A Two-Step quantile selection model for the safety analysis at signalized intersections

Abstract

The simultaneous estimation of crash frequency and severity has been studied for years, but most of the existing methodologies adopt mean regression models to estimate the parameters. This study presents the quantile selection model as a methodological alternative in analyzing crash rate and severity at different levels, focusing on addressing the heterogeneity and endogeneity issues so as to identify the influencing factors at signalized intersections. A two-step estimation procedure is carried out, in which the Heckman selection framework accommodates the endogenous relationship between crash rate and crash severity at different levels, while the quantile regression estimates various quantiles of crash rate instead of the mean regression, and accounts for the heterogeneity attributed to unobserved factors. Compare to the general Heckman selection model, the quantile approach is able to provide more comprehensive information about the impact of the influencing factors on crash rate. The model uses 555 observations from 262 signalized intersections in the Hong Kong metropolitan area, integrated with information on the traffic flow, geometric road design, road environment, traffic control and any crashes that occurred during two years. The proposed model reveals more detailed information in terms of different quantiles and improves the prediction accuracy.