... and I question the premise
And I would agree with that. As you point out, the situation is not analogous. My own thought is the OP is kind of getting ahead of him/herself. For autonomous cars to make an impact on insurance rates, you would have to have a significant portion of the vehicles be autonomous. Frankly, I can't see that happening for many years, as only a small portion of drivers can afford to go out and buy a brand new car. For buyers of used cars, it will be a long time before they can get their hands on an autonomous car. Doing a quick google for just the US, 2014 new car sales were on the order of 16 million cars. Used car sales were on the order of 41 million, about 2.5 times as many a year. Say new car owners kept their cars three years (arbitrary number) - it would be almost six years before enough cars were traded in for one year of used car sales. According to google, there are 254 million used cars in the US. Assuming six years before used cars start getting replaced, six years of new cars would be 96 million cars replaced, leaving 158 million to be replaced still. From that point figure 57 million get replaced every year, it would be another almost 5 years until all the cars are replaced. That's 9 years for what I would call a best case. I think it will likely be much longer than that, since a large number of used car buyers are buying cars more than three years old (so they'd have to wait that much longer before getting their hands on an autonomous car).