US Nonfarm Data "Squeezed of Moisture": Annual Revision Cuts 911,000 Jobs!

On September 9 local time, preliminary results of the annual benchmark revision released by the U.S. Bureau of Labor Statistics (BLS) showed that in the 12 months ending this March, the number of new nonfarm jobs in the U.S. was 911,000 less than previously estimated, equivalent to an average reduction of 76,000 jobs per month. This indicates that previous job growth was significantly "overestimated," and the actual expansion pace was much slower than initially reported.

Pre-revision non-seasonally adjusted data showed that in the 12 months ending in March, U.S. employers added nearly 1.8 million jobs in total, with an average monthly increase of 149,000. The latest revision nearly "halved" this number. The BLS stated that this revision is based on unemployment insurance contribution data, which has broader coverage and better reflects the true employment situation.

This is the largest annual downward revision since 2009, highlighting that the U.S. labor market had significantly cooled well before this spring.

James Knightley, Chief International Economist at ING, said, "The revision suggests that the loss of employment momentum occurred earlier and was more fragile than previously thought."

Samuel Tombs, an economist at Pantheon Macroeconomics, described the revision as "massive" and attributed it to the BLS's "birth-death model" overestimating the net job additions from new businesses.

In terms of statistical methodology, the BLS's monthly nonfarm employment data is primarily based on business sampling surveys, while the annual revision incorporates unemployment insurance tax data submitted by employers across states, which is more comprehensive but released with a lag. Experts believe this discrepancy explains why such a significant revision occurred in the short term and reflects how the pace of business openings and closures after the pandemic has increased the difficulty of model predictions.