Knowing the Future Could Make It Harder to Insure

We normally assume that a better forecast makes everyone better off. It gives us time to prepare and helps us avoid mistakes. Insurance reveals a complication. The ability to share a risk can depend on nobody knowing exactly who will suffer the loss. If a new prediction identifies the unlucky people before they arrange coverage, those people may find that the information has made their financial position much worse.

Consider a deliberately simplified town where every house has a one in ten chance of suffering $100,000 in damage next year. Ignoring expenses and profit, the expected loss is $10,000 per house. A sufficiently large insurance pool can spread that risk, allowing households to exchange an uncertain, potentially ruinous bill for a predictable premium. Now imagine a perfect forecast that identifies which houses will suffer the damage, and assume the damage cannot be prevented. For those houses, the $100,000 cost has become certain. A new policy covering it would need to be priced accordingly.

Nothing about the forecast has physically damaged a house. It has changed the terms on which people can share the loss. Before the prediction, each household could purchase protection against being unlucky. Afterward, the households known to be safe have little reason to contribute to a voluntary pool covering that particular loss. The identified households are left facing the cost themselves unless an earlier contract or another institution protects them. Information has revealed a difference that the original arrangement allowed people to insure against.

This example illustrates a problem associated with economist Jack Hirshleifer’s work on the social value of information. Learning more can reduce opportunities to share risk when the information arrives before people can commit to an agreement. It is a conditional result. Later research finds that the consequences depend on the information people already have and on the contracts they can sustain. Real forecasts are also imperfect, leaving plenty of uncertainty that insurers can still cover. The point is that better information can have costs as well as benefits.

The most obvious benefit is prevention. If a forecast lets a household reinforce a building or avoid a dangerous location, it can reduce the actual loss. That changes the calculation considerably. There is a useful distinction between information that helps us change an outcome and information that mostly identifies who will bear it. The first can make protection cheaper by reducing damage. The second can make pricing more precise while making coverage much harder to afford for people classified as dangerous risks.

One possible response is to arrange protection earlier, before the information is revealed, with terms that remain dependable afterward. Another is to spread some costs through a broader public arrangement. Both raise difficult questions about incentives and who should pay. Better forecasting does not answer those questions for us. As prediction improves, I think we will have to pay more attention to whether people can insure themselves against receiving bad news, as well as against the loss that follows.