Cities target police patrols and public services to control crime. What are the direct and spillover effects of such targeted state services? In 2016 the city of Bogotá, Colombia, experimented on an unprecedented scale. They randomly assigned 1,919 streets to either 8 months of doubled police patrols, greater municipal services, both, or neither. We study how crime responds to intensifying normal state presence in moderate- to high-crime streets, and what this implies about criminal behavior. Scale also brings challenges. Spatial spillovers in dense networks introduce bias and complicate variance estimation through “fuzzy clustering.” But a design-based approach and randomization inference produce valid hypothesis tests in such settings. We find that increasing state presence has modest direct impacts, even when focusing on the highest-crime “hot spots.” More intense state presence deters more crime. But in most cases, however, crime appears to displace to neighboring streets. Property crimes seem most easily displaced, while violent crimes may not be. One interpretation is that crimes with a more sustained motive are more likely to displace than crimes of passion, which state presence may more permanently deter.
