Combining Geographic and Social Proximity to Model Urban Domestic and Sexual Violence

• In Event: 4448 - New Directions and Innovation in the Study of Crime and Justice

Abstract

In order to understand the broader picture of the social dynamics affecting crime in urban areas, it is important to investigate not only the internal forces acting within neighborhoods but also the interactions between neighborhoods. Neighborhood studies of crime rate inference often use spatial statistical models such as spatially weighted regression to account for spatial correlation between neighborhoods. However, to obtain a more flexible model of crime across communities, we investigate the value of taking into account social proximity in addition to geographic proximity. In this paper, we develop techniques to combine geographic and social proximity in spatial generalized linear mixed models in order to estimate domestic and sexual violence in Detroit, Michigan and Arlington County, Virginia. We combine local and federal data sources such as the Police Data Initiative and American Community Survey. By comparing three types of conditional autoregressive (CAR) models, we find that incorporating information on social proximity to spatial models contributes to a more accurate estimation of crime. We also find that the sparse spatial generalized linear mixed model (SGLMM) has less variation than the Besag, York,and Mollie (BYM) and Leroux models - indicating a better fit to areal unit data. The results suggest that social ties between two neighborhoods improve our understanding of local crime. If such ties contribute to the transfer of ideas, customs, and behaviors between places, they may then transfer not only crime itself but also the effects of crime prevention efforts.

Authors

• Claire Kelling, Penn State University

• Gizem Korkmaz, University of Virginia

• Corina Graif, Pennsylvania State University

• Murali Haran, Penn State University