A latent segmentation based generalized ordered logit model to examine factors influencing driver injury severity

Abstract

This paper formulates and estimates an econometric model, referred to as the latent segmentation based generalized ordered logit (LSGOL) model, for examining driver injury severity. The proposed model probabilistically allocates drivers (involved in a crash) into different injury severity segments based on crash characteristics to recognize that the impacts of exogenous variables on driver injury severity level can vary across drivers based on both observed and unobserved crash characteristics. The proposed model is estimated using Victorian Crash Database from Australia for the years 2006 through 2010. The model estimation incorporates the influence of a comprehensive set of exogenous variables grouped into six broad categories: crash characteristics, driver characteristics, vehicle characteristics, roadway design attributes, environmental factors and situational factors. The results clearly highlight the need for segmentation based on crash characteristics. The crash characteristics that affect the allocation of drivers into segments include: collision object, trajectory of vehicle's motion and manner of collision. Further, the key factors resulting in severe driver injury severity are driver age 65 and above, driver ejection, not wearing seat belts and collision in a high speed zone. The factors reducing driver injury severity include the presence of pedestrian control, presence of roundabout, driving a panel van, unpaved road condition and the presence of passengers.