Summer 1997
by John Campbell

Earthquakes, hurricanes, blizzards, forest fires, floods, tornadoes, and volcanoes have always plagued humankind. Yet we continue to settle in harm's way. Cataclysmic events almost never hit the same location twice in a lifetime, so memories fade. The United States was fortunate to experience relatively few catastrophes from the 1960s through the 1980s, and even the property/casualty insurance industry, which pays for much of the damage, was lulled by years of relatively low losses. The industry, says actuary Ronald Kozlowski of the consulting firm Tillinghast-Towers Perrin, "largely lost the discipline of measuring and managing exposures susceptible to catastrophic loss." Premium rates fell to levels that allowed little to be set aside as a reserve for catastrophic losses. ¶ But perceptions have changed since 1989, when Hurricane Hugo struck the Southeast and the Loma Prieta earthquake shook the San Francisco area. Soon after, several worse disasters hit. The Mississippi River flooded adjacent cities in 1993. Hurricane Andrew in South Florida in 1992, and the Northridge earthquake two years later near Los Angeles caused over $45 billion in damage (1997 dollars), with the insured piece at $30 billion. ¶

These recent events introduced a new sense of scale to the calculations of insurers, academic observers, rating agencies, and regulators. The reinsurance market reacted first, by raising prices and restricting coverage. Now a consensus is growing that the frequency and severity of recent disasters are not aberrations, but indicate a far greater societal exposure than previously recognized. Not only do insurance authorities expect more disasters, but more of the population now lives in vulnerable locations, such as along major fault lines in California or along seacoasts exposed to hurricanes. From 1970 to 1990, the population of the southeast Atlantic coastal counties increased by nearly 75 percent, almost four times the increase for the nation as a whole. Within New England, development has intensified in recent years on Cape Cod, Cape Ann, and coastal Rhode Island, Connecticut, and southern Maine. Roughly half the nation's people now live in areas prone to hurricanes or earthquakes, and the value of exposed property has risen accordingly.

A greater sense of uncertainty also pervades the calculations of potential losses. Insurance companies no longer perceive the experience of the recent past as a reliable indication of the future. Recorded observations of global climate activity have been made for just over a century, which is a short period, climatically speaking. And weather patterns do not simply repeat. So there are high-intensity hurricanes awaiting us that have not been seen in the past century.

People in the industry are now more concerned about events so large that they threaten the solvency of insurance markets as a whole. If Andrew had taken a slightly more northerly track, through downtown Miami, insured damage might have reached $50 billion. A major earthquake under Tokyo could approach $1 trillion. As one measure of financial muscle, the total capital and surplus of U.S. property/casualty insurers runs to about $250 billion; reinsurers account for $37 billion more. And this capital is applied to all risks, not just natural catastrophes.

The awareness of greater exposure has opened fissures in the system of disaster risk management in the United States. The question of insurability has become a public policy issue, beyond the possibility of insolvency of individual insurance firms. Some in the industry would turn to the government as insurer of last resort, since the government already issues flood insurance and provides disaster aid. But if government protection were extended without passing on the true costs to those property owners who incur the risks, it could encourage greater exposure and larger losses when the next big one hits.

Most people don't think much about natural disasters, because we don't believe low-probability events affect us, explains Howard Kunreuther, director of the Wharton Risk Management and Decision Processes Center at the University of Pennsylvania. So we tend not to insure against such events. For example, the federal government set up a flood insurance program in 1968, but despite the availability of low-cost coverage, only one in five homeowners in flood-zone areas currently participates.

Nor do most homeowners or builders invest in measures that would mitigate damage, such as furnace tie-down straps, storm shutters, low-profile roofs, and stronger foundations. A survey of residents affected by the 1989 and 1994 earthquakes in California revealed that only one in ten homeowners invested in any type of structural measure that would lower their expected losses.

Businesses generally take a more objective view, but they too need to be prodded. Risk management for high-variance losses "is still not on the radar at many organizations," says Linda Ruthardt, who was a corporate risk manager before becoming Massachusetts Insurance Commissioner. "When it comes time to hand out the bonuses," Ruthardt says, "how do you value (before a disaster) the savings from mitigation?"

