Quantity or Spatial Coverage? Using Interpretable Machine Learning to Evaluate the Relationship Between Spatial Strategy of Police Stops and Crime in Crime Hotspots and Non-Hotspots

As a result of variations in offender activity patterns in crime hotspots and non-hotspots, the spatial strategies of policing (such as the quantity and spatial coverage) should differ. Existing literature has not examined the spatially heterogeneous effect of police stop strategies on crime. This study focuses on the police stops-crime nexus in China, using the case of a major city’s central district. We introduce the spatial dimension of police stops—spatial coverage—in addition to the commonly considered quantity of police stops. Our results indicate that the spatial coverage of police stops surpasses the number of police stops as the most important crime predictor. We further find spatial heterogeneity in the police stop-crime nexus. A larger amount of police stops and spatially focused stops are more effective in crime hotspots. In crime non-hotspots, higher spatial coverage of police is more important in deterring crime than a larger number of police stops.