A year-long evaluation by researchers at Portland State University’s Institute for Bicycle and Pedestrian Innovation has found that Portland’s bike boxes improve the safety of roads users on a number of levels. The research — funded through the Oregon Transportation Research and Education Consortium — complements another study from the University of Texas published last month that also found bike boxes to have a positive impact on traffic safety. City of Portland Bureau of Transportation officials say these studies confirm their own observations about the bike boxes and they are in the process of identifying new locations to add more of them.
Back in January, PSU released very preliminary data from their bike box research. The results were mixed and not very conclusive one way or another. When those results went public, several media outlets used them to question the overall efficacy of bike boxes. However, now that the full study is complete, the picture has grown more positive for bike boxes.
“Since the preliminary report, I am overall more confident branding the boxes as working.”
— Chris Monsere, researcher
According to PSU’s co-principal investigator Chris Monsere (who worked on the study with noted transportation researcher Jennifer Dill, Ph.D.) the additional analysis since January has convinced him of the traffic safety benefits of bike boxes. In an email to BikePortland yesterday, he wrote, “Since the preliminary report, I am overall more confident branding the boxes as working.”
Monsere, Dill, and their research team used a combination of before and after video data and surveys of people who used intersections where bike boxes had been installed. Here’s the breakdown of video analysis:
Of the 918 hours of total video, researchers analyzed 83 hours for “usage and compliance” of 5,315 people on bikes; 139 hours of bike/car conflicts (including 7,454 people on bikes); and 17 hours of yielding behaviors of 1,141 motor vehicle operators (of which 135 of those had interactions with bike operators that necessitated yielding).
“Having some academic review of what’s happening out there really helps us in our decision making.”
— Greg Raisman, PBOT traffic safety specialist
In addition, to find out if people understood the markings, researchers surveyed 468 people who bike through the bike boxes and 721 people who drive through the bike boxes were recruited via email to complete a survey.
The PSU researchers sought to answer three key questions: Do road users understand and comply with the pavement markings? Do the bike boxes improve safety (both actual and perceived)? and does the green color make a difference?
Do road users understand how to use the bike box?
Yes. Video analysis showed that 73% of motor vehicle operators stopped in the correct position behind the bike box and 86% of those surveyed said they understood the markings. 73% of people on bikes stopped ahead of the motor vehicle stop line, but interestingly, only 5% of bike riders positioned themselves in the bike box (out of the bike lane) in front of the motor vehicle stop line. That number jumped to 38% when someone on a bike was already in the box, showing how people were less timid to move in front of stopped motor vehicles if someone else had done it first.
Does the green color make a difference?
Researchers acknowledged that study limitations made it difficult to draw clear conclusions to this question. However, nearly 90% of motor vehicle operators said they preferred the color and people on bikes used colored boxes as intended more frequently, both of which the report said, “should increase their visibility and improve safety.”
Do the bike boxes improve safety?
This is the most important question. In terms of number of conflicts and yielding behaviors — yes. The research found that the number of conflicts decreased and the yielding behavior increased. According to the research, the number of observed conflicts decreased from 29 to 20 while the number of people on bikes increased 35% and the amount of motor vehicle right-turns increased by 7%. This led researchers to state:
“Controlling for differences in volumes of bicycles and right-turning vehicles, fewer bicycle-motor vehicle conflicts are expected for an intersection with a bike box.”
The positive safety results come with a caveat. Monsere says that since conflicts (especially major ones) are rare, “It’s hard to be as conclusive as we would prefer.” He adds that their results showed no negative safety effects and the fact that there was more yielding and fewer conflicts observed, “Over time,” he says, “one would expect fewer crashes”. Monsere also shared some reservations about placing bike boxes in suburban areas or in places where people aren’t as used to seeing bikes on the road.
In terms of people’s perception of safety, the results were positive and conclusive. Both motor vehicle and bicycle operators told researchers that the intersections were safer after the installation of bike boxes. Of particular interest is that 42% of motor vehicle operators who do not ride bicycles felt that driving through the intersections was safer with the bike boxes (compared to 14% who felt it was more dangerous). 77% of people who biked through the intersections felt they were safer with the bike boxes.
Last month, the Center for Transportation Research at the University of Texas also issued positive research findings on bike boxes. In their Executive Summary, they wrote, “Ultimately, bike box markings are recommended for installation at intersections where a majority of motorists do not turn right on red and the volume of bicyclists is high.”
