The Jobs Data Is Good, Bad and Trending Towards Recession. Got It? Me Neither

America has a serious jobs problem, in more ways than one. Coming into today’s highly anticipated jobs report from the Bureau of Labor Statistics (BLS), we already knew that hiring was basically flat in the second half of 2025, the worst non-recession jobs year since 2003. The struggle to get hired in America is very real, but it is not translating to firings. One big problem with all the labor market data is all the contradictory messages it sends. While job growth remains a serious(?) issue, BLS data proves that the unemployment rate remains low, at 4.3%, roughly about what it was in 2000 and 2006. Recession doomers will point out how the rate is currently slowly curling up like it did before it spiked into recessions after 2000 and 2006, but it’s not there yet. While some BLS truthers still remain out there, the data is trustworthy, and it all points in one direction of a robust labor force that isn’t adding a ton of new jobs. The employment to population ratio of 25- to 54-year-olds is near all-time highs at 80.9%. Total nonfarm layoffs and discharges are below its historic baseline. Despite the bad vibes and some of the data that backs them up, lots of people have jobs and the economy is not in a recession. We think.

One big problem with saying we are in a recession is that typically, we cannot say for certain until afterwards. It takes time to build enough data to make a conclusive case, and one really good month of January 2026 job data that far surpassed analysts’ expectations is not enough to overrule the gigantic downward revisions made to 2025 and 2024 job growth. Today’s release is perhaps the perfect example of how under Trump 2.0, there are conflicting narratives about the labor market within the same datasets, and it’s real. This good January figure is not team Trump messing with the numbers and doing insider trading off of it, they can do their usual insider trading without the immense surgical task of spoofing BLS data. Not to mention, why would they rig the BLS dataset to show 133,000 jobs created in January, but revise it to 553,000 fewer jobs in 2024 and 403,000 less than previously thought in an already bad 2025? Something difficult to understand is developing in our post-2020 economy, and you can see the changes in the data from that point forward everywhere you look. Part of the confusion around the jobs market is we have definitely anchored our expectations to a pre-2020 world that seems gone forever.

Another problem with predicting recessions is that a lot of people on Wall Street build reputations and bank accounts on being perpetually wrong doomers, because doom sells. It’s not as sexy to say “wow the economy was a lot more resilient than we thought in the face of Trump’s tariffs,” even though that proved to be objectively true in 2025. Most online audiences, especially our beloved antifascist political audience, want to hear that ol’ Donny nuked the economy overnight and is finally not going to wriggle out of this one when that’s just not how it works. It takes time to drop a nuke on the economy, and just because we can’t quite see the full blast zone today does not mean it cannot emerge tomorrow. Don’t you worry, ol’ Donny is doing everything he can to make the 1970s great again. If he gets what he wants at the Fed once Kevin Warsh takes over in May, political audiences will eventually get their wish. Jerome Powell has made that pretty clear the last several months.

How some things work never changes, but how job data works is clearly changing, and this is where the dreaded r-word rears its ugly head. But I’m not going to say it for fear of branding myself a Dalio doomer because I already write enough about all the warning signs I see popping up around us, so I’ll tag in Michael W. Horrigan, former associate commissioner with the BLS to say it for me. 

“The June [2025] downward revision of 133,000 was the sixth largest in absolute value since 2003, when probability-based sampling was introduced into the [Current Employment Statistics survey (CES)]. However, if you exclude the pandemic months, it was in fact the largest since 2003,” wrote Horrigan in September. “Nearly all the largest downward revisions from first to second closing have occurred when the economy was turning toward or in a recession: 1979–82, 1989–90, 2008–09. The brief recession of 2001 is an exception to this pattern. The June 2025 revision fits this historical profile, suggesting that it may be less an anomaly than a signal of a labor market weakening toward contraction.”

Now five months later, we have a jobs report revising already poor 2025 jobs data down even further. Far, far further than we previously thought. Strong GDP growth makes it very difficult to argue that the economy is in a recession, but all the labor market trends are clearly moving in that direction, save for January 2026.

Breaking this down by year, job growth in 2024 was revised down by more than a quarter (-553k jobs), and 2025 was revised down by close to 70% (-403k jobs). 2025 was the weakest year for job growth since the pandemic, and before that since the Great Recession.

[image or embed]

— Ben Casselman (@bencasselman.bsky.social) February 11, 2026 at 7:34 AM

The natural question to ask next is that if jobs data for 2025 had to be revised down nearly 70%, why should we trust that the January data is accurate? It’s a good question! Every initial jobs report should be taken with a grain of salt, because revisions are just part of how this data gets parsed out. And this is where we can debunk all the BLS truthers. It is very difficult to rig BLS data, in part because the businesses reporting to the BLS have copies of their responses too (revisions happen because smaller businesses with fewer resources typically report after the initial deadline). Erika McEntarfer, the head of the BLS fired by Trump because she wouldn’t massage the data to his liking, even told the Mueller She Wrote account who was making yet another baseless allegation that this one was baseless too.

Person who was fired here – you should still trust BLS data. The agency is being run by the same dedicated career staff who were running it while I was awaiting confirmation from the Senate. And the staff have made it clear that they are blowing a loud whistle if there is interference.

— Erika McEntarfer (@erikamcentarfer.bsky.social) February 11, 2026 at 9:25 AM

The problem with jobs data isn’t that it’s rigged, it’s that the reporting structure is decaying, and has been for some time, and Trump is making things worse. Sylvain Leduc and Luiz Edgard Oliveira wrote at the San Francisco Fed Blog that the response rate of the CES surveys “was hovering around 60% for the decade preceding the pandemic, it has since declined to less than 45%.” This in part explains why there have been outsized downward revisions that former associate BLS commissioner Michael W. Horrigan said fit recessionary patterns. Anyone who has messed around with Excel can tell you that less data is going to lead to more variance, and so that in part explains why some of these monthly jobs reports can be so wrong.

