> because they legitimately represent more risk Do they? Two equivalent ways of approaching the problem: Risk is the same for everyone and decreases with age

> because they legitimately represent more risk Do they? Two equivalent ways of approaching the problem: Risk is the same for everyone and decreases with age OR risk stays the same for every member of the population throughout their lives. What's interesting with fatal car crashes is that they can happen only once in a driver's life (you can't die twice). So when looking at an age cohort, the young male who dies at 18 can't die again at 36 and is therefore going to be excluded from his age cohort. Now the question is how could we predict the individual risk level (parent income? Education? Zip Code? Race?). Doing it accurately would probably be illegal, so we accept that the young careful drivers will just foot the bill for everyone.