A New Way to Bet on Disasters (Fortune, 1997) by David Stipp @FortuneMagazine November 4, 2012, 2:24 PM EDT E-mail Tweet Facebook Google Plus Linkedin Share icons Editor’s note: Every week, Fortune.com publishes a story from our magazine archives. This week, we turn to a 1997 article about changes in the U.S. insurance industry following Hurricane Andrew, one of the costliest disasters in American history. Earlier this week, Superstorm Sandy ripped through the nation’s East Coast, costing what some have estimated upward to $50 billion in property damage and lost business. If Andrew prompted insurers to change the way they view risks, what lessons might Sandy bring? Did the gods have it in for the insurance industry? Insurance executives trembled when Hurricane Andrew, the costliest disaster in American history, blasted Florida south of Miami on August 24, 1992. Resembling a 20-mile-wide tornado and unleashing the power of three Hiroshima-sized atomic bombs a minute, the storm took mercifully few lives–only about 40. But it showed no mercy for property. By the time Andrew had raked Louisiana and died, it had destroyed more than 60,000 homes, damaged thousands of businesses, and erased more than $16 billion from property insurers’ books. The nation’s largest home insurer, State Farm Fire & Casualty, was rendered insolvent by a $4 billion hit–its parent, the giant auto insurer, had to inject emergency capital to save it. Twelve insurers were completely wiped out, leaving $400 million in unpaid claims. The business got a little crazy after that. Insurers who had underwritten natural disaster policies realized they’d been rolling dice loaded with TNT, and many of the survivors began backing away from such risks. Others, frantic to limit their exposure to disasters, bid up premiums for backstop policies known as catastrophe reinsurance by as much as 100%. By the time hurricane season ended, demand for “cat” reinsurance had triggered one of the great gold rushes of the 1990s. Seemingly overnight, eight new cat reinsurers sprang up in, of all places, Bermuda. Mid Ocean Ltd. was first in this “class of ’93,” followed by Renaissance Reinsurance, CAT Ltd., and others. Their coffers were eagerly jammed with $4 billion of venture capital from Morgan Stanley, Warburg Pincus, and other investment houses. One of the Klondikers is James Stanard, founder and CEO of Renaissance Re. Three months after Andrew, Stanard, then an executive at Baltimore insurance firm USF&G, worked through his Thanksgiving holiday to draft a business plan. A few months later, he and a partner opened their cat shop in a 200-square-foot office in Hamilton, Bermuda, with two phones, two computers–and $141 million of seed capital. The phones immediately started ringing. “In the first two weeks, we got over 500 submissions” for policies, says Stanard, a reticent, no-nonsense actuary. “We had to triage them.” In 1994, its first full year, Renaissance racked up profits of $96 million and a 44% return on shareholders’ equity, about four times the industry norm. In 1995, the company went public, and a year later it catapulted onto the New York Stock Exchange. The business, which employs 31, recently had a market value of more than $1 billion. (That’s about $32 million per employee–if you like invidious comparisons, multiply that figure times your own company’s employment.) Half the class of ’93 is already on the Big Board; the upstarts have grabbed an estimated 25% of the cat market from entrenched giants like Lloyd’s of London, Germany’s Munich Reinsurance, and General Reinsurance in Stamford, Conn. All reinsurers sell policies to insurance companies as safeguards against major loss. A reinsurance group, for instance, might cover a home insurer’s losses from disasters above a trigger point of $100 million up to a limit of $400 million. Offloading such risks lets the “primary” insurer write many more policies than it otherwise could, since less of its capital must be tied up as reserves for rare, huge losses. Each year, reinsurers worldwide take in some $6 billion of premiums in return for covering about $50 billion of natural-disaster risks. While traditional reinsurers also cover other biggies–oil spills, shipwrecks, plane crashes, satellite explosions–the Bermuda newcomers are true moguls of doom, specializing in earthquakes and disastrous storms. The upstarts are hot because of their willingness to bet they have the answer to threats posed by the likes of Hurricane Andrew. They have adopted a remarkable tool known as “catastrophe modeling”–computer programs that estimate losses from future disasters. Cat models, as they’re called, predict how often hurricanes, as well as earthquakes, will strike different regions. Before Andrew, insurers sized up natural-disaster