One final note: an AI-generated analysis went viral today claiming the B-D model added +145k to January payrolls.

The actual number was -61k.

Wrong number, wrong sign, wrong framework.

Check bls.gov/web/empsit/c... before posting. It takes 5 seconds.

Did it drive the beat?

No.

The underlying employer survey data was strong, independent of the model.

Strip the B-D out entirely and job growth was still at or above normal.

So to summarize:

Was the B-D more favorable than usual in January?

Yes – both at the headline level and in Education & Health Services sector.

That's a fair critique and worth monitoring.

The labor market produced a normal January.

The seasonal adjustment turned that into a +130k SA headline.

The birth-death model's relative generosity didn't change the outcome.

For total nonfarm, the same exercise tells a more modest but still positive story. Ex-B-D job growth in January was roughly in line with the historical norm – not dramatically above it, but not weak either.

The underlying survey data – not the model – drove the outsized strength. The Birth-Death adjustment was along for the ride.

For education & health services, stripping out the B-D and comparing to the historical norm shows Jan ‘26 was approximately 90k above normal. The largest positive deviation since the post-COVID recovery.

This isolates actual survey responses – what employers reported directly to BLS – from the model's estimate of business births and deaths, along with seasonal adjustment challenges.

If the B-D were manufacturing the strength, the ex-B-D data should show weakness.

It doesn't.

Fine, it was a tailwind. But here's the test.

Strip the B-D adjustment out of the reported NSA gains entirely, then compare what's left to the pre- and post-COVID norm (average of 2017–2019 and 2023–2025).

The model assumed more net business creation in the sector than usual.

So, the question isn't whether the B-D was generous. It was. The question is whether it drove the beat.

The same is true for Education & Health Services, which I’ve also heard singled-out.

The January 2026 B-D of +85k was historically elevated – and indeed the largest tailwind of all sectors – with prior Jan adjustments ranging from roughly 5k to ~45k.

So yes, the B-D model was a relative tailwind compared to history.

That said, there's a fair observation buried in the bad math.

The Jan ‘26 total nonfarm B-D adjustment of -61k was less negative than the typical January reading, hence the chart at the top, which has ranged from roughly -100k to -250k over the prior decade.

They're in different statistical frameworks. The B-D is applied to *not seasonally adjusted* data. The headline is *seasonally adjusted* - mixing them is arithmetically meaningless.

First, the fundamental error: you cannot subtract the not seasonally adjusted (NSA) birth-death adjustment from the seasonally adjusted (SA) headline payroll number.

Here we go again…

A lot of analysis today claiming the birth-death model inflated the Jan job gain (130k).

Some are even subtracting the B-D contribution from the headline to argue the "real" number was negative.

The underlying logic is flawed – but there's a kernel of truth.

🧵
#EconSky

🫡 🙏

Going forward, the modified birth-death model should narrow this gap. But the lesson: treat any single monthly print with healthy skepticism, and watch the trend.

Bottom line: The Fed was making policy in 2025 based on labor market data that significantly overstated job growth. The revised data suggests the economy was cooling faster than anyone in the room knew.

The honest framing: the trend has turned higher, and that matters. But we won't know if this rebound is real until the next annual benchmark. Until then, treat the direction as signal and the magnitude as provisional.

There's no guarantee the recent acceleration isn't facing the same fate.

The BLS's modified birth-death model is designed to fix this. Early results are promising. But it's one methodological adjustment against a problem that's been compounding since at least 2023.

But can we trust the rebound? Two consecutive years of massive downward revisions (-598k in '24, -861k in '25) should give anyone pause. The monthly data has systematically overstated job growth for two straight years, and not by a small amount.

January's +130k total nonfarm job growth print is the first statistically significant month of job growth in a year. Not a blowout. But given where we've been, a meaningful inflection.

The narrative that matters for 2026 isn't "the BLS overstated jobs by 900k." It's that the labor market quietly absorbed an even more significant slowdown than we could see in real time… and came out the other side.

The 3-month average is even more emphatic, accelerating into year-end and through Jan ‘26.

The revised picture: the labor market was far weaker than reported, found a floor, and is now rebuilding momentum.

Here's where it gets interesting. Instead of dwelling on how weak 2025 was, look at the trajectory.

The 6m avg for private-sector job growth bottomed near zero around mid-2025. By Q4, it was clearly inflecting higher.

Without it, we'd be talking about a revision north of -1.1m. The fix is helping, but clearly not enough to eliminate the problem.

The original B-D model added 1,160k jobs over Apr–Dec '25. Revised: 917k. That's 243k in phantom jobs stripped out.

BLS already applied this modified approach to the 2024 post-benchmark period, where it pulled 228k out of B-D forecasts and reduced the Mar '25 benchmark revision by -229k.

A key methodological change buried in this release: BLS permanently modified the birth-death model to incorporate current sample information into monthly forecasts – not just during the annual post-benchmark recalculation.

What's left? Private Education & Health Services (averaging ~60k/mo) accounted for virtually all private-sector job growth in 2025. Strip that out and the private sector was contracting by an average of -30k/mo.