Today's January employment report came in at 130,000 jobs, well above the consensus estimate of somewhere between 55,000 and 75,000 (take your pick of forecasters). Stocks initially rallied. Then they didn't; all the major averages closed lower. For all the anticipation around this report — it was delayed a week by the partial government shutdown, after all — the market's inability to hold its gains tells you something. It could be telling you that investors are finally, slowly, learning a lesson that has been staring them in the face for years: the first release of any monthly jobs report is little more than an educated guess, and not a particularly good one. But probably not since they've been trading on these reports my entire career. Maybe there's another reason; more on that below.
Consider what else the BLS released today along with the January headline number. The annual benchmark revision — the once-a-year process where the BLS reconciles its survey estimates with actual payroll tax records — just told us that 2025 was far worse than we thought. Job creation was revised down by 403,000 for the year, dropping the monthly average from an already weak 50,000 to a nearly recessionary 15,000. The full-year total of 181,000 jobs is the weakest since 2003, outside of actual recessions. We didn't know that until today. We were trading on fiction.
And this isn't a new problem. Last year's benchmark revision knocked 898,000 jobs off the prior 12-month period. The year before that, 589,000 phantom jobs disappeared. Monthly revisions compound the issue — December's number was revised down today to 48,000 from 50,000, and November was cut to 41,000 from 56,000. The confidence interval on any single monthly release is plus or minus 122,000 jobs. That means a reported gain of 130,000 could, statistically, be anywhere from 8,000 to 252,000. We don't treat it that way, of course. We trade on the number as if it were fact.
The problem has gotten worse since COVID, and to be fair, this is not a political observation — it's a data quality one. Survey response rates at the BLS have fallen from around 60% before the pandemic to below 43% as of last spring. Less than half of surveyed businesses are responding. The models the BLS uses to fill in the gaps — including the controversial "birth-death" model that estimates jobs created by new businesses — have struggled to keep pace with a more dynamic, post-COVID economy.
Which brings us, inevitably, to the soap opera that has unfolded around the agency itself. Last August, the July jobs report came in at a disappointing 73,000, with steep downward revisions to May and June. President Trump, without evidence, declared the report "RIGGED" and fired commissioner Erika McEntarfer hours after its release — despite the fact that, as multiple former commissioners immediately pointed out, the BLS head doesn't compile the monthly reports and sees the numbers only shortly before the public does. McEntarfer had been confirmed by the Senate 86-to-6, hardly the profile of a partisan operative. Trump then nominated Antoni, a Heritage Foundation "economist" and Project 2025 contributor, to replace her. Antoni's nomination collapsed so spectacularly — he couldn't muster even Republican support in the Senate after his background and public statements received scrutiny — that the White House quietly withdrew it in September.
So where does that leave us? The BLS is currently operating without a Senate-confirmed commissioner, the agency's credibility has been publicly attacked by the administration it reports to, and its underlying data quality is genuinely deteriorating for structural reasons that have nothing to do with politics. When today's number came in surprisingly strong, the market's instinct to rally makes sense. But the failure of the rally makes sense too.
Who exactly trusts these numbers right now? If you believe the administration's past rhetoric, the data was being rigged to make Trump look bad. If you believe the data quality critics, the numbers are increasingly unreliable for methodological reasons. And if you're of a more cynical persuasion, you might wonder whether an agency whose previous commissioner was fired for producing inconvenient data might now feel some institutional pressure to produce convenient data. I'm not asserting that. I have no evidence of it. But the question is now hanging in the air, and markets hate unanswered questions.
The honest answer to all of this is the one I give regularly, to the occasional annoyance of readers who want more: I don't know what the jobs market is really doing, and neither does anyone else — including, apparently, the Bureau of Labor Statistics. What I do know is that trading on the first release of any monthly report, before revisions, before benchmarks, with a 43% response rate and a plus-or-minus 122,000 confidence interval, is not analysis. It's a coin flip dressed up in decimal points.
Watch the trends. Wait for the revisions. And maybe, just maybe, stop treating Jobs Friday — or in this case, Jobs Wednesday — like it's the word of God.
