Do bike lanes really make biking safer in NYC?
By: Jason Ramel, Mateos Shehu, Jonathan Gonzalez
Issue
Biking in New York City is an increasingly popular means of transportation that offers an efficient way to reach a destination relatively quickly in a crowded city while getting a little exercise in as well. Over time, we’ve seen advocates calling out for bike lanes to be introduced throughout the city under the notion that they enhance bike rider safety. It is this very notion that we sought to test for this project. Our goal was to understand what impact bike lanes might have on biker safety in New York City as a way to understand if they are, in fact, meeting one of their safety goals. Our approach to measuring safety entailed assessing how reported bike crashes have changed over time. Thus, our first research question was:
Does the introduction of a bike lane improve bike rider safety?
Following this question, we sought to understand how different types of bike lanes might differ in terms of safety. New York City classifies its bike lanes into three categories based on their design. Class one bike lanes are those with some sort of physical barrier between a passageway for cars and the bike lane. Class two lanes are those that are separated from the passageway for cars with just painted signage or indicators separating the two. Class three bike lanes essentially share road space with cars and can be found in the middle of passageways for vehicles. With these different types of lanes throughout the city, it was important for us to understand what relationship might exist between bike lane classifications and their safety. Thus, the next question we asked was:
Is there a relationship between actual bike lane safety, as determined by number of accidents, and bike lane classification?
With these questions in mind, we conducted an analysis that helped us gain insight into how bike lanes impact bike rider safety in New York City.
Analysis Approach & Methodology
We approached our first research question (Does the introduction of a bike lane improve bike rider safety?) by taking a look at bike lanes introduced in 2018 and comparing the number of reported bike crashes on those streets in 2019 to the number of reported crashes in 2017. Since the most recent bike crash data we have is for 2019, we chose to assess bike lanes introduced in 2018 so as to have a full years’ worth of the most recent data to work with. We chose to compare 2019 data to 2017 crash data in order to avoid any possible errors that could be borne from crash reports overlapping from one year to another, despite when they actually occurred. Our focus was on understanding the percent change in crashes once bike lanes were installed as a way to understand if the introduction of bike lanes may yield more or less accidents. We calculated percentages based on the difference between the absolute number of crashes reported in 2017 and 2019 for each street with a new bike lane.
In addressing our second question (Is there a relationship between actual bike lane safety, as determined by number of accidents, and bike lane classification?), we focused on looking at 2019 bike crash data by bike lane classification. This approach allowed us to understand how each classification’s safety compared to its peers. An important measure in this part of our analysis involved normalizing our data to take into account lane length. The basic premise for doing so was that it’s not enough to present the absolute number of crashes for any given lane since it is likely that the longer the lane, the more bike traffic, and accidents, it might see relative to others. Thus, our normalization involved accounting for the number of crashes reported per mile of bike lane coverage.
We leveraged RStudio and QGIS to conduct our analyses. RStudio was used to prepare our dataset to be run in QGIS. QGIS was used to prepare the maps used in our analysis. Click here to view our R Studio project; click here for the steps followed in QGIS.
Visuals Explained
The map above shows how the introduction of bike lanes impacted biker safety for the studied period. The purple and yellow lines represent the 111 bike lanes that were introduced in 2018 in New York City. Yellow represents a decrease and/or no change in the reported absolute number of bike crashes on that street following the introduction of a bike lane, purple represents an increase in bike crashes following the introduction of a bike lane. We did not normalize values in this part of our analysis because we were measuring the percentage change in crashes between two distinct years, making it unnecessary to normalize for variables like lane length (which, as a reminder, we account for in answering our second question). The map shows that among lanes introduced in 2018, there are 72 yellow lanes and 39 purple lanes. What this tells us is that introducing new bike lanes in 2018 more often than not either reduced crashes or caused no worsening of street safety conditions. While perhaps unsurprising to bike lane advocates, this is an important finding as it helps to validate the importance of bike lanes while pointing out that they don’t always yield an actual reduction in crashes.
