By Brendan Murphy, Lead Researcher, Accessibility Observatory, University of Minnesota
More people are biking or walking to work in North American cities each year, including here in the Twin Cities. With increased biking and walking, more opportunities for conflict with cars exist, and the safety of our more vulnerable road users becomes an increasingly important consideration. The rates of pedestrians and bicyclists killed in traffic crashes in the U.S. have both steadily increased since 2005, as has the total vehicle miles traveled (VMT) across the country.
The concept of “safety in numbers” for bicyclists and pedestrians has existed academically for quite some time and became more prevalent when Peter Jacobsen published his heavily-cited study, “Safety in numbers: more walkers and bicyclists, safer walking and bicycling,” in 2003. Safety in numbers (SIN) refers to the observable effect where an individual pedestrian or bicyclist’s safety is correlated with the number of pedestrians or bicyclists in an area—that is, in places in the city with more pedestrians or bicyclists, those users have a lower risk of getting hit by a car. However, we don’t fully understand which comes first, the safety or the numbers. If a place is safe for biking, more people may bike there (particularly if useful destinations are nearby); if more people are already biking somewhere, then drivers may be more on the lookout. The SIN effect is probably some combination of both.
In observance of National Bike Month and Bike to Work Week 2017, we’re looking at bicyclist safety in Minneapolis and exploring whether this SIN effect shows up in the city’s crash and traffic data. Studies have previously shown this effect on countrywide scales as well as within individual cities like New York City, but Minneapolis has been less well-studied. The goal of this study (funded by the Roadway Safety Institute) was to attempt to predict crash rates between cars and bicycles at street intersections—based on car and bike traffic levels—and then assess whether areas of the city exist that have much higher per-bicyclist crash rates.
We gathered and combined bike count data from 2007-2014 from the city, estimated bicycle count data published by Steve Hankey and Greg Lindsey’s group at the Humphrey School of Public Affairs, average annual daily [car] traffic (AADT) measurements between 2000-2013 from the city, and finally the city’s traffic crash records between 2000-2013. Using these data, we built a few statistical models describing the car–bike crash rates at Minneapolis intersections, in terms of the traffic levels flowing through them—an Ordinary Least Squares (OLS) model and a “two-part” model that accounts for the large number of intersections where zero crashes occurred.
Overall, the predictive accuracy was not very good, with an average error of 82.6% in trying to predict the number of crashes that would occur based on traffic volumes alone. We did, however, observe evidence of the SIN effect—for every 1% increase in bicycle traffic, there was only a 0.5% increase in the predicted number of crashes. Reasons for this low level of accuracy are manifold: more frequent, better data collection is needed first and foremost, since the study was based on estimated data extrapolated from fairly sparse count data. Additionally, models do better when they include variables pertaining to characteristics of the built environment, such as road geometry, signalization, and numbers of lanes. Lastly, crash reporting for bicycles is biased, since only the more severe crashes tend to be reported, leaving many minor crashes with no injuries unreported.
However, looking at how many crashes occurred, divided by the bicycle traffic level (giving crashes per bicyclist), yielded some useful information. The above map highlights intersections where, according to the data we used, bicyclists experience an elevated risk of getting hit by cars—that is, more crashes have occurred there per bicyclist than at other intersections with lower risk. Further study of individual intersections in this subset (e.g., Oak Street SE and Washington Avenue SE on the University of Minnesota campus) is needed to look at what other factors could be contributing to the elevated risk, but a quick visual assessment of the data can give a reasonable idea of where safety improvements could be implemented.
Bicycling is growing as a popular transport mode for getting around in cities without a car, and as more people regularly make active transportation choices, more people will be affected by road safety issues such as poor multimodal road design. It’s the responsibility of everyone to work toward a safer transportation network for bicyclists, and good quality data can help us get there.