Safety in numbers, gaps in risk environments, and bike helmets, Part II

If I’d known I could so thoroughly piss off the Smartest Boy Urbanists merely by suggesting that I like bike helmets AND better design, I’d have done so sooner. And louder. One was particularly outraged and sent me roughly 90 emails (ok, 7). I was told I was wrong roughly 490 times, which is interesting because I feel like didn’t really come to any conclusions other than “gosh, these data aren’t really all that helpful, and I like it when people do not get head injuries.” But one of the great skills of the Smartest Boy Urbanists is that they know you are wrong without reading what you wrote or listening to what you said, so we’re good.

I did suggest that I very much doubt that helmet-lovers like me are the reason that most American cities aren’t turning into Amsterdam particularly fast. I’d put that down to our general tendencies in urban and social policy (towards less competence rather than more) in every policy domain rather than dowdy old me with my helmet. And the very real possibility that land uses do not change as quickly as we might like.

There is a difference between these questions:

1. Are you, Bob the Cyclist, as an individual better off with or without a helmet, right now, in the streets you bike in, not necessarily the streets we would like you to have in order to lower your chance of a incident to virtually nothing so that anything like helmets designed to protect you in case of incidents you won’t have in that better bike environment become irrelevant?

2. Might mandating helmets suppress “Safety in Numbers” (SIN), because people do not like to wear helmets, so much so that their reticence cancels out the SIN effects (and worse, discourages cycling so much that people won’t cycle and thus get the exercise benefits)?

Keep in mind, I am not advocating for mandating helmets. I am asking questions, particularly, about #1: in the matter of “should I wear a helmet and/or have my kid put one, too”… should you equip like you are in the city you want, or should you equip for the city you have?

Because if the city you have is badly designed, I think it could be pretty irresponsible for urban experts to tell people that helmets are dumb because Amsterdam! Helmets may, in fact, be dumb because Amsterdam!, but most of us do not live there, and there is a big gap between many–not all certainly–but many existing cycling environments in US cities and what Amsterdamians enjoy.

If you tell people to get out there and bike “for numbers,” there’s a time period, as the numbers build, where you are pushing people into existing crash risks for your agenda and using them as a means to your end. Somewhere between “not enough bicyclists to reduce crash incidents so low helmets are irrelevant/current conditions” and “enough bicyclists to be visible, displace cars, and make life safer for cyclists” there is a gap of time where the risks might well go up–particularly for inexperienced cyclists, which includes kids–and who, exactly, is supposed to take those risks on for the sake of getting us the numbers we need?

Is that time period and subsequent gap in the risk environment important? I do not know. How much more cycling would we need in order to push through that gap? Dunno. We have some good modelings studies, but models are models and a well-designed empirical study can invalidate a million modeling studies.

Granted the needed bike density threshold, whatever that threshold might be in various contexts, how much time would it take to get it in local contexts? Dunno.

Now, before all y’all scream and yell that this gap is no reason not to pursue numbers, that is not what I am suggesting or exploring here.

But I do think that gap is something we can’t just assume away as unimportant while we look down our noses at helmets because in Amsterdam, things are so safe for bicyclists that helmets are silly. I personally approve of many things in and about the Netherlands we don’t have in the US for all the reasons we in the US can’t have nice things.

If that gap is meaningful, the policy and planning recommendation (as opposed to the individual equipment strategy) is probably to go after car drivers with policy (such as price floors on gasoline) to decrease volumes immediately to give bicyclists in existing urban environments a fighting chance to get out there, even with urban environments as they are. Policy changes can occur more quickly than built environment ones, but…I know. Getting such a radical policy change is beyond the scope of what I can imagine for political feasibility. It’s too bad.

We don’t talk about the policy environment in Denmark or the Netherlands very much, just the design and culture, and you do have to wonder how crucial the policy environment was in securing the ability to make the planning commitments they made to biking.That is, do you get Amsterdam as it is without a $3.80 per gallon excise tax on gasoline? (And do you maintain the ability to tax the daylights out of fuel when you have provided lots of other ways for people to get around?) While I see plenty of advocacy for changing design and increasing cycling, I see very little real movement to discourage car use at high levels, and that tepid approach to dampening car use strikes me as an important blind spot in the gap in the risk environment that I described. Now, I know most people who advocate for Amsterdamishness also think that lowering vehicle volumes is important, but that advocacy has been much less successful than the ability to get bike lanes built, and that means, to me, an evaluation of what is possible/credible/workable granted that sticky problem.

My musings the other day were rewarded by a very nice gentlemen showing me a dream bike and sending me an OECD study where I went through and read a bunch of the bibliographic references. It wasn’t great coverage on the OECD’s part, but they did cite a meta-analysis, and it’s always weird surveying the literature when somebody’s already done it. That said, the brain bucket question is waaaaaaaaay more interesting than you might think:

Hard shell helmets:

  • Risk of head injury decreases by 64 percent;
  • Risk of facial injuries decreases by 34 percent;
  • Risk of neck injuries increases by 36 percent

Soft shell helmets:

  • Risk of head injury decreases by 41 percent;
  • Risk of facial injuries decreases by 14 percent;

Again, none of these are important if you live somewhere where crash risks are so low that subsequent injury or fatality risk is practically nonexistent. But do you live there? I don’t. I live in Los Angeles. It’s going to take us awhile to become Amsterdamesque.

