You guys see the huge revision in the March employment estimate? US payrolls reported in March (number of people who were employed during March) revised down 911k - a record revision. I’ll try to make sense of it here:
“The preliminary estimate of the Current Employment Statistics (CES) national benchmark revision to total nonfarm employment for March 2025”: 911k, is 0.6% of the original estimate; the average error (or revision) has an absolute value of 0.2%. March 2025 surveys reported 158,506,000 (establishments) and 159,398,000 (households) people employed; with their black magic, the BLS base estimate was 158,487,000.
Each month, the BLS conducts surveys of households and businesses to estimate the number of people employed. The number released today (-911k) is the first revision to the employment estimate for the calendar year (March 2025). Most data for revisions are based on reports by Federal-State unemployment insurance programs; revisions are also based on annual surveys that collect results from one third of all businesses on a rotating basis - survey a particular business every three years.
Revisions are primarily attributed to the difference between monthly survey data and filings for unemployment insurance. If people are employed, but not reported via unemployment insurance filings, we should expect an overestimation of the initial employment estimate and a negative revision; undocumented workers reported on surveys for example. Revisions are also impacted by allowances for “nonresponse errors”.
https://www.bls.gov/news.release/archives/empsit_04042025.htm