I thought that Julien Berman’s Washington Post column on the “bleak economic picture emerging from the jobs numbers” was pretty good:
Despite constantly shifting tariff rates that have created enormous uncertainty for businesses, initial readings of the labor market’s health showed it holding strong. Throughout the first half of the year, unemployment remained relatively low and the economy kept adding jobs.
Turns out, this was an illusion.
On Friday, the Labor Department reported that employers added only 73,000 jobs in July — far fewer than the 115,000 forecasters had expected.
More important, the department published sweeping revisions to earlier reports that had made the job market look strong all spring. The estimated number of new jobs in May was lowered to 19,000, from the initial 144,000. June’s numbers fell to just 14,000, from 147,000. Together, these changes amount to an overall decrease in new jobs of almost 90 percent.
What’s worse, three-quarters of the added jobs were in just one sector: health care. Along with the lower numbers, this suggests that America is starting to see the effects of Trump’s tariffs ripple through the economy.
Basically, I think those results were obvious to anyone who had his ear to the ground while the previous results were counterintuitive to say the least. The post includes some nice graphs and charts including one of the number of jobs added monthly since January 2024 and one of the number of jobs added by sector. I had actually started drafting this post before reading this passage which surprised me:
A deeper problem here is that the monthly jobs reports might be getting less reliable. The Labor Department always modifies its estimates as new data comes in, but revisions as large as the one for the spring numbers raise concerns that the government’s statistical infrastructure is starting to buckle. Note that the Bureau of Economic Analysis has lost about 20 percent of its employees since the beginning of the year.
I don’t believe that the explanation he proffers is quite as good as he does. Between 1960 and 1970 the number of employees at the Bureau of Labor Statistics grew by just about 50%, roughly where it’s remained ever since. There is no obvious relationship between the number of BLS staff and BLS data or, indeed, U. S. population since the BLS contracts the actual conducting of surveys out to the Census Bureau which makes sense.
I also agree with him that President Trump acted impetuously in firing the chief of the Bureau:
Even more troubling is that in response to the weak numbers, Trump fired Erika McEntarfer, the commissioner of the Bureau of Labor Statistics, baselessly accusing her of manipulating the jobs report for political ends. His action undermines the independence and credibility of one of the government’s most important statistical agencies, and will cast doubt on the credibility of future data releases on employment, inflation, productivity and other key indicators.
“Partisan” and “political” are not synonymous and I don’t believe the BLS is partisan. The challenge doesn’t reside in finding management who’ll ensure that the BLS produces results the president likes more but in ensuring the the BLS produces better, more reliable data over time and being better suited the task of providing information needed to formulate good, timely policy.
Any organization could be better. The issue is more can they be better at a cost we want to pay. The BLS besides having its own experts had long had outside people from many specialties (business, economics and other academic fields like math/stats) giving them advice. The aggregated data they generate is all available online and the raw data is pretty easily, per a number of academics who write on the topic, obtainable. The 3 obvious ways to increase accuracy would be to have a larger sample taken and finding some way to make that larger sample more accurately represent the general population. They already put a lot of effort into that. Third would be to find some way to get businesses to report sooner.
I cant tell if they need more workers though it does seem odd that you would claim to want better results while firing 20% of the staff. I would also note that the BLS collects and analyzes lot of other numbers including CPI numbers, prices, productivity, worker safety, PPI numbers, occupational illness/injury numbers and probably some i am forgetting. Multiple sources have also reported that businesses are taking longer to report since the pandemic.
All that said, if people want to claim political bias they ought to show some numbers.
Steve
Tim Taylor covered this in more detail. Spending on statistical gathering has been decreasing for a while with Trump accelerating it. I didnt realize until reading his piece that the cuts in funding to Education were resulting in the drop of the age 17 NAEP test. We can stop complaining that our high school kids arent learning anything because we wont know. Problem solved. Trump is a genius.
https://conversableeconomist.com/2025/08/01/federal-statistics-and-false-economies/
Steve
Menzie provides some numbers. The large adjustment was not due to the seasonal adjustment as the seasonally adjusted numbers are about the same as the non-adjusted numbers. Also of note, the revised numbers more closely follow the ADP numbers.
https://econbrowser.com/archives/2025/08/is-it-seasonal-adjustment