Place-Based Interventions at Scale: The Direct and Spillover Effects of Policing and City Services on Crime[Clustering as a Design Problem]

In 2016, the city of Bogotá doubled police patrols and intensified city services on high-crime streets. They did so based on a criminological consensus that such place-based programs not only decrease crime, but also have beneficial spillovers for nearby streets. To test this, we worked with Bogotá to experiment on an unprecedented scale. We randomly assigned 1,919 streets to either 8 months of doubled police patrols, greater municipal services, both, or neither. Such scale brings econometric 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. In contrast to the consensus, we find intensifying state presence did not generate substantively meaningful or statistically significant reductions in crime. Our estimates suggest modest direct effects but with crime displaced nearby, especially property crimes. Confidence intervals suggest we can rule out total reductions in crime of more than 2% from the two interventions.