Ed Glaeser writes,

Consider the purely hypothetical case of a massive oil spill in the Gulf of Mexico. The traditional libertarian would argue that regulation is unnecessary because the tort system will hold the driller liable for any damage. But what if the leak is so vast that the driller doesn’t have the resources to pay? The libertarian would respond that the driller should have been forced to post a bond or pay for sufficient insurance to cover any conceivable spill. Perhaps, but then the government needs to regulate the insurance contract and the resources of the insurer.

I do not want to rely on torts. And I do not want to rely on government-regulated insurance.

Further thoughts are below.1. Suppose that instead of rare oil spills, we had oil leaking at a constant rate. Also, suppose that instead of complex interests we have only two interests to be balanced. That is, we have two property owners–an oil driller owns drilling rigs and a fishermen owns the fish (or the right to fish). There is an optimal amount that the driller should spend to reduce leakage, based on marginal benefits to fish of less leakage relative to marginal costs of the leakage-reduction technology. To put it another way, there will be an optimal amount of steady oil leakage, and that amount will not be zero.

Depending on the initial assignment of property rights either the driller pays the fisherman for fish damage or the fisherman pays the driller to install leak-reduction systems. The Coase theorem tells us that either way we get to the optimum.

2. In practice, things are complicated because there are many sorts of property rights that are affected by leakage. There are multiple fishermen. There are resources other than fish that are damaged. It is costly to organize a coalition of anti-leak interests and to deal with free-rider problems within that coalition. For example, if I own a nice beach but I do not want to pay for my share of the costs of the anti-leak coalition, I may still free ride on the anti-leak technology that is obtained by fishermen or other members of the coalition.

3. In practice, things are complicated because leakage is not a continual process. Instead, spills are large but rare events. That makes it harder to calculate the optimal amount of resources to devote to preventing oil spills. The optimum involves equating the marginal benefit of reducing the probability of a spill of size X to the marginal cost of using technology that produces such a reduction in probability. Note, however, that it is unlikely to be optimal to spend so much on spill prevention that the probability of a large spill falls to zero.

Therefore–and this is quite counterintuitive–one cannot necessarily conclude from the fact that there was a spill that not enough was spent on spill prevention. I have not seen this point made in the coverage of the BP spill, although I missed a lot of the coverage because I was on travel.

Again, this probably seems counterintuitive. Most people would tend to view the fact that the spill occurred as evidence that not enough was spent on spill prevention. But we cannot really know that. Yes, I am sure that ex post there are specific things that would have been cost-effective at this specific well. But it could be that, ex ante, if you had identified things that could have been done to reduce spill probabilities throughout the gulf, including but not limited to the things that would have happened to have prevented this specific spill, the cost-benefit analysis would have said not to implement those steps.

Another way of saying this is that, just as in the case of steady leakage there is an optimal amount of leakage that is not zero, in the case of rare spills there is an optimal probability of large spills that is not zero. So, although it seems intuitive that this spill constitutes an economic failure (I do not call it a market failure, because government regulation was a major determinant of the investment in anti-spill technology), and that intuition may be correct, it is not something that we know with certainty.

4. An oil spill insurance market might help to discover the point at which the marginal benefit of spill-prevention technology equals the marginal cost. Glaeser is implying that the insurance market is needed because BP lacks deep enough pockets to self-insure, but it is not clear that this is the case.

5. Overall, I think that the case for government involvement depends on showing that the costs of assembling the coalition of spill-affected interests is higher than the cost of having the government make mistakes in attempting to represent such a coalition. We have gotten quite used to the notion that the government regulates oil drilling, so that the potential members of the spill-affected coalition take it for granted that they do not need to organize among themselves, except to try to lobby the government.

However, the same factors that would make it difficult for the spill-affected interests to organize in a laissez-faire environment also make it difficult for them to organize to lobby the government. Hence, it would not surprise me to learn that the amount invested in spill prevention technology is below the optimum level, either in a market environment or with government involvement. Certainly, it is not surprising that spills take place under government regulation. However, we do not know whether this was a government failure–it could be that the regulators imposed the optimum amount of anti-spill technology, and this was just the luck of the draw.

There is a false choice between having spills or not having spills. The true choice is between market mechanisms for determining the amount to invest in anti-spill technology and government mechanisms for making the determination. We have not tried the market mechanism, and I have no basis for claiming that it would work better or worse than the government mechanism.