These are from a 2009 meta-analysis, so the study could probably be redone as helmets have changed, and bicycling environments have changed, at least in some places.

This meta-analysis in question appears here:

Elvik, R. et al. The Handbook of Road Safety Measures. Bingley, UK: Emerald, 2009.

I HAD TO GO TO THE LIBRARY IN SUMMER TO GET THIS BOOK. You people have no appreciation how much I sacrifice for you. I give and I give and I give. AND WHAT THANKS DO I GET? Sassing from the Smartest Boy Urbanists, that’s what I get.

I also found a bunch of really nice papers by Rune Elvik in general, and while meta-analysis is often hair-raisingly difficult to do well, accident researchers I find tend to do them in ways that make me say “damn I wish I were that competent” and Elvik does not seem to vary from this general impression.

(ok for some reason my autocorrect wants this name to be Elvis. I promise I have changed it back at least 7 times. So for the record, I am not talking about Elvis. Elvik. Elvis. Sigh.)

It’s never not interesting when there is a conflicting effect. They only had a handful of studies on the neck injury portion, but there are probably more studies now on the neck injury findings.

People have taken stabs at the other question above, but again…I am not convinced we really know much. That said, there are some really nice studies out there, but they show pretty mixed possibilities for Safety in Numbers using methods, the data that are really limited, and most of what we have are pretty speculative modeling studies. (You have to start somewhere, but it’s important to not confuse your start with well-validated empirical findings.)

The strongest claims on behalf of Safety in Numbers article I found was:

Jacobsen, Peter L. “Safety in Numbers: More Walkers and Bicyclists, Safer Walking and Bicycling.” Injury prevention 9, no. 3 (2003): 205-209 (here’s a pdf I found of it.)

This is a strong attempt with severely limited data, but he dug for some good stuff. He’s got walking and biking data for 68 California cities, walking and biking in 47 Danish towns, he’s got bicycling and walking data for 8 European countries, and he has fatality and injury outcome measures. He’s also got data for the UK going back to the 1950s, and he uses a power function to fit relationships between fatalities and relative risks. His best fit is not surprising, the California cities, where he has the most observations. Unfortunately there he’s got to use the percent of the journey to work mode share for biking and walking prevalence. He assumes that those mode shares hold for non-work trips, and while that’s debatable, I don’t really have a better idea for him, and I also don’t really have a better idea than using the percentages even though there is going to be more noise from small cities than we really want in a small sample.

I do have some problems with his power function as he has a couple big outliers in the dataset. A robustness check to address those outliers would have helped me. That said, most of the benefits in relative risk reduction on the California side accrue with really small changes in mode share. That conflicts with a later study, but it suggests that you do not have to have a bike utopia to have real gains as long as enough bicyclists get out there and ride. (Aha! That means we don’t have to accommodate bikes in design much at all, since we just need a few percentage point changes…. AHAHAHAHAHAHAHAHAHAH I AM JOKING. JOKING MIT YOU.)

And I’m not sure I can accept the assumptions about distance and injury; he plots distance per capita against relative risk, and distance is his use measure, and yes, but…I am not sure distance is a proxy for numbers, which is what he wants to show. Per capita distance can go up with relatively few cyclists or walkers who walk relatively long distances or with lots of people on the road making close-to-average-distance trips. Without seeing the distribution, you can’t tell. These are two separate things, and I suspect they might have different effects: the former has longer trips potentially displacing more VMT versus the latter having bigger clumps of people both potentially displacing cars physically and raising cyclists’ visibility to drivers because there are so many cyclists you can’t miss them.)

Again, aggregation to the city or nation-state level–he says he has intersection data but I don’t see any, how did that get passed reviewers and an editor? Am I reading that right?–does not help very much. It could be what we are seeing here is, simply, lower car use. I’m fine with that, btw.

The study that I thought made a nice case for SIN, but it’s an APM model, which means limited parameters and a lot hinges on the assumptions. This is:

Elvik, Rune. “The Non-linearity of Risk and the Promotion of Environmentally Sustainable Transport.” Accident; analysis and prevention 41, no. 4 (2009): doi:10.1016/j.aap.2009.04.009 (I couldn’t find a pdf of that one, sorry…)

Their outcome measure is number of accidents, to both pedestrians and bicyclists. The parameter coefficients come from prior studies. For doubling the number of bicyclists, they do not really see a big effect. The big payoff comes when they reduce traffic by 50 percent, and…that’s a big change, and it’s hard to compare that with what Jacobson finds in relative risk because Rune is examining counts. Do we need a big change in the numbers or a moderate change in the numbers? Rune suggests a pretty major shift in vehicle flow, not the smaller shifts in mode share that Jacobsen’s suggested.