risks mostly by rules of thumb. Led by the new breed on Bermuda, they’re now coolly calibrating nature’s deadly whims like dam engineers gauging deep canyons. That’s changing the property insurance business almost as much as Hurricane Andrew did. Five years after Andrew, quantifying disaster risks with cat models has abetted securitizing them–so now Wall Street’s high rollers can shoot craps with the gods. Teamed with investment banks, insurance companies have lately been hawking “catastrophe bonds,” high-risk debt that helps insurers boost their reserves for potential disaster losses. The Chicago Board of Trade even offers trading in “catastrophe insurance options.” Such instruments let insurers hand off disaster risks to investors, who get handsome returns if insured losses over specified periods are lower than forecast. Just picture hurricanes as huge spinning pork bellies. On second thought, don’t bother: Despite getting lots of media play, securitized cat risks are too dicey for most investors. The much bigger action revolves around insurance. To many people, putting price tags on future catastrophes falls somewhere between damned lies and statistics. In Florida, home insurers recently stirred consumer ire by proposing hefty rate hikes based on the computer forecasts–the companies say they need more to cover looming hurricane losses. Vows Bill Nelson, the state’s insurance commissioner: “I will not let the consumers of Florida become hostage to a computer.” Some scientists also question cat models, arguing that the past disaster patterns on which they are based may be a poor guide to the future. They cite evidence the global climate is changing, potentially causing more hurricanes and other bad storms. Still, “when people have massive amounts of money at stake, they get extremely efficient at digesting information and compounding it into prices,” says Russ Ray, a University of Louisville finance professor who has studied cat options. A remarkable 1984 study, for instance, showed orange-juice futures–prices bid for OJ deliveries in future months–can predict slightly better than the National Weather Service how low Florida temperatures will go on risky winter days. Massive sums indeed are at stake: From Texas to Maine, just the first tier of coastal counties contains over $3.4 trillion of insured property. Here’s how companies like Renaissance Re and CAT Ltd. work: Before reinsuring, say, an Alabama home insurer’s hurricane risks, they model their exposure to losses from the deal–if it seems too risky or the premiums don’t exceed projected losses plus the cost of carrying capital to cover the risks, they walk away. The models typically contain two kinds of data: One represents all the hurricanes expected to hit the U.S. for thousands of years to come–these “storm sets” include data on wind speeds, travel paths, and, crucially, probabilities that various storms will actually occur based on hurricane patterns over the past century. The second data set represents what happens to different kinds of buildings when they’re walloped by winds at various speeds and directions. (Earthquake models, using seismic data, are similar. But major quakes are much rarer and less amenable to statistical analysis than hurricanes, so forecasting losses from them is closer to blackjack than chess.) The programs calculate how storms that might be expected to hit, say, Alabama would damage particular properties there. Moreover, the models also show a reinsurer whether adding the Alabama coverage to its portfolio would dangerously concentrate its exposure along the paths of certain simulated storms–such “correlated risks” are a setup for insolvency, since even a smallish hurricane plowing through could cause a heap of losses. Thus, the models are like x-rays that reveal perilous risks hidden behind the wall of numbers in property insurance portfolios. CAT and Renaissance use half a dozen models, including internally developed software and off-the-shelf systems from various companies, most prominently Applied Insurance Research in Boston and Risk Management Solutions in Menlo Park, Cal. That helps give a better perspective on the risks, for different models tote up potential losses differently. Reinsurers concede their models haven’t undergone the acid test of correctly predicting losses over many years. But even if the models are significantly off–say, by systematically underestimating damage to frame houses from winds over 150 mph–they can still show the relative sizes of risks, just as a photograph of a man next to a giraffe can show their size ratio. That lets the reinsurers skim the cream of disaster bets. Says Charles Kline, CAT Ltd.’s president: “You get taken advantage of very quickly in this business if you don’t know what