Insurance is useful because it allows to prevent (at a price) the consequences of adverse, unpredictable events. However, the viability of insurance systems can be undone by asymmetric information, adverse selection and moral hazard. In the case of unemployment insurance, the insurer cannot reliably observe the search effort and in particular whether a job offer was turned down. The empirical evidence, starting with Robert Moffitt, shows that larger unemployment insurance benefits lead to longer unemployment spells, thus highlighting an apparent moral hazard issue.
Since Steven Shavell and Laurence Weiss, we know benefits should decrease with the length of the unemployment spell, and since Hugo Hopenhayn and Juan Pablo Nicolini we know that one should even get a wage subsidy during the first periods of a new employment spell, a subsidy that depends on the length of the preceding unemployment spell. The basic idea behind these schemes is that unemployed workers need to be encouraged to search and accept the first job opportunity.
Raj Chetty asks whether the patterns observed in the data by Moffitt and others are not a sign of moral hazard, but rather a sign of lack of liquidity. Households with little cash on hand need to accept any job when unemployment insurance benefits are low. With higher benefits, they can afford to look for a better matches, which lengthen the unemployment spells but leads to better outcomes, as previously argued by Daron Acemoglu and Robert Shimer. The key here is to allow households to accumulate assets, but not all can gather sufficient precautionary savings, as in the model pioneered by Gary Hansen and Ayse Imrohoroglu.
The novelty in Chetty's work is that he found a way to test this empirically and finds indeed that this liquidity effect is very important: 60% of the marginal effect of unemployment insurance benefits on unemployment duration. For example, severance payments lengthen unemployment duration, and cross-state variations of benefits indicate that this duration does increase with benefits for liquidity constrained households, but very little for others. The big consequence is high benefits in the range of 50% are now easier to justify, even if there potential for moral hazard issues, something Stéphane Pallage and Christian Zimmermann had already highlighted, but in the context of a political equilibrium.