The Battle for Baghdad: Testing Hypotheses About Insurgency From Risk Heterogeneity, Repeat Victimization, and Denial Policing Approaches

Abstract

The Iraqi Insurgency (2003–2011) has commonly been characterized as demonstrating the tendency for violence to cluster and diffuse at the local level. Recent research has demonstrated that insurgent attacks in Iraq cluster in time and space in a manner similar to that observed for the spread of a disease. The current study employs a variety of approaches common to the scientific study of criminal activities to advance our understanding of the correlates of observed patterns of the incidence and contagion of insurgent attacks. We hypothesize that the precise patterns will vary from one place to another, but that more attacks will occur in areas that are heavily populated, where coalition forces are active, and along road networks. To test these hypotheses, we use a fishnet to build a geographical model of Baghdad that disaggregates the city into more than 3000 grid cell locations. A number of logistic regression models with spatial and temporal lags are employed to explore patterns of local escalation and diffusion. These models demonstrate the validity of arguments under each of three models but suggest, overall, that risk heterogeneity arguments provide the most compelling and consistent account of the location of insurgency. In particular, the results demonstrate that violence is most likely at locations with greater population levels, higher density of roads, and military garrisons.