At 9:03 a.m. on September 11, 2001, Britt Newhouse stood in the lobby on the 52nd floor of the south tower of the World Trade Center. After United Airlines Flight 175 banked above the harbor behind him, it was pointed at the 50th floor. If not trimmed correctly, an airplane will rise as it accelerates, and the man who had killed and replaced the airplane’s pilot added power until he hit the south tower 24 floors above Newhouse. He doesn’t remember the sound of the impact.
At the time, Newhouse ran Guy Carpenter’s Americas operation. Guy Carpenter brokers reinsurance treaties, which protect insurers when a catastrophe—a hurricane or an earthquake—causes losses on a large number of policies. Reinsurance, in essence, is insurance for insurers.
Consumers don’t tend to know what reinsurance is because it never touches them directly. Reinsurers, massively capitalized and often named after the places where they were founded—Cologne Re, Hannover Re, Munich Re, Swiss Re—make their living thinking about the things that almost never happen and are devastating when they do. But even reinsurers can be surprised. And the insurers who make up their market put them on the hook for everything, for all the risks that stretch the limits of imagination. This is what the industry casually refers to as the “God clause”: Reinsurers are ultimately responsible for every new thing that God can come up with. As losses grew this decade, year by year, reinsurers have been working to figure out what they can do to make the God clause smaller, to reduce their exposure. They have billions of dollars at stake. They are very good at thinking about the world to come.
Lloyd’s, the London-based company that invented the modern profession of insurance, publishes a yearly list of what it calls “Realistic Disaster Scenarios.” The list imagines such events as two consecutive Atlantic seaboard windstorms or an earthquake at the New Madrid fault in the Mississippi Valley, either of which could strain or break an insurer’s balance sheet. By 2001, Lloyd’s had already envisioned two airliners colliding over a city, launching claims on hull insurance for the planes, property insurance and workman’s comp on the ground, and life insurance everywhere. But even Lloyd’s lacked the imagination to anticipate September 11. “People find it hard to believe in a risk unless they can see it in their mind,” says Trevor Maynard, head of exposure management for Lloyd’s. “When you try and describe a risk like this—some terrorists are going to teach themselves how to fly a plane, they will fly into property, the buildings will be weakened—by the time you get to your third point, peoples’ eyes are glazing over.”
Floors 49 through 54 of the south tower housed Guy Carpenter’s home office and its 750 employees. To any of Newhouse’s clients—insurers—professional employees in an office tower would have presented an attractive bet in terms of risk and reward. There are few slips and falls in white-collar work and not much dangerous equipment lying around. Newhouse indulges in mild profanity and, thinking like a broker, often lapses without warning into the voice of his customers as they think about risk. “If I insure 40 of these firms in two towers across 200 floors, probably the worst is that eight of them, maybe six of them, could be involved in the same loss,” he says. “Because the fire department can get there … usually a fire never spreads, historically, beyond six floors.”
The insurance industry describes these kinds of risks as “high frequency”: The more often a kind of accident occurs, the easier it is to guess its chance of happening and the easier it is to price insurance coverage. For slips, lawsuits, and fires, historical trends predict future probability. “What they hadn’t imagined,” says Newhouse, picking his words carefully, “was an intentional act of human causative agency.”
Guy Carpenter lost 29 employees on September 11. When the plane hit his building, most of Newhouse’s employees were well below him or out of the tower already. Guy Carpenter’s chief executive officer had traveled to an annual conference in Monte Carlo, leaving Newhouse as the most senior manager. After American Airlines Flight 11 hit the north tower, he told everyone to leave. The company had been in the World Trade Center in 1993 when a bomb exploded below the north tower. His employees did not need to be reminded that the world was a dangerous place, even if the building’s collapse still lay outside of the realm of imagination. By the time Newhouse reached the plaza level, a cordon of police officers waited to direct him through the mall below. When he got back to street level both towers were still standing. The hole in the north tower glowed, he says, like a dragon in a cave.
He saw the chaos on the ground beneath the towers and looked for a Hudson River ferry. He had no trouble getting on board; there was no rush. “There were people hanging around,” he says, “working their cell phones and watching.” They were still there in the streets, watching, when his ferry backed out and the floors of the south tower began to collapse and accelerate down. Until that second, they hadn’t believed that anything more could happen to them. No experience of fear drove them to get far away, as fast as they could, until it was too late.
