Determinants of Crash Fatalities in Urban Area: A Bayesian Ordered Probit with Sample Selection Model

Crash fatalities in urban area are a major public concern, and determinants are of major interest to the officials of traffic departments, policy makers, and researchers, among others. In this study, the Bayesian ordered probit with sample selection model is presented as a methodological alternative in analyzing crash fatalities so as to determine the influencing factors in urban area. The crash dataset from 2014 to 2017 maintained by Nevada Department of Transportation is employed to illustrate the performance of this model. A two-step estimation procedure is conducted, in which the Heckman sample selection model determines the sample selection process, while the Bayesian ordered probit model accounts for the probability of crash fatalities. By comparing with the Bayesian ordered probit model, the proposed model outperforms in terms of goodness-of-fit, and the determinants of crash fatalities are identified. The findings provide insights for the officials, planners, and policy makers.