Schrödinger’s bike lanes: Are they safe or not?
Cantabrigians may recall the famous thought experiment by physicist Erwin Schrödinger in which an imaginary cat is placed in an imaginary box, unobserved, with a vial of deadly poison and an unstable radioactive trigger. Without observation, we cannot know whether the cat is alive or dead. From a quantum mechanics standpoint, we might come to the conclusion that the cat is simultaneously alive and dead – so long as it remains unobserved. The moment we open the box to make an observation, reality asserts itself. The cat must either be alive or dead; it cannot be both.
Duality, of course, is common in nature. Light can be a wave; it can also be a particle. Peanuts are healthy for many; a deadly allergy for some. And bike lanes, it seems, can be safe and unsafe simultaneously, according to various analyses promulgated by Cambridge residents as well as other users of our roads. How can these seeming contradictions coexist?
Some of the confusion has been caused by simple errors in data analysis. The cycling advocacy group Cambridge Bicycle Safety recently wrote (“It’s no surprise Cambridge roads are getting safer,” May 22) claiming that Cambridge crashes resulting in EMS transport to the hospital had decreased by nearly 40 percent since 2016. Independent analysis shows this claim is factually incorrect.
According to the Cambridge Police Department crash log last updated May 31, there were 201 crashes in 2016 resulting in one or more hospitalizations, compared with 203 in 2022 – almost identical, and a far cry from “stark reduction” alleged by Cambridge Bicycle Safety. In fact, crashes resulting in hospitalization have averaged 205 each year since 2016, with very little variation (standard deviation ±4.2) aside from the pandemic years of 2020-2021.
Armed only with Microsoft Excel and an internet connection, any private citizen may verify these numbers for him or herself simply by downloading the latest version of the Cambridge Police Department Crash Log in CSV format; sorting by column J (hospitalizations) and deleting any entries of 0 to filter for only crashes with 1 or more people transported to the hospital by EMS; using the YEAR function to parse the year of occurrence from Column A (Date and Time); and finally using the FREQUENCY function to create a frequency table of crashes per year. This should result in the graph above, which unfortunately bears minimal resemblance to what was shared in Cambridge Day by Cambridge Bicycle Safety advocates:
Despite this data analysis flub by Cambridge Bicycle Safety, the overall goal of cyclist safety is noble and does not deserve to be overshadowed by the mistakes of any individual. Cyclists are an important segment of road users, accounting for nearly 7 percent of Cambridge commuters, according to U,.S. Census Bureau data from 2021. Whether one travels by motor vehicle, bus, train, foot, bicycle, scooter or any other mode of transportation, road safety should be the goal of any reasonable citizen. That’s why it’s so important for the City of Cambridge to investigate our current cycling infrastructure carefully and determine whether safety concerns remain.
At the heart of the bike lane conundrum is the presence of natural variation. Numerous scientific studies have demonstrated a trend of increased cyclist safety with the installation of cycling infrastructure such as separated bike lanes. For every average calculated, though, there is natural variation around the mean – just as genetic variations and environmental characteristics combine to make some of us skinnier or wider than the population average, some bike lane installations will inevitably perform better or worse than others due to the unique characteristics of each locale – road configuration; intersection density; even socioeconomic factors, according to some studies. News reports frequently ignore this nuance, yet measures of natural variation are present in almost all scientific papers.
For instance, a recent study by the Federal Highway Administration notes that the type of quick-build separated bike lanes being deployed around Cambridge (in which traditional bike lanes are replaced by separated bike lanes with a blend of flexible delineator posts and other vertical elements, such as parked cars) should theoretically reduce bicycle crashes by an average of 36 percent (crash modification factor of 0.640). In the very same data table, however, the researchers were careful to note that the standard error of their estimate was 0.203 – in other words, significant natural variation was observed in bike lane effectiveness, with some percentage of bike lanes expected to actually result in crash increases (i.e., crash modification factors above 1.000).
The natural variation reported by Federal Highway Administration researchers could explain the seeming paradox of recent observations by John Hanratty and the Cambridge Streets For All neighborhood group, who reported that some Cambridge neighborhoods appear to be experiencing higher levels of injury-causing crashes since bike lane installation.
