The Many New Indicators of an Epic Job Collapse
(Bloomberg Opinion) -- The job market is deteriorating rapidly, but no one knows by how much because government statistics are produced with a lag. Fortunately, several economists have created measures to light the darkness. While they disagree on some specifics, their overall conclusion is the same: The U.S. is experiencing an economic catastrophe.
The best “real time” government measure available for this unique circumstance is a weekly count of initial claims for unemployment insurance. The latest data were released this morning, and confirm that the labor market is deteriorating rapidly — 3.8 million workers filed for first-time unemployment benefits last week.
But data on most key U.S. labor market indicators is not as timely. The most recent data available for the unemployment rate and the workforce participation rate come from a Bureau of Labor Statistics survey of households about their activities during the week of March 8. The next snapshot will arrive on May 8, and will describe job-market activity in mid-April.
During normal times — or even during a normal recession — this regime of monthly updates that describe the labor market as it was a few weeks past is more than adequate. But the rapid rate of decline during the pandemic means we need something closer to real-time updates to understand economic conditions.
Economists Lisa B. Kahn, Fabian Lange and David Wiczer analyze data from Burning Glass Technologies, which collects information on job vacancies that are available online. They find a 30% collapse in vacancy postings, reflecting a massive decline in companies’ demand for workers.
Employers in all states pulled back, regardless of how hard they have been hit by the pandemic or how early they imposed lockdown orders. Labor demand held firm for nurses over this period, and increased significantly for essential retail stores like groceries and pharmacies. (Though not discussed in their paper, press reports suggest demand for delivery drivers and at distribution centers is increasing, as well.)
According to the researchers’ data, demand is in retreat in all other industries. Surprisingly, perhaps, occupations that lend themselves to telework saw a larger drop-off in vacancy postings than those that don’t.
Economists Alexander Bick and Adam Blandin attempt to fill in the picture further with a simple idea: Ask the same questions about workforce status that official government surveys ask, but ask them more often. Their “Real Time Population Survey (RPS)” will take place every other week, and has been administered twice.
Three key job-market indicators are the employment rate, defined as the share of adults with a job; the unemployment rate, which includes people who don’t have a job but are actively looking for one; and the combination of the two – the fraction of adults who are currently employed or who are unemployed but searching for jobs. What does RPS tell us about these measures?
The most recent government data are for mid-March, and report a 72.7% employment rate for 18- to 64-year-olds. As of mid-April, according to RPS, this rate dropped to 55.8%. That stunning 23% decline marks the lowest employment rate for the U.S. since 1962.
According to the government, 4.5% of workers ages 18-64 were unemployed the week of March 8. Three weeks later, RPS reported a Depression-level 20.2% unemployment rate for this group of workers. By the week of April 12, the survey found that this unemployment rate had dropped to 16.2%.
The RPS survey shows 76.1% of people ages 18-64 were participating in the workforce — they were either employed, or they were unemployed and looking for a job — during the week of March 29, the same rate as the government reported for the week of March 8. This suggests that the entire drop in employment during that period was picked up by an increase in unemployment — in other words, workers who got laid off didn’t leave the workforce entirely.
But between the weeks of March 29 and April 12, millions of people left the workforce, according to RPS. The labor force participation rate dropped by 12%, to 66.6%.
RPS finds that employment declined more sharply for women than for men from mid-March to mid-April. People ages 50-64 have seen their employment rate fall by 30%, while the decrease for those ages 18-29 was 11%.
Across all age categories, the pandemic is hitting every type of worker, but the less educated are bearing the brunt. Employment among college graduates has fallen by 18%. People without college degrees have had their employment rate fall by 30%. The Kahn, Lange and Wiczer results suggest that white-collar employment declines could accelerate if fewer vacancies translates into additional employment cuts.
The government’s weekly count of initial claims for unemployment insurance released this morning highlights the speed of decline. In the six-week period ending last week, over 30 million workers filed for unemployment benefits for the first time. To put that in context, the worst six-week period during the Great Recession ended on March 28, 2009, with 3.9 million initial benefit claims.
A survey by Olivier Coibion, Yuriy Gorodnichenko and Michael Weber finds millions more job losses than the government’s data on new unemployment claims would suggest. This is unsurprising, as the count of unemployment benefits reflects workers who successfully had their claims processed. State unemployment offices have been flooded with requests, and many can’t keep up with the historic volume.
Coibion, Gorodnichenko and Weber’s results suggest the unemployment rate in early April had only increased by 2 percentage points. If people are not actively looking for work, then they are not technically counted as unemployed by government statistics. In line with this, the three researchers also find a dramatic increase in the share of adults who are not in the workforce. These patterns reinforce that this is not a typical downturn.
Their findings imply that the official unemployment rate will be much lower than Bick and Blandin seem to think based on their RPS. Who will be right? It will depend on how households answer the government’s questions.
If you’re not at work but expect to be recalled, you should nonetheless be classified as unemployed. The last government report showed a large increase in this category of workers in March. But it also showed a large increase in the number of employed people who are absent from work. If this pattern repeats itself for April’s report, the official unemployment rate will be lower than actual job-market conditions suggest.
Another statistical challenge could come from people who have separated from their employer and want a job, but aren’t looking for work because they don’t want to get sick and many employers are temporarily shuttered. These people should be counted as out of the workforce, yet may show up in the unemployment figures.
Adding to the complications of figuring out who fits in what category, the pandemic has affected how surveys are conducted in the first place. The Census Bureau suspended in-person household interviews on March 20 and has relied only on telephone interviews to conduct the survey, rather than using both methods as is normally the case. The response rate for March fell substantially, from the low 80% range to 73%. This could distort the official statistics.
Congress and the Federal Reserve have performed admirably during this crisis, anticipating economic need before the official statistics reflected it. To continue crafting sound policy, they can’t be flying blind. Fortunately, they won’t be. But their dashboard will likely have many more indicators than they’re used to, and some will be harder to interpret than normal.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
Michael R. Strain is a Bloomberg Opinion columnist. He is director of economic policy studies and Arthur F. Burns Scholar in Political Economy at the American Enterprise Institute. He is the author of “The American Dream Is Not Dead: (But Populism Could Kill It).”
©2020 Bloomberg L.P.