When I want to know what those crazy people are thinking, I go to Zerohedge. On the Bureau of Labor Statistics employment release, I saw zerohedge [1] It is claimed that the number 336K is the result of “pursuing upward goals”. “Target search”, I have to look up, “…is the process of working backwards to find the right approach, knowing only the desired outcome.”
336K yes Well above the 170K consensus reported by Bloomberg. Zerohedge asserts that this is six sigma beats. However, Bureau of Labor Statistics Write that one standard deviation of employment growth is 80K; thus I get 2.1 standard deviation beats – still large, but not 6 sigma. In percentage terms (I think it is more appropriate to take into account the NFP growth trend), it is 1 sigma, considering the monthly change of 0.21%, the implied growth using consensus is about 0.1% (August NFP level was 170,000), and The BLS reports a relative standard error of 0.2%, with reported numbers well within +/- 1 standard deviation.
The rant in the article goes like this:
Let’s start with the household survey: New employment here was just 86,000, the lowest since May and the second-lowest number of 2017, not nearly as much as 336,000 jobs added (as was far less Accurate institutional investigation reports). 2023!
I don’t know where the author got the fact that the family series is more accurate than the institution. This is demonstrated through assertions.
Is it a seasonal estimation error? From Zero Hedge:
What about agency surveys? Well, it stinks here too. Yes, the overall data surge is huge, but the question is how much of it is purely seasonal.
Consider what Vanda Research currency trader Viraj Patel pointed out earlier: Official adjusted data showed the leisure and hospitality industry added 96,000 jobs. But unadjusted data showed the industry lost 466,000 jobs in September. This means unadjusted private sector employment is -399!
Seasonality is a concern. Estimating seasonality is challenging after a large hit to employment. However, it can be helpful to look at seasonally adjusted and unadjusted numbers.
figure 1: Nonfarm employment, seasonally adjusted (blue), nonfarm employment not seasonally adjusted (tan), and total employment covered by the Quarterly Census of Employment and Wages, not seasonally adjusted (green), All are on a logarithmic scale, with units of 000. Source: BLS, BLS-QCEW.
Note that September’s adjusted numbers are just unadjusted numbers, just like 2022 and 2021… Another way to assess the seasonal impact is to look at how the 12-month growth rate evolves. as shown in picture 2.
figure 2: Seasonally adjusted annual growth in nonfarm payroll employment (blue), not seasonally adjusted (tan), and employment covered by the Quarterly Census of Employment and Wages, not seasonally adjusted (green). Source: BLS, BLS-QCEW, and author’s calculations.
What about the example of Zerohedge in the leisure and hospitality industry? Adjusted employment increased by 99,000, but unadjusted employment fell by 466,000? Well, if 2023 is rigged, then it will happen in 2022, 2022, 2021…
image 3: Seasonally adjusted log ratio of leisure and hospitality employment to unseasonally adjusted (blue). Numbers represent logarithmic ratios in September of each year. NBER-defined recession peak-to-trough dates are in gray. Source: BLS, NBER, and author’s calculations.
This is just another example of Zerohedge’s data paranoia (see earlier discuss another example).
