Is cycling safer than driving?

When I first started bicycle-commuting, seven years ago, I was a bit worried about riding every day on roads designed only for, and filled with, multi-ton steel behemoths, driven at high speeds by stressed-out and distracted humans. I tried to find good, reliable traffic-safety statistics to comfort my wife... or bring me to my senses!

Well, at that time I could only find some high-level stats from the NHTSA saying that about 48 times as many people die in car crashes as die while cycling. We could just take that at face value and say cycling is 48 times as safe, but we all know that frequent cyclers are a tiny minority. Let's be fair to the drivers! More useful would be data about those of us who ride a lot, and compare us with the huge car-driver majority.

I found more NHTSA stats saying "14%" used a bike "20-31 days" per month in the summer. I also once managed to find data saying about 2.5% of the general population "ride regularly"... whatever that meant (and I can't locate the source any more either). Let's use that lower number of regular cyclists, to stack the odds in favor of cars (more riders would mean a lower rate of deaths per cyclist). So, assuming those dedicated cyclists making up 2.5% of the general population accounts for all of the cycling deaths, that's roughly 1 in 40 of the total population.

So the odds of being flattened while riding a bike seemed about the same as dying while driving a car (within the general population). How so, you ask? Well, there are 40 times as many drivers as regular cyclists, using my low number of 2.5% regular cyclists from above: 100% ÷ 2.5% = 40. Thus, with 48 times the fatalities for drivers, and 40 times as many drivers, the odds seem sort of closely related to the general population. OK, I know, drivers aren't 100% of the general population, but if we count passengers it gets close enough for my purposes here. (Hmmm... maybe I should count cycling passengers too. Got any stats for them?) So I figured that cycling was similarly dangerous, in simplistic terms at least, and maybe even slightly safer.

Some specific lessons I learned, though, about what groups were most at risk [though the below stats aren't necessarily "causal," that is, if females only cycle 13% as much as males, then their riding safety rat is similar to males]:
That made me feel even better since I wasn't in any of those risk groups, well mostly. Just 3 out of 5... uh...

But none of that data addressed deaths per mile traveled. I figured that skewed the odds heavily in favor of drivers since most cyclists ride far fewer miles. Maybe my maturity, helmet use and use of lights at night even the odds out into similar levels of danger between me riding to work versus me driving to work. Hopefully.

I let things rest there.

Then later on I found some Web pages for some group named "" that had more interesting statistics, thanks to my teammate Gary. Very interesting. Now, that Web site does have an axe to grind with regards to helmet use, but still has some neat info. For instance, the main conclusion I really liked to read was this one:
Apparently, staying sober while trying to move, or not trying to move while drunk, are both excellent ways of improving one's odds of survival.
Seriously, this data, from the main table, was pretty hopeful: Cyclists suffer "0.2" deaths per million miles traveled, versus "1.3" deaths per million miles traveled for motorists. Yay, cycling! That was some info I could really get to like!

But could cycling really be 6.5 times as safe, per million miles traveled, as driving a car? Where did that come from? Were the numbers and calculations real? The number of deaths is pretty reliable, but the number of miles we cyclists ride is critical to really comparing safety. How do they get at this: "3,000,000,000 miles cycled every year" (mentioned near the bottom of the page)? Sounded bogus to me, but:

If "2.5" cyclist deaths "per 1 million population" is correct (also from their main table), then we should have about 750 cycling deaths per year (assuming 300 million U.S. population x 2.5 cyclist deaths per million = 750 deaths). I don't see total deaths of cyclists on this page for comparison, but the US DOT says it was "622" in 2003, and the NSC says "926" in 2006, so that is right in the middle. They look right so far.

To get at the critical data for miles ridden by the cycling population, we'll have to do some number-crunching. We'd go: 750 deaths ÷ 0.2 deaths per million miles = 3,750 million miles (3,750,000,000 miles); close to the "3,000,000,000 miles" number they mention. Well, I'll be darned. Think of the number of tires we go through.

Anyway, just to get a reality-check, that means that your average U.S. citizen rides annually: 3,750 million ÷ 300 million = 14.4 miles, or less than I ride one-way on my way to work. OK, that seems about right! :-)

Let's figure out if the car-driver numbers are right, per these numbers. Well, I'm no mathematician, but I think I spotted an error in that main table from It says that there were "12.9" motorist deaths "per 1 million population." As bad as that looks, I think the decimal should be removed. The actual deaths from cars per year are usually in the 40,000 range; that would be closer to 154 motorist deaths per 1 million population (the NSC report here said 45,316 deaths from autos in 2006). In this case let's assume they meant "129" not "12.9" for a total of 38,700 deaths. Otherwise we'd have only 3,870 deaths per year from cars and even I might start driving!

