Is Crime Displacement Inevitable? Evidence from Police Crackdowns in Fortaleza, Brazil

Abstract: We evaluated one of the most common policing strategies in Brazil: the allocation of police blitzes. This place-based focused deterrence intervention has well-defined assignments, and 3,423 interventions were precisely recorded in Fortaleza-CE, Brazil, between 2012 and 2013. Our analysis takes advantage of the high spatiotemporal daily data resolution coming from an unprecedented longitudinal micro-Big Data (GPS and PING records) to make comparisons of small intervention areas, while controlling for common daily trends, deterrence (spatial and temporal), and diffusion; to show that an average police crackdown causes a 35% decrease in violent crime occurrences. There are diminishing returns of public safety to hours spent by police in a single area, corroborating what police officers know well from their own experience and discretionary behavior. Although crime increases by 6% immediately after the end of a blitz, we observe lasting deterrent effects (diffusion) after 2-3 days. The residual deterrence cancels the relocation of the crime, and the intervention does not generate significant temporal displacement. In addition, we do not find spatial displacement from crime in blocks up to 1.5 km from a blitz. This type of micropolicing tactics generates deterrence by being highly visible in a street segment for a short period and intermittently quasirandom in space-time, which produces uncertainty that might be crucial in minimizing the temporal and spatial displacement of crime. Of public policy interest, we show that the allocation of blitzes passes in an initial cost-benefit analysis.