A week and a half ago, Kevin Drum noted earlier the disjuncture between job creation as measured by the household survey (civilian employment) and establishment survey (nonfarm payroll employment). My conclusion was that the empirical evidence suggested putting most — if not all — weight on the establishment survey. Now, Torsten Slok compares the declines in response rates in the two surveys. The comparison further buttresses the case for using the establishment series.
Slok writes:
Survey response rates for employment and inflation have declined significantly during the pandemic, and this is introducing substantial measurement errors and uncertainty, see chart below. One such example is the widening gap between the establishment survey and the household survey in the employment report, with the establishment survey showing 2.6 mn jobs created over the past eight months and the household survey showing no job growth over the same period. …
Here’s the figure (where I’ve added a red line at March 2022, 8 months ago).
Source: T. Slok, 13 December 2022. Red line at March 2022.
Note that the household response rate relative to establishment rate drops precipitously starting around October of last year. The gap yawns more rapidly starting after March. And here’s the picture of the gap between nonfarm payroll employment and civilian employment adjusted to the NFP concept (this is a research series).
Figure 1: Nonfarm payroll employment official (sky blue), nonfarm payroll employment implied by preliminary benchmark (dark blue), and civilian employment adjusted to NFP concept (green), all seasonally adjusted, in 000’s. Figures in (parentheses) are net new jobs in millions, since 2022M03 (indicated by red dashed line). Source: BLS via FRED, and author’s calculations.
The gap in numbers widens exactly at the time the household response rate drops.
So: (1) I put more weight on the establishment series, and (2) the gap between the two series is more likely due to increasing, and biased, measurement error in the household series, rather than, for instance, primarily increases in multiple-job holders.
