It Turns Out I Was Right #1: Seasonal Adjustements Will Likely Exaggerate Job Growth
In an earlier blog post of mine, I predicted that the May and June jobs report would overstate job performance. My argument, in short, was that normally when summer starts, levels of employment normally fall for two reasons: 1) people in education lose jobs, 2) a lot of people (especially women) have to stay home and look after kids so they can’t work. Seasonal adjustments normally account for that so this disruption does not factor into seasonally adjusted data normally. I argued that because Covid has depressed seasonal employment in education and forces people to stay home from work anyway. Hence, the effect of summer starting won’t be as bad for employment as usual. Therefore, seasonal adjustments will over-account for the start of summer this year causing the employment reports to overstate job gains.
The May jobs report has just come out today and it shows I was right.
First, in seasonally adjusted terms we had very strong employment growth in education. The chart below shows how seasonally adjusted education employment dipped from August to September when seasonal adjustments overstated the effect of school opening up and how now from April to May education employment is dipping up.
A similar trend can be observed in the chart below which plots the level of female employment in seasonally adjusted terms divided by overall employment in seasonally adjusted terms. The fact that it is increasing this month is a sign that the end of summer has not depressed relative female employment as much as normal and the seasonal adjustments have over accounted for the end of summer.
Another piece of evidence that the job growth figures were overstated this month is to look through May’s report and compare two figures. One figure is the % growth of the seasonally adjusted overall employment level. The other figure is the % growth of the seasonally adjusted employment level for men in all industries except one designated “Education and Health Services” (the BLS does not publish the population of workers in education alone by gender, so I am selecting out the larger category).
One finds that the overall employment growth was 0.29% from April to May according to seasonally adjusted figures. If you drop out categories which would be especially impacted by seasonal adjustments, though, one finds an employment growth of only 0.04% from April to May.
This reduced figure does not tell us the “true” job growth rate as the employment levels in other industries may be impacted by seasonal adjustments (imagine employment levels in other industries go up more when summer starts because people leave education to find temporary work in other fields). Additionally, the male level of employment could still be impacted by school closing for summer because 1) they have less female job-market competitors (which could increase male employment during the summer) or, alternatively 2) men as well might drop out of the labor force to look after children. Hence, every category of the labor market is at least slightly impacted by the current error in seasonal adjustment. However, the 0.04% monthly growth rate is probably more accurate than the 0.29% rate.
Overall, May’s jobs report was considered to show disappointingly weak job growth. The true job growth is probably even weaker.