Can Unemployment Insurance explain the high unemployment rate?

10/16/2013 04:28:00 PM
Hagedorn et al. assert both sides of this river are the same. I don't see it.
Noah Smith (I think?) tweeted a fascinating paper by Hagedorn, Karahan, Manovski, and Mitman (2013), which purports to show that "most of the persistent increase in unemployment
during the Great Recession can be accounted for by the unprecedented extensions of unemployment benefit eligibility."

The paper talks about "macro" versus "micro" effects, but I think this is a misnomer--it's really all micro. But what it says is that many of the previous microeconomic studies have largely ignored a big important part of the microeconomic effects. I have only had time to give the paper a cursory reading, so I hope none of this commentary is premature.

When assessing the impact of unemployment insurance (UI) on the unemployment rate, economists typically use a search model, where workers apply, get a wage offer, and either accept or go on UI depending on the relative values of the offer and UI. A worker who is on UI is said to be "unemployed," though it is important to bear in mind that this "unemployment" is purely voluntary, in sharp contrast to the common narrative of people being unable to get a wage offer at all during a recession.

In search models, it turns out that the worker's optimal strategy is a "reference wage" which depends, in part, on the generosity of the UI benefit--if the wage offer they get is higher than their reference wage they will accept the job, whereas if the wage offer is less than their reference wage they will reject and take UI instead, hoping to apply again elsewhere and get a higher wage offer. Now, it should be pretty clear that the reference wage increases as UI becomes more generous. In what Hagedorn et al. call "micro" analysis--the standard analysis policy wonks usually rely on--we assume that the distribution of wage offers workers face is fixed, so UI has the effect of increasing unemployment by raising the reference wage, causing a higher fraction of individuals to reject their wage offers. What this leaves out is what Hagedorn et al call the "macro" effect (which is really also just microeconomics, but whatever) whereby the higher equilibrium wage causes movement up the labor demand curve--at a higher wage rate, firms are willing to hire fewer workers. Without doing empirical work, the direction of this macro effect is somewhat ambiguous: it could magnify UI's effect on unemployment rates by reducing the number of job offers in addition to decreasing the job offer acceptance rate, or it could mitigate the micro effect by lowering the reference wage somewhat due to poorer prospects of getting a higher wage offer.

The paper uses a clever empirical strategy to sort this out. It turns out that the generosity of UI vary somewhat state-by-state, since federal law gives states leeway in crafting their own UI eligibility criteria. You can't, however, merely compare changes in unemployment rates across states because of possible endogeneity of eligibility criteria, so what they did was compare changes between adjacent counties separated by state borders. Their reasoning is that a place like, say, Covington, Kentucky probably has a pretty similar population and economic conditions as Cincinnati, Ohio, because they are both part of the same metropolis despite the snootiness of Cincinnatians towards Kentucky the fact they are in different states with different UI policies. Thus, if either Ohio or Kentucky changed the generosity of UI eligibility in their sample, they could difference out the unemployment rates of the two counties to estimate the impact that policy change had on unemployment. The average of these difference is an (hopefully) unbiased estimate of the "local average treatment effect" of UI policies on the unemployment rate.

This approach has a couple of limitations. First of all, with extremely rare exceptions (such as Marrieta, Ohio, I have recently learned) cities that straddle state lines have totally separate governments on each side of the state border. This means that we can't totally rule out the possibility that their estimates reflect differences in city policies rather than the state-level UI changes--city policy could depend on state-level policy for a variety of reasons, violating their assumption that adjacent counties have similar economies. Additionally, and this is something they do address, there is always potential of endogenous movement across county borders as individuals and businesses move in response to the differential in UI benefits and or the tax policies that fund them.

That said, there is a glaring error in their theory of the "macro" impacts of the UI extension during the Great Recession. According to the theory they cite, increasing UI generosity increases unemployment by causing wages to rise. In ordinary times, given a reasonable reaction function for the Fed, that is a reasonable conclusion. The problem is that in a recession, that would actually be a net Keynesian stimulus, actually decreasing unemployment. This is a version of the "paradox of toil"--in a liquidity trap, anything that raises wages and/or decreases willingness to supply labor is expansionary, and anything that increases the willingness to supply labor is contractionary.

That is to say, the main limitation of this paper, despite their rhetoric to the contrary, is that it is all micro and no macro. I would like to see this issue explored in more depth. Are their empirical estimates wrong, or is the paradox of toil wrong?