Investigating Safety in Numbers in Cycling After the Entry of Dock-Based Bikeshare Programs in Three U.S. Cities

Abstract

This paper investigates the safety in numbers effect in cycling by comparing monthly bicycle crash trends before and after public bikeshare’s launch in Philadelphia and Pittsburgh, PA, from 2010 to 2019, and in Portland, OR, from 2014 to 2019. To estimate the bicycle crash trend in each city, the authors conduct time series analysis with a regression discontinuity design using generalized linear quasi-Poisson models. The entries of dock-based public bikeshare programs signal an increase in biking activity on the road and serve as the point of discontinuity in the analysis. For each city, the analysis models pre- and post-bikeshare bicycle crash trends inside and outside bikeshare program’s service area separately to compare crash trends at different biking activity levels within the same local context. Results show that before bikeshare, crash trends inside and outside service areas had similar patterns in Philadelphia and Portland while moving in opposite directions in Pittsburgh. The post-bikeshare crash trends differed across cities for both inside and outside of the service areas. The finding that bicycle crash trends changed in different patterns across the three cities indicates that there may be no single safety in numbers effect on bicycle crashes. The inconsistent patterns of bicycle crash trends within and across cities suggest confounding factors underlying the change in bicycle crashes.