Tuesday, February 22, 2011

Cycletracks, Safety, Numbers, and Statistics. What's a Poor Cyclist to Believe?

Montreal injury (not crash) data points (red boxes) from Lusk et al 2011 for non-cycletrack (lower left red box) vs cycletrack (upper right red box) roads. "Safety in Numbers" models are the result of a discussion between me and Bob Shanteau and based on Jacobsen-like calculations with different exponential factors between 0.3 and 0.7 (a factor of 1 would be a linear increase). I take responsibility for any errors or misinterpretation and thank Bob for thinking about this.
I've been a little concerned about a couple things looking like bandwagons lately. One has been the uncritical acceptance that bicycling safety is successfully addressed through safety in numbers; that hypothesis has arisen from Peter Jacobsen's 2003 paper. Critics call that the "More Butts on Bikes" approach, often assumed to work (but never proven) because motorists get used to seeing bicyclists all over the place and thus modify motorist behavior. I prefer to think of a hypothesis also involving more brains on bikes--more riders get more experience and stop competing for the Darwin Award. The other bandwagon is the hot topic of cycletracks. A recent paper by Lusk et al (Lusk, Furth, Morency, Miranda-Moreno, Willett, and Dennerlein, Risk of injury for bicycling on cycle tracks versus in the street, Injury Prevention 2011) has been quoted in Grist and elsewhere as giving unequivocal support for these facilities as reducing crash and injury rates. Well, if you completely separate cyclists and motorists they can never crash (into each other). But that ignores that nasty little word "intersections". It also ignores cyclists crashing without a car involved. The week I was in Calgary, Alberta in 2005 a cyclist was killed on a Calgary multiuse path. He sped through a curve and in trying to avoid a pedestrian, crashed head on into a concrete retaining wall. Do off-road facilities give a false sense of safety?

The Lusk study, at least in "brief report" form, worries me in that I don't see enough in depth study that would rule out any factors other than cycletracks vs not having cycletracks. That seems to beg the question. Perhaps other tests were done and not discussed.

But to push the discussion, lets ask this: Since the Montreal cycletracks apparently resulted in a 2.5 times increase in cycling over non-cycletrack streets, should the "safety in numbers" theory apply? Well maybe not. We don't know if Jacobsen's model or Smeed's law works on the microscale. But let's just say, for shits and grins, that it might. What happens then? How does one know if facility design, safety in numbers, or something else controls crash rates?

The Montreal cycletracks are quoted as having reduced the injury rate by 28% with an increase in ridership of 250% compared to non-cycletrack streets. Jacobsen's paper suggests that cycling crashes rise as the 0.31 power of an increase in cycling. Soren Jenson, in a more critical study of Copenhagen facilities, says Danish engineers often use a more complicated model, but with that exponential factor (or factors) between 0.5 and 0.7. He also states that Copenhagen cycletracks resulted in a 20% increase in cycling and a 10% increase in crashes. [Readers please note Comment #1 before reading further]. That would result in a power law of about 0.5. I've plotted a crash rate with exponents between 0.3 and 0.7, labelling the 0.5 curve "Spencer" because I don't want to blame anyone else for my assumptions.

Seems that the Montreal facilities track the higher of the three crash rates that one would get simply by applying the Safety in Numbers model and have a far higher crash rate than predicted by applying Jacobsen's model to Montreal cycletracks. Seems to me that if safety in numbers makes you safer and cycletracks make you safer, the crash/injury rate should be lower than we see in Montreal. Then again, I'm winging it....

Personally, I'd like to see someone actually push this a lot farther than my back of the envelope attempt by an amateur. The Jenson study is a good one, but over there rather than over here; I suspect cultural differences matter quite a bit. I'd also like to see a more detailed breakdown of injuries per mile of bike facility and of general-use facility (i.e. the road) in terms of injury severity. If a cycletrack has 2x the use-normalized injury rate but 0.25 times the serious injury rate (to be defined, but I would define "serious" injury as one that causes more than a few days of mobility loss, i.e., a sprained wrist vs. broken ribs or neck), is that an improvement?  If nothing else, cyclists need to be asking such questions rather than listening to the usual heated but untested arguments. It is, after all, your life, your road rights, and your tax dollars.

