This week’s expiration of supplemental unemployment insurance (UI) benefits will provide important insights into the U.S. labor market. With some findings already emerging, let’s carefully weigh the full body of evidence before reaching too many conclusions.
In response to the COVID-19 pandemic, Congress temporarily increased the size and duration of UI benefits. By early 2021, unemployed workers received an extra $300 per week, with benefits extended to a wider-than-usual range of workers. Many individuals could earn more by staying unemployed than by accepting a low-wage job. Some employers complained that they were unable to induce their former employees to return to work, and Congress declined to extend the program beyond this month.
Meanwhile, by June, widespread labor shortages had already spurred 26 states to stop paying the $300 weekly supplements. On Sept. 6, the entire supplemental pandemic UI program ended in the remaining states, although many unemployed workers still qualified for the standard UI program. The staggered timing of this phase-out creates a sort of natural experiment.
One recent study found the end of supplemental benefits to have less effect than expected, with employment rising by only 4.4 percentage points among people formerly receiving UI benefits. This would indicate that fewer people are basing their work decisions on expanded UI benefits than pre-pandemic studies would have us believe. For instance, we know that there is often a large spike in the number of people who find jobs after 26 weeks of being unemployed, which happens to be the time when UI benefits traditionally expire. So, there is little doubt that UI discourages employment, at least to some extent.
So why did this recent study reach a different conclusion than previous studies of unemployment insurance?
One possibility is that statistical techniques were not reliable in this particular case. The authors relied on a “difference-in-differences” analysis of the data, which means examining how the change in employment over time lined up with differences in states’ UI programs. In other words, did employment rise faster in states that eliminated the benefits in June compared to those that kept the supplemental benefits until September?
This statistical approach is better than simply looking at the 26 states that ended the program, without any comparison to a “control group.” All sorts of factors influence changes in employment during any given month. Indeed, it is basically the same statistical technique that medical researchers use to test new vaccines. But the approach will not work if the two groups of states are hit by a substantially different set of economic shocks at the time of the study, which appears to be the case here. The assumption of “other things equal” did not necessarily hold true.
That’s because the 26 states that ended the UI program in June are relatively conservative “red states,” and the other 24 are mainly bluer states. Soon after the program expired, there was a massive surge in COVID-19 cases and deaths, concentrated more in red states. Relative to population, COVID deaths in red Florida have been running more than five times higher than in blue California. Thus, red states were trying harder to gain jobs, but they were also facing more severe headwinds from the virus.
At the macro level, we can see how the COVID delta variant slowed employment during the same period. In July, total employment rose by 1.05 million, which is roughly five times the normal monthly increase, and more than pundits expected. Unfortunately, the pandemic resurgence prematurely ended this employment boomlet, and job growth fell to only 235,000 in August, much less than expected. Consumers cut back on travel and other activities that might expose them to risk. Thus, we should not be surprised that one study found lackluster job growth among a group of workers over the summer.
Here’s the lesson: There is nothing wrong with studies using the difference-in-differences approach to estimate the effect of a policy change. Rather, the problem is when experts give the impression that one single study offers a definitive answer to complex social science questions, and ignore the many previous studies that reached different conclusions.
Another lesson is that we should never be too quick to dismiss the basics of economic theory. If we pay people more on unemployment than they can earn on the job (particularly a job they don’t enjoy doing), many will respond by working less.
Scott Sumner is the Ralph G. Hawtrey Chair of Monetary Policy with the Mercatus Center at George Mason University and a professor emeritus at Bentley University.