Investigating bicycle crash frequency, severity, and safety in numbers at signalized intersections in Utah using crowdsourced data

Abstract

This study’s objectives were to (1) understand (geometric, traffic, operational, and other) factors associated with bicycle safety (crash frequency and severity) at signalized intersections; and (2) investigate whether the “safety in numbers” phenomenon applies to bicycling in the US. To accomplish these objectives, data for 2312 bicycle crashes over a ten-year (2010–2019) period were linked to crowdsourced Strava ridership data (as a measure of bicycle exposure) and other information at 2232 signalized intersections in Utah. Zero-inflated negative binomial models of bicycle crash frequencies and ordered logit models of bicycle crash severities were estimated, accounting for different levels of data availability. Also, an aggregate time period analysis compared bicycle crash rates and severity levels for different weekdays and months. Bicycle crashes were more frequent at signals with four legs, longer crossing distances, no channelized right turn lanes, more far-side bus stops, higher population densities, no places of worship, and in neighborhoods with lower incomes and greater shares of people of Hispanic or non-White race/ethnicity. Bicycle crashes were more severe when involving larger or left-turning vehicles, road users who disregarded the traffic control device, on arterial roadways, and at locations with vertical grades and without street lighting. Bicycle crash rates and the share of fatal/serious injury crashes were lower during the highest-ridership months of the year (May–August). Overall, the study found strong support for the “safety in numbers” effect, in which bicycle crash rates decrease with increasing bicycle volumes, when looking both across locations and over time.