Problems arise not just from those who reject insurance coverage, but also from those who embrace it. "Adverse selection" is one problem: Those more exposed to risk (such as the owners of homes on Cape Cod) are more likely to buy insurance, at a given price, than those who are not. A second problem is the "moral hazard" that policyholders, once insured, will tend to behave more carelessly, causing insurers to suffer higher losses.

Pricing aligned to the individual customer's risk can address the problem of adverse selection. But pricing today is inadequate. Because of the favorable loss experience of the '70s and '80s, premium rates are well below what are now considered to be actuarially sound levels. State regulators, who are either elected or appointed at the pleasure of the governor or a designee, have been loath to allow prices to rise much. And some firms, reluctant to cede growing markets to competitors, have decided to hold fast at the regulated, low prices.

Regulators also have tended to discourage pricing that differentiates by risk. Such proposals have been criticized as "aqualining," as if flood insurance price differentials were akin to racial discrimination. "We've attempted to put the cost where it belongs, to make a larger differential between coastal and inland homes," says Kenneth Amylon, senior vice-president of Amica Mutual Insurance Co. of Lincoln, Rhode Island. "But regulators don't like doing something like this all at once."

Faced with the true cost of building on a sandy shore or a flood plain, high-risk customers would have powerful incentives to change their behavior. And to that end, insurers are starting to ask for rate increases in various states, using a new type of argument to make their case. Traditional actuarial methods take a trend of past catastrophe losses and project it into the future. The new perceptions of risk and uncertainty find trend projections inadequate for rate-setting, given that the four or five decades of loss experience captured by insurers cannot hope to represent probable losses ahead. Rather than using an insurer's historical loss experience, the new approach takes data on the insurer's properties and uses computer models to simulate a large number of disaster scenarios. The results of these scenarios plot a range of probable losses.

At this early stage, however, the simulation approach poses challenges both for insurers and regulators. The quality of property data -- where a house is located, the characteristics of the structure, soil, trees, and so on -- varies by company. Some insurers can't even provide locations for the business properties they insure, because they often aggregate corporate accounts at a headquarters address.

Responding to the higher level of risk and uncertainty, and with pricing too low, some insurers have been trying to reduce the number of policies they write in exposed areas. Last year, Nationwide Insurance Co., based in Columbus, Ohio, curtailed sales of new policies in coastal areas in seventeen states from Maine to Texas. Two other firms, Commerce and Travelers, have pulled back from coastal properties in Massachusetts and Rhode Island, respectively.

Low pricing also helps explain why insurance firms don't offer premium discounts in return for mitigation measures. Discounts are common in other lines, such as fire and auto coverage. But if premiums are already below actuarially sound levels, insurers say that offering discounts could encourage more property owners, especially in hazard-prone areas, to purchase coverage from the company. And many insurers want precisely the opposite.

Curtailing coverage may help an insurance company. But it does not solve the broader social problem. Most regulators in high-risk states limit the portion of policies a firm can cancel or decline to renew. Many firms, moreover, have not tried to reduce excessive risk concentrations.

So development proceeds in high-risk areas, under construction standards that could generate losses far greater than the current premium structure can bear. Cross-subsidies abound. Inland residents subsidize those (generally wealthier) residents on the coast; highland homeowners subsidize those in flood plains; and policyholders subsidize people who don't buy flood insurance but receive government relief anyway. While cross-subsidies may be inevitable whenever risks are pooled, they are unusually large in catastrophe coverage, and encourage behavior that imposes major social costs.

Losses from a given catastrophe tend to be concentrated in a compact region. So a significant problem for property/casualty insurers, and for taxpayers in general, arises from the regional concentration of many insurers, especially the smaller ones. New England, in particular, has many companies with risk portfolios that are highly concentrated in a few metropolitan areas. The Insurance Services Office, an industry consultant, estimates that a catastrophe causing $60 billion in insured losses, such as a hurricane landing in a dense East Coast area, could bankrupt one-third of all insurers nationwide, and incur huge government costs as well. Simulating the impact of severe hurricanes and their effect on eighty insurer groups, the ISO found that the groups most vulnerable to insolvency have concentrated property in Massachusetts, New Jersey, New York, and Rhode Island. These firms are exposed both to their own policies and to state pools that require solvent insurers to cover the losses of those that are bankrupted.

