A good instinct, but the point is it's important to think about what the data means. The authors of the paper are just saying that the populations of people driving each type of car are different so looking at a percentage of fatalities is not particularly informative. If you took all those people who currently drive 20 year old fords at 10 under the speed limit and put them in teslas, maybe the tesla fatality rate per billion miles would start to look a bit better.