Measuring the colocation of crime hotspots

Abstract

Crimes tend to concentrate in high-risk places known as crime hotspots. While the size and locations of such hotspots vary between different types of crime as would the underlying conditions that trigger each crime, the extent of overlaps between their hotspots is understudied. Using crime data from Chicago aggregated at the community-area and the census-tract levels, this paper investigates the patterns of overlapping hotspots between different crime types to see whether a specific group of crime types regularly form a joint cluster. Specifically, we identify statistically significant hotspots for each crime and, using the frequent-pattern-growth algorithm, analyse the frequency of each combination of crimes sharing their hotspot locations across the study area. Results suggest that crime hotspots form stable multi-layered colocations and that each area holds its subset: namely, the pervasive, primary colocations consisting of assault, battery and criminal damage to property, which are frequently joined by 7 additional (e.g. street robbery, motor vehicle theft, weapons violation) crimes to comprise secondary colocations, some of which evolving to an even larger, tertiary colocation of hotspots with up to 11 additional crime types (e.g. homicide, criminal sexual assault, narcotics) to form crime-riddled neighbourhoods. This multi-layered structure of colocations as well as the crime colocation diagrams that show the most representative crimes at each colocation size would improve our understanding of the association between different crime types and the crime indicators of other crimes.