Determination of overlapping crime and collision hotzones for focused law enforcement

Abstract

This study uses a data-driven approach to crimes and traffic safety (DDACTS) approach to identify hotzones where focused law enforcement can reduce the number of crimes and number of collisions simultaneously. The authors applied macrolevel (zonal-level) crime and collision prediction models that use geographically weighted negative binomial regression coupled with the empirical Bayes method. Unlike early work, this approach takes spatial correlation into account. Our models are based on 5 years (2009-2013) of zonal level data for the City of Regina, Saskatchewan’s 244 traffic analysis zones. The authors used two response variables (total number of crimes and total number of collisions) and 18 explanatory variables. The explanatory variables include traffic volume, roadway, land use and sociodemographic data. The analysis identifies four DDACTS zones for total crimes and total collisions. These four zones cover only 1.4% of the City’s area but account for 10.9% of expected total crimes and 5.8% of expected total collisions. The approach offers an impartial scientific procedure that identifies areas for targeted law enforcement and opportunities to increase the effectiveness and efficiency of limited law enforcement resources.