OK, back to number-crunching: 300 million x 129 auto deaths per million = 38,700 deaths. Sure, that could be true in a really good year for cars. Then, that would mean 29,769 million miles traveled (29,769,000,000 miles, or 29.8 billion miles; yikes!).

But I don't want to assume any of those numbers are right, so...

According to the Federal Highway Administration, the average miles traveled for motor vehicles in the U.S. is about 12,000 miles a year. Assuming a 300-million population, as above, then we'd expect 3,600,000 million miles of auto travel (I think that's 3.6 trillion miles!). But the actual number (in 2006) is slightly lower at 3,014,116 million miles, no doubt because kids under 16 years old don't drive much, though they might add a few miles to the "3,750 million miles" of cycling calculated above.

Now then, for the 45,316 auto deaths in 2006, over the 3,014,116 million miles, that's 0.015 deaths per million miles (45,316 auto deaths ÷ 3,014,116 million miles traveled = 0.015). That number also matches well with data from the US DOT which says: "The fatality rate, computed per 100 million VMT, [was] 1.36 in 2007..." (or 0.0136 per million miles; pretty close to my rough 2006 calculation).

Uhhhh, but what happened to that very high "1.3" deaths per million miles mentioned? I was starting to think cycling is safer! Now cycling looks to be 13 times as dangerous per mile traveled as in a car! I was so hopeful.

Well, I won't trust any data too much, especially if it just comes off of some random Web page, but I won't stress out about it either. I can still rationalize that "my maturity, helmet use and use of lights at night even the odds out into similar levels of danger between me riding to work versus me driving to work," to quote myself. And since I ride fewer miles than I'd otherwise drive, about half as many (I ride about 10,000 miles a year; though that's kind of a lot I'd drive about 25,000 miles if I drove solo to work every day), my odds even out a bit there too. None of my calculations account for a rider's or driver's experience level, and I've heard that experienced cyclists have much safer records than, say, a 15-year-old boy. But since the same allowance could be applied to experienced motorists I will refrain from trying this.

Assuming the death rate truly is "0.2" deaths per million miles traveled by bike, and the 0.015 deaths per million miles traveled for motorists I calculated, and I use my personal mileage from above, then my odds of dying are:
  • Me, cycling: 0.002 chance per year (1 in 500 over 10,000 miles)
  • Me, driving: 0.000375 chance per year (1 in 2666 over 25,000 miles)
Hmmm, I believe that makes cycling 5.3 times as dangerous for me. Not what I want to hear! If you don't like these numbers, pick your own numbers. Let's dig a little deeper...

Annual cycling mileage is so key to comparing safety, but there's so little solid info (no offense to, because cycling use doesn't leave a clearly-identifiable trail like, say, gasoline use does. But I dug up some more info from the US DOT's Bureau of Transportation Statistics on bicycling mileage and ran those numbers for comparison. They say that "bicycling accounted for 0.2 percent (6.2 billion miles) in 2001."

If I use that higher cycling mileage from the DOT, instead of's number, the odds for cyclists improve by over twice. But if I recalculate the stats for cyclists with what I consider to be the better data I have used here, I get a slightly higher death-rate for cyclists per 1 million population: 300 million ÷ 926 deaths = 3.087. Still, that higher number of cycling miles yields a lower death rate of 0.149 deaths per million miles traveled by bike.
  • Me, cycling (recalculated): 0.00149 chance per year (1 in 671 over 10,000 miles)
I'd have to ride 10,000 miles every year for 671 years to make death likely! Better, but still 3.97 times as dangerous for me as if I drove. [Though I think the death rate per hour of travel may make more sense, which would make the risk very similar to driving.] Still, there are so many great reasons to ride a bike that I'll continue to ride even if it could be definitively proven less safe:
  • Stress-relief (never get stuck in annoying traffic!)
  • Fitness (ability to leap tall buildings in a single bound, and admiring glances from the opposite sex!)
  • Health (disease prevention; I like almost never getting sick)
  • Finances (I save about $3600 per year in gas alone, just by not driving to work, assuming $3.20/gallon)
  • Environment (car use is our single biggest contribution to global warming)
Yeah, I wish the numbers were more in my favor, but they're close enough. If I tried to factor in the risks of a sedentary lifestyle it might get too complex for me to handle!


Dave said…
"I'd have to ride 10,000 miles every year for 671 years to make death likely!"

This may or may not be true. What is almost guaranteed, though, is that several hundred cyclists, who won't have ridden 10,000 miles every year for 671 years, will indeed die in accidents each year.