The location of late night bars and alcohol-related crashes in Houston, Texas

Abstract

A study in the City of Houston, Texas, related the location of establishments primarily serving alcohol ("bars") after midnight to late night alcohol-related motor vehicle crashes. There were three data sets for 2007-09: 1) 764bars that were open after midnight; 2) 1660 alcohol-related crashes that occurred within the City of Houston between midnight and 6 am; and 3) 4689 modeling network road segments to which bars and alcohol-related crashes were assigned. Forty-five percent of the late night alcohol-related crashes were within a quarter mile of a late night bar. The bars were highly concentrated in 17 small bar clusters. Using the modeling network, Poisson-Gamma-CAR and Poisson-Lognormal-CAR spatial regression models showed a positive exponential relationship between late night alcohol-related crashes and the number of late nights bars and bar clusters, and a negative exponential relationship to distance to the nearest late night bar controlling for the type of road segment (freeway, principal arterial, minor arterial). A more general model dropped the bar cluster variable. Further, the Poisson-Gamma-CAR model appeared to produce a better representation than the Poisson-Lognormal-CAR model though the errors were different. The general Poisson-Gamma-CAR model showed that each late night bar increased the frequency of alcohol-related crashes on a segment by approximately 190%. For each mile closer a segment was to a late night bar, the likelihood increased by 42%.

Keywords: Alcohol-related crashes; Bars; Drunk driving; Spatial clusters; Spatial modeling.