Double Counting and Other Sins

In a previous post I mentioned this JPE paper on driving costs. As you can guess from my title, I believe the paper has some serious shortcomings.

The gist of the paper is that due to differences in average cost and marginal cost there is a negative externality cost that is not efficiently covered by insurance costs. The paper then goes on to recommend a Pigovian tax to correct the externality. My first thought on reading the summary was that the paper was about accident costs that are not typically covered by insurance - traffic delays to other drivers, time lost filing police reports, waiting for the towtruck, dealing with the car being in the shop, and so forth. But the authors quickly point out that they are not considering such costs, but rather are considering the marginal cost of an accident at twice the cost of the accident, based on the "it takes two to tango" argument, i.e. if either the at-fault car or the other car were not there then there would be no accident. The authors analyze the insurance data, conclude there is an externality, and then propose several Pigovian schemes.

The first problem I see is the apparent double counting of costs. From the paper:

In California, a very high-traf?c state, we estimate that a
typical additional driver increases the total statewide insurance costs by $1,725 +/-$817 to $3,239 +/-$1,068 each year, depending on the speci?cation. In contrast, in North Dakota, a very low-traf?c state, we estimate that others’ insurance costs are increased only slightly (and statistically insigni?cantly): $10 +/-$41 each year, as shown in speci?cation 10 of table 5 below. These estimates of accident externalities pertain only to insured accident costs and do not include the cost of injuries that are uncompensated or undercompensated by insurance, nor other accident costs such as traf?c delays after accidents.

Yes, and insurance rates in the S.F. Bay area and L.A. Basin reflect this quite well. Choosing to become a driver in either of these areas will cost the driver about that much adjusted for driving risk and insurance company profits. Part of the problem may be the mis-identification of the actual externalities created by each driver. The externality is the accident itself, but it is compensated for by the insurance or other payment to the other driver. If we then say that those insurance costs are an externality and add on a Pigovian tax, then another researcher ought come along and claim the necessity of the original Pigovian tax is an externality and we begin the descent into infinite regress.

In the category of Other Sins, the authors say:

Yet in many cases, from the vantage of causation, as distinct from negligence, economic fault will sum to more than 100 percent. Whenever it does, ef?cient driving incentives require that the drivers in a given accident should in aggregate be made to bear more than the total cost of the accident, with the balance going to a third party such as the government.

Actually, economic efficiency requires the "victim" of the negative externality be compensated by the person causing the negative externality by some agreement between the two parties. Giving the balance to the government creates perverse incentives. For example, a large Pigovian tax generating $66 Billion of revenue would encourage that government to eliminate funding, raise taxes, or otherwise create disincentives for alternative transportation in order to further increase revenues.

I know I said sins, plural, but quite frankly I've already devoted more time to this paper than it merits. After having my hopes raised by other bloggers I respect, I am very disappointed.

Share this