Safety in Numbers and Bicycle Safety: A Detailed Analysis of the Denver Metropolitan Area

Recently across the US, there has been a push to accommodate and encourage the viability of alternative modes of transportation—especially bicycling. Leaders across all levels of government, trade groups, advocacy and policy groups, and others are promoting different methods to make urban areas more bikeable. Now, as planning practice is moving towards implementing a transportation system that serves different types of travelers, the US faces challenges involved with retrofitting existing automobile-oriented streets. While implementing bicycle safety initiatives is becoming a popular movement among municipalities, there have been differing opinions on the best way to make cities more bikable in academic literature (Pucher & Buehler, 2012). There is an ongoing debate about what types of improvements will be the most effective at reducing crash rates and/or decreasing individual risk for cyclists. Since 2003, one of the key factors in this debate has been the phenomenon of “safety in numbers.” “Safety in numbers,” or SiN, describes the observed inverse correlation between bicycle ridership and cyclist risk (Jacobsen, 2003). As ridership numbers increase, the relative risk per cyclist is said to decrease (all else being equal). When examining large-scale datasets, such as national ridership counts and crash statistics, research suggests there is a significant negative, non-linear correlation (exponentially decreasing) between ridership and crashes per rider. This means that while the total number of crashes increases with ridership, the rate of crashes per rider decreases. While bicycle safety and SiN are well-researched topics, there are still many questions about the SiN effect that are still unclear. First, the full character of the SiN effect is not explicit in the existing literature. Nearly all studies of the phenomenon have been conducted with large units of analysis (cities, countries, etc.). No study to the researcher’s knowledge has considered the SiN effect at the individual street level with real data. Second, because SiN has not been studied with small units, there has not been a way to control for road conditions that also effect bicycle crash rates. And third, it is not clear how all of the factors that determine cyclist injury and fatalities—including SiN, bicycle infrastructure, speed limit, road design, congestion, etc.—interact with one another. These gaps in collective understanding about safety in numbers has led to disagreements among scholars about its nature and implications for practice. One of the major debates surrounding SiN and policy has been its use as an argument to dissuade investment in separated bicycle infrastructure. Some think that separated infrastructure may undermine some of the safety benefits that may affect cyclists because of SiN; the goal of this type of infrastructure is to limit motorists’ conflict points with cyclists, and because of this, separated infrastructure may actually endanger other cyclists on the road because fewer cyclists are interacting with drivers in mixed traffic, lessening drivers’ incentives to adjust their behavior (assuming that behavior modification underlies the SiN effect) (Thompson et al., 2017). Despite limited understanding about this topic, SiN is has been used to make policy justifications, specifically pitting policy-only solutions against infrastructure improvement ones (Bhatia & Wier, 2011; City of Berkeley, 2010). It is crucial, then to understand the SiN effect more fully. My research addresses these gaps in the literature and provides recommendation for practice. My research reports several major findings. First, the safety in numbers effect is reflected on the individual road segment level; using a Cragg double hurdle model, I showed that numbers are a significant predictor of crashes, even when other control variables—infrastructure, congestion measures, speed limit, functional class, median household income, and road length—are added to the model. Second, my research shows that the SiN effect is best characterized by a non-linear, exponentially decreasing mathematical model, even on the segment level. Third, my research created detailed predictions that quantify how the SiN effect changes under different conditions. The most notable of these findings was twofold. First, there was no significant difference in the predicted number of crashes for segments with or without bike lanes as the number of trips increased. And second, facilities with separated bike lanes also receive a safety benefit from increased exposure, but the benefit is not as strong as on segments without separated bike lanes. In summary, my research verified existence of SiN on the road segment level as well as characterizes the effect mathematically. I also suggest that practicing planners should encourage more biking to improve overall road user safety, but that this should be done in tandem with other measures such as bicycle infrastructure