Debate continues over whether the business survey mismeasured job growth for the April-June 2022 period, as the Federal Reserve Bank of Philadelphia The asserted net job growth was only 10,000, while the BLS reported about 1 million. The results of their calculations are shown in the figure below:
From what I understand, analysts drill down to the state level, make adjustments to match QCEW levels to NFP, make seasonal adjustments, and then aggregate to the national level using state level data.How well do my programs with them work over time (explain here), but this approach seems reasonable.
In contrast, I compared the high-level aggregated series to see if the NFP series reported by CES behaved anomalously relative to QCEW. Since the latter is non-seasonally adjusted, one would want to compare it with the non-seasonally adjusted NFP. Shown below are percentage comparisons (January 2022 gap of 0.36 percentage points, log) and absolute value (000’s) comparisons (gap of 581,000).
figure 1: U.S. nonfarm payrolls (blue) and total employment (red), unseasonally adjusted, log, 2022M03=0. Source: BLS and athor’s calculations.
figure 2: U.S. nonfarm payrolls (blue) and total employment (red), not seasonally adjusted, relative to March 2022, expressed in 000. Source: US Bureau of Labor Statistics, calculations.
If one argues that the QCEW is more accurate and nonfarm payrolls are inaccurate, one would want to see if the relationship between them has changed. In fact, with the exception of January 22, 2006, reported employment was below average for the past year.
image 3: Log ratio of nsa NFP to QCEW (blue) and 2001-19 mean (red dashed line). Dates of peak-to-trough recessions as defined by NBER are shaded in gray. Source: BLS via FRED, Bureau of Labor Statistics (as of December 18, 2022), NBER and author’s calculations.
As I mentioned earlier, for private NFPs, we have independent verification of trends through the ADP survey, which shows much higher job creation than the BLS series.
