Spatial Effects of Alcohol Availability on Drunk Driving Traffic Accidents

The spatial econometric model was used to conduct an empirical study of the spatial effects of alcohol outlets on drunk driving accidents. 3 356 cases data of alcohol-related crashes, social population and alcohol consumption related points of interest (POI) were collected based on the 2 114 traffic analysis zones (TAZ). After applying exploratory spatial data analysis, the traditional ordinary least square model, the spatial lag model, spatial error model and spatial durbin model were developed. The models were compared based on Lagrange multiplier, log likelihood and likelihood ratio test. Finally, the direct effect, spillover effect and total effect of explanatory variables were quantitatively analyzed. Results show that there is no spatial autocorrelation for the alcohol related crashes in urban district, but spatial autocorrelation in inner suburban district and outer suburban district. The incremental spatial autocorrelation shows that the most significant distance of alcohol-related road crashes is 1.54 km in inner suburban district and 7.95 km in outer suburban district, respectively. Compared with different spatial models, the SDM model obtains the best model fit and explanatory power. The spatial effects of explanatory variables in inner suburban district and outer suburban district are different; however, overall, retails density has positive direct effect and spillover effect, and the density of hotels and companies has negative direct effect. The conclusion can provide an important basis for traffic law enforcement, traffic management, land use, and so on.