I was standing there, too. I survived by dumb luck.
In the fall of 1998 I started a job in New York with one of the world’s largest reinsurers. Soon after, at a meeting in a conference room above Park Avenue, someone complained that the market for reinsurance had gone soft. “What we need,” he said, “is a good catastrophe.” It was a joke, but true, too. I offered that a cyclone had hit Bangladesh and it had been an active year for cyclonic storms. No, he explained, little of that value was insured. “Honestly,” said someone else, “I’ve never understood why those people don’t just leave. It’s a dangerous place.”
By definition, reinsurers work at the edge of suffering, and so have developed euphemisms to help them stand at a distance. A catastrophe is called a “loss event.” A catastrophe big enough to affect several reinsurers is called an “industry loss event.” Reinsurers both need and fear these. The first dedicated reinsurer, Cologne Re, formed after a fire leveled about a quarter of the city of Hamburg, Germany, in 1842. A loss event reminds insurers of the value of the product; reinsurers can raise rates, and the higher rates attract new capital. According to Dowling & Partners, an investment adviser for the insurance industry, after Hurricane Andrew ran across southern Florida in 1992, eight new reinsurers entered the market, together valued at $2.9 billion. Although hurricanes seem frequent, the insurance industry defines them as “low frequency” events; there are perhaps 10 a year and not tens of thousands. Unlike car accidents, you cannot look at the last decade of hurricane results and predict this year’s losses with any accuracy.
Andrew’s $25 billion in losses, along with the then-new availability of cheaper computing power, helped push insurers and reinsurers toward computer-driven modeling, which leaned on meteorology and seismology to more precisely define potential losses by estimating storm paths and categories and fault lines and tremor strengths. The models are sophisticated, but reality is wily.
After September 11, another eight new reinsurers followed the higher rates and entered the market, together valued at $8.6 billion. Hurricanes Katrina, Rita, and Wilma lured five more new reinsurers with $5 billion in capital. Between catastrophes, the new capacity drives premiums back down and reinsurers are forced to undervalue risks to stay in the market. Vincent J. Dowling of Dowling & Partners refers to this as the “cheating phase” of the cycle. That is, even though catastrophes present an existential threat to insurers, and the sober assessment of risk is a firm-defining competency, insurers, like people, can get complacent.
“The psychology piece dominates, even in boardrooms,” says David Bresch. “People measure against the perceived reality around them and not against possible futures.”
Bresch is in charge of sustainability and risk management for Swiss Re, founded in 1863 after a city fire in Glarus, Switzerland, and now the world’s second-largest reinsurer. “If the gap between modeled reality and perceived reality is too big,” says Bresch, “that tells you something about the market share you’d like to achieve.” In other words, what the models show as a best guess of a risk may not be what the insurers perceive the risk to be, and it can take more than just data alone to scare insurers into paying for risks at a rate that keeps reinsurers solvent over the long term. That takes an industry loss event. It takes a catastrophe.
Heike Trilovszky runs corporate underwriting for Munich Re, the world’s largest reinsurer. The afternoon of September 11, the press and the company’s shareholders needed to know what Munich Re’s losses would be. “To get answers we had to ask the brokers,” she says. “Aon Benfield, Guy Carpenter, they had offices in those buildings. It was odd. Is my counterpart still alive?” Before the end of the week, Munich had an initial estimate. In October, Trilovszky was at an off-site meeting when a board member arrived late. He had seen the revised loss estimate and he was pale. “We will survive,” he said, “but it’s serious.” Munich Re’s losses after September 11 ultimately came to $2.2 billion.
Reinsurance treaties are full of exclusions, but before September 11 terrorism was not considered significant enough to be one of them. Most reinsurance treaties renew on Jan. 1 and the meeting oriented itself around a new question: What do we do on 1/1/2002? “There were 15 managers, all with a background in property,” says Trilovszky. “We had flip chart paper and we put it on the floor. We were kneeling on the floor, working out concepts, brainstorming ideas … that was the Munich Re strategy for terrorism risk in the first days after 9/11.” In 2002, together with the rest of the industry, the company wrote terrorism risk out of any treaty with an insured value of greater than $50 million. Terrorism is what Trilovszky calls “nonfortuitous.” It stems from a few angry, motivated people, and nothing says there can’t be 10 World Trade Center events in a single year. “There cannot be a mathematical model,” she says, “for people like bin Laden.”
