"It is well known that unemployment benefits raise unemployment durations. This result has traditionally been interpreted as a substitution effect caused by a reduction in the price of leisure relative to consumption, generating a deadweight burden. This paper questions the validity of this interpretation by showing that unemployment benefits can also affect durations through a non-distortionary income effect for agents who face borrowing constraints. The empirical relevance of borrowing constraints and income effects is evaluated in two ways. First, I divide households into groups that are likely to be constrained and unconstrained based on their asset holdings, mortgage payments, and spouse's labor force status. I find that increases in unemployment benefits have small effects on durations in the unconstrained groups but large effects in the constrained groups. Second, I find that lump-sum severance payments granted at the time of job loss significantly increase durations among households that are likely to be constrained. These results suggest that temporary benefit programs have substantial income effects, challenging the prevailing view that social safety nets create large efficiency costs by reducing labor supply"--National Bureau of Economic Research web site.