That might seem like a logical point, but think about it a little more; if speeding is helping some people to not crash, then the point you’re effectively making is that speeding might actually make you even more than twice as likely to cause a fatal accident as a drunk driver, but sometimes helps enough that it evens out to be only twice as likely to cause a fatality.

It’s not true that non-crashes aren’t represented here- the NHTSA statistics capture the rate of crashes in terms like crashes per vehicle miles traveled, which represents both the probability of crashing and the probability of not crashing (one minus the probability of crashing). The fact that speeding is a huge factor in many crashes and that it increases the chance of death does in fact demonstrate that it’s overall less safe to speed. Speed might anecdotally help some people in some situations, but the actual stats do prove that the net effect of speeding is greater risk.

It's definitely harder to count the time spent speeding when that speeding doesn't end in a crash, but the NHTSA still tries to do it [0]. About 19% of vehicle miles are driven at more than 10 mph above the speed limit on highways, somewhat fewer on city streets.

[0] https://www.nhtsa.gov/sites/nhtsa.gov/files/documents/812485...