“Safe for all road users”
Debunking the commonly cited bullshit Marshall and Feranchak “study” that “protected” bicycle lanes make roads safer for all road users
“This paper provides an evidence-based approach to building safer cities,” begins the conclusion of the study, “Why cities with high bicycling rates are safer for all road users1” a study conducted by academics Wesley Marshall and Nicolas Feranchak. Published in 2019 in the Journal of Transport & Health, this study has become one of the go-to claims of evidence for the safety of “protected” bike lanes and the push to build them.
Here are a few instances of their “study” being cited as gospel in the wild:
Repeated studies consistently show that adding a protected bike lane to a road improves safety for all road users - including drivers, walkers, and bikers. Research shows that adding protected bike lanes reduces all collisions and injuries by 30-50%. The reason is simple: roads are easier to navigate when everyone is in their place. There can only be peace on the road, when everyone has a piece of the road. (Pasadena Complete Streets Coalition)
Physically separated bikeways, which have been shown to reduce fatalities for all road users, are an element of many Complete Streets designs. (CalBike)
“Bike lanes reduce crashes and fatalities for all road users and make the roadway more comfortable for both motor vehicle drivers and people riding bicycles.” (Coalition of Arizona Bicyclists)
“Cycling lanes, not cyclists, lower road fatalities” (CU Denver Press Release on the study)
Protected separated bike facilities was one of our biggest factors associated with lower fatalities and lower injuries for all road users. If you’re going out of your way to make your city safe for a broader range of cyclists … we’re finding that it ends up being a safer city for everyone. (Marshall himself in a Streetsblog interview)
Marshall even cites his own study in his recently released book Killed by a Traffic Engineer: Shattering the Delusion that Science Underlies Our Transportation System (review to come someday) by saying:
On average, you are about 20 times more likely to die when walking a mile than driving a mile. Bike a mile? That’s maybe ten times more dangerous than a mile in a car. But guess what happens in cities with lots of people walking or biking? They end up as our safest cities, far safer than conventional cities where everyone is conventionally driving.
Worst of all, the study was cited by former Encinitas Mayor and now California State Senator Catherine Blakespear when she was looking for political support for her atrocious Cardiff bikeway project.
“It’s abundantly clear to many of us who are recreational cyclists that we'd like to feel safe riding kids to school, the beach, dinner, or work. And the typical road infrastructure of painted bike lanes next to speeding traffic doesn’t make us feel safe enough to choose to ride a bike, especially with a child on board.
This perspective is backed up by studies. Last May, Separated Bike Lanes Means Safer Streets, Study Says summarized a study of 12 large metropolises, finding 44% fewer deaths in cities with protected and separated bike lanes:
’Researchers found that bike infrastructure, particularly physical barriers that separate bikes from speeding cars as opposed to shared or painted lanes, significantly lowered fatalities in cities that installed them.’ “ (link)
DownloadThe Marshall and Feranchak’s study contains more nonsense than truth. It’s laced with significant errors and the usual “garbage in, garbage out” bullshit from overly complicated non-repeatable models2. The only contribution it makes to the “science” of bicycling or traffic safety research is as an example of what not to do similar to the long laundry list of studies on bicycling safety by epidemiologists and public health activists M. Kary critiqued over ten years ago.
Yet, it’s this study that gets repeated as gospel by “protected” bicycle lane propogandists while taxpayers are being taken for a ride, both in funding these types of studies and the radical changes these activists want to do to the public road network but also the misinformation towards bicyclists, novice and experienced alike on safe cycling practices. It provides the perfect comeback for proponents of “protected” bike lanes when others make claims about their safety becoming an important part of The Science™. The study also gets repeated to the audience of non-cyclists. “See, this stuff makes things safer for you in your car too.”
But before demolishing this study, it’s necessary to review exactly what Marshall and Feranchak did in Why cities with high bicycling rates are safer for all road users.
Marshall and Ferenchak’s study investigates the supposed safety of cities with high bicycling rates (which is the goal of many bicycle activists, public health activists, and climate crisis activists who believe that bicycling is not common due to the lack of “safe” infrastructure which would make people feel safer) and examines whether these cities are safer due to the “safety-in-numbers” effect3 of more bicyclists or other factors such as the built environment or demographic and economic characteristics.
The study uses 13 years of data from 12 large U.S. cities, covering over 17,000 fatalities and 77,000 severe injuries across approximately 8,700 Census block groups. These crash data included whether those involved were bicyclists, pedestrians, or motorists (and presumably also motor vehicle occupants) along with their geographical location. For the bicyclist crashes, they looked at what type, if any, the crash site had for bicycling infrastructure. Interestingly missing from their analysis were any of the factors leading to the crashes, which is typical in these types of studies and is one of the key critiques M. Kary made in his commentary “Unsuitability of the Epidemiological Approach to Bicycle Transportation Injuries and Traffic Engineering Problems” on the various other studies that are supposed to support bicycle activism.4
The researchers used negative binomial regression models for these over-dispersed count data (fatalities and severe injuries for motorists, bicyclist, and pedestrians). They “tested” each hypothesis category (travel behavior, built environment, socio-demographic/economic differences) independently before combining them in full models. They used multilevel “modeling” to account for the hierarchical nature of the data (city and block group levels), addressing issues like spatial autocorrelation.
By punching all these data into the “models”, the results showed that the presence of more bicyclists did not significantly improve safety for all road users, countering the “safety-in-numbers” hypothesis.
Second, Marshall and Ferenchak’s results showed that the density of “protected bicycle lanes” were strongly correlated with fewer fatalities and severe injuries and that higher intersection density (the spacing between intersections), which they believed often led to lower speeds by motorists improved safety. It’s really this second item that the activists clench their teeth on in terms of their activism.