This positive research could help build momentum for the official adoption of bike boxes on a national level. The Federal Highway Administration is in the process of considering them for inclusion into the Manual of Uniform Traffic Control Devices. PSU’s Jennifer Dill says this report will be made available to the FHWA as part of the City of Portland’s ongoing “application to experiment” with bike boxes.
Momentum seems to be growing for more innovative federal engineering policy regarding bikeway design. With this positive research and the installation of bike boxes around the country, perhaps we’ll see this treatment become as common as the sharrow (which the FHWA finally adopted in 2009).
In Portland, we are certain to see more bike boxes in the months to come. PBOT installed our 12 existing bike boxes back in 2008. PBOT traffic safety staffer Greg Raisman says they’re happy to have research that backs up their observations. In a telephone interview today, he said, “Having some academic review of what’s happening out there really helps us in our decision making.” Raisman will be present at tonight’s Bicycle Advisory Committee meeting where he’ll roll out a process to identify new locations where PBOT will install bike boxes.
— Download the research referenced in this article: Evaluation of Bike Boxes at Signalized Intersections – PSU/IBPI (PDF); Effects of Bicycle Boxes on Bicyclist and Motorist Behavior at Intersections – City of Austin, (PDF)
If you have questions or feedback about this site or my work, feel free to contact me at @jonathan_maus on Twitter, via email at firstname.lastname@example.org, or phone/text at 503-706-8804. Also, if you read and appreciate this site, please become a supporter.
Thanks for that J. You KNOW I was about to scream bloody murder. Takes nads to disclaim one’s thing, and be forthright. I live on campus and observed with my own eyeballs a completely unscientific process of gathering information for their survey. The worst condition being that of primarily sampling folks who essentially called for the special infrastructure in the first place.
Now, about that lede…?
Now that the data show more box-positive results, I wonder how well the media outlets will report this?
(Ignoring post #1 to preserve my sanity)
bike boxes allow me to legally shore vehicles (not that that ever stopped me before).
crap…meant shoal. can’t keep my bike snob lingo straight.
can you please explain what you mean by shoaliing?
Um, Vance? “Observing with your own eyeballs” is a “completely unscientific process of gathering information” about the validity of the study.
Maybe I am naive, but I at least start from the position that professional, academic researchers will make every attempt to devise studies free from bias.
How did the researchers define “conflicts”?
Cool study. I am not surprised by the results, but am rather intrigued by the 14% of motorists who perceived intersections with bike boxes to be less safe to drive through. Huh? Perhaps they just don’t like them, but I have a hard time seeing how they would be more dangerous for anyone, let alone someone in a motor vehicle.
I’d be interested in knowing whether or not compliance increased over time. The bike boxes were still rather new when the study was initiated, I imagine that the bike boxes were confusing when they were first installed (to both drivers and cyclists), and I wouldn’t be terribly surprised to see compliance increase as the public became aware of what they were.
I rarely (or maybe never?) position myself “in the bike box (out of the bike lane) in front of the motor vehicle stop line” because there’s little point when I’m intending to continue through the intersection to the bike lane on the other side. I’m not surprised only 5% do this, and I suspect the number jumps to 38% when other cyclists are present because everybody is trying to make more room for bikes to get closer to the intersection or trying to position themselves for a fast take-off to get ahead of the pack.
I wonder whether the conclusion is that the bike box on NW Broadway @ Hoyt (as you come down B’way from the bridge) is a good place for a box. An awful lot of motorists turn right on a red there, even though they aren’t supposed to (pretty sure there’s a no-right-on-red sign there).
@chelsea #7 — Just a guess, but I bet the 14% (or many of them anyway) see the bike boxes as confusing, from which would follow the feeling that they are less safe to drive through. Confusion on the roads is never a good thing, not that that is (in and of itself) an argument against bike boxes. I was surprised that at least three different people who drive and know I get around by bike asked me how they were supposed to treat bike boxes. The set-up seemed clear enough to me, but some people really do find them confusing, or at least, aren’t sure if they’ve figured them out correctly.
1. Most drivers stop behind the bike box.
2. When drivers are already stopped behind the bike box, I don’t feel it necessary for me to ride into the bike box. They can still see me ahead of them, even though I’m not in the bike box.
