James Pethokoukis of Reuters flags a blog post at the Atlanta Fed, highlighting Fed research on how UI benefit extensions have affected the unemployment rate. Two Fed studies suggest that they may have contributed 0.4 to 1.7 percentage points to current unemployment. But a closer look at this research makes me skeptical that the effects have been so large.
The first study, from the San Francisco Fed, looks at average duration of employment for unemployed people eligible for UI benefits, compared to those who are ineligible because they left jobs willingly or are new workforce entrants. Obviously, for both groups, average unemployment duration is up sharply in the recession, but for the UI-eligible it has grown by an additional 1.6 weeks. Attributing this additional increase to the increased term of UI benefits implies a rise in unemployment of 0.4 percentage points.
Stepping back, everybody agrees that unemployment benefits that are more generous (either in amount or duration) do foster unemployment, by increasing the relative attractiveness of remaining unemployed and drawing benefits. The question is how strong this effect is, and how much its strength changes in recession.
In a normal economy, with fewer than 1.5 unemployed people per job listing, you would expect the effect to be strongest. If a significant number of those “job seekers” don’t actually want to go back to work, positions will go unfilled and unemployment will rise unnecessarily.
But today, there are more than 4 unemployed members of the workforce per job opening. Even if a significant number of unemployed people who are nominally in the workforce actually prefer to stay unemployed, other job seekers will easily fill those available slots. As such, UI extensions should do less to foster unemployment when the economy is weak than when it is strong.
This poses a problem for the San Francisco Fed study. Under the hypothesis where UI’s effects on employment approach zero in a weak labor market, you would still expect differential effects on unemployment duration between those who are UI eligible and those who are not. This is because none of the job seekers who are delaying their return to work because of UI policy are in the UI-ineligible cohort.
So you would still expect average duration of unemployment to rise above trend for the UI-eligible cohort and below trend for the ineligible cohort. In essence, you’d get results like those in the SF Fed study. It’s probable that this does not fully explain the 1.6 week gap. But I do think that the SF Fed study should be read as placing an upper bound* of 0.4 percentage points on the effect of UI extension on the unemployment rate, not as implying an effect at 0.4.
The other study, from the Chicago Fed,
reaches the 1.7 percentage point figure using a model based on the 1990 Katz and Meyer paper on this topic. [**Correction–see below.] However, Larry Katz himself has, just this spring, warned against using his 1990 results to estimate UI effects on unemployment in the current recession.
Katz notes that the effect of UI policy on unemployment in the 1970s and early 1980s (when the data for his 1990 paper were collected) was driven in part by corporate layoff-and-rehire cycles designed to maximize employees’ UI collections. Such practices are a much less significant feature of today’s economy, and as such more recent research has found less pronounced effects. Katz concludes that “unemployment insurance extensions are not a major source of current structural unemployment problems.”
The incentive effects of UI extension must also be weighed against the stimulative effects of paying UI benefits. For some reason it’s become almost taboo to note this on the Right, but UI recipients tend to be highly inclined to spend funds they receive immediately, meaning that more UI payments are likely to increase aggregate demand. UI extension also helps to avoid events like foreclosure, eviction and bankruptcy, which in addition to being personal disasters are also destructive of economic value.
As a result, I am inclined to favor further extension of UI benefits while the job market remains so weak. I am not concerned that this leads us down a slippery slope to permanent, indefinite unemployment benefits (which historically have been one of the drivers of high structural employment in continental Europe) as the United States has gone through many cycles of extending unemployment benefits in recession and then paring them back when the economy improves, under both Republican and Democratic leadership.
But we could eliminate these fears by making UI adjustment an automatic, rather than political, process. I haven’t seen any specific formulas proposed (if a reform is on the table, readers, please alert me) but in general UI should be extended when unemployment is high and/or rising, and contracted when it is low and/or falling. A formulaic adjustment program could mimic what Congress habitually does already, but without generating market uncertainty—or incurring risk that Congress will be too timid to pull the trigger on abbreviating UI benefits in recovery.
*Update: Upon reflection, “put an upper bound” is too strong a phrase here. There could be important aggregate differences between the UI-eligible and -ineligible cohorts; for example, the UI-eligible might tend to be in parts of the country where the economy is weakest, or the UI-ineligible may have less average work experience. It’s therefore possible that, absent any policy changes on UI, you would see different changes in unemployment duration between the two groups.
However, it’s unclear which direction these differences would point. Whereas, regardless of UI’s effect on total unemployment, extensions should raise unemployment duration for those eligible for UI even if they do not rise for the population as a whole, for the reasons I laid out. So, while the SF study should not be said to place an upper bound on UI extension’s effects on the unemployment, my inclination is to guess it would tend to overstate effects rather than understate them.
**Update 8/30: Bhashkar Mazumder, one of the authors of the Chicago Fed study, notes (in the comments at Marginal Revolution) that I have misinterpreted their paper. He’s right. Where the Atlanta blog post referred to a finding of a 1.7 percentage point effect on unemployment, I took that to be an unemployment rate figure imputed from Mazumder, et al’s estimate that UI extension could have increased unemployment durations by as much as 25 percent, based on the Katz and Meyer 1990 paper. However, looking back on the paper, that is clearly incorrect for a few reasons. Mazumder and his co-authors, as he notes, offer a direct estimate of a 0.7 percentage point effect on unemployment duration, which is not based on Katz and Meyer.
Katz’s admonition against using his 1990 paper to estimate current effects of UI benefits on unemployment would still apply to estimates of changes in duration. However, I should have noted that the Chicago Fed authors used the Katz figures as an upper bound, not a sole estimate, and that they also offered a separate estimate of the UI extension’s effect on the unemployment rate. My apologies for my the error.