The misclassification error in the May 2020 jobs report

on

The unemployment rate for May is higher than the 13.3 percent figure widely reported today.

Not noticed by me (and nearly everyone else) in today’s jobs report from the Bureau of Labor Statistics, part of the United States Department of Labor, is a note disclosing an error.

The error has to do with the way workers are classified: employed, unemployed, or not in the labor force. The upshot is this, as BLS describes: Without the error, the unemployment rate would have been three percentage points higher than the 13.3 percent widely reported today. (The three points is on a not seasonally adjusted basis.) Here’s what BLS noted: 1

If the workers who were recorded as employed but absent from work due to “other reasons” (over and above the number absent for other reasons in a typical May) had been classified as unemployed on temporary layoff, the overall unemployment rate would have been about 3 percentage points higher than reported (on a not seasonally adjusted basis). However, according to usual practice, the data from the household survey are accepted as recorded. To maintain data integrity, no ad hoc actions are taken to reclassify survey responses.

In a separate document, BLS offered a larger explanation: 2

Household survey: What would the unemployment rate be if these misclassified workers were included among the unemployed?

If the workers who were recorded as employed but not at work for the entire survey reference week had been classified as “unemployed on temporary layoff,” the overall unemployment rate would have been higher than reported. This kind of exercise requires some assumptions. For example, first one needs to determine how many workers might be misclassified. There were 5.4 million workers with a job but not at work who were included in the “other reasons” category in May 2020, about 4.9 million higher than the average for May 2016–2019. (While this category contains misclassified workers, not every person in this category was necessarily misclassified. The average for recent May estimates was 549,000 employed people with a job not at work for “other reasons.”)

One assumption might be that these additional 4.9 million workers who were included in the “other reasons” category should have been classified as unemployed on temporary layoff. If these workers were instead considered unemployed on temporary layoff, the number of unemployed people in May (on a not seasonally adjusted basis) would increase by 4.9 million from 20.5 million to 25.4 million. The number of people in the labor force would remain at 158.0 million in May (not seasonally adjusted) as people move from employed to unemployed but stay in the labor force. The resulting unemployment rate for May would be 16.1 percent (not seasonally adjusted), compared with the official estimate of 13.0 percent (not seasonally adjusted). Estimates of people with a job but not at work are not available on a seasonally adjusted basis, so seasonally adjusted data, such as the unemployment rate mentioned in The Employment Situation news release, are not used in this exercise. (Repeating this exercise, but combining the not seasonally adjusted data on additional people with a job but not at work in the “other reasons” category with the seasonally adjusted estimates reported in The Employment Situation news release yields a similar 3.1 percentage point increase in the unemployment rate for May — or 16.4 percent, compared with the official seasonally adjusted rate of 13.3 percent.)

More reporting is in the Washington Post at A ‘misclassification error’ made the May unemployment rate look better than it is. Here’s what happened.


Notes

  1. Bureau of Labor Statistics. The Employment Situation — May 2020. Available at https://www.bls.gov/news.release/archives/empsit_06052020.pdf.
  2. Bureau of Labor Statistics. Frequently asked questions: The impact of the coronavirus (COVID-19) pandemic on The Employment Situation for May 2020. Available at http://www.bls.gov/cps/employment-situation-covid19-faq-may-2020.pdf.

Comments

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.