Specifying "Space" in a Model of Crime in the Presence of Gangs: Geography or Social Networks?

George Tita, University of California, Irvine

Abstract This research builds upon existing models of crime that incorporate spatial effects by specifying dependency of events not simply as a function of geographic space, but also as a function of social space. I explore the spatial distribution of both violent and non-violent crime using techniques developed in spatial econometrics to identify whether the distribution of crime is a function of omitted explanatory variables (a spatial lag model) or due to a direct causal relationship among geographic units of analysis (census tracts.) The former is the conventional approach in the existing literature and is addressed by including a spatially lagged version of the dependent variable as an explanatory variable in the model. The latter requires one to specify a direct known relationship that links the units of analysis in some causal way. In this case, the relationship is specified as the social ties that define rivalries among gang contained within the census tracts. Thus, census tracts are "neighbors" when they contain the social space of rival gangs. In essence, a second lagged version of the dependent variable (a social lag variable) is created and also included in the model of crime. It is expected that the socially lagged variable will provide a more robust explanation of the spatial distribution of violent crime while the spatially lagged variable will better fit a model of property crime.

(Return to Program Resources)

Updated 05/20/2006