Analysis of vehicle damage severity in work zone accidents based on a random parameter logit model with heterogeneity in means and variances

Work zones are critical for infrastructure maintenance but pose heightened risks for traffic accidents. Current safety literature predominantly focused on occupant injury severity, leaving a critical knowledge gap regarding vehicle damage as a primary crash outcome. This study addresses this gap by investigating the determinants of vehicle damage severity in work zone crashes. Utilizing a comprehensive dataset of 7,333 vehicles from the U.S. Crash Report Sampling System (2016–2021), we develop a random parameter logit model with heterogeneity in means and variances (RPLHMV) to capture the context-dependent nature of crash factors. Empirical results demonstrate that the RPLHMV model significantly outperforms traditional fixed-parameter approaches by accounting for substantial unobserved heterogeneity. Significant random effects were found for truck involvement, speeding, and alcohol impairment, indicating their impacts are not uniform across all crashes. For instance, the effect of speeding on severe damage becomes significantly more volatile in maintenance-type work zones. Key fixed-effect determinants also included driver age, vehicle type, and crash configuration. The findings reveal that critical risk factors are highly context-sensitive, challenging the efficacy of one-size-fits-all safety policies.