Post-benchmark revision, here are changes in the BLS and ADP series:
Figure 1: Change in BLS private nonfarm payroll employment (bold black), +/- 2 mean absolute revision (gray +), and change in ADP private nonfarm payroll employment (teal), both in 000’s, s.a. MAR is at 50K, using total NFP for 2023-24 figures. Source: BLS and ADP via FRED, BLS, and author’s calculations.
The mean absolute revision for total nonfarm payroll employment going from first release to third release in 2023-24 is about 50,000. Using this figure, I can put an approximately 95% band on changes. The BLS series looks substantially higher than the corresponding ADP series (the standard deviation of the difference between changes in the two series is about 90K, mean error about zero).
In terms of cumulative change (in percent) since 2025M01, this is the picture.
Figure 2: Cumulative change in BLS private nonfarm payroll employment (bold black), +/- 2 mean absolute revision (gray +), and change in ADP private nonfarm payroll employment (teal), all in logs, 2025M01=0, s.a. MAR is at 50K, using total NFP for 2023-24 figures. Source: BLS and ADP via FRED, BLS, and author’s calculations.
Looks like divergences of this size are pretty common:
https://fred.stlouisfed.org/graph/?g=1RXBO
The shutdown and general turmoil within BLS raise questions about consisitency, but this is just one month’s data, and not crazily out of whack. If it keeps happening, there’ll be reason to question whether BLS is up to the job with current staffing.
Off topic – From the NYT:
“In Xi’s Purge of the Military, a Search for Absolute Loyalty”
“By reaching back to Maoist tactics of “rectification,” the Chinese leader is signaling that control over the gun requires a state of perpetual cleansing.”
https://www.nytimes.com/2026/02/16/world/asia/xi-military-zhang-youxia-mao.html?campaign_id=9&emc=edit_nn_20260216&instance_id=171166&nl=the-morning®i_id=97269976&segment_id=215349&user_id=c892abce8b05dc7ce4bc23d18c80e549
Does purging the top ranks of the military in an effort to assure personal loyalty sound like anyone we know?
As an investor non-economist, it puzzles me why BLS has to rely on surveys for the jobs data. Can’t payroll data be pulled monthly from the Treasury Dept and other Federal agencies? For that matter, can’t private payroll companies be required to submit aggregated monthly data?
JohnL: ADP has proprietary access to all the firms for which it processes payments. And yet this is an unweighted subsample of all firms in the economy. ADP in its estimates has to then weight, and it does so using QCEW (i.e., BLS) data. In principle, the government has a lot more data via tax and unemployment insurance data, via QCEW. As you can see, the QCEW data lags months behind the CES (survey) data in terms of reporting. That is the tradedoff in terms of timeliness vs. comprehensiveness.