Citibike's first fatality: How safe is bikesharing?

Citibike suffered it's first bike fatality ever yesterday.

The bicyclist, Dan Hanegby of Brooklyn, was riding on 26th Street between Seventh and Eighth Avenues in the Chelsea neighborhood around 8:15 a.m. when he swerved to go around a parked van, struck a bus next to him that was traveling in the same direction, tumbled off the bicycle and fell under the bus’s rear tires, the police said. Mr. Hanegby sustained severe trauma, the police said, and was pronounced dead at Bellevue Hospital Center.

It's been hard to determine how safe or dangerous biking is because while we have a good grasp on the number of fatalities, we don't have a good idea on the exposure (miles or minutes of biking). Bikeshare keeps track of this and gives us a better idea, though not perfect. Through 2016, there had been 88 million bikeshare trips lasting as average of 12-25 minutes (frustratingly, NACTO doesn't give the overall total but breaks it out into two types of users). And systems report usage of 1.2-2.5 miles per trip. So, some back of the envelope calculations.

1. Extrapolating from the 2015 and 2016 trip data, if growth is consistent we can expect ~33.8 million trips this year. We're about 5/12th in so that means we'43 about 14 million more trips for a total of 102 million trips since 2010.
2. At 12 minutes per trip that's 1.22 billion minutes of biking
3. At 1.2 miles per trip that's 122.4 million miles biked

Which means we have 1 fatality for every 61.2 million miles bikes or 1.6 per every 100 million miles. Driving results in 1.25 fatalities per every 100 million miles of driving, so it seems like bikesharing is less safe than driving. (Though I prefer the minutes of exposure metric, but that's not as easy to find). 

But... 1.2 is the number that came from Capital Bikeshare a few years ago (and may have been the "as the crow flies" number instead of actual miles). Citibike was getting a higher mileage per trip - around 1.4.  As you can see in the image above they began to become a larger share of the US bikeshare data. If we use 1.4 miles per trip the fatality rate drops to 1.4 per 100 million. So still not as safe as motoring, but close. But other indications are that it is safer than other bike riding:

In Washington DC, a total of 14 crashes were reported in the first year of operation, of which only one was serious in nature. Approximately one million trips were made during this same period for an injury crash rate of 0.83 injuries per million miles (the average trip length was approximately 1.2 miles per trip), which is lower than the injury rate of 7.3 injuries per million miles ridden for private bicycling in Washington, DC.

Comments

[Beltsville]

Ballpark safety is good enough for me. Using your rough approximations, you are only 30% more likely to die on a bike than die in a car, which is still a very small number. Once you consider the reduced fatality rate in other factors of life, cycling comes out WAY WAY ahead. I like saving money and being healthy.

[DE]

Sample size is getting better with time, so the numbers mean more as we go along.

"(Though I prefer the minutes of exposure metric, but that's not as easy to find)."

I assume for the reason that a car can go 70 miles in the same time that a cyclist might go 18, so comparing by miles would largely overinflate the dangers of cycling compared to time spent performing the activities. (Though since it's bikeshare being used here, the speeds would be closer to the amount of time spent in the activity--cars in the city aren't going 70, so maybe 25 mph versus 12 mph or so.)

[washcycle]

DE, I prefer the minutes metric because I think cyclists probably spend more time exposed, but not moving (since they don't cruise down the highway for hundreds of miles), and you don't have to be moving to get hit. As I like to say, the tree across the street from my old place was hit by cars three times, but it moved 0 miles.

Another factor is that we should really be comparing driver fatality rates - since I suspect that that is different than driver&passenger rates and since the choice is more like to be bike or drive, not bike or ride.

[scoot]

Also note that the bikesharing data is exclusively drawn from urban areas. Is the driving fatality rate a national statistic?