Toll road crash severity using mixed logit model incorporating heterogeneous mean structures

Abstract

The current study examined 1,465 crash observations (2017–2021) from Louisiana, identifying significant variables grouped into three major categories: drivers’, crash, and road characteristics. Considering crash injury severity as a dependent variable, we employed classic Multinomial Logit (MNL) model, and several other models to address unobserved heterogeneity in crash data including Random Parameter Logit (RPL), Random Parameter Logit with Heterogeneity in Means (RPLHM), and Random Parameter Logit with Heterogeneity in Means and Variance (RPLHMV). Our findings highlight the impact of factors such as driver gender, age, traffic violations, driver distractions, crash types, surface conditions, and roadway attributes on crash injury severity. These insights emphasise the complexity of toll road safety and inform targeted interventions to mitigate crash injury severity. Notably, male drivers and those under 25 years old increased property damage likelihood, while factors like driver distractions and lower posted speed limits reduced the likelihood of severe injuries or fatalities.