Opinion: A strong jobs report, built on a shakier foundation than it looks

The Bureau of Labor Statistics delivered an unambiguous surprise on Friday morning. October nonfarm payrolls rose by 165,000, more than double the roughly 65,000 that forecasters had penciled in, while the unemployment rate eased to 4.1 percent from 4.2 percent. Hiring was broad enough to look real: professional and business services added 70,000 jobs, health care 42,000, transportation and warehousing 25,000. Just as striking, the agency revised August's gain up to 186,000 from 133,000 and September's up to 60,000 from 29,000, adding a combined 84,000 jobs to a labor market that two months ago looked considerably softer than it does today.

Markets took the number as license to stop worrying about a stalling economy, and on the surface that reaction is reasonable. Job growth over the past three months has averaged 137,000, up from roughly 28,000 in the three months before that, according to an analysis of the release by the economics commentary site Hubbard O'Brien Economics. But the more useful question raised by this report is not whether the labor market is strong this month. It is how much confidence anyone should place in a single monthly reading, however good it looks, when the statistical machinery producing it has been quietly weakening for years.

A number that argues with itself

Start with an oddity inside Friday's release that got less attention than the headline. The payroll count, drawn from a survey of business establishments, has added 861,000 jobs since January. The household survey, which produces the unemployment rate, tells a different story: it shows a net loss of 377,000 jobs over the same stretch, even after October's household employment rose by 463,000. Both of these figures come from the same federal statistical agency, published the same morning, describing the same economy. The gap between them is not new — the two surveys always diverge somewhat, because they count different things in different ways — but a divergence this wide is exactly the sort of signal that should make anyone cautious about reading too much certainty into one number.

That caution is not a political complaint about whether the economy is secretly better or worse than advertised. It is a measurement problem, and the agency responsible for the data has said so itself, in writing, to Congress.

A 1981-sized sample for a 2026-sized country

In April, the Bureau of Labor Statistics sent lawmakers a report on modernizing the Current Population Survey, the household poll behind the unemployment rate, and its numbers are sobering. The survey's sample of roughly 60,000 eligible households is essentially unchanged from 1981, even though the population it is meant to represent has grown 61 percent since then. One survey respondent today effectively stands in for about 3,500 people, up from roughly 2,100 two decades ago. Response rates, meanwhile, have slid from the low-90-percent range to the upper 60s over the past fifteen years, as fewer households answer the phone or the door for government interviewers.

The consequence, in the bureau's own words, is that the margin for error has narrowed even as the stakes of reading it correctly have not:

A net change of fewer than 50 survey responses could move the headline unemployment rate by 0.1 percentage point.

The same report notes that a month-over-month shift in the unemployment rate of 0.18 points used to be statistically meaningful on its own. Today it takes two consecutive months of data pointing the same direction before statisticians can say the same thing with confidence. That is a quiet but significant erosion in what a single jobs report can actually tell policymakers, employers and households making real-time decisions.

The counterargument: the system still mostly works

None of this means the October numbers are fiction, or that the bureau is failing at its job. A Government Accountability Office review released in June found that the users of the jobs report — forecasters, central bankers, business economists — generally say the data still meets their needs, and that the bureau met its own internal targets for precision and the size of revisions from fiscal years 2020 through 2025. Large revisions, the GAO notes, are also a sign of a system doing what it is supposed to do: incorporating late-arriving reports from businesses and recalculated seasonal adjustments to replace an early estimate with a better one. October's upward revisions to August and September could just as easily be read as the data getting more accurate, not less.

It is also true, as the wire reporting on Friday's release made clear, that most of October's job growth landed in sectors — health care, professional services, logistics — that have been reliable sources of hiring for years, not in the kind of one-off categories that would raise suspicion about the number's durability. A strong report built on broad-based, plausible sectoral gains deserves more weight than one driven by a single volatile category.

The complication is that both things can be true at once. The jobs report can be the best tool available for reading the labor market and still be measurably less precise than it was a generation ago. The GAO's own recommendations acknowledge as much: it urged the bureau to publish a long-overdue assessment of how falling response rates affect the business survey's accuracy, and flagged that the Departments of Labor and Commerce eliminated three advisory committees in 2025 that had given outside statisticians and data users a formal channel to flag problems before they showed up in a jobs report.

• Nonfarm payrolls: +165,000 in October, versus a roughly 65,000 forecast
• Combined upward revisions to August and September: +84,000 jobs
• Household survey employment: net -377,000 since January, versus +861,000 in the payroll survey over the same period
• CPS sample size: roughly 60,000 households, unchanged since 1981, against a 61 percent larger population

Why the next five weeks matter

The practical stakes are not abstract. The next Employment Situation report is not due until December 4, a longer-than-usual gap that leaves businesses, investors and the Federal Reserve's December policy meeting to lean on a single data point that the bureau's own staff says carries a wider error band than it used to. The bureau and the Census Bureau have proposed adding an online response option to the household survey, but the GAO's review notes that funding constraints could push that fix past its planned 2027 rollout. Administrative data — unemployment insurance filings, tax records — have been floated by outside stakeholders as a partial substitute for survey responses the bureau can no longer reliably collect, but no replacement is imminent.

None of that is a reason to distrust Friday's number outright, and it would be a mistake to treat one strong report, or one skeptical column about the survey that produced it, as proof of anything larger than what it is. But it is a reason to hold any single month's reading — good or bad — a little more loosely than headlines encourage, and to treat the quiet argument over how America counts its own workforce as more than a technical footnote. The jobs report is one of the few pieces of economic infrastructure that both political parties, every business forecaster and the Federal Reserve rely on without dispute over its legitimacy. Keeping it reliable is unglamorous, underfunded work, and Friday's reassuring headline is a poor excuse to stop paying attention to it.