you’re doing.” As happened to many reinsurers, notably certain Lloyd’s of London affiliates, a few years ago. Nature set them up with an unprecedented string of catastrophes: Hurricane Hugo in 1989, windstorms that devastated parts of Europe in 1990, Typhoon Mireille in Japan in 1991, to name a few. As insurers turned out their pockets, it became clear that some had been played for patsies. Kline explains that many reinsurers had, in turn, purchased reinsurance, passing along risks to others with so-called retrocessional coverage: “You had situations where company A would recover from B, which would recover from C, which had put the risk back on A.” When big events hit, company A could wind up covering risks it thought it had passed to others, and face huge losses. “You won if you had the most protection or had gone outside the system for coverage,” Kline says. Worse, most reinsurers had gone “capless” by failing to set upper limits on coverage, a common form of recklessness in reinsurance that Hurricane Andrew punished. The practice stemmed from estimates by industry experts in the mid-1980s that insured losses from a worst-case disaster would be less than $10 billion–no sweat, reinsurers figured, they could absorb a few billion, and surely it wouldn’t happen for decades. The estimates also induced them to cover cat risks for what now seems way too little. “Then came Andrew, which forever changed people’s understanding of how big a loss could be,” says Kline. Soon after, the class of ’93 headed for Bermuda. At first glance, the island’s tidy capital seems an unlikely place to ponder terrible destruction. Awash in a playful surf of tourist dollars, Hamilton’s quaint streets are lined with pastel pink and yellow shops proffering Cuban cigars and lime cologne; on a recent day, three cruise ships are parallel-parked along the waterfront like stretch limos. Yet behind the minty pastels of Bermuda’s buildings are deadly serious cinder blocks–the place is a veritable hive of hurricane bunkers. Today, hurricanes have helped populate the island with insurance executives in jackets, ties, Bermuda shorts, and knee socks–the British territory’s curious compromise between propriety and perspiration. They ride mopeds to work, sober and erect, down winding floral byways, and write a palmy $23 billion of annual premiums. It’s just coincidence that “the Rock,” as some call their confining venue, is in hurricane alley. The insurers are here, of course, to avoid taxes. Cat reinsurers, especially, feel cheated by the usual corporate rates: Most years they generate lots of profit, since big natural disasters are rare, which is how they build capital to cover them. Thus, paying high-bracket taxes in good years seems like robbing their reserves–tax-loss benefits from disaster losses can’t be spread over enough years to compensate. Primary insurers, by contrast, tend to be big, purring profit engines that pay out a predictable share of premiums each year for many small losses. At least they were until Hurricane Andrew turned the insurance world upside down: Since reinsurers now cap their coverage, the primaries are the ones likely to be stuck with huge losses above the caps when the big one hits. And by managing risks with cat models, reinsurers hope to achieve the kind of predictable losses that were once the norm for primary insurers. Says Paul Hasse, CEO of CAT Ltd.: “We spread a lot of relatively small bets around the world, banking on the portfolio effect to minimize our losses from any one event.” The Bermuda reinsurers’ worst nightmare isn’t a major hurricane mowing down Miami; it’s getting hit by several mid-sized disasters at about the same time. Losses may be capped for each event, but a bunch at once could wipe out the reinsurer. This nightmare is deemed quite unlikely, though some scientists, including ones with major reinsurers such as Munich Re, say the chances it will occur are rising because of human-induced climate change. Other scientists, most prominently a hurricane expert named Bill Gray, say we’re probably due for a burst of hurricanes much sooner than any related to the greenhouse effect–because of forces totally beyond our control. Every disaster story must have a Jeremiah; Bill Gray is this one’s. He’s the leading forecaster of hurricane trends, a meteorologist at Colorado State University in Fort Collins, of all places. For ten years he’s been predicting a dramatic upswing in the frequency of Atlantic hurricanes. Sadly, he’s looking better every year: During the 1995 and 1996 hurricane seasons, roughly from June to October, 20 hurricanes formed, setting a two-year record. Four hit the U.S., causing more than $4 billion of damage. Says Gray: “It looks like we’re going to see hurricane damage like we’ve never