The BLS is set to report on Tuesday at 10 am its preliminary estimate of the upcoming benchmark revisions to payroll figures covering the 12-month period ending in March
Q: Why does the BLS do this?
The Labor Department publishes estimates every month on how many jobs employers added in the prior month based on a survey of employers. But more detailed data from state unemployment tax records subsequently become available, and the BLS uses these records to revise its estimates of recent job growth. They provide a preliminary estimate in August or September for the 12-month period ending the previous March, and then they issue a final revision six months later for the same period.
Q: What's expected to happen this time?
Based on quarterly data from these unemployment-tax records, job growth during the 12 months ended March is estimated to have been overstated by anywhere between 500,000 and 1 million. Currently, payroll employment has been measured as +1.758 million over this period, or around 146K jobs per month. A downward revision of 500,000 would lower the 12-month payroll gain to 1.258 million, or around 105,000 jobs per month for this period. A downward revision of 1 million would lower the 12-month payroll gain to 758K, or 63K jobs per month.
No existing monthly data will be updated in today’s release. That will happen in February, when the BLS releases its final benchmark revision to monthly employment from April 2024 to March 2025 based on the QCEW (state unemployment tax records). The BLS also revises data at that time through December 2025 to reflect changes in the birth-death model. Tuesday’s release of the preliminary revision is akin to a “preview” of the changes that will be published in February. The BLS does this every year. It is not new.
Q: Does this change the unemployment rate?
No, that comes from a different survey that won’t be affected by the QCEW revision.
Q: Why have revisions been so large?
The payroll surveys make certain assumptions about how many new businesses are being created and how many old ones are being shuttered, and it can be tricky to get those right. This was true in 2021-22, when the economy was adding more jobs (in part due to new business formations being elevated) and the initial numbers were later revised higher. While some people might point to such revisions as an example of "cooking the books," this lazy analysis betrays a fundamental misunderstanding of how the data is put together given that this is part of a regular process that happens every year. It's evidence that there's a high fidelity to making sure the numbers are as accurate as they can be given better quality data that isn't immediately available. The final benchmark revisions have in recent years pointed to a smaller revision than what was reported in the preliminary estimate. Last year, for example, the preliminary estimate said job growth for the 12-months ended March 2024 should be lower by 818,000 jobs. But in February, payroll employment for this period was revised down by 598,000.
Q: Is this new information for the Fed?
Not really. It doesn't say anything about job creation since March, per se. And even if it confirms that payroll growth has been overstated, Fed officials have been talking—out loud, in public, in speeches and press conferences—about how the administrative data (from the QCEW program) used for the benchmark process could in fact point to such an outcome. As an example, Fed Chair Jerome Powell’s speech at the Jackson Hole conference last month included a footnote that made the following observation: “In early September, the Bureau of Labor Statistics will publish a preliminary estimate of benchmark revisions to the level of nonfarm payrolls as of March 2025, based on data from the Quarterly Census of Employment and Wages. Data available to date suggest that the level of nonfarm payrolls will be revised down materially.”
Q: What is a normal-sized benchmark revision?
According to the BLS, over the last decade, absolute benchmark revisions have had a range of -0.4% to +0.3%, with an average absolute revision of +0.1%. The final revision for last year (of -0.4%) was the largest since 2009 (which was -0.7%). March nonfarm payroll employment is currently reported at 159.3 million, so a revision of the same magnitude as last year’s final revision would be something around -600K.
Q: Why does the BLS do this?
The Labor Department publishes estimates every month on how many jobs employers added in the prior month based on a survey of employers. But more detailed data from state unemployment tax records subsequently become available, and the BLS uses these records to revise its estimates of recent job growth. They provide a preliminary estimate in August or September for the 12-month period ending the previous March, and then they issue a final revision six months later for the same period.
Q: What's expected to happen this time?
Based on quarterly data from these unemployment-tax records, job growth during the 12 months ended March is estimated to have been overstated by anywhere between 500,000 and 1 million. Currently, payroll employment has been measured as +1.758 million over this period, or around 146K jobs per month. A downward revision of 500,000 would lower the 12-month payroll gain to 1.258 million, or around 105,000 jobs per month for this period. A downward revision of 1 million would lower the 12-month payroll gain to 758K, or 63K jobs per month.
No existing monthly data will be updated in today’s release. That will happen in February, when the BLS releases its final benchmark revision to monthly employment from April 2024 to March 2025 based on the QCEW (state unemployment tax records). The BLS also revises data at that time through December 2025 to reflect changes in the birth-death model. Tuesday’s release of the preliminary revision is akin to a “preview” of the changes that will be published in February. The BLS does this every year. It is not new.
Q: Does this change the unemployment rate?
No, that comes from a different survey that won’t be affected by the QCEW revision.
Q: Why have revisions been so large?
The payroll surveys make certain assumptions about how many new businesses are being created and how many old ones are being shuttered, and it can be tricky to get those right. This was true in 2021-22, when the economy was adding more jobs (in part due to new business formations being elevated) and the initial numbers were later revised higher. While some people might point to such revisions as an example of "cooking the books," this lazy analysis betrays a fundamental misunderstanding of how the data is put together given that this is part of a regular process that happens every year. It's evidence that there's a high fidelity to making sure the numbers are as accurate as they can be given better quality data that isn't immediately available. The final benchmark revisions have in recent years pointed to a smaller revision than what was reported in the preliminary estimate. Last year, for example, the preliminary estimate said job growth for the 12-months ended March 2024 should be lower by 818,000 jobs. But in February, payroll employment for this period was revised down by 598,000.
Q: Is this new information for the Fed?
Not really. It doesn't say anything about job creation since March, per se. And even if it confirms that payroll growth has been overstated, Fed officials have been talking—out loud, in public, in speeches and press conferences—about how the administrative data (from the QCEW program) used for the benchmark process could in fact point to such an outcome. As an example, Fed Chair Jerome Powell’s speech at the Jackson Hole conference last month included a footnote that made the following observation: “In early September, the Bureau of Labor Statistics will publish a preliminary estimate of benchmark revisions to the level of nonfarm payrolls as of March 2025, based on data from the Quarterly Census of Employment and Wages. Data available to date suggest that the level of nonfarm payrolls will be revised down materially.”
Q: What is a normal-sized benchmark revision?
According to the BLS, over the last decade, absolute benchmark revisions have had a range of -0.4% to +0.3%, with an average absolute revision of +0.1%. The final revision for last year (of -0.4%) was the largest since 2009 (which was -0.7%). March nonfarm payroll employment is currently reported at 159.3 million, so a revision of the same magnitude as last year’s final revision would be something around -600K.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
