without unemployment lurching. September offered a piece of that and a piece of the opposite. Hiring on the payroll measure nearly stalled. Unemployment rose a tenth. Wages decelerated. Participation stabilized after a nasty July. You can build a hopeful paragraph out of those pieces. You can also build a worried one. I lean toward the muddle, not the miracle. Labor markets rarely glide. They step down, look fine for a month, then surprise you.

Think of the jobs market as a heavy truck on a long grade. Speed drops before the engine fails. A 29,000 print is the speed dropping. Health care still adding jobs, construction holding its average, manufacturing off the lows: those are the engine still turning. Financial activities down 129,000 since May is a cylinder that has already gone quiet. Trucks can coast a long way on a grade. Coasting is not the same as climbing.

Recent labor-market research has spent years on this exact ambiguity. Establishment data and household data diverge most when the cycle is turning, because new firms, closing firms, and shifts into self-employment all hit the two surveys on different lags. According to labor economists who track the series closely, the first print is a draft. The third print is closer to a document. Trading the draft as if it were the document is a choice. It is not a requirement.

How Households Actually Feel a Number Like This

A payroll miss does not arrive in a living room as “29,000.” It arrives as a slower reply from a recruiter, a posted role that sits open, a raise conversation that gets pushed to next quarter. Wage growth at 3 percent is not a crisis if prices are calming. It is a tighter budget if rent and insurance are not calming with it. The report cannot see your lease. It can only see the average hourly line, and that line just got a little less generous.

The jump in the labor force is the part I keep coming back to. Nearly half a million more people counted as in the market for work. Some of that is seasonal noise. Some of it is people who sat out in July deciding the cushion was gone. When participation rises and payrolls do not, the unemployment rate is the exhaust valve. It rose. That is arithmetic, not mood. Mood shows up later, in spending, in quit rates, in how long people tolerate a job they do not like because the next one is no longer obvious.

1. Check the revision, not only the headline. July and August just lost 60,000 jobs between them.
2. Separate payrolls from the household count. They diverged hard this month.
3. Watch wages and hours together. Pay cooled, the workweek did not crack.
4. Look past health care. Breadth was flat, and finance has been shrinking since May.
5. Treat the first print as a draft. The next two releases will edit it.

Rates, Positioning, and the Squeeze That Followed

Bond shorts had a story. Growth was fine, the labor market was fine, supply was heavy, and yields belonged higher. A payroll number under every estimate does not kill that story. It interrupts it. Interruptions are expensive when the position is crowded. The move from 5.22 percent to 5.16 percent is not a regime change on its own. It is a reminder that the regime was leaning on a labor narrative that just took a hit. Seasonals flipping from August into the fall only added kindling. The jobs miss was the spark.

I have sat through enough of these mornings to distrust the first hour. Liquidity is thin, headlines are blunt, and the household survey has not been read yet by half the people hitting the button. By the afternoon the nuance usually creeps back in. Strong household employment. Stable hours. Health care still positive. Those details do not unwind a squeeze. They cap how far people are willing to chase it. If you are allocating rather than scalping, the useful question is not “did yields drop.” It is “does a 29,000 payroll trend, if it sticks, change the path of policy.” On today’s evidence, it should at least slow the people arguing for more restriction.

What Would Make the Next Print Matter More

One soft month is a weather report. Two soft months are a climate note. Three, with revisions still pointing down, are a change in the furniture. The details I will be watching are dull on purpose. Does health care reaccelerate toward that 33,000 average, or does 17,000 become the new run rate? Does finance stop leaking jobs? Does the workweek slip under 34.4? Does the household surge of 406,000 get revised or repeated? A repeated surge with weak payrolls would say the establishment survey is missing something real. A vanished surge would say September’s optimism was noise.

Black unemployment at 7.0 percent deserves a second look next month, not a shrug. Group rates that jump while others hold can be sampling. They can also be the early cut of a slowdown that has not reached the median worker yet. Teen unemployment at 14.5 percent is high in the way teen unemployment is often high. It is still a barometer for how easily the newest workers are being absorbed. Little change is fine. A grind higher would not be.

There is also the boring institutional point. If July had been published as negative the first time, the conversation around the prior policy step would have sounded different. Data dependence is only as good as the data. A series that revises by tens of thousands of jobs, in both directions, within sixty days is a series that should be held loosely. Holding it loosely is not the same as ignoring it. The 29,000 figure is still information. It is just not a verdict.

A Practical Read for Anyone Allocating Capital

You do not need to become a labor economist to use this report. You need a short list of what changed and what did not. What changed: the level of recent job growth, now lower after revisions; the unemployment rate, a tenth higher; wage growth, a touch cooler; the bond market’s willingness to stay short. What did not change: the workweek, most industry trends outside a slower health-care month and a still-shrinking finance sector, the long-term unemployed count. A portfolio that was priced for endless labor strength just got a mark. A portfolio that was priced for an imminent collapse did not get confirmation.

In my experience the expensive mistake is updating too far on one release. The cheap mistake is not updating at all. Split the difference. Trim the confidence you had in the August headline. Raise the odds that the next policy step is a pause rather than another tightening. Leave room for the household survey to have been early rather than wrong. And keep an eye on insurance and other financial jobs, because a 90,000 decline concentrated in one corner of finance is the sort of slow leak that never leads the evening news and still shows up in earnings six months later.

Simple lens: payroll trend + revisions + wages + hours. If three of four cool, the labor bid for higher yields is weaker than last month.

The Mixed Report, Without the Spin

Call it mixed and you will not be wrong. The household survey painted a stronger picture than the establishment survey, which is exactly why unemployment could rise while people reported more work. Markets cared more about payrolls, which is why bonds squeezed. Both reactions can be rational. They answer different questions. Households answer “are people working.” Payrolls answer “are firms adding names to the books they report.” Policy and rates have a long habit of listening to the second question first.

Foreign-born hiring outpacing native-born hiring puts the cycle back in a familiar groove. Full-time and part-time both rose, with part-time rising more. Health care cooled without reversing. Construction matched its trend. Manufacturing edged up from a higher base than December. Finance kept losing ground. None of that is a slogan. It is a labor market losing speed in public while still employing more people than it did in January on the survey that asks workers directly. Holding both facts at once is the whole job.

So what do you do with a morning like this? You mark the miss. You mark the revisions. You refuse the urge to declare the landing achieved or the landing failed. Twenty-nine thousand jobs is a small number in a workforce above 160 million. Small numbers, repeated, are how large trends begin. The next release will tell us whether September was a pothole or the start of the grade. Until then, the honest position is the unfashionable one: slower than advertised, not yet broken, and already enough to squeeze anyone who was sure the labor market could not surprise them.

I will be reading the next revision before I trust this one. That is not cynicism. That is the tuition this series keeps charging.