| dc.description.abstract | This thesis examines the experiences of unemployed workers. The chapters use administrative, survey, and self-collected data to understand how unemployment and the government insurance that covers it impact workers. I argue that unemployment insurance increases subsequent unemployment less than previously thought, and many earnings-relevant skills would likely hold steady during any additional unemployment. Collectively, the chapters show that the costs of unemployment insurance expansions are smaller than the existing consensus suggests.
The first chapter, written jointly with Geoffrey Schnorr, identifies the employment effects of approving additional unemployment insurance claims whose eligibility is in doubt. To be eligible for unemployment insurance in the United States and most other countries, workers must have lost their job through no fault of their own. We use administrative data on the universe of unemployment insurance claimants in California since 2002 to estimate how adjusting the leniency of these criteria would affect claimants' subsequent employment. By comparing claimants whose applications are as-good-as randomly assigned to relatively strict government employees with those assigned to relatively lenient government employees, we isolate the causal effect of approving the marginal cases. Empirically, we find that unemployment duration increases by approximately two weeks. Based on a theoretical model of optimal UI, we map this employment effect to the efficiency cost of unemployment insurance. We find that the efficiency cost of approving these marginal cases are much lower than the efficiency costs for other types of UI benefit expansions.
The second chapter, written jointly with Andrew Johnston and Attila Lindner, documents how survey-based skills evolve as workers remain unemployed. In addition to the psychological and financial costs that unemployment imposes on workers themselves, a central concern to policymakers is that unemployment may also harm long-term productivity if workers lose skills while unemployed. We provide direct evidence that this does not appear to be the case for a set of general skills using a linked panel of skill elicitations and administrative employment records among newly unemployed German workers in the late-2000s. We validate that the measured skills are meaningful: baseline measurements predict prior earnings and panel measurements change around certain significant life events. Despite large falls in both subjective well-being and reemployment earnings as workers remain unemployed, we find that all of the earnings-relevant cognitive and noncognitive skills in the survey remain constant. We provide evidence that the lack of observed skill depreciation cannot be explained by various types of measurement error.
The third chapter, written jointly with Peter Ganong, aggregates the existing academic literature measuring the effect of UI benefit generosity on unemployment duration. Building on the first chapter, we hand-collect a comprehensive dataset of published studies that estimate this effect. While the theoretical and empirical consensus is that UI benefit generosity increases unemployment duration, we argue the magnitude is overstated due to publication bias: statistically insignificant or negatively-signed estimates are much less likely to be published. Correcting for publication bias decreases the overall average elasticity of unemployment duration with respect to UI benefit generosity by one-fifth to one-half. Aggregating evidence across studies, the predicted elasticity is approximately 0.3 in a policy regime similar to the typical U.S. state. However, the elasticity is larger by a factor of two or more under typical potential benefit duration extensions that occur during recessions.
JELClassification:J24,J64,J65 | |
