UPDATE ON PAYROLL EMPLOYMENT GAINS DATA REVISIONS

The following information was released by the Federal Reserve Bank of San Francisco:

The latest revisions to monthly payroll employment data issued in August by the Bureau of Labor Statistics (BLS) were large, renewing concerns about data reliability. However, our updated analysis shows that these revisions remain similar to other large revisions when compared with historical data over the last 60 years. This suggests that incoming data are not generally subject to greater fluctuationsand thus may not reflect greater uncertaintythan in the past.

Data reliability continues to be questioned

In a March 2025 study (Leduc, Oliveira, and Paulson 2025), we assessed short-term revisions in payroll employment and consumer price index (CPI) inflation data over past decades. We found that these revisions were roughly the same size over the two to three years ending in 2024 as they were in the years preceding the pandemic, on average.

Our findings ease some concerns about the reliability of the data and that monetary policy may become too gradual. During times of high uncertainty, policymakers may put less weight on their forecasts and let the evolution of incoming data have more influence on their actions (see, for example, Daly 2023). If policymakers also perceive the data as more uncertain, they may need more evidence that the economy is moving in a particular direction before adjusting policy, which runs the risk of falling behind the curve.

In the current environment of continuing elevated uncertainty, questions about the reliability of incoming economic data have persisted. For example, in June, The Economist published an article exploring the possibility of U.S. economic data becoming murkier (The Economist 2025). In July, the BLS issued a note addressing data collection reductions around the information underlying CPI data releases (Bureau of Labor Statistics 2025a).

More recently, the discussion resurfaced following the August release of payroll employment data (Bureau of Labor Statistics 2025b). In its release, the BLS says that data revisions for the months of May and June were "larger than normal" and that the gains in total nonfarm payroll employment were revised down by 125,000 jobs and 133,000 jobs, respectively. The joint downward revisions of 258,000 jobs brought the overall gains for each of these two months to fewer than 20,000 jobs. Alongside the reading of 73,000 jobs gained in July, the latest revisions to payroll employment numbers spurred concerns that the labor market has been weaker than previously thought.

Recent revisions to payroll employment gains data in historical context

Following the methodology in our March analysis (Leduc, Oliveira, and Paulson 2025), we assess payroll employment data volatility by looking at data revisions using thePhiladelphia Fed's Real-Time Data Set. This data set reports how data have been revised over time as statistical agencies obtain more information.

Figure 1 shows the average monthly revision to payroll employment gains per year since 1990, using the absolute value of the change between the first release and revised second release of the data. The second release provides a refined estimate for a given month usually one month after the initial release. The figure updates our March analysis to include data on revisions through June 2025.

The figure shows that, on an annual basis, first revisions to payroll employment gains data in 2025 through June are within the historical range. These 2025 revisions have been a bit higher than those in 2023 and 2024 on average, but they are still generally comparable to the pre-pandemic historical average (red dashed line) and many other average annual readings.

To assess how the August revisions have affected the average reading for the year, we next look at month-to-month revisions. Figure 2 breaks down the annual averages into their individual monthly revisions, focusing on the period since January 2020.