Shaking Criminal Incentives

Abstract: We study criminal incentives exploiting a historically unique source of exogenous variation, the unanticipated 1995 Kobe earthquake, which influenced several Japanese municipalities with thousands of deaths and building damages, while it left others unaffected. Natural experimental evidence between 1990-2000 indicates that the decline in burglaries post-earthquake is disproportionally large for affected municipalities, even after controlling for other key determinants, such as labor market conditions and police forces, indicating the response of housebreakers to damages that reduce the value of prospective takings. Yet, other crime types remain unchanged, eliminating the possibility of generalized effects that influence crime at large or of substitutions across different crime types, implying the existence of adjustment costs to changes in criminal specialization. For robustness we measure housing damage using both fully and partly destroyed houses, as the destruction of houses may mechanically decrease burglaries by simply reducing the number of targets. To account for endogenous reporting, we use not only reported crimes but also actual arrests and cleared-up events. We tackle further endogeneity concerns by instrumenting the extent of damage with the distance from the earthquake epicentre, which arguably influences burglaries exclusively through devaluing the potential loot. Importantly, our estimated elasticity is robust and consistent with burglary value elasticities derived in the non-experimental literature. This is the first comprehensive natural experimental study that explores both the direct impact and potential substitution effects across different crime types in response to the changing economic value of prospective criminal takings, key but overlooked aspects of illegal behavior.

Joint with Yu Aoki (University of Aberdeen & IZA).