Evaluating traffic safety performance of countries using data envelopment analysis and accident prediction models

Abstract

Road safety is an issue of global importance, receiving both national and international attention. According to the World Health Organization, road traffic injuries are extrapolated to become the fifth leading cause of death in the world by 2030. Studies conducted to gain better insight into how countries can improve their road safety performance levels often use one single variable – the number of fatalities per million inhabitants – and focus predominantly on European countries. This thesis looks to develop and analyze models incorporating a wider range of countries as well as a wider range of road safety performance indicators using data envelopment analysis and accident prediction models. The first method, initially calculate the efficiency scores using three input variables (percentage of seatbelt use in front seat, road density, and total health expenditure as percentage of GDP) and two output variables (number of fatalities per million inhabitants and fatalities per million passenger cars). It was found that the addition of the percentage of seatbelt use in rear seats (fourth input variable) and the percentage of roads paved (fifth input variable) improved the efficiency scores and rankings. Overall, the percentage of seat belt use in front seats and the total health expenditure variables had the greatest importance. The second method developed three accident prediction models using the generalized linear modeling approach with the negative binomial error structure. The elasticity analysis revealed that, for Model 1 and Model 2, the health expenditure variable had the greatest impact on the number of fatalities. For Model 3, the seatbelt wearing rate in front seats and the seatbelt wearing rate in rear seats had the greatest effect on the number of fatalities.