Scientists have identified some (but likely not all) of the risk factors that raise bicycle lane crash rates. For instance, in the same Federal Highway Administration study, researchers determined that bicycle crashes are more likely when bike lanes are installed at locations with mixed land use (for instance, commercial-residential neighborhoods such as Hampshire Street, Main Street or North Massachusetts Avenue). Another recent study in the journal Accident Analysis & Prevention provides additional variables to examine, identifying features such as frequent driveway and road intersections, lack of continuous separation and increased complexity for turning drivers as major risk factors for protected bike lanes. And a 2022 piece in Transport Reviews reported that single-bicycle crashes, which account for 52 percent to 85 percent of cyclist injuries (based on hospital and emergency department data), are most commonly caused by road maintenance issues such as uncleared ice and snow, fallen leaves or potholes.
In the context of natural variation, a location-specific or neighborhood-specific approach becomes key. Whereas aggregated multicity studies provide generalized information on cycling infrastructure effects averaged across a wide area, localized data allows city planners to home in on specific problem areas and come up with neighborhood-specific solutions. The analysis done by Hanratty and Cambridge Streets For All takes some steps toward this approach, but could be significantly refined with additional information such as monthly location-by-location traffic counts for motorists, cyclists and pedestrians. Such data is not available to the Cambridge public, in large part because the city has not collected it.
For a city such as ours, populated by so many world-class scientists, engineers and experts of every stripe, it seems a travesty that we should leave any issue unresolved for lack of data – particularly, an issue so pressing as public safety. Yet today, in Cambridge, it is not the welfare of an imaginary cat that we have failed to check in on, but rather the lives and livelihoods of our own residents, as well as the many commuters, customers, service providers and others who use our streets each day. Quantum superposition does not work at the city level – we cannot remain both safe and unsafe while debating endlessly with incomplete data; we must open Schrödinger’s box, make the necessary observations and collect the necessary information to determine just what is going on on our streets. And from there, we can fix it, because one thing Cambridge is not short on is invention and ideas. We just have to take that crucial first step.
Serenus Hua, Hampshire Street
It’s simply a matter of political will. The argument has been presented as either you are pro bike lane and therefore anti-car murder or you are against bike lanes and thus pro-car murder. I would need more exact data but most business owners don’t live in Cambridge and can’t vote here this as a constituency they are traditionally very weak.
I would just like to go on record as stating that I am pro-car murder. There are FAR too many humans. Car murder thins the herd of the slow and the not-all-that-bright.
I wrote some of the CBS report, and I actually noticed this discrepancy and took into account: the Police annual report doesn’t match the crash log. (There were other discrepancies as well, e.g. entries were someone was marked as being transported to hospital but zero people were marked as injured.) So I talked to the Police Department analysts, and they explained the difference.
The data in the crash log is inaccurate, because it’s transcribed from written reports, and things get missed when it gets added to the computer.
Thus, every year, when the Police Department does their crash report, they manually go through every written report and make sure it’s accurate. As a result, the Police Department annual reports are more accurate than the crash log.
This is explained in the full report, in the last paragraph: https://www.cambridgebikesafety.org/2023/05/16/the-experts-were-right-cambridge-roads-are-getting-safer/
(I would’ve asked if they can go back and make the crash log more accurate, but apparently they’ve been trying and failing to hire a traffic analyst for 3 years so they are very understaffed.)
This is why you don’t just spend 5 minutes opening an excel spreadsheet to analyze data. You need to actually get a sense of how accurate it is (_every_ data set has garbage for a variety of reasons), which means you need to dig in and get a feel for its limitations and inconsistencies.
Collecting more data is of course always valuable, and it would be great to have, and I hope the City collects more. But it’s pretty clear this is just the excuse de jour for why we can’t make the city more safe.
Luckily we have the ability to observe the subject and upon doing so we can see that separated bike lanes do increase safety, comfort, and ridership.
A very thought provoking article. Where is the city administration on this issue of separated bike lane safety? Sitting on the sidelines letting residents do the analysis the city should be doing. Or are they relying on the oft quoted studies from other cities that separated bike lanes are making our streets safer. Moving forward with installations without taking an objective, unbiased look at safety is really bad government particularly when $50 million is budgeted for more separated bike lanes on Mass Ave. Councillors are quick to point out that most of the budget is for median removal and underground relocation of utilities but NONE of it would be happening if it were not for the mandate to install separated bike lanes. Cambridge residents need to pay attention to how our tax money is being spent and what we are getting in return. The City Council told us our streets would be safer and more persons would feel `comfortable’ cycling. So where is the proof?