Stay tuned.

This is another APM study:

Schepers, J P, and E Heinen. “How Does a Modal Shift From Short Car Trips to Cycling Affect Road Safety?” Accident; analysis and prevention 50 (2013): doi:10.1016/j.aap.2012.09.004. (No pdf again, sorry. I am a Googling failure).

Schepers works for the Ministry of Infrastructure and the Environment, which strikes me as being better than the Ministry of Silly Walks but not nearly as good as the Ministry of Magic.

This study, like Rune 2009, uses an accident prevention model again, but they have decent data, and APM models are not that data-hungry. They find using the data on crashes from Dutch cities that a mode shift for short car trips to bike trips would have no effect on fatalities and some increase on injuries. This is a good finding for the SIN advocates: without a safety in numbers effect and displacing cars, increasing biking would mean more crashes. It didn’t in this APM. And that’s good because you get the benefits of more biking with no real losses in fatality. If you want gains, then see Rune 2009, above.

Hooray. But. And this is a big one.

They do estimate a substantial increase in the risk of serious injuries from single bike crashes. These can be pretty serious, and a helmet could help with these ones. I suspect one effect embedded in the SIN effect shown in priors is that with good design and welcoming spaces, inexperienced bicyclists get out and ride, which is great, but that includes children, and then we need to get real about what risk and whose risks change, why, and what we should do about it. Kids learn and grow up, but there are new kids coming in and learning the system. Insulating them from major injury, even if it is a single-bike crash, is life-enhancing.

One last one I liked:

Bhatia, Rajiv, and Megan Wier. “Safety in Numbers” Re-examined: Can We Make Valid or Practical Inferences From Available Evidence?” Accident analysis and prevention 43, no. 1 (2011): doi:10.1016/j.aap.2010.08.015.

Basically, they say what I said when I stirred up trouble on Fboo and on the blog. APMs and aggregate spatial units are not an optimal way to figure this out. APMs tend to use spare specifications with volume data for the modes, that’s about it. Aggregate geographic spatial units are not really relevant to biking-motorist interactions in any meaningful way, and so we have no good information about design, only vehicle volumes (which can proxy for design if your design is lowering volumes.)

Only these authors say these things by sounding smart and with evidence and whatnot.

Some excellent papers I read that you might like too:

Wei, Feng, and Gordon Lovegrove. “An Empirical Tool to Evaluate the Safety of Cyclists: Community Based, Macro-level Collision Prediction Models Using Negative Binomial Regression.” Accident; analysis and prevention 61 (2013): doi:10.1016/j.aap.2012.05.018

Wegman, Fred, Fan Zhang, and Atze Dijkstra. “How to Make More Cycling Good for Road Safety?” Accident; analysis and prevention 44, no. 1 (2012): doi:10.1016/j.aap.2010.11.010 FOUND A PDF FOR THIS ONE

This one looks to be very promising in handling the murky mechanism problem I noted with SIN assumptions because they have a cool natural experiment around “cycling season”:

Fyhri, Aslak, and Torkel Bjørnskau. “Safety in Numbers – Uncovering the mechanisms of interplay in urban
transport with survey data.” AVAILABLE ONLINE!! YAY!

Somebody asked me about moral hazard with wearing helmets, so here’s some on those (people are naughty):

Gamble, Tim, and Ian Walker. “Wearing a Bicycle Helmet Can Increase Risk Taking and Sensation Seeking in Adults.” Psychological science 27, no. 2 (2016): doi:10.1177/0956797615620784 (PDF LINK)

This suggests that people compensate with other safety behaviors when they do wear helmets (people are smart):

Phillips, Ross Owen, Aslak Fyhri, and Fridulv Sagberg. “Risk Compensation and Bicycle Helmets.” Risk analysis : an official publication of the Society for Risk Analysis 31, no. 8 (2011): doi:10.1111/j.1539-6924.2011.01589.x

And just to piss everybody else off more…I don’t understand why everybody in planning loves roundabouts; when I am walking around them, all I can think is “Why am I going around this dumb circle instead of just cutting through it? Why did people forget that part of why DuPont Circle is cool? I hate everybody.” This study agrees with me so IT MUST BE BRILLIANT (I am joking, joking with you…)

Parkin, John, Mark Wardman, and Matthew Page. “Models of Perceived Cycling Risk and Route Acceptability.” Accident; analysis and prevention 39, no. 2 (2007): doi:10.1016/j.aap.2006.08.007

So now for those of you who wish to tell me I am wrong, TERRIBLY WRONG…I will summarize my conclusions here:

I got nothing I’d bet my first edition Varian on. But y’all haters do not have much in the way of evidence in support of your “helmets are not a solution, becoming Amsterdam is the solution” position, either. Maybe someday, you will, if people keep poking at the questions and changing streets.

Oh, but I do hate roundabouts.