Quantifying regional heterogeneity effect on drivers’ speeding behavior using SHRP2 naturalistic driving data: A multilevel modeling approach

Abstract

Disparities in traffic safety among different regions and driver populations have been investigated in previous studies. These studies found significant differences in crashes between population groups and geographical locations with various socio-economic characteristics. However, driver-behavioral factors, specifically speeding behavior, which is a critical aspect of traffic safety, have received less attention in case of analyzing the impact of local characteristics on driver-behavioral choices that might increase the risk of crashes. The impact of local characteristics on various driver behaviors is getting even more important as the data from the connected and automated vehicles as well as similar second-by-second trajectory level data from naturalistic driving studies are becoming more available. In fact, neglecting mentioned impact might lead to erroneous inferences due to the disparities in socioeconomic characteristics in different regions. Therefore, this paper, for the first time, utilized multilevel logistic regression modeling approach to evaluate the effect of driver's locality-related factors on driver speeding behavior using naturalistic driving data collected from the SHRP2 project in six US states. The results showed that the geographical location of drivers accounted for about 7.7% of the variability in the likelihood of a driver driving over the posted speed, regardless of other explanatory variables. The methodology and the results from this study can pave the road for future human factor studies utilizing trajectory-level data from different geographical locations to reduce the heterogeneity and increase the transferability of the results without introducing a bias in inferences.