In 2004, I was working a summer job as a bike messenger in Boston when I was struck by a car in an intersection. I ended up in the hospital, undergoing extensive surgery to have a metal plate inserted from my left wrist up my forearm—I can still feel it whenever I type. Lying there on the operating room table, I wondered if this would have happened if there were better protections for cyclists.
The whole premise of bike lanes is that giving cyclists a dedicated space will allow them to ride safely and coexist with vehicle traffic. And yet, bike lanes don’t always make cyclists safer—and sometimes they offer no protections at all. New York City, for example, has reported 34,112 collisions between cyclists and motorists since 2014, despite the addition or enhancement of about 330 miles of bike lanes (with 27 cyclists killed in 2019, the most of any year since 2000).
As COVID-19 inspires more people to use bikes as a transportation option, it seems likely that accidents—and fatalities—will rise as well. We need to examine whether the bike lane improvements that cities are investing in actually make roads safer for cyclists. And if they’re not, why? How can cities design them to be them safer?
Crunching the numbers on crashes
I teamed up with Smart Design’s engineering intern Anna Bernbaum to analyze bike lanes using open data sets from three major cities: New York, Boston, and San Francisco. We looked at each city’s number of collisions before and after bike lanes were installed, adjusting to account for the overall increase in ridership (which of course would make the number of collisions go up).
We also wanted to know if bike lane design was a factor. During our study, we were able to identify three main types of bike lane improvement: conventional lanes designated with painted white lines, protected bike lanes separated from car lanes by a physical buffer, and painted whole lanes (usually green) with icons called “sharrows” indicating where cars and bikes share the lane. Would the data show a significant difference between these improvements?
Finally, we looked at whether collisions tended to occur in intersections or on streets. We combined collision data since 2012 with bike lane improvements over the same period. (You can explore the relationship between collisions and bike lanes in our online interactive heatmap.)
The results of our study were undeniably compelling. As we expected, lanes protected by physical barriers resulted in a decrease in collisions over time, while the most common improvements—conventional painted lanes—did little to affect collisions much either way. But there were some surprises, as well: While collisions in San Francisco decreased a fair amount after bike lane improvements, New York City’s bike lanes experienced a marked increase in collisions over time—especially, it seemed, in its “sharrow” lanes. How could that be?
What’s more, while all three cities tended to see the bulk of their collisions occur around intersections, Boston’s intersections suddenly saw a dramatic decrease in collisions after 2015. When we drilled down further, we found out why.
How bad design leads to more collisions
At Smart, we always conduct both quantitative research to find out what’s happening and qualitative research to see why it’s happening. We knew from the quantitative data that collisions had increased along New York City’s sharrow roads. But to find out why, we used Google Streetview to see what was happening. When we did, we discovered that Boston painted the sharrow icon in the center of the road, while New York City painted it on the side of the lane, where only cyclists would see it. In other words, where the design was geared toward motorists, collisions went down; where the design was geared toward cyclists, collisions increased.
Our investigation also showed that Boston had developed a design system for its intersections called “bike boxes,” with areas painted a vivid green and marked with icons to show where bicyclists could make turns or wait for the light to change. This was a pretty compelling explanation for why collisions in Boston’s intersections dropped.
Last, we honed in on some of New York City’s most problematic intersections to see if there might be a design explanation. Google Streetview has a feature that let us see an intersection over time, which we correlated to rises and falls in its numbers of collisions. We often saw intersections that had been repaved and then the bike lanes were only partially repainted or not repainted at all. It seemed apparent that New York was not considering bike infrastructure when creating construction plans, resulting in confusing or absent markings in intersections, and more collisions.
The importance of smart data collection
There was another important discovery we made: While collision data for intersections in Boston and San Francisco was straightforward, New York City proved difficult. The reason being that, where the other two cities specified intersection versus street in their collision reporting, New York only gave longitude and latitude of each crash. Fortunately, Bernbaum found a clever workaround using an algorithm method called “R trees” to draw a sort of bubble around each route. Any collision that happened within that bubble could now be assigned to a corresponding bike route. We were able to see a more comprehensive picture this way, but it was a reminder that solid, reliable data depends on cities collecting as many pertinent details as possible in its reporting.
So are bike lanes making cyclists safer? The answer is: It depends. And design choices have everything to do with it. Based on our qualitative and quantitative data analysis, here are three fairly simple design moves cities could make right now in order to decrease collisions.
- Repaint “sharrows” in the middle of the road, not on the side, so motorists see them.
- Prioritize bicyclists with dedicated, clear markings through intersections.
- Always update bike lane markings in intersections after construction.
Safety measures don’t have to be a guessing game. By collecting good, detailed data, cities can empower more studies like ours and come up with actionable, effective solutions that better protect people.
John Anderson is director of technology at Smart Design and a technologist-in-residence at the Berkley Center for Entrepreneurship at NYU Stern. Anna Bernbaum is a design engineering intern at Smart Design and recent graduate of Imperial College London.