Study indicates that CaBi is safer than personal biking, but not motoring; and provides little safety-in-numbers effect

A recent study of bikesharing safety that used data from Capital Bikeshare and other systems found that "bikesharing appears to have a lower nonfatal injury rate and fatality rate than personal cycling." Despite the lower injury rate, CaBi still appears to be more risky then motoring. In addition, the study determined that CaBi was not contributing much, if any, to a safety-in-numbers effect. The data comes from CaBi and the authors note that

Capital Bikeshare does record non-injury collisions when they are reported. Capital Bikeshare reports confirmed crashes, as long as the location is known, even if there is no police involvement or hospital visit. But like Nice Ride Minnesota, Capital Bike Share does not search police reports or emergency room data to find incidents. Those incidents must be reported back to the operator by police, emergency medical services, or a member. There do exist cases of “discovered damage,” in which incidents appear to have occurred but no report was made.

Over nearly 5 years, CaBi reported 150 collisions and 55 hospital injuries

Compared to personal cycling

The good news is that CaBi seems to be safer than personal cycling. CaBi users had a non-fatal injury rate that was only 65% of the national rate for all cyclists.  

In addition, CaBi has had no fatalities (in fact there have been none in the United States). Though the sample size is likely too small for accurate comparison, if there had been  1 CaBi fatality by mid-2015, the fatality rate for CaBi users would be roughly half the national rate for cyclists (10.6 per 100M trips as opposed to 21).

Of course, comparing national data to data from a bike-friendly city brings in a little bit of an apples to oranges issue, but it still is encouraging.

For reasons unknown, DC's has a higher crash and injury rate than the San Francisco Bay Area does (Nice Ride doesn't record this data).

But it's safer on a per-mile basis

Comparing per-mile data for vehicle-collisions from British Columbia, CaBi's rate of 277 per 100M miles compares favorably to that rate of 425 per 100M miles. 

The primary reason for bikesharing being safer seems to be the design of the bikes themselves, but experts note that their presence exclusively within urban environments - with slower speeds - could also be a factor.

Compared to motoring

Despite improved safety, CaBi's injury rate of 919 per 100M trips was still higher than the rate for passenger-vehicle occupants which is 803 per 100M trips. 

Safety in Numbers

 

 

The study authors suggest that DC is not benefiting from the "Safety in Numbers" effect.

The high linear correlation between the rise in bicycle commuters and collisions provides an early suggestion that a rise in bicycle activity may have led to a proportional rise in collisions during this period in at least two of the three case study areas [Minneapolis being the exception].

They further looked at safety-in-numbers by comparing data based on zip code and again it doesn't appear to be much benefit.

The results, presented in Table 10 suggest that in Washington DC, there is little to no relationship between the relative change in collisions within a zip code and the number of bikesharing trips in that zip code. This result suggests that there is little evidence in the collision and activity data to support a safety-in-numbers effect resulting from Capital Bikeshare. At the same time, it also does not suggest that the presence of Capital Bikeshare is contributing to an increase in collisions within high-volume zip codes. The number of Capital Bikeshare trips within an area is not found to alter the localized change in bicycle collisions from the broader trend that is underway. That is, a growth in collisions occurred in the evaluated region (Washington DC), but as shown earlier in Figure 15, this growth was broadly correlated with a general increase in cycling activity.

Other Stuff

They also created a set of crash heat maps and concluded that "some change in the concentration [of crashes] has occurred, but in general the spatial pattern of collisions was found to be rather stable over time."

Unrelated to safety, but interesting nonetheless is a comparison of ridership data for the systems in DC, Minnesota and the Bay Area. Minnesota shuts its system down from November until April, but comparing DC's to San Francisco's does show how the weather differs, and how that impacts usage. 

Comments

[DE]

"For reasons unknown, DC's has a higher crash and injury rate than the San Francisco Bay Area does (Nice Ride doesn't record this data)."

Because in SF, it's mostly the cab drivers who drive like they are insane, whereas here, that's just standard driving procedure? Having ridden both places, that's my impression. YMMV.

I do still feel like there might be a safety in numbers effect, even if it hasn't been measured (or hasn't happened) yet.

[xmal]

Thank you for the writeup, but want to check the per-mile and per-trip numbers. You use 10 million as your denominator, but the tables say 100 million.

Are the vehicle stats also per 100 million, or are they really per 10 million, in which case bikeshare is an order of magnitude safer? Thanks for clarifying!

[washcycle]

xmal, sorry that was my mistake. All numbers are in terms of 100 million. I 'll correct it.

[washcycle]

DE, that's one possible explanation. It would be easy to test by determining the crash rate for pedestrians and comparing it in both cities we could determine if "different drivers" was the cause.

Another theory of mine is that the Bay Area system is smaller and thus found only in the most bike-friendly parts of town, while CaBi expands out into the suburbs.

One possible explanation for the lack of a safety-in-numbers effect could be cyclist experience. It's likely that as the number of cyclists increases - especially if it happens rapidly - that the average experience goes down. Given a group of cyclists with 10 years of bike commuting experience and another with 2, I'd expect the former to be in fewer crashes. So if average experience is going down, but crashes are remaining level, then cycling is actually becoming safer, because less experience is needed to get the same effect.

Anyway, there are lots of variables and teasing it all out will take years. But I predicted that bikeshareing would created a wealth of new data for researchers and this study shows how that data can help us understand all of these factors better.

[DE]

That's a good idea. Got me looking and I haven't found what I'm looking for yet. But interestingly, DC's pedestrian *fatality* rate was actually lower than California's as a whole for 2013 (1.39 vs. 1.83 per 100k poulation). Maryland's was 1.82; Virginia's was 0.91.

http://webcache.googleusercontent.com/search?q=cache:ZgRbWZxDENcJ:www-nrd.nhtsa.dot.gov/Pubs/812124.pdf+&cd=1&hl=en&ct=clnk&gl=us

In 2013, DC had 9 pedestrian fatalities; according to the SF police department, SF had 21. DC has about a 20% smaller population than SF.

Hope to find pedestrian crash rates if work time allows.

[Crickey7]

Interesting stats. The not-safer-than-car trips assertion may be true, though one needs to note that a "trip" in each has wildly different characteristics. Pretty much any baseline you would use for that statistic would have other issues, though.

[DE]

Mostly finding data for deaths. In 2012, SF had a ped death rate of 1.70 per 100k population; DC had a rate of 1.11.

http://webcache.googleusercontent.com/search?q=cache:NixvHiry2Z0J:www-nrd.nhtsa.dot.gov/Pubs/811888.pdf+&cd=3&hl=en&ct=clnk&gl=us

NHTSA does have this little gem for 2013:

"The individual State percentages of pedestrian fatalities by total
traffic fatalities ranged from a low of 0.7 percent (North Dakota)
to a high of 45.0 percent (District of Columbia), compared to the
national average of 14.5 percent."

Could be skewed since DC is the only "state" that is entirely a city.

[washcycle]

Also, the authors chose to compare only bike-vehicle crashes for cyclists to vehicle crashes for motorists. But I'm not sure excluding crashes that don't involved a motor vehicle makes sense. Certainly the motorist crash data includes single-car crashes. I'd really compare CaBi's 1412 crashes per 100M trips to 803 per 100M trips for motorists. Which makes it much less safe - on a per trip basis (with all the issues that involves).