The other issue is response rates climb dramatically after the first round. The January figure everyone is celebrating today will very likely be revised lower in the coming months, unless we are to break this multi-year trend. “About 60 percent of firms respond by the first closing (the basis for the initial release), 90 percent by the second, and 95 percent by the third,” wrote Horrigan. “First-closing response rates have slipped from around 70 percent pre-pandemic to 60 percent today.” These wild and confusing swings make complete statistical sense given the falling response trends. That’s what variance is.

But the San Fransico Fed authors don’t believe that this increased variance on the path to the final number means the jobs data “may have become more uncertain and less reliable than in the past,” the same way that declining household and business response rates have clearly borked things like the Michigan Consumer Sentiment survey that’s currently at all-time lows. “Our Letter shows that this is not the case for important labor market and inflation measures,” wrote SF Fed Blog. “Over the past two to three years, the revisions to payroll numbers and CPI inflation rates have been in line with their pre-pandemic averages.”

So if lower response rates creating more noise in the monthly data releases aren’t to blame for the increasingly bleak jobs market picture, what case do the labor market bulls have? This is an economy that has recently demanded over 100,000 new jobs per month to keep the unemployment rate steady, but now we’re learning that basically every industry except for healthcare and social assistance had flat or negative job growth in 2025, yet the unemployment rate is still low? What’s going on?

Yikes here are the numbers for the past 12 months. Everything is pretty much dead except healthcare/social assistance. Note that healthcare growth is largely not about the business cycle — it’s about the aging of the population and more elderly people needing care.

[image or embed]

— Catherine Rampell (@crampell.bsky.social) February 11, 2026 at 6:47 AM

Manufacturing overall has LOST jobs over the past year (-83k).

Pretty much every manuf subsector is flat or down, with modest exceptions for fabricated metal products, electrical equipment, food.

[image or embed]

— Catherine Rampell (@crampell.bsky.social) February 11, 2026 at 7:19 AM

I think the Dallas Fed wrote a paper in October that provides the best explanation for the seeming contradiction between the strong economy and a weakening labor market, usually a lagging indicator that is one of the last ones to deteriorate going into a recession. The article is titled “Break-even employment declined after immigration changes.” This is where the political readers get their economic catharsis, as the Fed has been crystal clear about this since September: they think Trump’s racist immigration policies are a big part of the story about the worsening labor market, and Jerome Powell said in September that is why the Fed cut rates, not Trump’s tariffs which have inflationary fears attached to them.

“A new, high-frequency estimate of break-even employment shows a dramatic reversal in immigration flows, combined with cyclical shifts in labor force participation, has caused the monthly break-even requirement to collapse from a peak of approximately 250,000 in 2023 to about 30,000 in mid-2025,” wrote Anton Cheremukhin at the Dallas Fed in October. Break-even employment is a simple concept; it’s just the pace of job growth needed to hold the unemployment rate constant. The best explanation I see for how we can seemingly have all this jobs data trending towards the dreaded r-word, yet unemployment rates and GDP are far, far away from anything resembling it, is that the bar has been lowered for the labor market. We don’t need a ton of job growth anymore.

Here’s the harrowing chart from that Dallas Fed paper showing the new racist reality that Navarro is saying is good for the economy. The author calculated that just 30k jobs a month now is enough to keep the unemployment rate steady. www.dallasfed.org/research/eco…

[image or embed]

— Jacob Weindling (@jakeweindling.bsky.social) February 10, 2026 at 9:53 AM

Another way you can trust that this jobs data is not rigged is that the day before it came out, White House advisor Peter Navarro was on Fox Business saying “We have to revise our expectations down significantly for what a monthly job number should look like…Wall Street has to adjust for the fact that we’re deporting millions of illegals out of the job market.” Does that sound like a guy who thinks January job growth is going to nearly double analysts’ expectations? 

The latter part of Navarro’s quote is a bald-faced racist lie that is this administration’s standard policy in every aspect of their degeneracy, but the former is getting at what the Dallas Fed wrote about in October. Something is happening to the economy where we just need fewer new jobs to sustain a healthy unemployment rate. Yet again in the Trump 2 era, one of the worst people you know made a great point.

Which is why I finally put the dreaded r-word in a non-theoretical title today, because a BLS commissioner said it before last year’s bad jobs figure was revised down 70%, and fewer jobs implies lower growth and the definition of a recession is two consecutive quarters of negative GDP growth. Something has broken in the labor market and it’s not clear what. It’s probably a combination of a lot of things; ICE terrorizing workers off the job and sending the farm economy into a measurable recession in places like Iowa and Nebraska; fewer responses leading to higher month to month noise in the data; slowing immigration trends that preceded Trump and accelerated under him; AI killing many entry-level jobs, and surely many more blind spots we have yet to identify.

It seems like we are barreling towards one of two not so great realities. Either there is a new, far lower baseline where the economy simply does not need to hire as many workers to sustain itself, which trends towards lower long-term growth (and you can see fears of this in increased long-term borrowing costs across the entire western world and Japan), or perhaps the reason why a former associate BLS commissioner thinks all these aggressive downward job growth revisions “fit this historical profile” of when “the economy was turning toward or in a recession” is because we are currently turning toward or in a recession. I genuinely do not know, but none of this feels like it’s good news.