After two years of big data revisions that shook public confidence in its estimates and drew political scrutiny, the Bureau of Labor Statistics appears to have gotten the recent jobs numbers pretty much right.
There were about 79,000 fewer employees on U.S. payrolls in March than previously believed, the agency said on Friday. That suggests that employers added about 16,000 jobs per month on average in the last three quarters of 2025 and the first quarter of 2026, rather than the 23,000 jobs per month that had been indicated in monthly jobs reports.
The revisions are part of a longstanding annual process known as benchmarking — a procedure the government uses to reconcile monthly estimates, with more accurate but less timely data from state unemployment offices.
The updates used to get relatively little attention, but large downward revisions in 2024 and 2025 led some economists to question the reliability of the monthly estimates, which are based on surveys. Policymakers, including Jerome H. Powell, then the chair of the Federal Reserve, said last year that they believed the monthly reports were consistently overstating the amount of hiring, and that the labor market was weaker than the numbers suggested.
The revisions also became a political problem for the Bureau of Labor Statistics. President Trump cited the large 2024 revision when he fired the head of the agency, Erika McEntarfer, last summer, arguing without evidence that it showed she was biased against him. Economists across the ideological spectrum rejected that claim, noting that the agency made revisions in both directions under presidents of both political parties.
The 2026 revisions, however, are on track to be the smallest since 2021, representing a downward adjustment of just 0.1 percent of the total work force. That could help ease concerns that the government is struggling to measure the economy accurately.
Some of the largest downward revisions this year were in the wholesale and retail trade sectors, where employment was overstated by a combined 240,000 jobs. Employment in construction and in transportation and warehousing was understated, however, and is set to be revised up by nearly 200,000 jobs.
The numbers released on Friday are preliminary. Final figures will be released and incorporated into the government’s official jobs numbers early next year.
Economists aren’t sure why the jobs numbers drifted so far off course in recent years, or why they seem to have gotten better more recently. The agency this year updated its procedure for estimating the number of jobs created by new businesses and lost by those that shut down. That could help explain some of the recent improvement but probably not all of it, economists said.
Perhaps a larger factor is simply that the labor market has settled into a more normal — and easier to measure — period after years of large disruptions. The pandemic and its aftermath led to huge shifts in seasonal hiring patterns, business formation activity and other once-predictable behavior. Big swings in immigration policy during the Biden and Trump administrations further jumbled employment patterns.
Such rapid shifts can upend assumptions that are baked into the statistical models the Bureau of Labor Statistics uses to produce its monthly estimates. The smaller revision announced Friday could suggest that hiring is settling into a more predictable pattern again.
“Part of it maybe is we’re in more of a steady state,” said Guy Berger, a labor economist. “Maybe some of what caused these models to malfunction has stopped.”
The problem, Mr. Berger said, is that problems could re-emerge if hiring patterns shift again.
“We don’t know whether it’s going to work in perpetuity,” he said. “Let’s not get too excited about having solved this problem once and for all.”
