This paper employs Bayesian negative binomial models to examine the effects of curb extensions on traffic safety through their implementation as part of a stormwater management program in the City of Philadelphia. Although widely deployed to promote traffic safety, the effects of curb extensions on traffic safety are poorly understood. Between 2011 and 2023, the Philadelphia Water Department installed 156 vegetated curb extensions that protruded into the street and initiated the design of 342 curb extensions to be built by 2029. Although pre-existing crash data featured in the selection of treated locations, the focus on stormwater management and long list of treated and untreated intersections provide unique, quasi-experimental conditions under which to examine the effects of curb extensions. After matching these curb extensions to intersections and removing midblock interventions, we created a dataset of 94 treated intersections and 269 control intersections installed between 2013 and 2023. Each intersection included a single curb extension. We matched these spatially to annually aggregated crash data from 2012 to 2024. We also matched these to data on the roadway width, roadway class, signalization at the intersection, and expected acres of drainage associated with stormwater management. Analysis suggested that curb extensions reduced the incidence of reported crashes by around 13% at a typical intersection. However, the 95% credible interval ranged from a 35% decrease to an 18% increase. Findings suggest that a single curb extension may reduce collisions and injuries but that additional studies are needed from fully treated intersections.
Guerra et al. (Tue,) studied this question.