By Robert Eisenbeis, Ph.D.
On Friday, the Dow bounced over 750 points, reacting to the May employment report. In truth, it seems that participants never read beyond the first sentence, which stated:
“Total nonfarm payroll employment rose by 25 million in May (private sector employment increased 3.1 million while government employment declined by 585 thousand) and the unemployment rate declined to 13.3%….”
This report was truly a surprise, especially to economists who predicted the unemployment rate to hit 20% and the economy to lose another 8.5 million jobs. That’s a big miss, to say the least.
Digging deeper, there were clearly some other positives in the report besides the headline numbers, especially in the Establishment Survey, which looks at where the jobs were created. On a seasonally adjusted basis, 22% of the jobs (669K) were created in the goods-producing sector, the bulk of which were in construction. But a whopping 78% of the jobs were created in the service sector, with the leisure and hospitality segment, which had been hard hit by the pandemic, adding 1.239 million jobs. Other gains were reported in retail (367K) and in education and health services (427K), while government employment declined by 487K. Additionally, average hours worked increased slightly, as did average weekly earnings.
The picture was not so optimistic when it comes to who was and was not employed. While the overall reported unemployment rate declined from 14.1% to 13.3%, the unemployment rate for black Americans increased from 16.7% to 16.8% and increased for Asian Americans from 14.5% to 15.0%, but declined for Hispanics and Latinos from 18.9% to 17.6%, a rate that exceeds that for black Americans, in particular.
Digging even deeper, our chairman, David Kotok, pointed out some interesting and sobering information at the very bottom of the report about the impact of COVID-19 on employment data collection and possible implications for the confidence intervals surrounding the numbers. In particular, the report notes that, because of the virus, adjustments had to be made that reduced the survey response rates, which were 69% for the establishment survey, below its historical average, and 67% for the household survey, 15 percentage points below its pre-pandemic rate, implying an increase in the confidence intervals. To illustrate, the average reported monthly job creation from the establishment survey from January 2019 through February 2020 was 185.6K, with a 90% confidence interval of plus or minus 110K. In other words, a reported number of 185.6K could be as large as 295.6K or as small as 75.6K. If comparable intervals were applied to the reported 2.5 million new jobs for May, then the actual number could be as large as 4.25 million or as small as 750K.
Similarly, the 95% confidence interval for the reported unemployment rate before the pandemic was plus or minus 0.2 percentage points. The report states, however, that furloughed workers were considered to be unemployed. At the same time, the large number of workers who were classified as employed but absent from work were probably misclassified. If these people had also been considered unemployed, as the furloughed workers were, then the reported unemployment rate would have been 16.3% for May and 19.3 percent for April, close to the 20% that economists had predicted.
The bottom line is that digging deeper reveals that the employment situation may not be as rosy as the headline numbers suggest, due to reporting and classification errors that reduce the confidence one can have in the reported numbers. In other words, it is risky to put too much weight on a single data point when the confidence intervals are large (and likely to be even larger than history has suggested) when we are in unsettled times.
Editor’s Note: The summary bullets for this article were chosen by Seeking Alpha editors.