I see a lot of ebullience in the wake of the March headline NFP number. I’m not so convinced (see here), and looking at other indicators buttresses my wariness.

Figure 1: Nonfarm payroll employment (bold black), private NFP (blue), private NFP from ADP (red), QCEW covered employment, seasonally adjusted by author using Census X-13 (green), and civilian employment, 3 month centered moving average, smoothed population controls (brown), all in logs, 2025M03=0. Source: BLS, ADP via FRED, BLS, and author’s calculations. 

Covered QCEW is hard to interpret to the extent I applied a seasonal adjustment to the reported aggregate series, whereas a more careful approach would apply the seasonal adjustment to individual components, and then sum up to aggregate. In addition, since QCEW utilizes tax records presumably for documented workers, it might undercount actual workers.

With respect to the household series, the civilian employment series is relatively noisy because of its relatively small sample size; I try to mitigate that issue by taking a three month centered moving average. However, the more challenging issue — not so readily remedied — is the sensitivity to the estimates of population controls. In an era where a tremendous amount of fear has been engendered by the internment/removals program of the Administration, the household survey may be even more affected by measurement error (both unbiased and biased).