Cost Benefit and the Government Machine

I have just been reading an interesting post on Marginal Revolution on the benefits or otherwise of Regulation. Tyler Cowen (why does anyone called Tyler always remind me immediately of Tyler Durden?) says in part:

1. Government regulations have a very large aggregate net benefit relative to their costs; rules for clean gas, taken alone, might be more valuable than all the other regulatory costs we bear. If you don't believe me, try visiting Mexico City in November.

2. Many government regulations are simply unnecessary.

3. No one has come up with a good algorithm for weeding out the bad regulations from the good ones. Nonetheless cost-benefit analysis, for all its philosophic flaws, can serve this function.

Which is fair enough. But what do we mean by cost benefit analysis? Consider a very unlikely event with quite a severe impact: say cost $1B and probability 1 in a million. Cost benefit analysis would suggest it is worth spending no more than $10 to prevent this happening, but if you owned a $1B building, you would probably pay more than $10 to insure it against a one in a million earthquake. I certainly would. (And of course the whole insurance industry is based on paying more for protection than it is 'worth' in a pure cost-benefit sense.)

Really then we should perform cost benefit analysis relative to a utility function, and part of the reason we might disagree about regulation is a simple disagreement about what this function should be.