3. When I arrive at an intersection first, I stop in the bike box, so that no driver will roll through the bike box and stop at the crosswalk. This ensures that i will still be positioned ahead of any drivers that come along.
4. When I’m approaching an intersection and the light is green, most drivers will stop and wait for me to pass, even though I’m behind them. I don’t feel that it’s necessary for them to stop and wait for me when they have the right of way, but I do appreciate that they are conscious enough to look before turning– and I think that the green paint continuing through the intersection is a visual reminder to them to stop and look before turning.
5. No right-hook deaths in Portland since 2007.
is this like the global warming studies where they just keep adjusting the data untill they get what looks good to them?
did they ever ask a truck driver that has a long hood what they thought of a bike box? they can’t see a bike sitting in front of them. when the light turns green that bike had better go right away or he might be a statistic. truck drivers often look at their paper work for address’s, maps… at red lights so they are looking down and don’t see a bike cut in front of them.
Did anyone learn any lessons from the recent accidents with bikes and buses? dont put yourself in danger. I think the bike box is just foolish planing by fools who have no business trying to plan anything
M.M. – “Just a guess, but I bet the 14% (or many of them anyway) see the bike boxes as confusing, from which would follow the feeling that they are less safe to drive through…”
my thought exactly.
Jim, before bike boxes, two deaths and one serious injury due to right hooks within the space of three weeks. After bike boxes, no deaths due to right hooks.
So, lesson learned.
right hooks are still happening all the time. we have just been fortunate that nobody was killed. I saw a guy today that was nearly right hooked, he must have been a little rattled, the next block he almost knocked over a jaywalker. that was odd because i just saw a video on youtube of a bike running over a jaywalker
Sure, right hooks still happen. Motorists still pass cyclists and then suddenly turn, and cyclists still attempt to pass motorists who have begun to make a turn.
But nobody has died since 2007.
And the fatality that brought us bike boxes involved a cyclist who was stopped at a red light next to a cement truck; when the light changed, the truck driver turned right, completely unaware that a cyclist was to his right. The bike boxes help prevent those kind of right hooks, by positioning the cyclist ahead of the driver (whether directly in front of the driver, or in front and to the right, the cyclist is still positioned ahead of the driver), where the driver can see the cyclist. The bike box informs both the driver and the cyclist where they should stop at a traffic signal, and the green paint that continues through the intersection serves as a visual reminder to the driver to look before turning.
Since 2007, we haven’t had a single fatality from the type of right hook that took Tracey Sparling, so as I said, lesson learned.
fred f #3 – I’m not necessarily impugning their plan. On the contrary, I was impugning it’s implementation. For instance, I observed dozens of college-aged adults, on a college campus, and concluded that they were college students. It was this type of individual that was ‘interviewed’ predominately. There is a circumstance whereby you start with an attractive, fit, 19 year-old white girl that weighs 105 lbs., and send her out to collect data. Lots, and lots, of 140 lbs., white, half-men of a college age get ‘interviewed’, while the 40-something black-guy on the construction crew, weighing in at 320 lbs., gets purposefully avoided.
This is what I mean when I assert that essentially the people calling for the infrastructure are actually providing most of the basis for this survey’s conclusions. Of course a sampling from within the PSU bubble is going to yield positive results. Take that survey to La Grande, OR, Medford, Burns, John Day, anywhere, and tell me again bike-boxes would fair well. For that matter, enter a brown-person neighborhood for the first time in your life, and sell your can’t-drive-a-car-gotta-ride-a-bike message there.
I’m sure modern statistical data collection, and analysis, are beyond my ignorance. I’m just as sure that no matter how great your plan is, if you send out a bunch of incompetents in order to deploy your plan, what does it matter how good the plan was?
What of this other? If I observe you contaminate a sample in a lab, and comment upon it, what difference does it make whether or not I followed scientific methodology, let alone established any sort of empirical evidence to support the conclusions drawn from my observation? I’m under no such mandates commenting about my observations of their data sampling techniques. Riddle me this. I’ve lived right smack-dab in the middle of the PSU campus for over 6 years. I rode thirteen three last year, prolly more this one. I ride around that campus on numerous round trips, seven days a week. I made eye contact, while saddled up, no less than 50 times, with the survey takers. How is it that their planning worked out in a way that I slipped through their survey? Why is it that when I quiz some of the other area curmudgeons, they have had a similar experience? Yeah, your bike-boxes work awesome. Especially if you only ask people things about them, that already like them regardless.
