The debate over whether the facility survey is mismeasuring employment development over the April-June 2022 period continues, with the Philadelphia Fed asserting the net task development was only 10 thousand, vs. the approximately 1 million reported by BLS. The outcomes of their computations are provided in the following figure:
Figure 3: Log ratio of nsa NFP to QCEW (blue), and 2001-19 average (red dashed line). NBER defined peak-to-trough economic downturn dates shaded gray. Source: BLS by means of FRED, BLS (since 12/18/2022), NBER, and authors estimations..
As I kept in mind previously, for personal NFP, we have independent confirmation of trends through the ADP study, which reveals much greater task production than the BLS series.
As I comprehend it, the experts go down to the state level, adapt to match QCEW levels to NFP, seasonally change, and then aggregate up to national level utilizing the state level information. I have no great sensation for how well their treatment works over time (explained here), but the approach appears plausible.
In contrast, I have compared the series at a high level of aggregation to see if the NFP series as reported by the CES is behaving in an anomalous way with regard to the QCEW. Since the latter is in not-seasonally adjusted terms, then one would desire to compare against the NFP in not-seasonally changed terms. Revealed below are contrasts in percent terms (gap at 2022M06 is 0.36 ppts, log terms), and in absolute (000s) terms (gap is 581 thousand).
Figure 1: Nonfarm payroll employment (blue), and total covered employment in United States (red), not seasonally changed, all in logs, 2022M03= 0. Source: BLS, and athors calculations
Figure 2: Nonfarm payroll work (blue), and overall covered employment in US (red), not seasonally adjusted, relative to 2022M03, in 000s. Source: BLS, calculations.
If one thought that QCEW was more precise, and nonfarm payroll incorrect, then one would wish to see if the relationship in between them had actually changed. In point of reality, reported work has been lower than average in the previous year, and with the exception of 22M06.
In contrast, I have actually compared the series at a high level of aggregation to see if the NFP series as reported by the CES is acting in an anomalous way with regard to the QCEW. Because the latter is in not-seasonally changed terms, then one would want to compare versus the NFP in not-seasonally changed terms. Shown below are contrasts in percent terms (gap at 2022M06 is 0.36 ppts, log terms), and in absolute (000s) terms (space is 581 thousand).