During recessions, the government enacts a range of policies to dampen their negative effects and to stimulate the economy. One such policy is to extend unemployment insurance durations. This policy is quite controversial. On the one hand, recessions are periods when it is hard to find a new job. Workers that lose their job, thus, have a need for more insurance. But on the other hand, many people worry that unemployment insurance (UI) slows down the recovery from recessions by reducing the incentives workers face to find new jobs. This has led to a vigorous public debate regarding UI extensions during the last few recessions.
Such debates are partly ideological. But they are also partly about factual matters that we economists can provide evidence on. In the case of the debate about UI extensions during recessions, a crucial question is: how much does a UI extension increase the unemployment rate? If UI extensions increase the unemployment rate by a lot, this must be registered as an important downside of such a policy that then needs to be weighed against the benefit of providing the unemployed with greater insurance during the recession.
As is common in economics, it is quite complicated to estimate the effect of UI extensions on unemployment. UI extensions are enacted when unemployment is high and rising for other reasons (i.e., when the economy is going into a recession). Looking at whether unemployment is high at the time of UI extensions therefore runs the risk of confusing cause and effect.
At the micro level, there exists a substantial and quite convincing literature that estimates the effect of longer UI on individual workers. These studies in many cases leverage large discontinuities in the UI system across otherwise similar workers. For example, UI might discontinuously become more generous for workers when they turn 45 years old. In this case, researchers can compare 44 year old unemployed workers with 45 year old unemployed workers.
But the macro effects of a UI extensions may differ from the micro effects for a number of reasons. One reason is that macro UI extensions provide fiscal stimulus which may lower unemployment. Another is that macro UI extensions improve the bargaining power of workers and may through this channel raise their reservation wages and discourage firms from creating jobs. For these and other reasons, it is important to be able to estimate the effects of UI extensions at the macro level.
In recent work, we tackle this challenging task. Our approach to overcoming the “reverse causality” problem discussed above is to use a quirk in the design of the UI system in the United States. Ever since the early 1970s, the US has had an Extended Benefits program that leads to automatic UI extensions under certain circumstances. The quirk in the system that we exploit is that certain parts of this program are optional at the state level. States can adopt these “options” or they can not adopt them. These optional parts allow states to qualify for Extended Benefits more easily. Many states have adopted these options. But other states have chosen not to adopt them, presumably because they are more skeptical about the wisdom of extending benefits during recessions.
The fact that states have made these different choices allows us to estimate the effect of UI extensions by comparing states that have the same qualifying status (i.e., that satisfy the same trigger rules for extended UI) but have chosen to adopt different rules and therefore differ in whether they actually get a UI extension.
Along with co-authors Miguel Acosta at the Federal Reserve Board and Andreas Mueller at UT Austin, we constructed a dataset back to the mid-1970s with information about all the trigger rules for extended benefits and real-time data on all the trigger variables for all US states. This was a difficult task that involved scraping together information from dozens of sources for the period before digital recordkeeping began.
With this dataset in hand, we are able to estimate the effect of UI extensions on unemployment (and other related variables). We choose to do this separately for periods when the initial duration of UI is modest (less than 60 weeks) and for period when initial durations are long (more than 60 weeks).
Our first main finding is that for the short duration period the effect of a standard 13-week UI extension on the unemployment rate is the raise the rate by 0.28 percentage points. This estimate turns out to be quite consistent with micro estimates of the effect of extra UI on the duration of unemployment of individual workers. In this sense, our estimates imply that macro “general equilibrium” effects are small in this context (or cancel each other out).
Our second main finding is that for the long duration period the effect of a standard 13-week UI extension is close to zero. This estimate turns out to be quite consistent with prior macro estimates most of which focused on the Great Recession period. Our baseline estimates exclude the Covid recession. When we include this recession, the size of our estimates increases somewhat.
So, what do we conclude about the wisdom of extending UI during recessions? Our work does not directly answer this question. It provides one of the key ingredients to answering this question. But this ingredient must be combined with estimates of the value of extra insurance. Our finding is that the cost of UI extensions in terms of added unemployment is significant when the initial duration of UI is modest. These increases in unemployment must be weighed against the value of UI extensions in providing insurance to unemployed workers.