Evaluation of Not-At-Fault Assumption in Quasi-Induced Exposure Method using Traffic Crash Data at Varied Geographical Levels

Abstract

Acquiring real-world driver distribution data on roadways is a challenge. The quasi-induced exposure (QIE) method is a promising alternative as it only requires the available crash data. The question to be answered through this study is whether the not-at-fault driver assumption of the QIE still holds when the population is broken down to smaller geographical levels, such as counties, towns, or routes. This is important because the result will provide statistical support to choose for or against the application of QIE at disaggregate levels. In this study, the distributions of driver gender, age, and vehicle type between four groups of drivers in the crash data were examined, using data obtained from the state of Connecticut from 2015 to 2017. Namely, they are the not-at-fault drivers and at-fault drivers in two-vehicle crashes (NF2 and AF2) and the not-at-fault drivers and at-fault drivers in three-or-more vehicle crashes (NF3 and AF3). Chi-square tests and Wilcoxon Mann–Whitney tests were used to provide statistical evidence of whether the driver groups come from the same population. The evidence shows that there are no statistical differences between the distributions of NF2 and NF3. The QIE assumption of not-at-fault drivers is valid at all tested geographical levels. Driver characteristic distribution in the NF2 (and NF3) groups in the crash data should be a good representation of the driving population. The results also revealed the similarities of distributions between AF2 and AF3 and the significant differences between the not-at-fault drivers (NF2 and NF3) and at-fault-drivers (AF2 and AF3).