The Bureau of Labor Statistics (BLS) released a stunning revision on September 9, 2025: the U.S. economy added 911,000 fewer jobs than initially reported in the 12 months ending March 2025. This marks the largest downward revision on record, dating back to 2000.
The BLS attributed the error to business reporting issues and other discrepancies. While revisions are part of the agency’s normal process, this magnitude is unusual. The final numbers will be confirmed early next year once the BLS finalizes its comparison of monthly employer surveys against state tax records.
The political fallout was swift. White House press secretary Karoline Leavitt declared,
“The BLS is broken. This is exactly why we need new leadership to restore trust and confidence in the BLS’s data.”
Critics warn that framing technical revisions as political failures risks undermining confidence in one of the country’s most important economic institutions.
Markets weren’t blindsided. Goldman Sachs economists predicted a revision in the 550,000–950,000 range, and Treasury Secretary Scott Bessent suggested as much as 800,000 in advance. Although the revisions are significant, the markets had anticipated it. This suggests that the data collection process is a concern, rather than the political intentions of the BLS.
Why it Matters
The U.S. labor market has been remarkably resilient, but the revision shows that the “hard data” is beginning to align with the “soft data.” Soft data, which reflects people's actual feelings about the economy, has indicated that the economy was far worse than what the hard data reported. For months, sentiment has suggested more weakness than official numbers revealed. This adjustment closes that gap.
For the Federal Reserve, it’s another data point as policymakers weigh when to cut interest rates. If job creation is weaker than believed, the case for easing becomes stronger.
As I told Marketplace’s Kai Ryssdal this week, the most significant risk is erosion of trust in the data itself. Once the public doubts the numbers, it gets harder to anchor expectations, guide policy, and maintain confidence in the economy.
The Bottom Line
Revisions like this remind us that the economy is never as precise as the headlines suggest. Data gets revised, models get updated, and perceptions often run ahead of the numbers. However, the signal here is clear: the labor market is weaker than we thought, as both hard and soft data are aligning to signal a weaker economy, and trust in the data that guides our economy is under pressure.
What do you think? Does this revision change how you see the strength of the labor market—or is it just business as usual for economic data? Leave a comment and share your thoughts.
On Marketplace
from had an excellent segment on Marketplace Radio about this topic. The segment, by Sofia Terenzio, is titled Revisions in jobs data were more than expected. Marketplace host talks with Kathryn Anne Edwards, labor economist and host of "Optimist Economy," about the annual BLS release of preliminary revisions for a full employment count.On the same episode, the segment "Trust is a concern" says economics professor on BLS data confidence by Sean McHenry aired. Marketplace host
talks to Abdullah Al-Bahrani, associate dean at Northern Kentucky University, about how he utilizes data for lesson planning and what happens when that data becomes unreliable or disappears.
OpenAI GPT 4.o Mini says: Recent significant revisions in BLS data are typically caused by factors such as:
Delayed or Updated Data Submissions: When sources provide additional or corrected information after initial release, revisions become necessary to reflect the more accurate data.
Methodological Changes: Updates in survey designs, sampling techniques, or seasonal adjustment methods can lead to substantial revisions.
Benchmarking Processes: Revisions often follow periodic updates to benchmark datasets, which realign earlier estimates with more comprehensive data.
Introduction of New Data Sources: Incorporating newer administrative data or survey results can lead to revisions of historical figures.
Errors or Approximations in Preliminary Data: Initial estimates may rely on incomplete or provisional data, requiring significant revisions once full data is available.
How to potentially reduce or avoid large revisions in the future:
Enhanced Data Collection and Validation: Improving the timeliness and accuracy of data collection reduces reliance on later corrections.
Faster, More Comprehensive Surveys: Increasing the frequency and scope of surveys can reduce data gaps that lead to revisions.
Implementing Real-Time Data Monitoring: Utilizing advanced analytics and administrative records can provide near real-time updates that minimize revisions.
Clear Communication: Transparently communicating provisional estimates and revision policies can help manage expectations and interpretations.
NBC says: https://www.nbcnews.com/business/business-news/trump-administration-ramps-pressure-labor-department-data-probe-rcna230342