Is there something really wrong with the nation’s employment counts?
Questions about the quality of the jobs data came into the spotlight this month when President Donald Trump fired the boss of the Bureau of Labor Statistics after its latest report showed, among numerous disappointing results, a sharp revision to June’s initial jobs tally. Trump insists, without proof, that the bureau produces numbers biased against him.
One flash point in the statistical dust-up was a 133,000 revision downward to what was previously reported about June’s count of seasonally adjusted, nonfarm employment. So, gains were cut to 14,000 from 147,000.
In an instant-analysis era, thin job creation transformed the economic storyline.
The original data suggested the new administration’s unorthodox business agenda was boosting employment. The revisions raised fears that Trump’s new policies are cooling hiring.
And the president’s wild claims of massive inaccuracies aren’t just political theater. Seeding doubts about government data reliability can hurt business and consumer confidence and damage the overall economy.
Minuscule slice
I’m a numbers person, and I know this job market math can be really geeky.
With that warning stated, here’s what my trusty spreadsheet found looking at the BLS’s own tracking of revisions. The historically non-partisan statistical bureau is quite open about its methodology, as the BLS details its revisions back to 1979.
We’ll focus on June 2025. Its revisions wiped out 90% of the previously reported gain. Only 13 months since 1979 had larger downward first revisions.
And look back over the last 47 years. The typical month’s first revision, up or down, reflected an average swing of 41,000 job changes. That’s roughly one-third of the average 125,000 jobs added per month since 1979.
So, June was definitely an eye-catcher.
Yet try another lens to view these changes. Think about revisions as a slice of all U.S. workers who have an employer and are not farmers. By the way, the self-employed are in this tally.
June’s 133,000 downward revision is not even 0.1% of the month’s 160 million total U.S. jobs. And that share is on par with average results back to 1979.
Revisions are statistical refinements equaling a minuscule slice of the overall job market. These adjustments only seem large when looking at the one-month swing in employment.
So, you could argue the first draft numbers are 99.9% correct.
It’s one of many reasons economists of all political stripes have said in recent days that the U.S. is the gold standard of jobs data. They’ve all decried the BLS chief’s firing.
Additionally, let’s admit we all have a problem with economic data. Worrying about a single month’s gyrations of any business variable doesn’t give anyone a solid idea of what’s going on – good, bad, or indifferent.
Quick work
Revisions are not admissions of error. They’re part of the job tracking process.
Think about it. Ever wonder how the national employment count is released days after the end of a month?
These numbers are based on a survey of roughly 600,000 U.S. worksites. One wrinkle: not all bosses respond quickly. Lower and slower response rates are an issue across all polling, regardless of who conducts it or the topic being surveyed.
Please understand that the first-Friday-of-the-month jobs report we all chat about reflects only what the BLS gathers in roughly the first 12 days of the previous month. Typically, the initial study covers roughly three-quarters of the employers in the survey.
So the BLS’s first estimate of June employment – released July 3 – was simply quick math.
On Aug. 1, BLS gave us a second peek at June, with more survey results coming in. That means it was a better snapshot of the employment picture. And the trend wasn’t pretty.
By the way, there will be at least two more revisions to June’s numbers.
The first will be in the following monthly jobs report, scheduled for Sept. 5. Since 1979, the second revisions generate an average of 31,000 changes, up or down. And it’s down 36% of the time.
Trump’s BLS critique was also fueled by May’s second revision issues at the same time as une’s first one. It saw a 125,000 downgrade of the job count, the fifth largest since 1979.
June’s next revision is scheduled for February 2026, when the BLS conducts its annual “benchmarking,” which involves retooling the employers’ survey with the help of a study of plus unemployment insurance records covering 95% of all jobs.
Want to see what those other unemployment-insurance-based stats say about the second quarter’s job market, which includes May and June’s debate-provoking data? That report comes out in December.
Basically, tracking the job market – at least how the federal government has done it for decades – involves making swift estimates followed by improving the data as new information flows in.
Where’s the beef?
Does Trump have a legitimate gripe that job counters aren’t fair to him?
Look at all 53 months he’s been in office, through his two terms.
There have been 26 downward first-month revisions in job tallies. That’s 49% of the time. Those revisions have changed the job counts – up or down – on average by 40,200 jobs. The net result of the revisions has been to add an average of 100 jobs to the total employment.
Compare that track record to the 557 months of first-month revisions since 1979: 267 downward revisions or 48%. Average swing was 41,000 with an average net addition of 1,800.
It’s hard to see much bias.
Jonathan Lansner is the business columnist for the Southern California News Group. He can be reached at jlansner@scng.com