Thoughts on unemployment insurance and the mean duration of unemployment

Harvard economics professor Greg Mankiw provides his thoughts on the optimal duration of unemployment insurance (UI) during an economic downturn:

UI has pros and cons. The pros are that it reduces households' income uncertainty and that it props up aggregate demand when the economy goes into a downturn. The cons are that it has a budgetary cost (and thus, other things equal, means higher tax rates now or later) and that it reduces the job search efforts of the unemployed. To me, all these pros and cons seem significant. I have yet to see a compelling quantitative analysis of the pros and cons that informs me about how generous the optimal system would be.

So when I hear economists advocate the extension of UI to 99 weeks, I am tempted to ask, would you also favor a further extension to 199 weeks, or 299 weeks, or 1099 weeks? If 99 weeks is better than 26 weeks, but 199 is too much, how do you know?

It is plausible to me that UI benefits should last longer when the economy is weak. The need for increased aggregate demand is greater, and the impact on job search may be weaker. But this conclusion is hardly enough to tell us whether 99 weeks is too much, too little, or about right.

The thought I've been kicking around recently is that the length of unemployment insurance should automatically adjust based on either the mean or median duration of unemployment. My hypothesis is that the economic optimum is likely within the range of 1x and 2x the mean (average) duration of unemployment.

The median duration of unemployment is currently 21.6 weeks. That means 50% of all unemployed workers are finding jobs within that span of time. (By comparison, the mean is currently 33.8 weeks.) When half of all unemployed workers find jobs within 21.6 weeks, providing 99 weeks of unemployment insurance seems a bit excessive to me. Sure, some people are naturally more employable than others. The less education people have, the higher their chances of being unemployed. College graduates probably find new jobs quickly while high school dropouts take much longer. For this reason, the mean is probably a better guide than the median.

I'll suggest a simple policy rule:

The length of unemployment insurance should automatically adjust throughout the business cycle. It should provide full benefits until 1x the mean duration of unemployment, then decline [not necessarily linearly] until being discontinued at 2x the mean duration of unemployment.

Because the mean duration of unemployment is considerably longer than the median, this will provide most unemployed workers with full unemployment benefits during the entire time they're unemployed. Those workers who don't find jobs will gradually be given stronger economic incentives to look harder and lower their standards. Workers who are still unemployed after 2x the mean duration of unemployment are significant outliers, and would be cut off, although welfare may still be available to them if they truly need it.

A downside to the rule above as stated is that it will tend to hit poorly-educated workers the hardest, because they are less employable than most workers. I'm not dogmatic about the rule. It's really just a suggestion. A simpler and more generous version would provide full benefits until 2x the mean duration of unemployment, ending with a sharp cutoff point. The key is that the length of unemployment insurance benefits should automatically adjust throughout the business cycle, using mean (preferably) or median duration of unemployment as a guide.

Let me also suggest a second policy rule:

The amount (i.e. dollars per week) of unemployment insurance benefits people receive should also vary throughout the business cycle. The amount they get should automatically go up or down in inverse proportion to the output gap, with the default level assuming an output gap of zero.

With this second policy rule, when the economy is weak payment amounts would go up, providing an extra economic stimulus; when the economy is strong, payment amounts would go down to incentivize the unemployed to take one of the many jobs available.

Update: Bentley University economics professor Scott Sumner writes in:

I favor a self insurance approach. Each worker contributes 10% of their income into a private UI account, which they can draw from when unemployed. Any funds not used can be allocated to retirement. This avoids the problem of UI acting as a disincentive to find work.