ACCIDENT RISK AT A RAILWAY LEVEL CROSSING

Abstract

The purpose of this research is to estimate risk models, which can assess the safety at a railway level crossing. The accident risk, in terms of equivalent fatalities in a period of time, is decomposed into two parts: the accident likelihood, in terms of number of accidents per period, and the accident impact, in terms of equivalent fatalities per accident. Each of the risk dimensions is investigated, using nonlinear regression, Poisson regression, and negative binomial regression, and considering the effect of exposure variables, highway characteristics, railway characteristics, and the control devices, at railway level crossings. Empirical results indicate that Poisson regression is good for the estimation of accident likelihood; and negative binomial regression is good for the estimation of accident risk and accident impact.