The insurance industry deals with excessive risk concentration by pooling or handing off exposures to large catastrophes. This occurs through "treaties," or contracts, between primary insurers and reinsurers. Most reinsurance treaties are written in one of two types: The reinsurer assumes a share of the risk in exchange for a share of the premiums, or it assumes a layer of risk, say for losses between $1 million and $5 million. In these ways, reinsurance spreads risk from the primary insurer to other parties.

But as economist Kenneth Froot of the Harvard Business School points out, the pass-through of risks has been only partial. Given the current prices of reinsurance contracts, Froot finds, insurers have bought very little reinsurance for events causing industry-wide losses greater than $5 billion -- the very events whose risks most need to be shared. For a $50 billion event, the great portion of the last $45 billion in losses is not reinsured.

Many insurers have been seeking more reinsurance coverage, but find it tough to get at prices they feel they can afford. Reinsurance premiums generally run considerably higher than estimates of actuarially expected losses (loss times the probability of the reinsurance being triggered). But after major catastrophes, such as Andrew and Northridge, the perceived level of risk and uncertainty rises, and so do premiums. In a recent purchase of reinsurance by the California Earthquake Authority from a subsidiary of Berkshire Hathaway, Froot calculates, the annual premium was set at more than six times the expected loss. The pricing of this deal is typical in the marketplace today.

Losses from major catastrophes (and from environmental liability claims, including asbestos-related diseases) forced some reinsurance firms into bankruptcy and eroded the capital and surplus of many others. New capital does not flow quickly into the reinsurance industry in part because of the barriers to entering the business, explains Robert Klein, director of the Risk and Insurance Research Center at Georgia State University. "You need knowledgeable people, you have to raise capital, you have to establish the company and get regulatory approval -- there are all these transactions that take time and expertise," Klein says.

Because of these difficulties, many observers and industry managers are now looking at alternatives fashioned in the capital markets. One promising alternative is the "act of God" bond, which provides a payout to an insurer in the event of a catastrophe. United Services Automobile Association, a large home insurer, recently sold $477 million of one-year bonds through a special-purpose trust, to hedge against a major hurricane striking the East Coast this hurricane season. If USAA experiences losses of at least $1.5 billion from a single hurricane, one class of investors forfeits their interest and entire principal; if the loss is between $1 billion and $1.5 billion, they forfeit a portion of the principal. For that gamble, investors get a return that's currently about 11.8 percent. A second class of investors, who risk only their interest on the bonds, receives roughly 8.6 percent. (Both rates fluctuate with London bank lending rates.)

Rating agencies were able to rate the bonds by estimating, using computer models, the probability of disaster, which is comparable to default on the bonds. They calculated the likelihood that USAA would be hit with a hurricane costing the company more than $1 billion as less than 1 percent (or once in one hundred years). They thus rate the bonds that put principal at risk as equivalent to highly rated junk bonds, and the principal-protected bonds as investment grade.

The security thus is a composite of a bond and a reinsurance contract. The additional interest payment for the bonds where the principal is at risk is the reinsurance premium for the additional amount at risk -- the principal.

These securities will appeal to large institutional investors such as pension funds and mutual funds, observes Christopher Lewis, a senior manager at Ernst & Young LLP. A main attraction, besides the high yield, is that catastrophe risk fluctuates with a different pattern than do returns in stock and bond markets. And "act of God" bonds unbundle the catastrophe risk from other types of risks, especially asbestos and environmental liability taken on by a reinsurance firm. So investors don't have to bear these other huge and even less predictable risks.

Of course, investors will have to be savvy enough not to overpay for "act of God" bonds. Investors have responded slowly, thus far, to these and other types of catastrophe securities. They are still uncertain about the nature of the risks, and they lack good standards with which to evaluate such securities. Better standards and improvements in the quality of the underlying information are being developed. If a market does mature, it could provide a significant public service, stabilizing the supply and lowering the cost of catastrophe risk insurance.

