Identifying Dynamic Spillovers in Criminal Behavior

Abstract In this paper we provide evidence on the effectiveness and efficiency of law enforcement policies that target different types of crimes. Our approach acknowledges that past crimes may affect future crimes and we use it to calculate these dynamic spillovers both within and across a broad selection of crimes. Our identification strategy leverages a highly detailed data set that contains the exact location and time of all reported crimes together with the police response to these crimes in Dallas from 2000-2008. Our identification strategy isolates the causal behavioral effect of prior crimes on future crimes and is robust to a variety of sources of endogeneity. We find that policies targeting lighter crimes (which are, for instance, suggested by the "Broken Windows" theory of crime) are ineffective at reducing violent crimes in the future. JEL Codes: K42 , R23