Testing Determinants of Crime in Philadelphia: A Statistical Spatial Analysis Using Geographic Weighted Poisson Regression

Keywords: crime, broken windows theory, poisson regression, geographic weighted poisson regression
Session Type: Virtual Paper Abstract
Day: Sunday
Session Start / End Time: 2/27/2022 08:00 AM (Eastern Time (US & Canada)) - 2/27/2022 09:20 AM (Eastern Time (US & Canada))
Room: Virtual 3

Abstract

This paper examines the geography of non-violent crime in Philadelphia, Pennsylvania. Determinants of crime were assembled based on modern Broken Windows Theory literature that emphasizes social disorder, physical disorder, and other ecological factors. Though there have been many studies linking ecological factors to crime in Philadelphia, few have included multiple environmental and social factors to get at a more complete model of crime determinants. Even less has been studied on how determinants can vary in influence across neighborhoods (i.e., whether relationships are nonstationary). This research will test an annual average of non-violent crime counts from the years 2015-2019 aggregated at the census block group level against social and ecological determinants of crime. General linear modeling (GLM) assuming a Poisson distribution will provide a well-specified global, but aspatial, baseline model. Using a Geographically Weighted Poisson Regression model (GWPR) will capture these relationships while considering spatial heterogeneity. Based on reduced spatial autocorrelation among residuals, results show that the GWPR model is a better fit than the GLM. The model indicates ecological determinants have a strong influence in areas of north-eastern, central, and southern Philadelphia.