I just read that the external social cost of having another driver in California — due to accidents and not even citing congestion — ranges from $1,725 to $3,239 a year.
The number seems high to me (I couldn’t spot any problem in the paper), but I learned a new argument for market failure.
An externality arises from the difference between the average and marginal costs of accidents. Each driver pays, on average, half the cost of an accident involving two people. But in Coasean terms, the "marginal product" of either driver was to cause the full accident (this seems to ignore that on some days, certain drivers will crash with someone else, virtually no matter what, but of course this does not cover every case either).
So should you go drive in Idaho instead? Or can the argument be flipped to create a comparable uninternalized benefit from driving? When you drive to someone or somewhere else, you will bring some consumer surplus or producer surplus to the other person. Or is the "complementarity of the match" more potent on the downside, with accidents?
That is all from Aaron Edlin and Pinar Karaca-Mandic, in the October 2006 Journal of Political Economy. The authors also claim that an optimal Pigouvian tax on driving could raise $66 billion in the state of California, more than the other state taxes put together.