Joint estimation of fatal and injury crash models with controlling for contemporaneous correlation between the disturbance terms

Numerous crash prediction models have been developed using a variety of statistical methods to understand the relationships between crash occurrence and crash covariates. However, none of these models has considered contemporaneous correlation between the disturbance terms across crash models, hence resulting in the separate estimation of crash models. Neglecting contemporaneous correlation may yield consistent but inefficient estimators because the assumption that the disturbance terms are independent across equations is violated. This paper describes the development of crash outcome models to investigate the kinds of factor contributing to the likelihood of fatal and injury crashes occurring with a seemingly unrelated probit approach, which can account for the correlation between the disturbance terms. The results showed that two different crash severity outcomes such as fatal and injury crashes are contemporaneously correlated, and thus, crash prediction models for fatal and injury crashes should be simultaneously estimated to obtain more efficient estimators.