Monday, July 14, 2008

Portland bike crash data

Update: added a couple of new graphs with data on licensed OR drivers by age group

I've been sifting through the Oregon DOT crash database recently and thought I'd post a few of the more interesting diggings here. Maybe there's something to be learned. Keep in mind that the figures come from reported crashes, and bike crashes are substantially under reported. All data is for Multnomah County (Portland city, basically) city streets (no freeways).

Maybe the most puzzling results were the ages of drivers in bike/car crashes compared with other car crashes. Click the chart below for big.

copyright Joseph Broach, 2008

The distribution for non-bike crashes is what I'd expect. Lousy young drivers who steadily improve with experience from mid-20s on. As reaction times and vision deteriorate, most people also drive less and more cautiously.

But, for bike crashes, that learning/wisdom seems to be delayed, and things don't start to improve until mid to late 40s. Maybe there just aren't enough bike/car interactions to learn from, even in Portland. Any ideas? That blip around 80 years old may be just noise, or maybe that's when most people honestly can't see a cyclist any more.

I've often heard that fault runs about 50-50 in bike/car crashes. The Portland data on severe injury/fatality crashes suggests that cyclists are more likely to be cited for errors. Click for big.

copyright Joseph Broach, 2008

The most common cyclist errors in serious crashes are running lights, stop signs, and proceeding without the right-of-way. I guess the bright side is that riding smart may reduce your risk of getting hurt quite a bit.

Men are more dangerous behind the wheel in general, and the likelihood of getting smacked by a guy is even greater on a bike. Just over 60% of bike/car injury crashes involve a male driver, compared with a bit under 56% of non-bike crashes. Are guys just more aggressive drivers?

copyright Joseph Broach, 2008

Stay safe out there. I know I'll be giving any middle-aged guys turning left a wide berth.


Some good comments here and elsewhere suggested that it would be nice to have some idea of "exposure" by age. Ideally this would be miles driven in Portland. As it is, I'll settle for the number of licensed drivers by age group in Oregon--not perfect, but better than the nothing I had before. The new data comes from the FHWA's Highway Statistics 2000. If drivers of all ages were equal and A) drove the same number of miles B) in the same conditions C) with the same degree of skill and care, then we would expect crashes to track licensed drivers very closely.

In each graph below*, the proportion of licensed drivers in each cohort (the data are in 5 year chunks) is plotted along with the smoothed proportion of crashes by driver age. What stands out is that the age of drivers in bike crashes follows our "If all drivers were equal" assumptions pretty closely. The relationship is much weaker in non-bike crashes. One scenario that would lead to this discrepancy is that drivers learn how to avoid crashes with cars but not bikes.

copyright Joseph Broach, 2008

* You may notice that the non-bike crash curve is slightly different from the first graph. I took another good suggestion to remove single-vehicle car crashes (rollovers, fixed objects), since the equivalent bike crashes almost never get reported. The result is a slightly flattened initial hump, which suggests young drivers are overrepresented in single vehicle accidents.


Blogger Everett Volk said...


It would be interesting to see the injury crash distributions compared against the gross number of drivers distribution. Or, to describe it a different way, what does the per capita injury crash distribution look like when "per capita" is the universe of drivers in each age cohort.


9:13 AM  
Blogger p. squiddy said...

Joe, Could explain what is graphed in the first chart? I'm not sure what "density" is.

I imagine part of the 80 plateau is that the graph is smoothed and there are probably vastly fewer data points out that far.

6:47 PM  
Anonymous Ron Richings said...

Could be that your first graph simply reflects the demographics of the 'baby boomer bulge' in the general population.

This might also correlate to some degree with who can afford to drive a car -- ie boomers.

Ron Richings

7:34 PM  
Blogger Joe said...

Thanks for the comments.

Everett and Ron: I added some data on roughly how many drivers are in each age range.

p. squiddy: Density here is basically the proportion of crashes involving a driver of a given age. The values don't matter as much as the shape. You're correct that it's smoothed between ages. The raw line is difficult to read.

9:48 PM  
Blogger madlib said...


I realize it is a long shot that you will see this since you last updated this back in July of last year. My name is Cody McGrath and I was a student in a couple of your intro to microeconomics and macroeconomics classes a few years ago when you were teaching at the University of Montana. Back then I was a business major with little interest in econ, though I found your classes very intriguing and enjoyed your teaching very much. I didn't know it then, but after completing my coursework for my undergrad I realized you were easily the best teacher I had at U of M. I want to thank you for that. Since then I have become very interested in economics. I was wondering if I might pick your brain on a few issues that I have been wondering about. If you ever see this and wouldn't mind answering a few of my questions I would love to hear from you. My e-mail address is Again thanks for being such a great teacher and I hope to hear from you. If not best wishes.

Cody McGrath

9:19 PM  
Anonymous Tucson Injury said...

Interesting data. I believe the accidents does not entirely depend on the age or gender of the cyclist. I think it is more of how responsible they are when they are on the road.

5:24 AM  

Post a Comment

<< Home