Suppose I am a utility-maximizing, risk-averse trophy husband/wife. Currently I have a generous monthly allowance, but due to a lack of foresight (or the blindness of love), I do not have a prenuptial agreement. There has been no trouble in the marriage yet, but what if things go south? Perhaps I should put part of my allowance toward premiums on divorce insurance, which one company now claims to offer. Previously, the initial allocation of divorce risk could not be altered; now, perhaps, it can. Whether risk can be reallocated is an issue that is present in far less tawdry situations than this example; in fact, it is everywhere.
On Wednesday, October 22, Professor Lee Fennell gave a talk entitled "Risk Reversals" as part of the Chicago's Best Ideas lecture series. (The first CBI presentation on October 1 was a fascinating talk by Mary Anne Case about "Why Evangelical Protestants are Right When They Say that State Recognition of Same-Sex Marriages Threatens Their Marriages and What the Law Should Do About It.")
The "stickiness" of risk allocations motivates Fennell's analysis. Many scholars have proposed new mechanisms for shifting risk. Fennell's investigation, however, tackles the analysis at a different level; rather than ask whether any particular risk-shifting scheme is desirable, she asks how easy or difficult, given an initial allocation of risk, transacting away from that allocation should be. Buying life insurance is easy, but selling my ability to bring a tort claim if my spouse dies in an industrial accident is difficult. I must buy car insurance, but I cannot insure against damage claims from accidents that have already occurred (although settlement is a form of insurance). How close are we to a world of Coasean risk shifting, and how close should we be?
The lecture did not advance a thesis about risk reversals so much as highlight that risk reversal is a pervasive consideration for any risk allocation scheme. Fennell thus proceeded to give some taxonomies of risk reallocation, drawing attention to coverage gaps and difficulties along the way. One distinction is whether a third-party facilitates the transaction or party A just sells the risk to party B. Another distinction is between insurance versus reverse insurance. Many of the examples of risk that cannot be shifted come from reverse insurance. Reverse insurance unwinds existing insurance or puts risk on a person that had been allocated elsewhere. Selling unmatured tort claims is an example of a gap in reverse insurance. Suppose I often jog along the highway, I want to sell my right to recover in tort if I get hit by a car. The cost of the risk is held by the driver (or his insurer), but if such reverse insurance was available, now I, the jogger, could hold that risk. Insurance is more common than reverse insurance, but insurance has its share of gaps. For instance, insuring my house against price declines is difficult but now possible. (Fennell has also researched homeownership risk more generally (see Part I.C).) It is extremely difficult to unload the two biggest sources of undiversified risk for Americans: their houses and their jobs. Selling shares in future earnings could diversify away job risk, but so far this practice is rare: a minor-league baseball player and a novelist have attempted it. Potential legal barriers may explain the relative absence of this scheme and some of the scholarly proposals above.
Taking stickiness into account can change whatever is the "traditional" story. The traditional story for adverse selection goes like this: An insurer cannot tell whether purchasers are "good risk" or "bad risk," but purchasers know which type they are. Policies are a better deal for the "bad risk" group, so as more "bad risk" people sign up, the price increases and the "good risk" group drops out, leaving the "good risk" group uninsured and the "bad risk" group insured at a high price. Suppose instead that the "good risk" group has low inertia and the "bad risk" group has high inertia. ("Inertia" is the amount of nudging needed for a person to change whether she owns insurance.) Whether adverse selection unravels now depends on the default state, that is, whether the system is opt in or opt out. If the default state requires opting in, the "good risk" group quickly opts into the system, and the "bad risk" group drifts slowly into the system with many not opting in at all. If the default state requires opting out, however, the system unravels quickly as the "good risk" group swiftly exits but the "bad risk" group remains in the system. Although an opt-out system fails for the same reason as the traditional story, an opt-in system works. The takeaway is that myriad factors affect risk analysis.
The type of risk reversibility can affect its desirability. A risk improvement can be a backfill, for instance traditional insurance, which brings a person back up to a zero state; gravy, for instance a lottery ticket, which improves a person already in a neutral state; or a hybrid, for instance products liability coverage, which recovers from a loss with compensatory damages and goes beyond a zero state with punitive damages. In theory, each risk improvement can be unwound, but should unwinding a backfill be treated differently than unwinding gravy? One reason for differential treatment is that a large loss on an individual could trigger social insurance; selling an unmatured tort claim could put a person on welfare if the harm comes to pass. The seller does not internalize this cost, causing too much assumption of risk by individuals. This problem, however, could be separable from the risk type by restricting how much of a claim a person can sell.
One student asked about a problem with Fennell's reallocation system, and Fennell corrected her, "Well, it's not really my system." The existing system is not really her--or anyone's--system, and that system has its share of problems. Professor Fennell prompts us to question whether the present system is really the best we can have.