An exploratory analysis on city characteristics likely to affect autonomous vehicle legislation enactment across the United States

Abstract

As Autonomous Vehicle (AV) technology enters developmental maturity, the need for real-life testing is crucial. AV testing and deployment in a city or State must be preceded by favorable legislation in that jurisdiction. However, research on which city characteristics affect the likelihood of adopting AV legislation remains in question. In this paper, we investigate city characteristics of different urbanized areas (UZAs) in States across the US. Panel data was collected from 29 UZAs and 23 States over the span of eight years (2011–2018) on variables such as existing AV legislations, electric vehicle score, congestion, and vehicle miles traveled (VMT). A method of fixed effects logistic regression is implemented to overcome omitted variable bias and the limitation of datasets available. Features are selected through the leaps and bounds algorithm, and time and State fixed-effects logistic regression models are tested to predict a binary variable for AV legislation enactment in UZAs. Our results yield a five-predictor State fixed effects model as the most robust by variable statistical significance and concordance. The findings reveal that an increase in electric vehicle use, GDP per capita, freeway VMT, and land use score of a UZA increase likelihood of AV testing adoption, while an increase in fatality cases negatively impact the likelihood of adoption (Ceteris paribus).