To close, I'd like to remind readers that to some degree, we fear things in proportion to how we have manufactured our fears. Or as Dave Horton, a Sociologist at Lancaster University says so well in a series of essays posted on his own site and on on Copenhagenize:

"...Fear of cycling belongs to a fearful culture (Glassner 2000; Massumi 1993). UK sociologist Frank Furedi (2002) argues that western societies have become dominated by a ‘culture of fear’. We have never been so safe, yet never have we been so fearful. ‘“Be careful” dominates our cultural imagination’ (ibid.). We belong to ‘a culture that continually inflates the danger and risks facing people’ (ibid.). ‘Activities that were hitherto seen as healthy and fun … are now declared to be major health risks’ (ibid.). What is more, ‘to ignore safety advice is to transgress the new moral consensus’ (ibid.).[1]

Our fears are produced (Sandercock 2002), which is why they are subject to such variation. Obviously, some fears take more work to produce than others. Most people fear a lunging shadow down a dark alleyway. Fewer people fear waste incinerators, nanotechnologies or the policies of the World Trade Organisation (Goodwin et al 2001, 13) because those fears are more difficult to produce. Fear of cycling is neither inevitable nor ‘natural’ and needs similarly to be produced. It also always exists relative to other fears...."

Thanks muchly to Bob Shanteau for doing most of the good thinking here. I take full responsibility for any brain farts.


Cyclist_Lorax said...

The Copenhagen report gives a 10% increase in crashes to all modes, not bicycling crashes, for a 20% bicycling increase. It is imp[ortant to note that traffic volumes and the 0.7 exponent are already taken into account, so a 0% increase means no change in safety, so a 10% change is bad, The bicycling crashes, when viewed against the 20% reduction tell a much worse story (>30% increase over no treatment with the increased volume). Jensen, whether intentionally or not, elected, in contrast to every other bicycling study I know of, NOT to show the bicycling crashes and compare that to the expected mode share increase. Since his data was published, I took the liberty of doing some simple % change analysis in bicycling crashes in the following publicly available FaceBook photo album. I suggest you take a took that these images and the explanations:


Khal said...

Thank you, Dan (I assume that is you?) for the clarification.

Jon said...

I have met Anne Lusk and had some interaction with Peter Furth. Dr. Lusk's background is in, IIRC, landscape architecture and she comes to interest in cycling by way of public health, so she's coming at the problem with a different set of skills and priorities than the propeller-beanie set. Prof. Furth has more of an engineering background and, with his students, has come up with some innovative designs that have the potential to solve some real connection problems around Boston, including a simple pair of contraflow lanes on one-way streets that I used nearly daily. He's also come up with some designs that scare the spanner out of me.

VC advocacy both has a serious image problem and has been completely left behind by traffic planning in this country. A published rebuttal of "safety in numbers" would only partially address the problem; we need to provide a compelling alternative story. (I also wonder about error bars....)

Khal said...

I too would like to see more error bars, but they are only as good as one's underlying model assumptions; at times, they give a false sense of uncertainty. One needs good data to test uncertainty models and the more I read about traffic modeling (such as the Jenson paper) the more I am convinced its a challenge not for the faint of heart.

VC has worked because the non-vehicular facilities built in the U.S. were hideous, the roads better, and with clear and predictable rules. A competent cyclist is in fact operating a vehicle within a well-established paradigm.

The challenge to the bicycle specific design-build establishment is to design these new facilities in a way that demonstrably proves they are better (in terms of safety and efficiency to cyclists), not just "feel" better. I think folks like John Allen try to avoid both extreme camps and simply ask "what works, and why?". I concur.

Khal said...

As far as uncertainty, for example:

John Allen said...

Concerning safety in numbers, please also see


John Allen said...

Oh, and on the validity of Jacobsen's study and the more general issue of the validity of safety in numbers, please see


and especially the ocmments.

anto9us said...

Kahl. I did not know where else to contact you. I would like to ask you a question about bike awareness development in Los Alamos.