Temporal stability of pedestrian injury severity in pedestrian-vehicle crashes: New insights from random parameter logit model with heterogeneity in means and variances

Pedestrians can be categorized as the most vulnerable road users since they have less protection compared to other road users, which makes their safety of utmost importance for transportation agencies and safety researchers. To improve the safety of pedestrians and to reduce the associated costs, it is an important task to identify the factors that affect pedestrian injury severities in pedestrian-invovled crashes. Several studies have been conducted in this field, but quantitative studies have not examined the temporal stability and transferability of the variables influencing pedestrian injury severity over the years. In this research, using Los Angeles crash data from 2012 to 2017, a random parameters logit model was employed to determine the variables that significantly affect the degree of pedestrian injury and to investigate their stability over time. Moreover, to consider different layers of unobserved heterogeneity and to obtain better statistical fit, the distributions of random parameters are allowed to vary across the observations. Pedestrian injury severity levels are divided into severe, minor, no injuries. Two types of likelihood ratio tests were used to test the transferability of the estimated models over the seven years. The results obtained from the model estimation and likelihood ratio tests revealed that variables affecting the pedestrian injury severity over these years have changed significantly and are not stable. The instability of the variables affecting the pedestrian injury severities shows that it is a necessity to dynamically analyze the crash data and to consider the potential variations over different time periods. (C) 2021 Elsevier Ltd. All rights reserved.