previously seen. Florida is a sitting duck. There’s just no way out.” Gray was mostly ignored in the 1980s when he began saying things like this. He’s a stubborn empiricist in a field increasingly dominated by theorists who simulate the global climate on supercomputers. When he first sounded the alarm, hurricanes were rare: A striking lull in their annual activity began about 1970 and continued into the 1990s. Ironically, the storms dwindled just as millions of people heard the siren call of the coast–Florida’s population has more than doubled since 1970. Between 1988 and 1993 alone, insured property values on the Sunshine State’s coast shot up 54%, according to the Insurance Institute for Property Loss Reduction. Here’s what got Gray going: If you look long enough at a chart of annual hurricane frequencies since 1890, you’ll see a kind of sine wave cycling over 40 to 60 years. Look closer, and you’ll see that a crest seems due this decade. Further, Gray had noticed a number of striking correlations. One is that hurricane activity rises during especially rainy years in the sub-Saharan area of West Africa. Another coincidence: Slow hurricane seasons tend to coincide with El Nino events, periodic flows of warm water into the eastern equatorial Pacific. Such correlations may be manifestations of a vast atmospheric engine that churns out hurricanes. If so, the storms’ frequencies probably aren’t random. Oddly, Hurricane Andrew didn’t lend weight to Gray’s jeremiads: For all its havoc, it hit in a slow season. The following season, 1993, also was light, as was the next. Had Gray been crying wolf? Then all hell broke loose: 19 tropical storms in 1995, the most since 1933. Two turned into hurricanes that hit the U.S. To Gray, everything was falling into place. The surge was preceded by a shift in the “great conveyor belt,” an oceanwide gyre that moves Atlantic waters in a roughly circular way. Its flow seemingly influences everything from sea temperatures off Africa to El Ninos in the Pacific. And, if Gray and his collaborators are right, it is the hurricane engine’s throttle. About three years ago, he says, it shifted to high. Well, maybe–much of this is still conjecture. But many of the details fit with what’s known about hurricanes. They often begin forming off West Africa when waves of hot air from the Sahara collide with cooler, moist air along the coast. The roiling that results can turn into a self-feeding process: Warm seawater evaporates, rises, and cools, releasing heat that fuels violent updrafts and further evaporation. This chimneylike structure spins up into a hurricane when conditions are right. One aid is warm ocean temperatures in the path of the roiling air–an ingredient the Atlantic conveyor supplies. Hurricanes also need relatively light crosswinds at high altitudes–otherwise their tall storm clouds get decapitated. That explains the link with El Ninos, which kick up hurricane-suppressing winds high over the Atlantic. Insurers find all this riveting. Just think, maybe such correlations can be fed into cat models to better predict how many hurricanes are headed for their balance sheets in the coming season–preferably before annual policies are renewed. Primary insurers may be blocked from “stormlining,” or refusing to cover homes near the coast, but if they know a bad season is coming, they can buy more reinsurance. Similarly, a tipped-off reinsurer can reduce coverage in high-risk areas. It might also pass off risks by buying retrocessional coverage from rival reinsurers that lack its knowledge edge. Several reinsurance executives, requesting anonymity, say their companies are quietly funding studies aimed at getting this sort of advantage. Lips are sealed about the details. But the goals aren’t hard to guess. For years, Gray has used El Nino and other leading indicators to issue fairly accurate December forecasts of the upcoming season’s tropical storm count. What insurers really need now is some idea of how many hurricanes will hit land where they do business. On Aug. 1 last year, Florida State University meteorologist Jim Elsner stirred excitement by making the first crude forecast along those lines: By analyzing correlations between regional landfalls of past hurricanes and indicators such as El Nino, he predicted the U.S. East Coast was about to get hit. On Sept. 5, Hurricane Fran slammed into Cape Fear, N.C. Soon afterward, Elsner says, he got a six-figure offer to join an insurer as a captive augur. No dice: “I didn’t want to lose my academic freedom,” he says. In any case, researchers agree that breakthroughs in hurricane forecasting will require more data on the storms’ past patterns. “You can’t tell a sine wave is a sine wave until you’ve seen more than one crest and one trough,” says Richard