It’s a good thing payrolls are (allegedly, purportedly, suddenly) growing again in America, because if past is precedent, the sharp deceleration in US jobs growth last year almost invariably presages a recession.
I don’t know whether to take the BLS data seriously anymore, and neither does anybody else. Between post-pandemic dynamics (e.g., low survey response rates), staff shortages, politicization and what I have to believe is low morale amid rolling government shutdowns and the (apparently very real) threat of being fired for reporting numbers that might reflect poorly on the executive branch, you’ll be forgiven for suggesting the bureau’s no longer the gold standard. For anything.
Revisions are supposed to be an exercise in refining data such that it’s more accurate. That, in turn, should build confidence in the numbers. But if they’re too large (the revisions), such exercises can paradoxically undermine credibility as the public wonders, “If you were that far off, why’d you even bother?” And, “Why should we trust the numbers we’re seeing now when we know they’re going to be revised such that the final figures bear no resemblance to the initial estimates?”
With that mind, it’s worth noting that the post-benchmark revision total for US hiring in 2025 — which, you’re reminded, reflected a net 181,000-job gain, down an absurd 70% from the initially reported figures — was the fourth-lowest ever for a non-recession year.
There’s the chart. The grey-shaded areas are (obviously) recessions.
The drop-off in 2025 (the first year of Donald Trump’s second term, and the onset of what he’s still describing as a new American “golden age”) versus 2024 (the last year of Joe Biden’s only term) was almost 90%.
As BMO’s Ian Lyngen remarked, 2025 “marks the lowest annual total for a non-recession year since 2003” and, disconcertingly, “all three years in which cumulative annual payrolls totaled less than [2025’s] 181,000 immediately preceded or succeeded a recession.”
BLS needs to rethink the jobs report. The revisions are too large for the data to be credible.
It seems the error sources include:
– Declining response rates. More accurately, responses are slower so the response rate at the first preliminary report is low.
– Undocumented workers. BLS surveys do not exclude undocumented workers but the QCEW data used for benchmarking does, and BLS’ corrective methodology assumes steady immigration patterns.
– Gig workers. Neither BLS nor QCEW data capture gig work, which has grown a lot.
– Birth-death model. I’ve read this tends to be most inaccurate at economic turning points.
– Annual benchmarking. Quarterly benchmarking could help reduce sample error stacking and the birth-death model error.
I would think that with access to all state and federal agency data, and a sufficient budget and essential status, the jobs report data could and should be better.
Hopefully this doesn’t get fixed until after 2028, because everything Trump touches becomes a mess or worse.
On the article topic, I still think that the economy excluding AI investment is already tipping into recession. In a bifur-K-ated economy, recessions might develop more slowly and be harder to define and call. “Slowly, then quickly”.
The apparently likely effects, of AI on (human) employment would seem only to be in the early stages – so hard to see anything but further declines on the horizon. At least the Fed shouldn’t have to be exhorted, manipulated and threatened to ease…
Plot real GDP growth and payroll growth. Pretty clear relationship.
https://fred.stlouisfed.org/graph/fredgraph.png?g=1RX29&height=490
Some argue AI is driving productivity higher, so those curves will decouple – in a secular, sustained way, not the cyclical pattern and short blips we’ve always had, that are apparent in the chart.
https://www.ft.com/content/4b51d0b4-bbfe-4f05-b50a-1d485d419dc5?
I am doubtful we’ll see that decoupling in the near term. Among other reasons, if AI did displace that much labor that quickly, consumer sentiment and demand would be hit and with it GDP.
Oh don’t be such a worry wort. Those right-sized employees will simply pivot into well-paying careers as nurses/PAs and working in construction jobs building datacenters and prisons. Didn’t Elon and Jensen recently say something to that effect?