Published December 1, 2006 | December 2006 issue
The views expressed are those of the authors, not necessarily of the Minneapolis Fed. For a more technical discussion, see "A General Equilibrium Interpretation of Damage-Contingent Securities," by R. Anton Braun, Richard M. Todd and Neil Wallace, in Journal of Risk and Insurance 66 (December 1999) 583-95.
Hard winds have been blowing through Florida's homeowners' insurance industry of late. Following disastrous hurricane seasons in 2004 and 2005, private insurers doubled or tripled premiums in many areas of the state, and some refused to renew policies. Citizens Property Insurance Corp., a quasi-public entity that provides insurance to property owners who feel priced out of the private market, has become the largest home insurer in Florida, with more than 1.2 million policies. This year, Citizens was obliged to take on over 300,000 new policies after three insurers folded under the strain of hurricane claims. With Citizens facing a $1.7 billion deficit from payouts in 2005, the executive director resigned in September, citing physical and mental exhaustion.
The turmoil in private windstorm insurance in Florida (insurers are also refusing coverage or charging high premiums along the Gulf Coast and up the Eastern Seaboard) is the latest in a series of U.S. property insurance price spikes since the early 1990s. Each time a major catastrophe has occurred—Hurricane Andrew in 1992; the Northridge, Calif., earthquake in 1994; the terrorist attacks on New York and Washington, D.C., in 2001; and the powerful U.S. hurricanes, culminating in Katrina's record-setting devastation of New Orleans and coastal Mississippi last year—the market for disaster insurance has reacted sharply, restricting supply and dramatically raising the price of the limited coverage that remained.
After each disaster, government has stepped in, attempting to pick up the pieces. To provide state-backed insurance of last resort, Florida's state government merged two existing insurance plans to form Citizens in 2002. The California Earthquake Authority was created by that state's legislature in 1996. The Terrorism Insurance Program, established after 9/11, provides a federal backstop for terrorism policies written by private insurers.
Government intervention into insurance markets in the wake of natural and manmade disasters has sparked lively debate over the past 15 years among economists and policymakers. Much of the debate focuses on whether the private insurance industry can ever be capable of insuring property against catastrophes, such as hurricanes, earthquakes, floods and terrorist attacks, without running for cover from Uncle Sam when the "big one" hits.
Given the public's familiarity with insurance for more common risks like fire and car accidents, this focus on insurance markets seems natural. But the economics of catastrophe risk are fundamentally different from those of risks covered by car, fire and other standard insurance contracts. In addition, consumer misperception of risk can distort demand for insurance. And government intervention itself can distort both demand and supply. (See "Flaws in the Ointment".) With these complications, it's little wonder that dealing with catastrophes remains highly problematic not only for those whose lives are caught in their destructive paths, but for the insurers, policymakers and academics who hope to mitigate their consequences in the future.
To understand the market's role in providing for catastrophe risks, it's essential to understand how sharing catastrophe risks differs from the familiar risk-pooling arrangements in most other property insurance markets.
Let's start with auto insurance, simplifying reality to highlight central points. Consider a relatively homogeneous group of drivers, perhaps 45-to-54-year-old rural residents with good driving records. Two key facts are that they each face about the same degree of risk of having a claim and that not all of them will actually have a claim. That is, they are all in the same position beforehand, but different positions after the fact. This symmetry in their beforehand risk exposure makes it easy to see that they might wish to find a way to share whatever losses befall some of them. Conventional insurance—with each member of the group paying a premium into a pool and those with accidents receiving a payout from the pool—is an effective way to share risks. The insurance industry, including its reinsurance companies, has a long history of successfully arranging and managing risk pooling along these lines.
Catastrophe risks differ in critical ways. Note, however, that size alone is not necessarily the critical difference. While the fact that catastrophes can cause large losses is important, catastrophe loss risks are not necessarily higher than the other losses that insurers routinely cover with few problems. For example, the total insured losses from Hurricane Katrina (about $60 billion as of August 2006, according to the National Oceanic and Atmospheric Administration) were about half of the total claims paid out by auto insurers in 2005 ($114 billion, per the Insurance Information Institute). Even with uninsured losses included, damage from Katrina ($125 billion) was only slightly larger than auto insurance claims.
Of course, the large losses from Katrina were much harder to predict than the roughly equivalent losses from auto accidents in 2005. Total auto insurance losses vary only slightly from year to year, making it simple to set total premiums and allowing the auto insurance system to function largely as a pay-as-you-go system, with only relatively small reserves of funds carried over from year to year. Is that a crucial difference? To answer that question, we have to explore further.
Start the exploration with a trip to an imaginary island. Everyone on this island lives on the coast (the interior is uninhabitable). Most years, the island is placid and reliably productive. However, the inhabitants all know that occasionally, perhaps every 10 to 20 years (although the precise number doesn't matter), a large storm will devastate part of the coast. These storms can come from any direction, so that each part of the coast is equally at risk. That is, the risks are still ever-present and symmetric beforehand, as in our auto insurance example above, but by contrast the total amount of actual losses varies a lot from year to year. For a degree of realism, we'll also assume that timing and severity of these storms can't be known in advance, but that's not essential to the story. Can the island's inhabitants effectively deal with these issues, or does the large variability and unpredictable nature of yearly losses preclude a good solution?
