September 9, 2025 – Today’s preliminary benchmark announcement from the Bureau of Labor Statistics (BLS) reveals weaker job growth between March 2024 and March 2025 than when it was first reported based on survey data. These numbers are likely to anger President Trump and the White House who incorrectly view revised data as political manipulation. Trump has already lashed out at BLS, including firing the agency’s commissioner because a jobs report showed a rapidly weakening labor market. But these BLS data revisions are not corrections of mistakes. Revisions are part of the regular, transparent process to update employment counts with the most comprehensive data possible.
In today’s release, BLS provided preliminary estimates—during which no data will actually be revised—as a window into its eventual benchmark revisions that will be implemented in the beginning of 2026. According to the data, average monthly job growth between March 2024 and March 2025 may have been only half the pace that was initially estimated, about 70,600 jobs per month rather than 146,500. These preliminary estimates are consistent with other signs of slowing job growth in late 2024 and the beginning of 2025. The bulk of these revisions reflect 2024 data—in fact, despite the predictable angst they will generate from the White House, today’s revisions tell us very little about the state of Trump’s economy since he wasn’t president in 2024.
Instead, the preliminary benchmark revisions released this morning are simply part of regular BLS communication regarding the best available employment data. Monthly payroll employment estimates are based on a large sample with a fast turnaround; data are regularly updated for two subsequent months as new survey results come in, and then the data are revised again annually in February to reflect administrative records, which are comprehensive but less timely.
Any political retaliation due to today’s release will harm the ability for BLS to provide timely and unbiased statistics, either because the Trump administration is intending to undermine data integrity, or because political attacks on the dedicated public servants at BLS limit their ability to collect, process, and release these statistics. The latest economic data—which are wholly unaffected by today’s preliminary revisions—suggest the labor market is weakening for all workers. Job growth has been especially weak since May. Household survey rates also point to falling prime-age Black employment and higher unemployment for U.S.-born workers. Punishing the messenger will only further damage the federal data infrastructure and cloud our ability to understand the state of the economy.
Why does BLS revise employment estimates?
Every month, the BLS reports two critical sources of employment data: monthly employment levels and changes from the Current Employment Statistics (CES), also known as the establishment survey or payroll survey; and unemployment rates and employment-to-population ratios from the Current Population Survey, also known as the household survey. To update CES-based employment estimates with the most accurate data, BLS implements a well-documented and regular set of revisions.
The preliminary benchmark revisions are a preview of possible revisions to the employment counts and monthly job growth estimates from the CES. In February 2026, BLS will release final benchmark revisions to ensure that its data are as accurate and comprehensive as possible and, only at that point, will they revise historical CES data. The CES is only a sample of total employment in the United States, and as part of its regular benchmarking since 1949, the BLS incorporates much more comprehensive data based on unemployment insurance tax filings by employers. These near-universal payroll records from the Quarterly Census of Employment and Wages (QCEW) are more accurate than the CES, but not as timely.
Today, BLS reports that more comprehensive tax records suggest March 2025 employment levels may have been 911,000 lower than the current published value, but it will provide the final benchmark estimate in February 2026 when there are more available data. At that point, the benchmark will be implemented and historical data on payroll employment as well as wages and hours will be updated. If the benchmark adjustments follow recent patterns, the final benchmark revision for payroll employment will be slightly less negative than the preliminary revisions.
The reason for the overestimated job growth by the CES, and hence the negative revisions, is because the CES survey sample is becoming less representative—perhaps because of slower net immigration and slower net business creation in 2024. Through the benchmarking process, BLS corrects the CES sample frame and the survey’s underlying assumptions about business creation and destruction.
These regular and transparent steps by BLS to ensure that its data are as accurate and informative as possible are one of the hallmarks of the federal statistics infrastructure. Any effort to undermine them should be soundly rejected.
The Economic Policy Institute (EPI) is a nonprofit, nonpartisan think tank working for the last 30 years to counter rising inequality, low wages and weak benefits for working people, slower economic growth, unacceptable employment conditions, and a widening racial wage gap. We intentionally center low- and middle-income working families in economic policy discussions at the federal, state, and local levels as we fight for a world where every worker has access to a good job with fair pay, affordable health care, retirement security, and a union. www.epi.org