Isolating Opportunity from Demographics: A Case Study of Motor Vehicle Theft in Philadelphia

Abstract

Considering the extent of the motor vehicle theft (MVT) problem, it is surprising that there is such a dearth of studies on these crimes at the neighborhood level. In 2008, nearly one million vehicles, valued at 6.4 billion dollars, were reported stolen in the United States. Additionally, only half of these stolen vehicles were ever recovered. The purpose of this study is to increase the limited knowledge base on the characteristics of neighborhoods that predict MVT levels. Its focus is on the identification of specific types of land use that increase MVT levels, net the impact of sociodemographics, as posited by the opportunity theories of rational choice, routine activities, and crime pattern theory. The study site is Philadelphia, Pennsylvania with its 1816 census block groups serving as the unit of analysis. The percentage of total land area for each block group utilized by various theorized criminogenic land uses is determined by Geographic Information System (GIS). Evaluated land uses include shopping centers, bars, high schools, colleges, parking lots, youth hangouts, and single family homes. A `proximity space' variable is also computed consisting of the percentage of block group area that is located within one street block of the combined criminogenic land uses. Its usefulness is in determining whether the impact of crime-producing land uses spreads into the surrounding neighborhood. Negative binomial regression models test various hypotheses around the general research question "After controlling for socio-structural correlates, is the presence of certain land uses predictive of MVT levels found at the neighborhood level?" Results demonstrate that land use, both independently and through neighborhood demographic structure, promotes or suppresses MVT levels, a finding consistent with opportunity theories. Specifically, the percentage of land use in block groups utilized by shopping centers, bars, and commercial parking lots is related to higher MVT counts, but that utilized by colleges and single family homes predicts lower MVT counts, net the impact of neighborhood sociodemographics. Interaction models demonstrate that SES, racial heterogeneity, single-parent families, and percentage 15-24 year olds moderate the impact of land use on MVT. Findings are relevant to urban planners, crime practitioners, and crime theorists.