Exploring the association between speed and safety: A path analysis approach

Abstract

Road safety is influenced by many factors; these factors include characteristics of the road, climate, traffic and, most importantly, vehicle speeds. Previous research shows that increases in speed are typically associated with an increased collision risk. Moreover, previous studies have also found relationships between road and traffic characteristics and collisions. In addition, these features have also been found to affect speeds. This paper aims to model all the aforementioned relationships simultaneously using a Structural Equation Modelling approach. More specifically, the paper attempts to model the relationship between average speed and collision frequency, while taking into account the effects of factors that confound the relationship. Moreover, the analysis attempts to assess the mediated effects that some variables have on collisions through their effects on speed. The data used in this study originated from 353 two-lane urban roads in the city of Edmonton, Canada. The average speeds were obtained from 35 million speed survey observations collected over a five-year period. The speed data are linked to the crash frequency at each location during the same time frame, along with the other factors (road, traffic and climate). The results show that, among others, average speed, volume, segment length, medians and horizontal curves all have statistically significant effects on collisions. On the other hand, shoulders, speed limits and vehicle-lengths are some variables that significantly influence speeds. The results also show that the effects of some variables on safety are indeed mediated through speeds (both partial and full mediation is observed). These findings provide valuable insight that may assist decision makers in choosing and developing alternative speed management strategies, which, in turn, could help improve safety.

Keywords: Causal mediation; Indirect effects; Mediation analysis; Path analysis; Speed-Safety relationship; Structural equation modelling.