As the bill for damage from natural disasters grows, and conventional insurance struggles to cope with the new scale of risk, pressure intensifies for the federal government to step in as reinsurer of last resort. That's not surprising, since the government's role has been expanding for roughly three decades. David Moss, an economic historian at the Harvard Business School, characterizes the dramatic expansion of federal disaster relief after 1960 as part of a new, broader risk management role for government. With the rapid rise in income after World War II, providing security for the individual citizen (as
opposed to security for businesses or workers) became a primary function of the federal government. Social regulation to make the individual safer proliferated in the areas of highway travel, consumer products, and environmental pollution. Disaster relief likewise emerged as a federal function, and the government now commonly covers half of uninsured losses.

Government is uniquely positioned to offer insurance against major catastrophes. Through disaster aid, it implicitly offers insurance -- after the fact -- at a price essentially equal to the actual loss, far less than what reinsurers or the capital markets charge. And it can raise funds through its taxing authority.

National governments in several other countries make this role explicit, structured as a public reinsurance scheme. Some reformers believe a similar system would benefit the United States. In France, the government requires that every insurance policy include comprehensive disaster coverage along with a corresponding surcharge. Private insurers can pass on, at their discretion, a portion of their catastrophe risk (and the premiums) to a state-guaranteed insurer.

Moss argues that a modified version of the French system makes sense for the United States, because "it would transform what is now paralyzing uncertainty into manageable risk."

The big problem with government reinsurance is moral hazard. Governments are inexperienced at assessing and pricing catastrophe risks, and federal intervention could reduce the pressure on insurers to diversify their risks and price them more appropriately. So the most effective national reinsurance scheme would enhance private mechanisms for internalizing more risk. Some observers, for instance, propose that the U.S. government, like France, require the purchase of all-hazards insurance. Christopher Lewis also suggests that the government sell reinsurance only covering layers currently unavailable in the private market. To minimize moral hazard, the program would be based on industry losses, not losses for an individual insurer.

Most observers also would insist that insurers align premium rates more closely, if not completely, with the property's risk. Both homebuyer and builder would then have greater incentives to invest in mitigation.

State and local governments could also help limit moral hazard, and catastrophe damage, through better zoning and building regulations. A substantial amount of the insured losses from Andrew could have been prevented by tougher enforcement of building codes already on the books. Getting tough on enforcement is politically difficult, because local officials fear they'll lose development to their neighbors. But it is not impossible. Thus the Insurance Institute for Property Loss Reduction, an industry trade group, has been working with Rhode Island state and local agencies to incorporate mitigation measures into land-use policies.

The Institute is also spurring the private sector to offer customers sufficient incentives to reduce risks. A homeowner who invests in mitigation measures would get a "seal of approval" that could lead to a series of discounts, such as a lower insurance premium, better mortgage terms, perhaps even a discount on building supplies. A dozen national insurance firms to date have signed on to the concept.

These proposals work best in concert, for risk management policies and decisions are "nested," in Howard Kunreuther's phrase. Make cost-effective mitigation available, and it will be attractive to property owners. If these actions reduce expected losses, they would then increase the availability and affordability of insurance and reinsurance. Allow insurers to weight premiums by risk, and they will be willing to promote mitigation.

None of these proposals will go uncontested, given the many people who stand to lose their subsidies. And it would be unfair, and politically impossible, to double premiums overnight for high-risk properties such as those along the shore. But it is appropriate, over the long run, to shift people out of harm's way, or to compel them to pay the true cost of living there.

YANKEE TEMBLORS

Although most property owners in New England don't buy earthquake insurance, a significant risk exists. The strongest known earthquake in the region occurred in 1755, centered off Cape Ann, Massachusetts, and geologists estimate that it corresponded to 6.25 on the Richter Scale, a fairly destructive quake. Walls collapsed, objects were flung from shelves, and fallen chimneys rendered some streets in Boston impassable.

In this century, more moderate earthquakes occurred in 1904 near Eastport, Maine, and in 1940 near Ossi pee, New Hampshire, and several quakes have shaken southeastern Canada over the past two decades.

Geologists put the odds of another 6.25 quake occurring somewhere in New England, in any one year, at one in three hundred. The 1755 quake today would cause $5-$6 billion in property damage within Route 128 alone, according to an estimate prepared by the Massachusetts Civil Defense Agency. Older, unreinforced masonry buildings and areas built on fill, such as Boston's Back Bay, could suffer the worst damage.