When accounting for each bike lane type overall, our results show that class one lanes saw an increase in crashes per mile of nearly 39% when comparing 2019 to 2017. Class two lanes saw a ~3% increase in crashes, class three lanes saw an 11% increase. Assuming bike lanes continue to be added and taking into consideration a 2016 study conducted by the National Association of City Transportation Officials that found bike ridership increases when protected bike lanes are introduced, we find that when just accounting for increases in crashes, class two lanes might be the best investment to consider¹. This is because of their lower requirement for physical infrastructure and their relatively low amount of increased crashes. Notably, while class one lanes provide the greatest increase in infrastructure, they also seem to result in the greatest increase in crashes. Class three lanes saw a lower increase in crashes relative to class one, but it still wasn’t as low as class two’s increase and, importantly, this lane class doesn’t accommodate rider safety as much as class two. Class two lanes balance the provision of reasonable infrastructure with a fairly low overall increase in crashes when taking into account an expected increase in ridership.
It’s important to note that our initial observations showed a greater number of streets that saw either a decrease or no change in crashes. However, when accounting for the three lane classes, we observed overall increases in the incidence of crashes when comparing 2019 to 2017. This can be explained by the fact that the initial analysis focused on the change within individual streets, while the latter analysis aggregated changes in bike crashes by each lane class. Thus, when comparing 2019 to 2017, a larger quantity of streets with new bike lanes saw a decrease or no change in crashes but the percent change in crashes by lane class increased overall, likely due to the increase in ridership.
The second map (featured above) shows how each bike lane classification differs from a safety perspective for 2019. The purple lines represent class one lanes, green represents class two, and yellow represents class three. The thickness of the lines represents the number of reported bike crashes relative to the length of the specific lane. What the map shows is that class one lanes, represented in purple, experienced the most amount of crashes for their lane length since they have the most amount of thick lines compared to their peers. Class two bike lanes have the second most crashes relative to lane length, followed by class three bike lanes. These results may be surprising given that class one lanes are built with relatively greater separation between bike riders and cars, and separation gradually decreases from class two to three. As a way to test these findings further, we took a look at the total number of crashes reported for every bike lane class and compared it to each lane type’s total street coverage. We found that class two bike lanes (green lanes) are, overall, seeing more accidents relative to their total street coverage, followed by class three, then class one (see table below). What this tells us is that there are bike lanes that have more dangerous segments than others, even when a bike lane might be more dangerous overall.
Reflections
It is important to note that these findings are not conclusive and do not represent the totality of the bike riding experience in New York City. By virtue of assessing specific moments in time, our analysis manages to capture how bike riding may or may not in fact be safe depending on the built environment around it. With that in mind, it’s also important to consider that our findings show that there are certain types of bike lanes that appear to have more dangerous segments than others. This is important to consider as it supports a localized approach when it comes to deciding what type of bike lane would serve a specific area best. When making decisions at a higher level, broader sets of data might imply that certain design approaches might be better than others when that may not necessarily be the case. For example, a higher level review of bike lanes in New York City might have inferred that class one bike lanes are most safe, and that class three bike lanes aren’t. However, the localized view of Manhattan showed that class one lanes actually have more unsafe segments compared to other lanes. This, again, highlights the need for a localized approach when assessing how a built environment, especially for bike lanes, should be designed. It’s also important to consider the trends observed amongst the different bike lanes when leveraging a localized design approach. So, while class two bike lanes were more dangerous overall, they are less dangerous at a more local level (compared to class three), and they saw the lowest increase in crashes, even as bike lines are presumably being added throughout the city. Not to mention, they also offer relatively reasonable infrastructure requirements.
All this said, it is worth noting that our findings do not determine the direction of the relationship between bike lane safety and bike lane classifications. For example, it could be the case that class one lanes are on streets with lots more reckless drivers. Importantly, however, we must note that our review of the introduction of bike lanes in 2018 generally saw more decreases in reported crashes than increases. While this finding is only as reliable as a snapshot in time is, it’s worth considering when making street design decisions and speaks considerably to the benefit of adding (the right kind of) bike lanes.