Other things to know about the police crash logs:
* There’s actually two different ones in the Open Data Portal. One has transports to hospital, the other is much more detailed and matches the state reporting requirements, but lacks the hospital transports.
* The two logs are a little inconsistent; there’s a bunch of post-processing done to get the pedestrian/bicyclist injury counts for the simplified one, and it’s a rather ugly data format. And has some… subjective choices officers on the spot have to make.
* Sometimes mistakes happen; I’ve noticed a bike crash that was catalogued as a pedestrian crash (or maybe vice versa, been a while).
* Councilors get emailed crashes so sometimes that provides a lot more data and can help catch inconsistencies. I think that’s how I noticed the one above.
* There’s another data source useful for cross-validation of individual instances: the police daily log. It doesn’t record everything though, I don’t really understand which crashes end up here.
Anyway, as far as I can tell the police annual report (due out for 2022 any day now) tends to be the most accurate for traffic crash purposes because of the manual validation of all data.
And another constituency ignored by the city on bike lanes is the pedestrians, especially the slow walking seniors and those with mobility issues. The new bike lane configurations, if they keep parking at all, they often put cars into the middle of the street. To cross we walkers have to dodge bike lanes first, and then see around cars parked in the middle, to finally get across. Its double jeopardy for walkers
@MC Resident, the study cited here is new and was developed by transportation experts at the Federal Highway Administration. It is developed using data that includes the recent quick build separated bike lanes in Cambridge. The report importantly includes enough data to distinguish between signal and noise (due to natural variation or other biases that Serenus mentions). Could you share any specific concerns that would prevent you from believing the well-produced and federal transportation study on this exact topic?
https://highways.dot.gov/sites/fhwa.dot.gov/files/FHWA-HRT-23-025.pdf
Yikes. Probably should have left this one in the draft folder
The city has finally started prioritizing everyone’s safety over the convenience of car owners. Like indoor smoking, we aren’t going back.
Here is what I don’t get: So many comments from folks against redesigning roads to make them safer for bikers and pedestrians, say or hint at this being unpopular or being pushed through by some motivated minority over the wishes of “the many”. This really should require some evidence. As I understand it, the policy was endorsed in a referendum and councilors that are pro making our roads safer have gotten re-elected. I recognize that municipal elections are a low turnout affair, but it’s the best data I have seen. Would love to see some backup behind the idea that this is even unpopular.
NC Walker- I’m a walker myself and I just… cross at the crosswalks. Obviously still the risk from bikers running red lights, but no parked cars in the way!
@cportus, you are correct, there are many many supportive residents who would like to see safer streets and protected bike lanes. The city runs carefully organized community surveys about a variety of topics every two years, including city services, roads, healthcare, safety, etc.
2/3 of Cambridge residents said that they want to see protected bike lanes installed to make our streets safer [Question 56: http://cambridgema.iqm2.com/Citizens/FileOpen.aspx?Type=4&ID=6136%5D. There are also many people who walk and drive who want to see the roads be safer in general for every user, not just people who bike.
This is a scientific survey — the city calls individual residents or provide them an online link, to ensure that the results are representative across various demographic factors.
The data around protected bike lanes and bike use in general, to include safety, is spotty at best. The only really relevant question is “Do they make the world safer?” In my experience, as someone whose family bikes year-round all over Cambridge, they clearly make bicycling more dangerous. That is different from saying the status quo was acceptable, but the protected bike lane program is a failure when it comes to improving bike safety.
What to do? Enforce traffic laws and remove street hazards. Those are the two big things the City is doing neither. We probably need peak hour speed limits for bikes and things like bikes, too. As an older, slower cyclists, I am terrified of getting knocked over by someone passing me inches from my handlebars or smashing into me when they ignore the red light I stopped for.
Since the protected bike lanes have been put in, these things have happened to me for the first time, all while I’ve been riding in a protected lane:
1. Last week I was doored in a protected lane. An uber pulled into a driveway and the passenger opened the door as I was passing. In 30+ years of riding in Cambridge, I had never been doored before. It was scary but fortunately I was not hurt.
