Evaluation of traffic warning signs on truck safety considering endogeneity, a copula-based method

Abstract

It is hypothesized that traffic warning signs are installed at crash prone locations. However, all past studies model the impact of traffic warning signs on crash count independently. Failure to account for the interrelationships might result in biased models’ estimates. That shortcoming could be accounted by simultaneously modeling traffic warning signs’ installation, and crash counts . In this study, the copula abased method was proposed to model traffic warning signs on crash count occurrence while considering the correlation across the two models error terms. Only observations related to truck downgrade crashes were considered in this study. Also, only downgrade warning signs related to those crashes were employed. We accounted for the endogeneity effect by considering the dependent variable of presence of warning sign from one model as an additional regressor, endogenous predictor, in the main discrete count model . Standard joint model with various Archimedean copulas were considered, and Ali-Mikhail-Haq (AMH) copula was identified as a best perfromed copula. The model estimation results highlighted the importance of accommodating the endogeneity in modeling the impact of warning signs on traffic safety, beside accounting for correlation across the error terms. A detailed discussion about the employed methodological approach is provided in the content of this manuscript.