Geospatial Modeling and Simulation of Property Crime in Urban Neighborhoods: An Example Model with Foreclosure

Abstract

Based on neighborhood life cycles, this paper describes the development and the functions of an Urban Crime Simulator (UCS) that are based on a concept that neighborhood goes through cycles from newly established and energetic neighborhoods to matured and stabilized ones and then to deteriorated neighborhoods that await for new stimuli for revitalization. The UCS was developed to estimate changes in property crime rates as induced by changes in the socioeconomic characteristics of urban neighborhoods. UCS is fully integrated with geographically referenced data and is operational in GIS environment. It offers flexibility in the inclusion of neighborhood attributes that may best fit a specific localized context and knowledge of local neighborhoods and neighborhood attributes as suggested by criminological literature.

With UCS, urban neighborhoods are profiled by a selected set of attributes as defined by users. These neighborhoods are first classified into clusters by a hierarchical cluster analysis, which minimizes in-cluster differences and maximizes between-cluster differences. When attribute values of a target neighborhood are updated with projected or planned changes, UCS searches the entire area to find a reference neighborhood with an attribute profile that is the closest to that of the target neighborhood. Once the reference neighborhood is found, all neighborhoods in the reference neighborhood’s cluster are statistically analyzed to yield an estimate for what a new crime rate may be for the target neighborhood with the projected changes.

UCS has a set of tools to assist its users. Correlation among included attributes can be easily calculated to detect if there is any issue of co-linearity. Global and localized spatial autocorrelation can be calculated to evaluate if any spatial dependency among their data would cause any concern in the simulations. Finally, global and localized regression models enable UCS users to assess how appropriate the selected attributes are with respect to explaining the variation in crime rates among the neighborhoods.

UCS is software designed for practical use by law-enforcement agencies that may not be able to take the necessary time to assemble a detailed comprehensive database as other modeling approaches require before carrying out such simulations.