Spatial Networks of Neighborhood Violence

Purpose: We propose a network approach to studying neighbourhood violence that shifts the focus away from explaining crime levels of individual neighbourhoods towards models that explain citywide networks of crime correlations. Our conceptualization places the network of inter-neighbourhood crime correlations as the phenomenon to be explained: why some pairs of neighbourhoods have crime rates that are highly correlated, and others do not. Methods: We use Exponential Random Graph Models (ERGMs) to implement this framework empirically. ERGMS are applied to correlated trends in shooting incidents across neighbourhoods in Chicago. Our models attempt to explain inter-neighbourhood crime correlations in terms of three mechanisms: spatial proximity, neighbourhood homophily (neighbourhoods are more likely to be connected in terms of crime correlations if they share underlying characteristics associated with violence), and flows of people across neighbourhoods based on 2019 Safegraph mobile phone GPS daily mobility data. Results: Whilst spatial proximity of neighbourhoods plays a role in explaining correlations in shooting between neighbourhoods, we also find crime correlations between distant neighbourhoods, driven by socioeconomic proximity (similarity of neighbourhoods in terms of their socioeconomic attributes) and people flows. Conclusion: Our findings support the conceptualisation of neighbourhood crime as an ecological network, rather than as purely neighbourhood-level or spatial phenomenon. The policy implication is that a focus on the violence levels in one neighbourhood may be insufficient to reduce its rates of violence if its position in the citywide network of crime connections is overlooked.