The New York Times Review of Books takes a look at Richard A. Posner's book "Catastrophe". Posner makes an economic argument that we should do more to prepare for improbable catastrophes, such as asteroids colliding with earth, global warming and even particle accelerators that run amok.

The expected costs of a future event are the costs of that event, if it should happen, divided by the probability that it will happen. Thus, if I offer you $1,000 if a tossed coin turns up heads, the expected cost of my offer is $500. (Suppose I offer you $100,000 if a card drawn at random from a full pack is the ace of spades. Would you prefer that offer to $1,000 tied to the toss of the coin? Anyone interested in maximizing his assets would: the expected cost of that offer to me -- and hence the expected value to you -- is $1,923.)

When a catastrophe is really catastrophic -- and Posner, it should be emphasized, isn't writing about ''minor'' disasters like the terrorist attacks of 9/11 -- it can have a significant expected cost, even if the event is extremely improbable. Consider, for example, the risk that a high-energy particle accelerator will produce a ''strange matter'' disaster. The official risk-assessment team for one of these accelerators, at the Brookhaven National Laboratory, offered a series of estimates, one of which puts the annual risk of a disaster at one in five million. That seems a very small risk. But since the disaster would kill six billion people, that estimate gives it an expected cost of 1,200 lives per year. Even if the risk is estimated more conservatively at one in a billion, it has an expected annual cost of six lives. Would we build such an accelerator if we knew that six people would die every year in which it operates?