Multivariate analysis can help disentangle the effects of community disorganization and violence from potential confounding factors. Unfortunately, it is a very common modeling strategy in this line of research for scholars to include numerous indicators of community factors in the same model. As we have discussed before, we see this as a form of model overspecification4 and so we expect that partial coefficients in these models are underestimated due to overlap between measures. For example, in studies by Bellair (1997), Barnett and Mecken (2002), and others, the authors include a long series of neighborhood variables in the same model (e.g., population size, population density, percent urban, percent youth, population change, resource disadvantage, neighborliness/neighbor interaction, heterogeneity, residential stability, percent single parent families). Hipp (2011) includes percent Black, percent Latino, heterogeneity and segregation, residential stability, percent occupied units, percent crowded households and age of buildings, among other factors, in the same models. While this choice may have been by design for the authors’ research needs, these statistical models are probably overly conservative for ours. Thus, we focus on multivariate models that we believe provide a “fair” test of associations between community factors and crime, to assess whether those associations are more consistent for violent crime.