|
Abstract
|
Wesley E. Marshall and Nicholas N. Ferenchak (Why cities with high bicycling rates are safer for all road users) claim that "improving bike infrastructure with more protected/separated bike facilities is significantly associated with fewer fatalities and better road safety outcomes for all road users." There are a number of roadblocks to accepting this conclusion based on the research presented in the paper.
First, although the study has been widely interpreted as giving evidence for the benefits of "protected bike lanes" along city streets, the authors' "protected/separated bike facilities" category includes "trails/shared-use paths," per an email from Professor Ferenchak. Three of the twelve cities (San Francisco, Portland, and Austin) had a combined total of 12.5 miles of "protected bike lanes" at the end of the study period and the others had none, according to the People for Bikes Inventory of Protected Bike Lanes (https://peopleforbikes.org/green-lane-project/inventory-protected-bike-lanes/). By comparison, there were more than 1300 miles of "multi-use paths" in the 12 cities according to the source used by Marshall and Ferenchak (Milne, A., Melin, M., 2014. Bicycling and Walking in the United States: 2014 Benchmarking Report. Alliance for Biking & Walking, Washington, D.C., p. 152).
Second, ordinary bike lanes apparently had the same effect on traffic safety as the "protected/separated facilities" (section 4.1.2), but this result was left out of the tables and the conclusions.
Third, a significant p-value does not imply a causal relationship. With 112,918 observations, it is not difficult to find coefficients that pass conventional significance tests.
Fourth, there are confounding variables left out of the model. The authors say that previous work shows that "denser, more urban areas generally experience lower road fatality rates than more suburban or rural environments." Yet the models do not control for population density, nor for the amount and speed of driving, the two most important factors in explaining fatalities and serious injuries. The omission of these variables is particularly problematic given that the selected cities fall into two distinct groups: moderately to very dense (San Francisco, Chicago, Portland, Minneapolis, Seattle, Denver) and with many low-density areas within the city limits (Houston, Dallas, Austin, Oklahoma City, Memphis, Kansas City). We would expect a lot more and faster driving per capita in the latter group, and therefore higher serious and fatal injury crash rates.
Fifth, the model is misspecified: it does not account for the panel structure (repeated observations of the same unit over time). Using OLS on panel data as if it were cross-sectional underestimates the standard errors, potentially leading to spurious findings. Including city-specific fixed effects would be one way to control for the substantial differences between the cities.
Finally, the authors suggest that "better road safety outcomes for all road users" means greater bicyclist safety ("it stands to reason that such infrastructure may help improve bicyclist safety"). Even if there were evidence to support the claim of greater safety for all, it does not follow that there is evidence of greater safety for bicyclists, given that bicyclists account for only 2% of U.S. traffic injuries and fatalities (U.S. DOT NHTSA, Quick Facts, 2017). For example, between 2000 and 2017, total US motor vehicle-related fatalities decreased by 11% while bicyclist fatalities increased by 13% (https://www.iihs.org/topics/fatality-statistics/detail/yearly-snapshot).
Language: en |