2. A few months ago I got in a head-on collision with someone riding the wrong way in a protected bike lane in Harvard Square. We managed to slow enough so the collision wasn’t horrible, but it was still pretty unnerving.
3. About a year ago, a wrong-way cyclist forced me to a stop in a protected bike lane in Porter Square to avoid hitting him. When I pointed out he was going the wrong way, he threatened to kick my ass. Who needs that on their way to work?
In my bicycle group, one member has fallen not once but twice after hitting broken bollards that so frequently litter the protected lanes. As he put it, he was not commenting on whether the protected lanes made biking safer or not, but these two falls only happened because the City has put in, and poorly maintained, separated lanes.
The many close calls both my wife and I have had while biking and walking through Inman Square, the ice that chokes many of the protected lanes during much of the winter, the broken pieces of bollards that stick up out of the ground on the left side of protected lanes like tank traps, the challenges of making a left hand turn from inside a protected lane and many more issues are all clear indicators that our bike infrastructure and safety policies need a deep review. Even the dimensions of our separated lanes are dubious, being narrower than state guidelines in many places and not reflecting that design safety standards places like Portland, OR are realizing are needed because of the increase in use of large, fast, battery operated “bicycles,” mopeds and scooters.
We are putting in infrastructure that is going to last a very long time. We should be doing a lot more work to make sure that what we’re installing is the best way to make our streets safer for everyone. To me it is very clear that it not only isn’t making my life as a cyclist safer, it’s making things more dangerous.
Has CambridgeDay just become an outlet for the NextDoor crazies to have a bigger platform? This person has been laughed off of that platform for showing they have very little knowledge on how to appropriately vet and analyze data. As Itamar Turner-Trauring pointed out, just pulling random stats without background knowledge of the data collection process is sloppy and irresponsible.
The best quality data we have is from the FHSA showing even quick build lanes increase safety.
CambridgeDay needs to stop giving these people a platform to peddle their conspiracy theories or rants (like Barbara Anthony’s baseless ranting last week).
Seriously, do better.
There is a reason why people go to school to learn how to analyze data and what happens when you don’t.
The author used an inaccurate, incomplete data set and didn’t bother to check if better data is available. There is better data available. You just have to look for it.
The author mentions the Federal Highway Administration report but misapplies the data from it. The author misuses standard error. It doesn’t make sense to say “Standard Error shows that there is variation”. Every measure has variation.
Standard error to see if what you are measuring is real (i.e., it is statistically “significant”). The FHA numbers provided by the letter’s author show that the reduction in bike crashes (0.640) is, in fact, larger than the natural variation (0.203). In other words, bike lanes do reduce crashes.
The FHA report shows that this is statistically significant at the >95% confidence level. The author left that important detail out.
In fact, the FHA report showed that many different types of bike lane infrastructures significantly reduced crashes. That was not mentioned in the letter.
The author referenced the John Hanratty and the Cambridge Streets For All report. The serious flaws in that report have been well-discussed in this forum and others. They failed to take into account basic stuff like changes in traffic patterns. That is why their conclusions were at odds with many, many other properly conducted studies.
I agree with the author’s statement that “confusion has been caused by simple errors”. This letter is a good example.
What I don’t see in the above data is the important question that needs to be applied (and hasn’t) is are the crashes in questioning happening on the streets that there has been a redesign or application of new bike lanes?
Looking at the street data specifically involved rather than a city wide data point would be more important in telling whether there has been any change in results.
Right now we have data that neither proves nor disproves anything pertaining to the subject of the redesigns and bike lanes.
Itamar, the headline statistic of your analysis (40% crash reduction since 2016) relies on the very same automated Cambridge Police Department crash log that you have just spent so much time trying to discredit.
You switched sources right in the middle of your analysis – manual crash report data for 2016-2021; automated crash log data for 2022. Note that the manually-tabulated crash report data is consistently higher than the automated crash log data, by about 20-50% each year. So, it’s obvious 2022 is lower only because you switched data sources for that year alone.
It’s pretty disingenuous to claim that a data source is inaccurate for every year (2016, 2017, 2018, 2019, 2020, 2021) *except* for the one year you need to your make your headline statistic.
