Homicide hotspots in Chicago: disentangling background persistent risk from contagion dynamics

Crime-reduction strategies often assume that hotspot crime is driven by near-repeat victimization, overlooking other mechanisms that may produce similar patterns. This study underscores the need to design prevention strategies based on prior identification of the processes that generate crime concentration. We demonstrate how Cox–Hawkes point process models can distinguish between competing mechanisms underlying crime concentration. We apply a parametric Cox–Hawkes point process using the Bayesian Spatiotemporal Point Process (BSTPP) package in Python, which jointly models the persistent background intensity and the self-exciting component. The analysis draws on 2087 homicide incidents recorded in Chicago between 1 January 2018 and 31 December 2021. Both event dependence and risk heterogeneity contribute to the formation of homicide hotspots in Chicago, with the estimated intensity divided approximately equally between the persistent background and the self-exciting component. The contagion effect is limited to a few hundred metres and decays relatively rapidly over time. Findings demonstrate that homicide hotspots cannot be addressed through a single prevention logic. Strategies premised exclusively on near-repeat or retaliatory dynamics risk misalignment where violence reflects enduring persistent risk in some areas of Chicago. Effective prevention therefore requires identifying the dominant mechanism operating in each context and tailoring interventions accordingly.