In the discussion on yesterday's post on transportation safety we started looking at this claim that "general aviation" (small planes) is much less safe than airlines and that even airlines are a bit less safe than driving, all on a per-mile basis. Looking more, I think that link is confused.

They calculate the following accident rates, all in fatalities per billion vehicle-miles:

driving: 14.7
airlines: 15.7
GA: 131

The problem is, these are per vehicle-mile when the actually useful information is per passenger-mile. If we found that 100-passenger vehicles had 10x the fatalities-per-vehicle-mile rate than 1-passenger vehicles, this wouldn't mean the 1-passenger vehicles were safer. Risk of death per mile is predicted passenger fatalities/passenger miles.

I recalculated these statistics (work and sources are here) in terms of passenger-miles. These are all fatalities per billion passenger-miles:

passenger cars: 12.49
us airlines: 0.12
motorcycles: 239

Even in the worst years for airlines (like 2001) they still are only at 0.75 fatalities per billion-passenger miles.

(I didn't recalculate the GA numbers because I'm less interested in them, but my understanding is that their occupancy numbers are quite low, and so switching from vehicle miles to passenger miles should only bring them down by 2x at most.)

(Because an airplane doesn't carry 1000 people, there has to be something else involved in why my numbers for airplanes have them 1000x safer than the meretrix ones. I've re-checked my calculations but either of us could have something wrong. Other people looking at my numbers would be great. Alex points out in the comments that 12 is 100x 0.12, not 1000x. Which means my numbers are in agreement with the earlier ones after all.)