Reinsurers prefer to underwrite risks they can name; such a treaty is called a “named perils” cover. Far more often, however, the market forces them to sign an all-perils cover. Although reinsurers can exclude risks they already know about from an all-perils treaty—an act of terrorism, for example—they cannot exclude what Donald Rumsfeld might call an unknown unknown. That’s the God clause.
Models exist for some low-frequency risks such as hurricanes, earthquakes, and oil spills, but they don’t exist for every one. Trilovszky doesn’t have a model for airplane crashes. “Would it hurt us if crashes became more frequent? It would,” she says. “As it is now, I’m not sure we ever had a property loss from an aircraft crash. It’s theoretical. It lives within the error of the models we have.” The second something unexpected happens, though, it’s no longer theoretical or unimaginable.
“The history of the industry,” says Newhouse, the broker, “is we cover everything in a catastrophe until after the catastrophe. Then we rewrite it.” Until the Sixties there was no hours clause—a time limit—on damage that can be claimed to be from a hurricane. Now the standard is 96 hours. Newhouse has been a reinsurance broker for 35 years. Every new industry loss event—every new catastrophe—is the biggest in his career, he likes to say, and he knows by now that every catastrophe teaches the industry something. “If you want to be able to price [an event], if you want to be able to build an industry and a capital base that can handle it … then you’d better figure out a way to define it. Then capital can be accumulated to deal with it.”
I ask him why the reinsurance industry failed after the World Trade Center bombing of 1993 to either raise prices for terrorism coverage or to exclude acts of terrorism from coverage altogether. He looks at me as if he would like for me to figure it out for myself so he doesn’t have to say it. “Because it wasn’t big enough,” he says, pausing before adding, “economically.” The cost of an event, he explains, dictates how much the industry spends on understanding and preventing it. “I’m not saying people didn’t react to it,” he adds, “but you know, if your company’s life is threatened by the size of the loss, you’re going to react much differently.”
Newhouse keeps in his office a collection of plaster monkeys, the kind that come in sets of three with hands over ears, mouths, and eyes. They remind him when approaching a reinsurer about a client to hear, speak, and say no evil. The monkeys are all gifts, given to him to rebuild a much larger collection that was lost when his last office disappeared.
Swiss Re’s global headquarters face Lake Zurich, overlooking a small yacht harbor. Bresch and a colleague, Andreas Schraft, sometimes walk the 20 minutes to the train station together after work, past more yachts, an arboretum, and a series of bridges. In September 2005, probably on one of these walks, the two began to discuss what they now call “Faktor K,” for “Kultur”: the culture factor. Losses from Hurricanes Katrina, Rita, and Wilma had been much higher than expected in ways the existing windstorm models hadn’t predicted, and it wasn’t because they were far off on wind velocities.
The problem had to do more with how people on the Gulf Coast were assessing windstorm risk as a group. Mangrove swamps on the Louisiana coast had been cut down and used as fertilizer, stripping away a barrier that could have sapped the storm of some of its energy. Levees were underbuilt, not overbuilt. Reinsurers and modeling firms had focused on technology and the natural sciences; they were missing lessons from economists and social scientists. “We can’t just add another bell and whistle to the model,” says Bresch, “It’s about how societies tolerate risk.”
“We approach a lot of things as much as we can from the point of statistics and hard data,” says David Smith, head of model development for Eqecat, a natural hazards modeling firm. “It’s not the perfect expression.” The discrepancy between the loss his firm modeled for Katrina and the ultimate claims-based loss number for his clients was the largest Smith had seen. Like others in the industry, Eqecat had failed to anticipate the extent of levee failure. Construction quality in the Gulf states before Katrina was poorer than anticipated, and Eqecat was surprised by a surge in demand after the storm that inflated prices for labor and materials to rebuild. Smith recognizes that these are questions for sociologists and economists as well as engineers, and he consults with the softer sciences to get his models right. But his own market has its demands, too. “The more we can base the model on empirical data,” he says, “the more defendable it is.”
After their walk around the lake in 2005, Swiss Re’s Bresch and Schraft began meeting with social scientists and laying out two goals. First, they wanted to better understand the culture factor and, ultimately, the risks they were underwriting. Second, they wanted to use that understanding to help the insured prevent losses before they had to be paid for.
