11.522: UIS Research Seminar (Fall
2009) - Discussion notes
Cyclist Safety and
Street/Intersection Design
Eric Vallabh Minikel · October 6, 2009 5:15p – 6:00p
Goals and challenges
My goal is to explore the impact of street and intersection design on cyclist safety. Cyclist (and driver) behavior will also be an integral part of the study, since it is in large part the mechanism by which design has an impact.
Scarcity of data will be a major challenge. Suppose you want a collision rate per cycle trip. There are fewer cyclists than drivers, and most accidents are minor and go unreported—therefore, it takes a long time to amass enough data for a numerator. Cyclists don’t buy gas, register their vehicles or get smog checks, and are almost never counted by automatic counters, so it’s quite hard to get a reliable denominator as well.
Studies of the subject take a variety of perspectives. One is the “macro” viewpoint: compare two countries with good and poor cyclist safety records, say, the Netherlands and the United States, and ask what they are doing differently. Another is to use several years of citywide collision data and compare one street to another, one cyclist demographic to another, and tease out trends. Finally, there is the before and after study: collect data for the same street before and after a design change—say, a new bike lane. Each of these approaches has its own limitations and they often contradict one another. One strong and oft-cited phenomenon to be aware of is “safety in numbers”—the more cyclists ,the fewer accidents per cyclist.
The Macro Viewpoint: Pucher
and Dijkstra, 2000.
On a fatalities per million trips basis, walking and cycling in the U.S. are three times more dangerous than driving in the U.S., and several times more dangerous than walking and cycling in Germany or the Netherlands. The authors acknowledge that it is “impossible to undertake rigorous statistical analysis to isolate the impact each measure has had on safety improvements” (p. 18). Yet clearly these countries are doing something right, probably several things, and the authors assert that superior walking and cycling facilities (i.e. sidewalks, bike paths/lanes/tracks) must be part of the story.
Before-and-after: Jensen, 2007.
This data-rich study examines a total of 24,000 crashes before and after the installation of new cycle tracks (i.e. raised, sidewalk-level bike lanes) and bike lanes in Copenhagen. The author offers no explanation for the bizarre finding that the crash rate per cyclist actually increased after the installations. The number of cyclists increased as well—apparently people incorrectly perceived the new lanes/tracks as being safer—and the author concludes that the net public health effect was probably positive (since cycling is healthy). Frustratingly, no distinction is made between different severities of crashes. Cycle tracks, for example, decreased overtaking accidents (obviously) but increased accidents at intersections. The Toronto study (below) suggests that overtaking accidents may be more often fatal than other accidents, so even if the cycle tracks increased the number of accidents they may have decreased the severity.
Citywide data: Toronto, 2003
In this study of three years of cyclist-motor vehicle collisions in Toronto, a preponderance of variables—cyclist age, behavior, weather and light conditions, time of day, etc.—are considered. Yet the “Key Findings” section is fairly devoid of conclusions or recommendations. One point which is mentioned repeatedly is that many accidents involve cyclists riding on the sidewalk. This might be taken as a recommendation against sidewalk cycling, but consider some other possible explanations: (1) mostly inexperienced cyclists ride on the sidewalk; (2) people ride on the sidewalk if the street is really dangerous because it is safer to ride on the sidewalk but still not very safe, or (3) sidewalk riding makes minor collisions at driveways more likely but major collisions less likely. This is a typical example of how hard it is to make a policy prescription out of data.
Safety in Numbers: Jacobsen, 2003
This is one of several studies that finds that as the absolute number of non-motorized users increases, the absolute number of accidents increases much more slowly. This has been found at all scales, whether comparing country to country, city to city, or before-and-after on one street. This study does a poor job of establishing causation, but one explanation is that when there are more non-motorized users, drivers are more aware of them, and so more careful. This suggests that getting more people to walk and bike will benefit all, and that a perception of safety may be a self-fulfilling prophecy.
Areas for study
I’d like to study the impact of street characteristics (bike lanes vs. cycle tracks vs. none; parallel parking or none, width, speed, one or two way) and intersection characteristics on cyclist safety. I have data for 430 bike crashes in Cambridge during 2004-8, and may soon have access to data from Alameda County, CA and Boulder, CO. So some sort of citywide analysis may be possible, but with relatively few data points and no reliable exposure measure (i.e. number of cyclists). As an aside, it may be possible to do a before-and-after study of a redesign or signal change Mass Ave. and Vassar St.
Topics for discussion
1. How lengthy a time, or large an area, does one need in order to establish statistical significance? How can I handle the tradeoff between specificity and sample size? (Looking at all characteristics of a street means an even smaller n for each street type).
2. If you could conduct a before-and-after study of a redesign of our very own Mass Ave. and Vassar St. intersection, how could you gather enough data points to have any interesting results by May 2010?
3. If cyclist safety measures have a certain ‘cost’ in terms of travel time for motorists, how can improvements best be sold to the public?
References
Jensen, Soren Underlien. Bicycle Tracks and Lanes: a Before-After Study. TRB 2007
Back to discussion list.
Back to 11.522 home page.