Developing Targeted Safety Strategies Based on Traffic Safety Culture Indexes Identified in Stratified Fatality Prediction Models

Abstract

The south korea transportation safety authority (KTSA) conducts the special traffic safety culture investigation (STSCI) every year to assist local governments in promoting traffic safety. To address the issue of diversity, the local agencies were grouped into four regions by administrative district unit and offered region-specific safety promotion strategies. However, it is unclear if such a classification truly reflects the underlying differences that contribute to traffic safety. The goal of this study is to identify the most relevant attributes that affect the safety performance of local agencies (called traffic safety culture indexes in the current study) so that targeted safety promotion strategies can be recommended. To accomplish the goal, latent class cluster-based negative binomial regressions were applied for a comprehensive list of factors such as demographics, socio-economic features, roadway conditions, traffic violations and road user driver behavior; resulting in seven clusters of local governments. The following indexes were found to significantly and strongly affect crash fatalities in the clusters: rate of wearing helmet, rate of pedestrian's signal compliance, the number of unlicensed driving violations, total paved road length, province, ratio of male to female, and population density. Further, stratified negative binomial regression models were developed to identify statistically significant factors for predicting fatal crashes within each cluster. These cluster-specific features allow the KTSA to design targeted strategies for effective safety promotion.