UC Davis

This dissertation contributes to literatures on the design and effects of two important public policies: Unemployment insurance and the minimum legal drinking age. I provide two unique perspectives on the design of unemployment insurance programs. First, by studying an oft-ignored policy parameter which influences benefit generosity, and second by analyzing an understudied margin along which these programs affect work incentives. I extend the literature on the minimum legal drinking age by empirically testing the hypothesis that increases in alcohol consumption at the drinking age have causal spillover effects on the alcohol consumption of younger peers.

In chapter 1, I analyze a previously unexamined policy parameter used in unemployment insurance (and other social insurance) programs. Since unemployment insurance replaces a percentage of prior earnings while a claimant is out of work, policymakers must define a base period from which prior earnings are measured. A base period structure common in the Unemployment Insurance (UI) program in the United States has two important implications. First, for claimants with volatile enough earnings, this base period structure creates ``benefit risk''---a job loss at the wrong time implies lower benefit amounts. Second, since base periods are determined by the claim filing date, claimants can partially avoid the negative effects of this risk by strategically timing their claims. Using several new sources of administrative data from California’s UI program, I provide four key results on this topic. First, I demonstrate that exposure to benefit risk is widespread. Of roughly 22 million claimants in my sample, over 8 million are exposed to some level of benefit risk. Second, using a bunching approach I demonstrate that roughly 3% of affected claimants strategically delay their claims after a job loss in order to receive higher benefits. Third, I provide evidence that information frictions are a key barrier preventing more widespread use of this strategic response. Finally, I use a dynamic model of job search and Unemployment Insurance to show that the private welfare costs of benefit risk are large. After accounting for claim-timing responses, the average claimant would trade 4% of their expected Unemployment Insurance benefits to eliminate exposure to benefit risk. This number rises substantially among young and especially low-income claimants.

In chapter 2, which is joint work with Lester Lusher and Rebecca L.C. Taylor, we provide causal evidence of an ex ante moral hazard effect of UI by matching plausibly exogenous changes in UI benefit duration across state-weeks during the Great Recession to high-frequency productivity measures from individual supermarket cashiers. Estimating models with day and cashier-register fixed effects, we identify a modest but statistically significant negative relationship between UI benefits and worker productivity. This effect is strongest for more experienced and less productive cashiers, for whom UI expansions are especially relevant. Additional analyses from the American Time Use Survey reveal a similar increase in shirking during periods with increased UI benefit durations.

Chapter 3, written with Eunju Lee, uses data on sibling pairs near the minimum legal drinking age to provide causal estimates of peer effects in alcohol consumption. Following prior work on other outcomes, we exploit the discontinuous increase in alcohol consumption of the older sibling at the legal drinking age in a regression discontinuity design. Our preferred point estimates imply that the number of binge drinking days reported by the younger sibling decreases by 27% of the mean at the cutoff. While our estimates are somewhat imprecise, we are able to consistently rule out leading positive estimates of peer effects in alcohol consumption. Our research design provides estimates which are interpretable as the causal effect of the peer's alcohol consumption. This is in contrast to most prior work which instead identifies the causal effect of exposure to the peer. We explain how this distinction matters for policy.