It depends how you group the age cohorts, but my recollection is that the second oldest cohort (something like 65-75) actually has one of the lowest accident ra

It depends how you group the age cohorts, but my recollection is that the second oldest cohort (something like 65-75) actually has one of the lowest accident rates, and the oldest one has a higher accident rate per mile but tend to drive fewer miles total (in particular, because they're not commuting to work.) So the benefits of taking away licenses from the elderly might be fairly limited compared to say, more strict testing for teenagers and tighter DUI rules. Also, a good point someone was made is that it may be easier for technology to solve the problems of elderly drivers who still have good cognition. E.g. you can make up for their limited neck flexibility with blind spot radar and backup cameras, and poor reaction times with collision detection systems. It's harder to do that for a teen who wants to drive too fast or make aggressive lane changes.