I find it interesting that motor-vehicle right turns increased by 7%…
and it’s good to hear that they found a cheap way to increase safety…
I’ve been “right-hooked” by a TriMet bus at an intersection with a bike box. The box accomplished absolutely nothing. And, yes, I did complain to TriMet with the bus number, route number, date and exact time. I received a message saying the supervisor would “talk with the driver.”
I concur with the observation that there is little if any need to move into the spot in front of a motorist if one is continuing straight. In my opinion, that maneuver does nothing but delay a motorist unnecessarily.
I agree with Vance that, without sharing the sampling methodology, any conclusions are highly suspect to say the least. Trying to gather data from riders and motorists is going to be a highly non-random audience even assuming total neutrality on the part of the surveyors, which is pollyannaish to the extreme.
I think it is safe to say that most bicyclists perceive they are safer in a bike box but the opinion of motorists and the reality of that perception are entirely unproven.
My common sense perception is that I am safer in the few boxes I encounter on a frequent basis. My more logical opinion tells me that my perception is not necessarily based in fact.
Vance Longwell –
I wonder if you are referring to another study we are undertaking this summer about the cycletrack on SW Broadway, which runs through PSU. For the bike box survey, we intercepted cyclists last year at five of the bike boxes, none of which are on a college campus. We asked every cyclist who stopped to complete the survey.
re comment 6, this is exactly the question i asked chris monsere when he and prof. dill presented the preliminary report months ago.
the answer is that they turned the research assistants loose on the raw video with the instruction to flag any interactions between motor vehicle and bicycle that did not appear “typical,” and then he and prof. dill sifted through those results, ranking conflicts as “major, substantial, minor, or no conflict.” for purposes of that metric, a “conflict” was any situation that would have led to a collision if one or the other player did not take evasive action. not clear to me why the research assistants were not given this definition.
severity went from “precautionary braking or change of direction” to “emergency braking or change of direction” to “full stop.”
the total number of conflicts recorded using these methods on 139 hours of video featuring something like 7,500 cyclists was 29 pre- and 20 post-, none “severe.”
don’t know much about statistics, but these numbers do not look very meaningful to me.
Actually, confusion can be a good thing. It makes people slow down and think through a “confusing” zone.
The decrease in fatalities due to right hooks is a statistically insignificant. This number of fatalities is too small and inconsistent over the years to mean a thing.
are #22: a drop from 29 to 20 is a reduction of over 30%. That’s pretty substantial. And from glancing through the study, I believe it was statistically significant. “meaningful”? That is a different question and very subjective. What do you mean by “meaningful?” If I’m one of the few who didn’t get right hooked because of the boxes, then I’d say it’s meaningful.
Jim #12. I appreciate your commentary and what it brings to the discussion (usually). But dont’ make things up about comparisons to global warming studies and suggestions of falsified data. Read the bike box study. Read the methods. Read the results. Then see for yourself if the conclusions fit the data.
1. I hate shoaling.
2. If I get to the intersection first, I tend to stop where the bike box would be; if I get there later than first I pull in behind whomever is in front of me.
3. I suppose that if bike infrastructure use is mandatory and that it puts me in a position to be right hooked from a light turnign gree then maybe the bike box is a good idea, but see #1.
4. in support of Vance: “All data are theory-laden.”
-Norwood Hanson, Patterns of Discovery: An Inquiry into the Conceptual Foundations of Science. Cambridge University Press, 1958.