With a bit of reflection, it's clear that a good solution would be for the island's inhabitants to pool their risks in a simple pay-as-you-go system. The inhabitants devise an arrangement such that, when one of the catastrophic storms hits, those who suffer losses receive payments from those who don't. The amount of the payments should be set so that everyone shares equally in the total loss. In other words, a pay-as-you-go risk pool is effective even though total risk varies significantly and randomly from year to year. The symmetry of the risk (beforehand) among the island's inhabitants is the key to this conclusion. Note that there is no need to maintain large "reserves."
(The question of whether the islanders would benefit from storing up an inventory of goods that could be drawn upon in a disaster year is a separate question. Let's set it aside for now by assuming that all of the output on this island is completely perishable and cannot be stored away for a rainy day—papayas, say, not coconuts. A more detailed answer to the storage question would involve the degree of perishability of output and the frequency of storms. But it wouldn't overturn the basic elements of the simple risk-pooling arrangement described here.)
Note also that this solution might be implemented with or without "insurance companies." For example, with a mutual insurance arrangement, all inhabitants would be assessed a high but equal premium each year. In good years, their entire premium would be refunded at the end of the year. In storm years, the premiums would be used to pay losses, and any remaining surpluses would be rebated pro rata, so that all inhabitants suffer equally on net. A possibly simpler way to achieve the same end might look a lot like government "disaster payments" paid for by taxes on those not suffering losses. In fact, if we were all equally vulnerable, and if there were nothing we could do to change that (say, by mitigating risks), then the oft-maligned system of government disaster payments might, in fact, be optimal. Unfortunately, these assumptions seem very unrealistic. Under realistic assumptions, traditional disaster payments are inappropriate and conventional risk pooling is at best inadequate, as we'll see in the next example.
Our visit to the mythical island has shown that the sporadic and unpredictable nature of catastrophes does not in itself mean that simple risk-pooling arrangements are inadequate for managing the losses they inflict. But remember that our island is mythical, and the key myth is that everyone is equally exposed to catastrophic risk beforehand. This is not very realistic. In reality, the odds of major catastrophe seem lower in North Dakota than on some of our coastlines or earthquake faults. This asymmetry of risk exposure gives rise to additional potential gains from trade between the more exposed and the less exposed regions.
However, those beneficial trades are fundamentally different from the simple risk pooling that works well in auto insurance or on our mythical island. Using traditional insurance concepts to implement or even talk about this very different type of risk trading may be problematic, confusing and misleading. Let's first understand how to handle large, asymmetric risks effectively and then return to the question of whether something that looks like traditional insurance has an appropriate role.
To simplify again, let's take the extreme view that the world consists of Florida, which risks large, sporadic hurricane losses that damage all individual residents to the same degree, and North Dakota, which has no risk of catastrophe. (Alternatively, like our mythical islanders, Floridians are equally at risk beforehand and use an insurance pool to guarantee that they are also equally hurt after the fact. But unlike our islanders, they have an opportunity to reduce their collective after-the-fact pain by trading with a different region.) People can choose to live in either place, and when they choose to live in Florida, they are increasing the world's total exposure to risk. Under these circumstances, risk pooling can't achieve the full benefits of trade. Given our assumptions, Floridians have no risks (left) to pool among themselves, because they all suffer the same amount of loss after a hurricane hits. And they can't pool risks with the North Dakotans, because there are no risks in North Dakota.
Nonetheless, assuming that both Floridians and North Dakotans prefer to avoid risks, they can make risk-based trades that benefit each side. For example, Floridians might agree to pay a certain amount of goods to North Dakotans each year if, in return, North Dakotans agree to pay a larger amount of goods to Floridians every time a hurricane hits Florida. If the amounts are set appropriately, on net Floridians will benefit because they receive goods when they need them most. And North Dakotans will also benefit, as the present value of the payments they receive in normal years will be high enough, relative to the present value of the payments they make in hurricane years, to compensate them for taking on additional risk.
Assuming that North Dakotans and Floridians have similar tastes and similar subjective assessments about the odds of hurricane damage, equilibrium in the market for trades between them is likely to result in payments to North Dakotans being more than actuarially fair. That is, the expected present value of the payments they receive will exceed the expected present value of the payments they make. This need not be true if the tastes and views of North Dakotans and Floridians differ more, for example, if North Dakotans put lower odds on catastrophe losses. For the moment, however, assume that North Dakotans and Floridians have similar tastes and expectations. Because Floridians are risk averse, they will be willing to pay more than actuarially fair premiums in order to reduce the yearly variability in their net (after storm losses and payments to/from North Dakota) income. In addition, the fact that Floridians may have to pay in excess of "actuarially fair" premiums provides an appropriate incentive for people to think twice before adding to the world's risk exposure by moving from North Dakota to Florida. (Traditional disaster payments do the opposite: They provide an inappropriate incentive to move to the risky region.)
In the real world, this type of risk trading is most clearly seen in the emerging catastrophe bond market. ("Cat-bond" investors, in return for a premium interest rate, accept the risk of losing their principal if a defined catastrophe occurs.) There may be other less obvious arrangements to carry out the same basic exchange. Whatever the method of implementation, the key unconventional features needed to deal with large, time-varying, asymmetric risks are
In principle, then, it should be possible for society to work out suitable arrangements for managing catastrophic risks. However, it's not clear that we want to call those arrangements "insurance," at least not if we use insurance to mean risk pooling. And it may be that in using only insurance models and insurance language to implement and talk about catastrophe risk management, we end up confusing the issues and ourselves.