Moral Hazard vs. Liquidity in Unemployment Insurance

It is well known that higher unemployment bene…ts lead to longer unemployment durations. This result has been interpreted as evidence of “moral hazard”–a behavioral response to distorted marginal incentives to search. This paper shows that unemployment bene…ts also raise durations through a “liquidity”e¤ect for households who cannot smooth consumption perfectly. The empirical importance of the liquidity e¤ect is documented in two ways. First, state-level increases in unemployment bene…ts have larger e¤ects on durations for households that are likely to be liquidity constrained (e.g., those with low assets). Second, lump-sum severance payments signi…cantly increase durations among constrained households. Together, the empirical estimates imply that 60% of the e¤ect of unemployment bene…ts on durations is due to the liquidity e¤ect. To evaluate the welfare implications of this …nding, I derive a new formula for the optimal level of unemployment bene…ts in terms of the ratio of the liquidity e¤ect relative to moral hazard. Implementing this formula using the empirical estimates implies that the optimal unemployment bene…t level exceeds 50% of the previous wage. Keywords: unemployment durations, optimal bene…ts, wealth e¤ects, borrowing constraints E-mail: [email protected]. I have bene…ted from discussions with Alan Auerbach, Olivier Blanchard, David Card, Martin Feldstein, Amy Finkelstein, Jon Gruber, Jerry Hausman, Caroline Hoxby, Juan Jimeno, Larry Katz, Bruce Meyer, Emmanuel Saez, Jesse Shapiro, Robert Shimer, Adam Szeidl, Ivan Werning, anonymous referees, and numerous seminar participants. Philippe Bouzaglou, Greg Bruich, David Lee, Ity Shurtz, Jim Sly, and Philippe Wingender provided excellent research assistance. I am very grateful to Julie Cullen and Jon Gruber for sharing their unemployment bene…t calculator, and to Suzanne Simonetta and Loryn Lancaster at the Dept. of Labor for assistance with state UI laws. Funding from the National Science Foundation and NBER is gratefully acknowledged. The data and code used for the empirical analysis are available on the author’s website. An earlier version of this paper with additional results is available as NBER working paper 11760.