“The fallacy has been to cram ‘good medicine’ down people’s throat because the ’experts’
thought it was good for them.” Fisher, E. G., and R. Reeder. Vehicle Traffic Law. Traffic Institute, Northwestern University, Evanston, IL, 1974, 20, 29.
re comment 25. the “before” count was taken during winter months. the “after” count was taken during spring months. the kids who reviewed the tapes were given no instruction what to look for. and they couldn’t even see what the signal phase was. in hundreds of hours of tape, looking at thousands of cyclists, they found whatever they found, and then chris monsere sifted through that to come up with only twenty incidents here and twenty-something there. some stuff the kids had identified ended up on the cutting room floor, because monsere thought it just looked like the motorist and the cyclist “negotiating” priority. i think if you read the report closely you will see that the numbers are “meaningless.” not using that as a technical term, but descriptively, in its common usage.
bike box markings are recommended for installation at intersections where a majority of motorists do not turn right on red and the volume of bicyclists is high
The problem with Portland’s use of bike boxes is that PBOT is trying to use them to solve right hook problems at intersections with a much higher percentage of motorists are turning right.
this sentence from the austin report is precious: “The installation of the ‘No Right Turn on Red’ sign did not discourage motorists from turning right on red.”
re comment 25 the statistics also indicate that there was an increase of cyclists, 35%, and motorists making right turns , 7%. This would indicate to me that the percentage drop in the number of incidents is over 50% when taking these numbers into account. Not insignificant even when a moderate amount of error is taken into consideration.
what the austin report shows. one, you put down a green box with a green approach lane, more cyclists will ride to the right. same might be said of sheep. two, if the cyclist arrives during the red signal phase, the green box does allow the cyclist to get to the head of the queue. three, after dark, more cyclists will assert the travel lane (with or without the box) in order to make themselves more visible to motorists. but we are not measuring that. four, while bicyclists will do pretty much what you tell them to do (see item one above), motorists tend not to cooperate with any intersection treatment whatsoever. five, although we are interested in measuring “conflicts” before and after, we have no numbers or really anything to say about that in this report, thanks.
re comment 31. listen carefully. the word “conflict” does not have a reliable definition here. we don’t know what the kids who reviewed the tapes saw or did not see, flagged or did not flag. we don’t know exactly how monsere decided to disregard some of the stuff they flagged. the report reaches a predetermined set of conclusions.
#28 Thanks for the reply. After reading the methods and results more carfully I think I’ll stick to my original statements. The criteria for a “conflict” is defined and if applied consistently across all video should spread bias evenly among treatments. And then, as Don Y #31 mentions, there was an increase in the number of bikes and right turning cars in the “after” treatment (number of hours controlled)so the difference, before vs after, appears even greater.
Their model predicts fewer conflicts with boxes. Fine. But we are not sure if the reduced conflicts are due to reductions in bicyclist avoidance behavior due to increased confidence (due to increased numbers of cyclists or perceived improved safety from bike box)versus changes in car drivers behavior. Stated another way, maybe the cyclists in the “after” condition are simply riding with more confidence and not reacting negatively to innocuous car movements.
also — and again, i know nothing about statistics — but isn’t there such a thing as a minimally significant sample size? twenty? twenty-nine? for real?
the number of conflicts observed is not a big sample size… because there simply weren’t a lot of conflicts observed. Did you want them to wait a few years before several hundred conflict had happened?
They state clearly in the research that one of the main caveats of the evaluation is that the occurrence of conflict is rare, thus making it more difficult to draw conclusions than they’d prefer.
the report reaches a predetermined set of conclusions.
This can’t be said enough.
What PBOT wants PBOT gets.
Advisory committees and public comments notwithstanding.
are #33 and BURR #37 What you are suggesting is nothing but insulting (and a bit naive). You honestly think that a grad student and professor would put their credibility and careers on the line simply to benefit PBOT? Read the report. They lay out their objevtives, the methods and results very clearly. The conclusions follow the data fairly well. And they describe possible limitations to their study.
Finally, it is listed as a draft. I’m sure the authors would appreciate any constructive feedback they would recieve.
It’s almost impossible to do a truly quantitative study of human behavioural phenomena in a situation like this, so there is a lot of qualitative evaluation required; and, as a result, a lot of flexibility in what your conclusions do or don’t say.
1) The use of conflict analysis in traffic safety studies (for motor vehicles) dates to the early 1960s. Researchers generally found that the occurrence of specific conflicts are a good indicator of actual crashes. However, given the time and expense of collecting these data, most motor vehicle safety studies rely on police reported crashes. There is a large body of research focused on estimating the effect of highway safety improvements using crash data.
For analysis of bicycle or pedestrian interventions, reported crashes are very infrequent. Thus, nearly all studies of these treatments use some type of conflict-based approach. The definition of conflicts in our study is consistent with work of other researchers in the field (see the report’s reference list).