The business of insurers and reinsurers rests on balancing a risk between two extremes. If the risk isn’t probable enough, or the potential loss isn’t expensive enough, there’s no reason for anyone to buy insurance for it. If it’s too probable and the loss too expensive, the premium will be unaffordable. This is bad for both the insured and the insurer. So the insurance industry has an interest in what it calls “loss mitigation.” It encourages potential customers to keep their property from being destroyed in the first place. If Swiss Re is trying to affect the behavior of the property owners it underwrites, it’s sending a signal: Some behavior is so risky that it’s hard to price. Keep it up, and you’ll have no insurance and we’ll have no business. That’s bad for everyone.
To that end, Swiss Re has started speaking about climate risk, not climate change. That the climate is changing has been established in the eyes of the industry. “For a long time,” says Bresch, “people thought we only needed to do detailed modeling to truly understand in a specific region how the climate will change. … You can do that forever.” In many places, he says, climate change is only part of the story. The other part is economic development. In other words, we’re building in the wrong places in the wrong way, so wrong that what we build often isn’t even insurable. In an interview published by Swiss Re, Wolf Dombrowsky, of the Disaster Research Center at Kiel University in Germany, points out that it’s wrong to say that a natural disaster destroyed something; the destruction was not nature’s fault but our own.
In 1888 the city of Sundsvall in Sweden, built of wood, burned to the ground. A group of reinsurers, Swiss Re among them, let Sweden’s insurers know there was going to be a limit in the future on losses from wooden houses, and it was going to be low. Sweden began building with stone. Reinsurance is a product, but also a carrot in the negotiation between culture and reality; it lets societies know what habits are unsustainable.
More recently, the company has been working with McKinsey & Co., the European Commission, and several environmental groups to develop a methodology it calls the “economics of climate adaptation,” a way to encourage city planners to build in a way that will be insurable in the future. A study of the U.K. port of Hull looks at potential losses by 2030 under several different climate scenarios. Even under the most extreme, losses were expected to grow by $17 million due to climate change and by $23 million due to economic growth. How Hull builds in the next two decades matters more to it than the levels of carbon dioxide in the air. A similar study for Entergy, a New Orleans-based utility, concluded that adaptations on the Gulf Coast—such as tightening building codes, restoring wetlands and barrier islands, building levees around chemical plants, and requiring that new homes in high-risk areas be elevated—could almost completely offset the predicted cost of 100-year storms happening every 40 years.
As with Sweden’s stone houses, all of these adaptations cost more money in the short run, but reinsurers must take the long view, and they can drag development along with them. The public, whom the reinsurers refer to as “the original insured,” should be concerned by these hints. Even when they are forced to sign all-perils covers, reinsurers are writing more known risks out of their treaties. Swiss Re publishes an annual report on catastrophe losses; since the 1970s losses have been increasing exponentially. It’s this graph that gives reinsurers pause. The God clause includes less each year because every loss event—every catastrophe—reminds them of the hubris of thinking God doesn’t have any surprises left.
Munich Re’s Trilovszky offers a similar point from a different perspective: The companies that buy insurance are complaining, she says, that insurers and reinsurers have capital but aren’t willing enough to take on new business risks, such as the risk of a compromised brand or stolen intellectual property.
Like many reinsurers, Munich Re underwrites a line called “contingent business interruption.” Essentially, if for some reason a supplier fails to produce a crucial part for a factory, business interruption insurance covers the factory’s owner. The earthquake off the coast of Japan this March happened on a Friday; on Saturday, Trilovszky was doing research on an initial loss assessment when she discovered that a technology company in Louisiana had already stopped operations. It had run out of chips. “That,” she says, “was my eye-opener.”
Reinsurers are already exposed to a loss after a natural catastrophe. But given the complexity of global supply chains, it’s hard to tell exactly how much more loss will come from business interruptions elsewhere in the world. Two seemingly uncorrelated risks—an earthquake in Japan and a tech firm in Louisiana—are now related. Losses gather mass and complexity on their way toward reinsurers, which serve as a garbage bin for systemic risk. Reinsurers haven’t yet figured out how to respond to the problem of contingent business interruption after a catastrophe. “The easy thing to do is, we don’t cover this,” Trilovszky says. “The question is to adjust the product so it’s still useful for the insured but possible for the insurer.”
You could think of politicians as underwriters, too. Like Bresch and Trilovszky, they must compare the probabilities of low-frequency events, weighing the risk of a downturn against a sovereign default, or of Chinese imperial ambition against more small wars in the Middle East, or of climate change against the cost of regulation.
