BLS will release (w/delay) on Wednesday. Until then, we have ADP (which underwent benchmarking). Using ADP data (2022M01-2025M12) to extrapolate the implied preliminary benchmark, we have the following picture:
Figure 1: Implied benchmark revision of private nonfarm payroll employment (bold black), ADP regression based nowcast (blue square), +/- 1 std error band (gray line), Bloomberg consensus of 2/7/2026 (inverted red triangle), all in 000’s, s.a. Source: BLS December release, BLS, ADP January release, and author’s calculations.
The goodness of fit of a regression in first differences deteriorates substantially with the revision in the ADP series (the adj-R2 is about 0.50 vs. 0.58 before; the slope coefficient drops from essentially unity to 0.61.
It’ll be difficult to assess the reliability of the levels predictions because a (final) benchmark revision will be incorporated into the private NFP series, so it may be useful to focus on changes. The implied change from Bloomberg is +70K, while the ADP regression based is +63K (a simple ARIMA(1,1,1) yields +56K).
