Expert Commentary

The BLS error that’s made unemployment look lower than it really is for 3 months straight

The unemployment rate in the U.S. has been percentage points higher than officially reported since the coronavirus pandemic shut down the economy. We break down the misclassification error behind the underreporting.

(Anastasiia Chepinska / Unsplash)

On May 8, the Bureau of Labor Statistics reported that the U.S. unemployment rate for April stood at 14.7%. On June 5, the BLS reported an unexpected improvement in the unemployment rate for May, to 13.3%. While unemployment remains historically high, some economists had predicted a spike to 20%.

News outlets questioned how economists could have missed the mark so badly. Surely some prognostications were way off, but those widely reported official unemployment numbers didn’t reflect reality. Tucked away in the May jobs report: detail of a misclassification error making the official unemployment rate lower than it should have been.

In the midst of an unprecedented economic shutdown due to the coronavirus pandemic, survey takers had misclassified some workers as “employed but absent from work,” rather than “unemployed on temporary layoff,” according to the BLS. The actual unemployment rate for April might have been north of 19% — up to 19.5%. The rate for May likely breached 16%. The misclassification happened during the March survey too. Unemployment that month was roughly 5.3%, not 4.4% as the BLS first reported.

Those higher rates represent an “upper bound” — a worst-case scenario in which everyone misclassified really was on temporary layoff and unemployed — BLS Commissioner William Beach explains in a June 29 blog post. Last month, the BLS said that along with the Census Bureau they were “investigating why this misclassification error continues to occur and are making changes for the June collection.”

“Before the June collection, the Census Bureau provided more training to review the guidance to the interviewers,” Beach writes. “The interviewers also received extra training aids.”

Journalist’s Resource asked the BLS what their investigation has uncovered. We’ll update this article if we hear back before July 2, when the June jobs report is due for release. Beach’s blog post doesn’t say whether the error would affect data for June.

Last month’s jobs report had a section several pages deep on how the coronavirus pandemic affected calculations. That’ll be where to look to see if the error persisted in June. If the error continues it will mean four straight months of misclassified national unemployment data. The error has gotten some news coverage, but if it’s still an issue for June it will be important for news outlets to explain the error alongside the official unemployment figure, so readers and viewers have a truer sense of the scope of job loss in the U.S. during the ongoing coronavirus recession.

So what happened?

Normally, a respondent might have told BLS survey takers they had a steady job but still didn’t work at all during the survey period. The BLS would classify this person as employed — maybe they were on vacation or had jury duty, Beach writes.

But these are not normal economic times.

Think of a typical sit-down restaurant. As the U.S. economy shut down in April and May and states banned in-person dining, most of that restaurant’s staff would have lost their jobs, at least temporarily. Chefs and servers and bussers might have expected to return to work when the economy re-opened, but they weren’t working when the BLS conducted surveys for the unemployment rate these past months. Some 4.9 million workers who were on temporary layoff in May should have been counted as unemployed, but instead were counted as employed, according to Beach.

The error is important context to bring to coverage of projections of the health of the U.S. economy. Consider April and May. Estimating that unemployment would reach 20% from 19.5% requires less of a leap than 20% from 14.7%.

The two surveys behind the U.S. unemployment rate

The jobs report is drawn from two national surveys BLS conducts monthly: one of households — called the Current Population Survey — and one of employers called the payroll survey. BLS conducts the household survey jointly with the Census Bureau. It includes 60,000 households survey takers reach through telephone and in-person interviews.

When it comes to calculating unemployment, BLS wants to know the number of people over age 16 in a household who worked at least one hour during the week that includes the 12th day of the month. “If a person worked one hour on the 8th and were laid off the rest of the month, they would still, for the unemployment situation, be considered employed,” former BLS Commissioner Erica Groshen told JR in April.

It’s the same basic idea for the payroll survey, which seeks feedback from 145,000 business and government agencies. Those establishments report anyone on their payroll — including people receiving severance — during the pay period that includes the 12th day of the month. Anyone that the surveyed establishments paid during that period is counted as employed.

For more on the data behind the employment situation, check out our explainer 7 big things you should know about the monthly jobs report from June 2019.

About The Author