Gordon, a modeling expert at Risk Management. Which raises a critical point: Cat models are based on the assumption that we’ve seen it all, with respect to hurricanes, over the past century–that’s all the detailed storm info there is from sources like ships’ logs and newspapers. But what if hurricanes have longer cycles, and a huge crest is bearing down? That, in fact, is one of the main questions under study at the Risk Prediction Initiative, a Bermuda think tank that Renaissance Re, CAT, and other reinsurers launched three years ago. “We’re pretty sure we can get decent data on hurricane landfalls going back at least a couple of thousand years,” says Anthony Knap, director of the Bermuda Biological Station for Research, an institute overseeing the work. Some intriguing–and worrisome–hints about long-term landfall trends already have been unearthed by Kam-biu Liu, a Louisiana State University researcher funded by the initiative. Liu’s studies are based on the fact that intense hurricanes are accompanied by tsunami-like “storm surges” that carry sand from beaches to coastal lakes. In Alabama and Florida, he has sunk clear plastic tubes into the beds of such lakes to extract sediment cores, which show layers of gunk deposited over the centuries interspersed with telltale layers of beach sand. Liu can tell when a hurricane put a sandy layer down via carbon dating of the surrounding organic stuff–plants contain traces of radioactive carbon, which acts as a clock by turning into “normal” carbon at a known rate after they die. His most recent findings suggest that about 1,000 years ago, the number of hurricanes hitting the Florida panhandle dropped dramatically. Specifically, between 2,000 and 1,000 years ago, five times as many intense hurricanes apparently landed near Florida’s Western Lake than during the past millennium. “It was really startling,” he says, “but it’s too early to talk about thousand-year cycles.” Still, if more data confirm the pattern, it would seem the hurricane engine has a throttle setting above “high” that no one imagined: “apocalypse now.” Strangely, though, Bermuda’s reinsurers have suffered lately from a more mundane problem: excess competition and price warfare. “They really need a hurricane,” says Charles Vere, a second vice president at General Reinsurance. Perverse as that sounds, it’s true. Globally, insured losses from hurricanes have been low since Andrew, despite multiple U.S. hits. That’s helped make the cat business look hugely profitable, luring back reinsurers burned in 1992. Lloyd’s, in particular, is aggressively trying to regain market share lost while regrouping after Hurricane Andrew. Surging competition has led to sharply lower pricing of cat risks–premium rates have plunged 10% to 20% this year, according to IBNR Insurance Weekly, a newsletter by Dowling & Partners Securities in Hartford. It will take a hard knock like Hurricane Andrew to boost premiums and ease the profit squeeze. The Rock’s quants are irritated by what’s happening: “We’re seeing competitors selling a lot of cat reinsurance below their expected loss rate” from the risks, says William Riker, senior vice president at Renaissance Re. Of course, Bermuda’s reinsurers don’t really want disaster visited on their fellow humans–they want it to hit these actuarial clowns who are ruining everything. For now, they have to wonder: What if the gods get truly perverse and suspend major hurricane losses for years? One answer is to diversify. Mid Ocean, the class of ’93’s first member, already has pared its cat business dramatically in favor of more traditional marine and aviation reinsurance. Other Bermuda reinsurers are still betting that they can ace rivals with superior cat modeling. To retain market share, for instance, they can knowingly underprice risk coverage along with cutthroat competitors, then hand the lousy bets to them via retrocessional coverage. But such “risk arbitrage” is perilously complex and isn’t likely to pay high returns. So, like vying casinos, Bermuda’s reinsurers are inventing new betting modes, most remarkably, “live cat”: short-term coverage sold to primary insurers to cover their losses from an approaching storm. Renaissance and CAT recently began offering such policies between the first sighting of a proto-hurricane and its landfall–as the storm approaches land, the bids fly ever faster, and prices change to reflect what’s known about it. If the reinsurers can predict hurricane paths better than their customers–a big if–they can win by selling live cat to property insurers unlikely to get hit. Reinsurers won’t say how they price live-cat risks–naturally, it’s a trade secret. But recently a visitor in Kline’s office stumbled on a possible clue hidden behind his office door: a dart board.