Except you then divide the total cost by the number of people that bear it, and the term cancels out. Your reasoning implies that two governments merging justifies an increase in the number of restrictions simply because the numbers are bigger.
'the public' is the group of people that bear the cost. And no, it doesn't cancel out; the cost of avoidable accidents to the individual members of the population is small, but not that small. AAA (which is in the sinsurance business, admittedly) estimates, based on Federal Highway Administration data, that fatal accidents cost about $6m on average, and that the total cost to Americans is >$1500 per person per year: http://newsroom.aaa.com/wp-content/uploads/2011/11/2011_AAA_...
This is actually a great study - comprehensive, rigorous, and quite recent. The overall estimate of accident costs is just shy of $300bn/year, which is something like 1% of GDP. Even if we assume only 1/10th of fatal accidents involve innocent victims of careless drivers - probably somewhat low - that's still a substantial burden on the economy, which the public has a collective interest in mitigating.
I was trying to compare the expected societal cost from one person driving without a seatbelt, to the actual cost to the victim in one instance, which seems wrong in retrospect. Honestly you took me back with that $6M number. But then I looked into it, and 98% of that is "work-loss costs", which are clearly borne by the victim. The actual costs to the rest of the group are nowhere near this amount (roughly $120k), while the victim is sacrificing $6M of self-value. The victim is by far the loser, so the idea that the victim somehow externalized his costs and cheated society is preposterous!