Reinsurers, however, have no incentive to mislead. Their choices on risk, with billions of dollars at stake, are necessarily aligned with the pursuit of truth. If a reinsurer is more scared of a risk than it should be, its shareholders will punish it. If it is less scared than it should be, the world, eventually, will break it. There are rewards for politicians, corporations, think tanks, and activists who dissemble about risk. There are none for reinsurers. If they’re taking on less of it than their insurers would like them to, then the world is more dangerous than we’re willing to admit. What a reinsurer will underwrite, then, offers a marker, one edge of what Bresch calls the gap between modeled and perceived reality. For instance, no event of the last 10 years—the invasion of Iraq, the reelection of George W. Bush, the surge, the election of Barack Obama, or the death of Osama bin Laden—has persuaded the reinsurance industry to budge from its stance on terrorism risk on Jan. 1, 2002. For those with billions waged on getting it right, terrorism is now a constant, and almost impossible to cover.
By 2001, I was working as a consultant for another reinsurer, Converium. The company kept its U.S. front office on the 47th floor of One Chase Manhattan Plaza at the corner of Nassau and Cedar streets. I did several stupid things on September 11. I already knew an airplane had hit a tower when I got on the subway to go to work. I learned of the second plane while I was still on the train, got off at my stop, and walked four blocks closer. I bumped into my 9 a.m. meeting in the lobby of One Chase. He told me we had been sent home and started walking north. I rode the elevators 47 stories up, since I figured I would need my laptop. I called my mother and told her I was safe.
I rode the elevator back down and walked a block closer until I was separated from the south tower by a small park. As Newhouse stepped onto a boat for New Jersey, I stood and listened to news on the radio of a limousine parked with its windows down, and I watched. When the south tower began to fall the ash was thick, black, and moving fast, and I assumed, briefly, that I was going to die. And then I didn’t. Maps of September 11 victims show bodies recovered on that block at the corner of Cedar and Broadway. I ran, found a hole in the dust that was the door to Two Chase Plaza, and I lived. Until the tower began to fall, it just didn’t occur to me that anything worse could happen. It’s a simple story. I wasn’t worried about terrorism. And now I am.
In December 2001, I was working late, back up on the 47th floor, when the fire alarm rang. I bolted from my desk and past a break room where two janitors were still sitting. “Don’t we need to leave?” I asked. “They’ll tell us if we have to,” they said. I ran, alone, down 47 flights of stairs and emerged on the street without wallet, coat, or keys. There was no fire. I spent the night with friends. I had overpriced the risk of terrorism.
Catastrophes do not surprise all societies equally. Greg Bankoff, an historian at the University of Hull in Britain, has defined for the Philippines a “culture of risk,” a set of adaptations built around the constant threat of natural disaster. Agricultural systems in the Philippines focus on minimizing loss rather than maximizing yield. The islands developed a kind of low, buttressed “earthquake baroque” style for stone buildings. Communities are quick to relocate out of danger.
In a paper this year on risk culture in Western Europe, historian Christian Pfister of the University of Bern in Switzerland details how towns pass on memories of disaster. A stone house in Wertheim, Germany, sits on the Tauber River, painted with the dates and levels of floods back to 1621. Storm tide markers dot the North Sea coast. In 1606 the farmers of Italy’s Aosta Valley built a chapel near the glacier that delivered regular floods and began a procession to it every July. And on Japan’s east coast, stone tablets mark the high-water marks of tsunamis. They all send the same message: When we are gone, remember this flood. And prepare for the next one.
From Newhouse’s office you can see, near the receptionist, a monument to the 29 Guy Carpenter employees who were killed on September 11. It stands about five feet high, in polished wood, inscribed with 29 names and topped with a crystal flame. Newhouse is now Guy Carpenter’s chairman. When I tell him that I was downtown on September 11, he reaches into a box below his desk and gives me a pin he’d had made for staff and clients: the towers of the World Trade Center, wrapped in a single black ribbon. I mention the tsunami stones, many of which had been ignored in the burst of economic activity in Japan that followed World War II. Commerce grows along dangerous places, on rivers and coastlines. Is it possible, I ask, that there’s a value to forgetting? He pauses. “If the stone is there the stone is there,” he says, “but 500 years of upside, with the absolute certainty that it’s going to be the generation after me, not me, that gets drowned—that’s human nature, isn’t it?”