Crash injury severity analyses with multilevel thresholds of change modelling approach for at-fault out-of-state drivers

Abstract Crash injuries are major concerns in the transportation area due to the severe social and economic impacts. In this study, instead of focusing on the total crashes, crash injury severities caused by out-of-state drivers (drivers whose licenses were issued by other states or countries) were targeted. The different influencing factors between crashes caused by in-state and out-of-state drivers were investigated, and the impacts of drivers’ residency on the injury severity levels were identified. Besides, to simultaneously considering the unobserved heterogeneity issue of crash observations and relaxing the parallel lines assumption (constant parameter estimates across different severity levels), a multilevel thresholds of change modeling (MTCM) approach was proposed. Then, empirical analyses were conducted based on the data from state of Florida, where crashes occurred on roadway segments and intersections were analyzed separately. The analyses results showed that MTCMs provided better model goodness of fits comparing to the traditional ordered logit models and random effects ordered logit models. In addition, interesting findings have been concluded through comparing the variables affecting the two driver population, and suggestions were made to improve the crash injury severity from the perspectives of police enforcement, driver education, and roadway management.