Three Essays in Labor Economics and the Economics of Crime

Abstract

This dissertation examines how information shocks, labor demand disruptions, and policy interventions affect crime, labor markets, and racial inequality. Each chapter uses quasi-experimental methods and administrative or panel data to document unintended consequences of social media trends, mass layoffs, and employment regulations. The first chapter studies a viral social media trend that spread information on how to steal certain Kia and Hyundai models. Using a difference-in-differences framework, I show that thefts of these vehicles surged more than sixfold relative to other cars. Cities more socially connected to Milwaukee—the trend’s origin—experienced markedly greater increases in thefts, alongside higher vehicle theft arrests. The results suggest that social media accelerated the diffusion of criminal skills, imposing large costs through victim damages and criminal justice expenditures. The second chapter examines the effects of manufacturing mass layoffs using data on workers certified for Trade Adjustment Assistance over two decades. I find that large layoffs—those eliminating at least one percent of a county’s labor force—lead to long-lasting declines in labor force participation and per capita income, accompanied by sustained increases in violent and property crime. These effects are most pronounced in areas hit by layoffs in low-wage manufacturing sectors and do not appear driven by reductions in police deterrence. The third chapter analyzes Ban-the-Box (BtB) laws, which delay employer access to criminal records. Using a stacked difference-in-differences approach with county-level data, I find that BtB significantly reduces employment for Black workers, driven entirely by a decline in hiring. There are no comparable effects for White or Hispanic workers. Employment losses are largest in counties with high rates of Black crime and large Black–White crime gaps, consistent with employers relying on group-level proxies when individualized information is unavailable.