Third, their results indicated that areas with higher median income and a greater percentage of white residents tended to have lower fatality and injury rates. The authors believe these findings “are suggestive of equity disparities that deserve additional research.” Given that bicycling advocacy is immensely woke and driven by racist, sexist, and ageist AWFLS, this too is probably another item for teeth clenching.
Well dear reader, if you clicked on the bios for either academics above, you’d notice these results are precisely what they likely wanted to find. Marshall and Ferenchak, both degreed Civil Engineers (an embarrassment to this author’s profession) and are far closer to activists than anything resembling engineers or scientists in hard science fields. Personally I wouldn’t trust these two to mix even a bag of Quickrete and thank whatever benevolent being or force you believe in that they are not structural engineers.
And to get to the meat of the criticism, doesn’t it seem odd that one type of infrastructure (“protected bicycle lanes’) can have such a large effect on the behavior of a variety of road users?
This oddity, and likely many more issues with the study struck Dr. Paul Schimek, a fellow traveler in the skepticism of “protected bicycle lanes.”
Schimek is known in Principled Bicycling circles for his excellent evergreen The Dilemmas of Bicycle Planning article, informative articles geared towards motorists about safe and legal bicycling behavior, a guide to improving bicycling laws (an issue the Platitude-driven and Cluster B(ike) activist fails to address), and comments about children cycling which abolish the fake concern troll arguments of activists jokes such as
. Among Schimek’s strongest points are in his harsh critiques of “bicycle friendly” infrastructure such as door zone bike lanes, and last but not least, “protected” bicycle lanes. His long form bicycle collision studies for both the states of Oregon and Pennsylvania are must reads being solid examples of how to study crashes the correct and objective way. As an academic researcher, one of the good ones, he also has the credentials and ability to pierce the world of academic journals.Paul Schimek’s critique, published as a letter5 in the same journal which published Marshall and Feranchak’s study highlights several methodological flaws. These flaws are good enough for the average activist and for the corrupted pay-to-publish journals common in so much of “Science” and academia, but he proves the study doesn’t stand up to reality.
Among the issues was how Marshall and Feranchak defined “protected bicycle facilties” The study’s category of “protected/separated bike facilities” included both trails/shared-use paths which tend to be located off public highways and in parks and protected bike lanes, which are not the same. Some of the cities of which they asserted the installation of “protected bicycle facilites” were the driver of said city’s elevated safety “for all users” had minimal protected bike lanes (only 12.5 miles across three cities), while trails accounted for more than 1,300 miles. This blending of different facility types by Marshall and Feranchak is nearly impossible to be accidental (they claimed they cross checked the data manually) and misleads readers, as trails and bike lanes serve different functions and contexts. Crashes on off-highway trails also do not typically get recorded in official traffic collision databases and thus would not make it into their study. This alone, if the peer reviewers and the journal were actually doing their job instead of being activist hacks, would get this paper rejected.
Second was, oops, Marshall and Feranchak’s fancy pants model also found that “regular” bicycle lanes, the ones that offer no “protection” had the same effect as their category of “protected” types. This is buried in the data and results but missing from the abstract, conclusion, and the press releases. It was only Schimek who noticed this.
Third is an issue with Marshall and Feranchak failing to understand basic statistics. (
could perhaps help them out with that). Schimek in his letter points out that, with the study’s large dataset of over 112,000 observations, finding statistically significant results can be easier but does not imply causation. The study’s findings of significance could result from large sample size rather than meaningful relationships.Schimek wasn’t done with the dirty stats stuff. He noted the models did not control for critical confounding factors such as population density and driving speed, both major influences on road safety. The selected cities varied significantly in density and driving habits, which could possibly affect crash rates. An example he cited was that denser cities generally have lower fatal crash rates than those with sprawling, low-density layouts.
Another critiques comes from elementary statistics in that the study’s model treats data as if cross-sectional, ignoring the repeated observations over time, which can bias results by underestimating standard errors. Using fixed effects for each city could better account for city-specific differences. Schimeck also noted the author’s low p-values, a common trick used by “researchers” to make their data appear valid against their hypothesis.
Schimek then noted the obvious in that the study suggests improved safety for all road users implies greater safety for bicyclists, yet bicyclists are only a small fraction of total traffic injuries and fatalities. Schimek argues that safer outcomes for other road users do not necessarily indicate safer conditions for bicyclists, especially given that while overall traffic fatalities decreased in recent years, bicyclist fatalities have actually risen. That means there may be other variable involved that surprise, Marshall and Feranchak left out of their model.
His entire critique is below.
And should accompany any rebuttal to anybody who cites this bullshit study.
Speaking of bullshit studies, if this one wasn’t enough, check out the takedown of Ian Walker’s “carbrain” study, featured on a recent episode of the Dark Horse Podcast hosted by
and Bret Weinstein. The goods start at about the 44 minute mark.
The “safety in numbers” hypothesis posits that the more bicyclists there are, the safety said bicyclists are. “Safety in numbers” was popularized as gospel into the bicycling advocacy world by Peter Jacobson but Jacobson provided no real evidence for the phenomenon and John Forester and others discovered the same trend may be due to “faulty math” and random chance.
This is also a common divide between the Platitude-Driven/Cluster B(ike) approach to crash analysis (which doesn’t exist because these camps focus on feelings and platitudes) and that of the Principled Bicyclists/”Vehicular Cyclists”/Bicycle Drivers who approach bicycle crashes with objectivity. As
might put it, one approach is fueled by trait empathisation while the other by trait systemization.