Again, I applaud your passion for cycling, but please try to use more rigor in your future statistical analyses.
@ssrhua For all your talk of “rigor”, your letter is totally misleading and incorrect.
You talk about “natural variation” but neglect to point out that the FHA report found that bike lanes reduced accidents *above and beyond the natural variation*.
Standard error doesn’t show “significant” variation. It shows variation. You then compare the standard error to your effect o see if the effect is real (larger than the standard error). If it is, the effect is statistically *significant*.
That is what significance means in statistics. That’s how you use the standard error.
The FHA report found that the bike lanes resulted in a *significant* reduction in accidents. It’s right there, in the report.
In other words, the reduction in accidents was *larger* than the “natural variance”. To turn that around to argue that it shows bike lanes are not effective in reducing accidents is wrong. Head-spinning wrong.
Your talk of variation explains nothing.
@FrankD exactly, either they aren’t aware how illogical and misrepresentative their analysis is, or their intentionally misrepresenting what the data says to support their argument.
Either way this letter shouldn’t have been given attention by CambridgeDay.
@shedrovemehere. Indeed. This letter by Serenus Hua reflects a lack of understanding of basic statistics. The author was also involved in the Cambridge Streets For All report that was widely mocked for its many flaws.
These people simply don’t know what they are doing. They seem to think that can just wander into data analysis without knowing anything.
Another commenter said that Serenus was widely panned and critized on NextDoor for sloppy and error-ridden misapplication of data. I believe it.
Cambridge Day needs more quality control for their letters section.
The author of the letter (Hua) also makes other mistakes.
The Standard Error is not a simple measure of variance. It also depends on the amount of data you have. The more data, the lower the Standard Error.
Thus, you cannot use it as a stand-in for “natural variability” as Hua claims. That is plain wrong.
The Standard Error is used for comparison purposes. It tells you the probability that the effect (e.g. the average reduction in accidents) differs from the population average (no effect).
You can’t and don’t use it as a reflection of “natural variability” as Hua did.
The FHA study he refers to shows clearly that there is an effect. Bike lanes reduce accidents with a confidence level of >95% to >99%.
Maybe Cambridge Day can find someone who actually knows statistics, that can recognize poorly done, misapplied data methods. Then they can avoid publishing such nonsense.
“FrankD”, “AvgJoe” – I’m sorry to be the one to tell you, but your understanding of statistics needs a bit of work. In particular, your explanations of how to interpret standard error would be problematic in most situations, but is especially incorrect in the context of the FHA study, which uses a Poisson regression (which is why they use the very specific term, ‘standard error of the estimate’). Comparing 0.640 > 0.203 to draw conclusions is also completely inaccurate, for a number of reasons that would take quite a while to fully unpack and explain, but the most simple of which is that a CMF of 0.640 corresponds to 64% as many crashes, not a 64% reduction in crashes (that would be a CMF of 0.360, instead).
It’s probably a bit out of scope here to provide a statistics tutorial; however, in lay terms, measures such as standard deviation or standard error of the estimate tell you *how much* natural variation you have (or simply variation, if you prefer). Only ~68% of the population is within 1 standard deviation of the mean; ~95% within 2 standard deviations; ~99.7% within 3 standard deviations. The larger the standard deviation, the wider the spread of outcomes in that population. Any reader may do their own calculations to figure out just how many standard deviations would be required to get to a crash modification factor of >1.000, given an average CMF of 0.640 and a standard error of the estimate of 0.203.
As I already stated above, the average separated bike lane may well be safer than the average traditional bike lane. That does not prevent some traditional bike lanes from being safer than their peers, or some separated bike lanes from being less safe than their peers. Populations naturally vary.
Perhaps “FrankD”, “AvgJoe”, or “shedrovemehere” would like to share their real names and professional qualifications, so that readers can judge whether they have any relevant statistics or data analysis experience. My own professional qualifications are a matter of public record, as is my field of research (which inherently involves extensive use of statistics) and the numerous peer-reviewed research papers as well as other scientific advances resulting from it.
@Serenus: Yet, somehow you keep offering flawed analyses.
The standard error values you talk about are the same numbers used in the FHA to show that bike lanes *significantly* reduce accidents. Yet, you twist the argument to conclude that they don’t.
