Developing Nested Logit Probability Models for Rollover Accident Occurrence and Driver Injury Severity

Traffic accidents impose costs and expenditures on society, and rollover accident is one of the most dangerous accident types with a high fatality rate. Most rollovers take place in rural roads due to the particular roadway conditions, environmental considerations and high speed of vehicles. This research aims to determine the important factors that cause rollover accidents and increase the severity of driver injury through modeling the causes and severity of accidents. Nested Logit methodology is utilized which fits this type of accidents both logically and structurally. It accretes the ability of investigation causes of occurrence in the upper nest and injury severity in the lower nest. A broad range of variables including characteristics of the vehicle, driver, traffic-way flow, roadway, environment and collision are taken into account. The model has been proposed based on real world information which can predict 79 percent of observations correctly. In addition, elasticity function has been applied to evaluate the relative importance of variables. According to the final results, the effective parameters in occurrence of rollover accidents are light condition, number of lanes, roadway surface condition, drunk driving and driver drowsiness. On the other hand, factors such as use of airbags and seat belts, vehicle type, driver ejection and the sequence of events are found to be effective with regards to the driver injury severity.