Abstract: In order to deeply analyze the risk factors and their heterogenous impacts on severity of rollover accidents, a mixed logit model was developed based on 5 aspects of safety factors being human, vehicles, roads, environment and accidents. Model parameters were estimated by using the Monte Carlo method and elasticity values of the statistically significant variables were calculated. The model inputs were decided from 22 independent variables with statistical hypothesis testing of 5% significance level. The statistical results indicate that the severity is correlated with driver's gender, age, seatbelt usage, airbag deployment, ejection, alcohol use, drug use, risk driving behaviors, vehicle type, left-turning action, roadway section type, speed limit, accident location, accident date, adverse weather condition, light condition, and pavement condition, and that there is heterogeneity in the influencies of young and male drivers on the severity.

Key words: rollover accidents, injury severity, mixed logit model, heterogeneity, Monte Carlo method