There must be a reason why your letter and the Hanratty report come to conclusions that are the opposite of studies done by the professionals like the FHA. I don’t think it is because the experts are wrong.
All over the US and world, cities have installed protected bike lanes. They reduced accidents. Over a few years, cycling rates soared, doubling in places like London and Paris. Why? Because people felt safer cycling.
Yet with this letter, we are somehow supposed to believe here, in Cambridge alone, bike lanes don’t do what they do everywhere else.
This is despite a recent Federal Highway Administration study showing that bike lanes reduced accidents here in Cambridge. And despite all those thousands people in other cities that then felt safer cycling. And it is despite common sense telling you that separating bikes from cars is obviously safer.
Do I have that right?
I’m sorry “FrankD” but you do not have that right. You seem really confused about the actual text and conclusions of my letter, despite them being printed in black and white directly above this discussion.
Quoting directly from the letter: “Numerous scientific studies have demonstrated a trend of increased cyclist safety with the installation of cycling infrastructure such as separated bike lanes.” Could you please let me know what part of that statement you disagree with? Silly me, I thought I agreed with you on that point, but I guess you’ve figured out my innermost thoughts and feelings better than I have myself =)
Here’s the central point of the entire letter: It can be simultaneously true that *the average* bike lane offers safety benefits, and also that *some bike lanes* are more dangerous. Just like it can be simultaneously true that the climate is getting warmer each year, and yet winter still happens.
Location-specific analyses would allow us to pinpoint which bike lanes are average, below average, or above average with regard to safety. Below average bike lanes could then be redesigned so as to perform better, while average or above average bike lanes could be left alone.
It’s a mystery to me why anyone would oppose allowing the City to collect additional data to improve bike lane performance.
@serenus “We need more data” is a classic stalling technique. Whenever someone wants to kill a project, that’s what they say. “We need another study”. This is not fooling anyone.
Bike lanes have been studied extensively all over the US and the world. The FHA studied our bike lanes, here in Cambridge. The answer is clear: Bike lanes reduce accidents.
Your argument is a red herring. Because data has some variability, not every single lane is safe? Nonsense.
All data has variability. You can’t draw that conclusion.
It is well-established that bike lanes reduce accidents. That is universally accepted among actual scholars and professionals who study this for a living.
Bike lanes have also proven themselves in the real world. There are dozens, if not hundreds of bike lane success stories: London, Paris, NYC, Minneapolis, Toronto, DC, etc.
All of these places found a reduction in accidents and an increase in cycling because people now feel safe cycling. The FHA found that our Cambridge bike lanes, like everywhere else, significantly reduced accidents by a substantial degree.
Why would anyone oppose more data? Becaue we already have the data. We also have real-world outcomes with many, many bike lane success stories.
“We need for data” is just an excuse to stop bike lanes. “We need more data” is a weapon that people commonly use to kill projects they don’t like.
For example, when bike lanes were first proposed here, some people said, “they may work other places but we need a study here in Cambridge”.
We now we have a study here showing that our bike lanes work. So now you’re going to move the goalposts to say, “we need to study every inch of the bike lanes”?
Sorry, not buying it. We don’t need more data. We have the data we need. What we need is a vocal minority to stop trying to kill a project that has wide support among Cambridge residents.
@ssrhua I thought about it, and there were two options: omit 2022, or use preliminary data. I decided including 2022 data from the log as preliminary data was acceptable because:
(1) the error size has been declining over the years, it was much higher in 2016 than in 2021. So recent crash log data is less likely to be massively revised upwards when validated.
(2) even if the error is much higher this year for some reason, it would still have Cambridge being much safer given how badly the state (and the country as a whole…) are doing post-pandemic, unless the error was suddenly much bigger than anytime in the previous 7 years.
I also made sure to note that 2022 data is preliminary and likely to revised higher.
If you disagree with my decision you can just look at the data up to 2021… which supports my argument. Or just wait a few weeks and the official data from the police will be out.
@Itamar Turner-Trauring
Nice, reasoned arguments that show a deep understanding of the underlying data and broader context in which it was collected.
Sadly, that will not persuade people who pull numbers out of context, ignore co-variate factors, and use flawed methods to support thier agenda.