It is not true that the research assistants were not given instructions; they were instructed to identify all interactions between motor vehicles and bicycles that could potentially be a conflict. These clips were isolated, then reviewed in a panel setting (by Prof Dill, myself, and our lead graduate assistant, Nathan McNeil). First each panelist made an independent coding of the video clip. The clip was then reviewed and discussed and a consensus coding reached. This approach improves consistency and repeatability of our review of the conflicts. We have a log of every clip that was reviewed (including those that were coded as “no conflict”).
2) As others have pointed out, the raw conflict counts must be adjusted since both cycle volumes and right-turning vehicle movements increased from the before-to-after conditions. To control for these, changes we employed a modeling technique. The models meet statistical checks, but clearly more data are better, thus my comment about not being as conclusive as we would prefer. To be sure, we can also not possibly control for every possible difference between the before-and-after conditions. In my experience, volumes of the conflicting traffic is most often the significant contributing variable to control.
BURR, I agree that behavioral studies (human or not) are exceedingly difficult because the continuum of response types and the difficulty in defining behaviors. But behavior studies happen all the time and advances have been made over the last 20 years to make them more scientifically defensable.
And that’s different than suggesting that someone fudged their data to make PBOT happy.
Also, there is very little flexibility in catagorizing their conclusions in any scientifically defensible way. The data defines the conclusions. Not the other way around.
There is weakness/limitations to the study (e.g. non-blind video analysis). But the data is what it is and the conclusions fly direcly from the data.
Thanks for the clarifications Chris. You really should add those details (independent coding, etc) to your methods.
I’m with an earlier commenter, in that I generally don’t get in front of cars in the box (instead sticking to the side) so that I don’t slow people down unnecessarily.
However, there are two exceptions:
One is if the car is signaling a right turn. I damn well want them to see me. Of course, not everyone signals.
The other is if a car has pulled forward into the bike box. Do they not see the “wait here”? Or the big bike? If a car is stopped in the bike box, I will stop smack dab in front of them. And when the light turns green, I take my time accelerating (“oh gosh, where are my pedals?”).
Immature, yeah. Satisfying, definitely.
re comment 40.
quoting from page 16 of the report:
“all potential conflicts were flagged by the research assistants who were instructed to liberally define these events as any motor vehicle–bicycle interaction that did not appear typical.”
if this does not actually mean (as it certainly appears to mean) that the definition of “conflict” ultimately used by the senior researchers, i.e., precautionary or emergency braking or evasive maneuvers, was _not_ given to the research assistants, then this sentence should be rewritten.
no mention in the report of “log of every clip reviewed,” etc.
when you presented the report to the lunch crowd back in january, you said some of the stuff identified by the research assistants was determined by you not to be a “conflict,” because you judged it instead to be a matter of the motorist and the cyclist “negotiating priority.” which of course would not be necessary if the cyclist was behind the motorist, in the travel lane. you showed a clip of a cyclist and a delivery truck going through a sort of “you first, no you first” dance. people laughed, as you apparently intended them to do. no cyclists were harmed in the making of the film.
to accept at face value that the cyclist is appropriately placed to the right of a motorist who might turn right and to treat the resulting confusion as “not a conflict,” and to then report that the boxes seem to be functioning “as intended” . . .
is to report a predetermined conclusion. i do not think i am impugning anyone’s integrity to say so, i am pointing out a blind spot.
also, if i may. since you are likely to be doing this again in connection with some of the boxes greg raisman is about to put in (including at the now infamous couch and grand). let’s get some very specific criteria for the research assistants to code, with simple yes or no, one or zero, marks. let’s get the cameras positioned so you can clearly see the signal phase. and let’s use as controls a few intersections that do not have any striped bike lane at all, and that have substantial right turning motor traffic and through bike traffic. (pretty soon you may not be able to find any of these in portland.)
If that incedent with the cement truck had a bike box way back when… If the bike was sitting in front of a cement truck- he probably would not have been able to see the bike because of the big and high hood on the truck, in the end it might have ended up the same. The bike box is still a theory, nothing proved
” in that I generally don’t get in front of cars in the box (instead sticking to the side)”
sorry but that is what they are for. if you are hanging on the side you are completely defeating the purpose of the box.
You hit the nail right on the head. They are poorly thought out and not as safe as we were led to believe.