Then there came flashes of lightning, rumblings, peals of thunder and a severe earthquake. No earthquake like it has ever occurred since man has been on earth, so tremendous was the quake. The great city split into three parts, and the cities of the nations collapsed. Every island fled away and the mountains could not be found. From the sky huge hailstones of about a hundred pounds each fell upon men.
--Chapter 16, The Revelation to John, New Testament, New International Version
What's the return period on Armageddon? Yes, thanks to catastrophe models and ratings agencies, insurers can't stop thinking of disaster in terms of probability. The storms of 2004 and 2005 provided stark evidence, too, that severe events could happen more frequently than once suspected--and that worse could still be in store. The lesson--there isn't a catastrophe too improbable, or too severe, for risk management.
Models can produce a probable maximum loss (PML) curve--a chart that's a function of the worst amount an insurer is set to lose along a continuum of return periods--as far back as typically the 10,000-year point. And to what does that tail end of the continuum translate in loss totals?
"A lot of people want to know how up is up," says Tom Larsen, senior vice president of product management, Eqecat Inc., "what do we think is the worst possible event."
The up, he says, is $1 trillion, and rising.
The question comes to mind, though, if risk managers at insurance companies should even be concerned with such points at the tail end of the curve--apocalypses of biblical proportions with annual probabilities in the hundredths or thousandths of a percentage.
"Quite frankly, it doesn't make sense for an insurance company to have enough capital to pay every conceivable claim. There's no way they could make any money," says Paul Budde, executive vice president, product development, at reinsurance intermediary Benfield. Protecting a company against its 10,000-year event--with its annual probability of 0.01 percent--would be "ludicrous," he says.
But Budde says there is value in widening the risk radar to detect disasters that, though not New Testament bad, are quite awful and quite possible.
John Beckman, president at Carvill's ReAdvisory service, is in this camp. "It seemed to be too many people were looking (before Katrina) at the one-in-100 loss as their worst-case event, which is not what it is," he says. "It is just a relatively infrequent event, but it could still happen, and you need to know you're exposed to that. But you should also look at what's your one in 250, what's your one in 500, what's your one in 1,000," Beckman says.
Insurers used to focus on two points on the curve--the 100-year for hurricanes and 250-year for earthquake--in part because ratings agencies honed in on these two PML points.
And in part, the reason that insurance company risk managers are looking beyond these two PML points is, again, because of the ratings agencies.
"It's not just, 'Give me your 250-year earthquake loss and your 100-year hurricane loss' anymore," says Don Windeler, earthquake practice lead for modeler Risk Management Solutions Inc. "All the ratings agencies are asking much tougher questions and being much more specific about what they ask."
"We've been asking these questions, but now we're requiring the documentation of that detail," says Robert DeRose, assistant vice president at A.M. Best Co. Inc. "Before, these are questions that we would ask in management meetings, but now we're asking for the information in actual text and documentation."
Sure, the 100-year and 250-year numbers still factor into its equation. But last year, Best revised its stressed BCAR calculation, which simulates how an insurer's surplus would stand up to multiple events, to now consider a rare double-whammy year of two 100-year (or one 100- and one 250-year) events.
The agency also examines the tail end of the PML curve out to the 10,000-year event, which is a highly improbable event occurrence, but one that's modeled.
In the case of another ratings agency, Fitch, its new questions involve the so-called tail-value-at risk, or T-VaR, formula.
"We've kind of changed our approach a bit, in that, we're not just looking at points on the PML curve for CAT exposures," says Jim Auden, managing director of Fitch's insurance group. "We're looking at the whole curve."
Well, actually the average of the whole curve--from the 250-year event to the end of the tail.
"That's the concept--factor the extreme events into the analysis. Then we're taking an average of those," he says.
And major insurers, for their part, have learned to look at their whole PML curve.
"People have expanded their view of exposure beyond just the one-in-100 loss to include more severe events," says Beckman, "and some companies have found that, as they go up, the size of their event doesn't change much and others have found that one in 500 is multiples of one in 100."
"And that's the kind of number that could cause them problems if that did occur," Beckman adds.
But there are doubters.
"Some companies just automatically as a matter of policy--in fact most--simply look at a 100- or 250-year event," says Louis Jacobs, assistant vice president, natural hazard perils, at primary property insurer FM Global. "That's going to miss just about all the big stuff."
"I think the ratings agencies will sort of twist their arms in order to run those analyses and report it to them. That still doesn't mean that those insurance companies are using those results internally to make their own business judgments," he says.
"I don't mean they're fudging the numbers. They could still be reporting, let's say, a 500- or 1,000-year to a ratings agency, and making their internal business decisions still on the 250-year event," he says.
DeRose at Best agrees that such a discrepancy was apparent after the 2005 tropical cyclone season.
"It could be poor risk management on the company's part, or it could be an attempt to maintain top line and meet shareholder expectations, maybe never thinking an event that severe would happen--and one did," adds DeRose. "They were taking a bet."
"It's always possible for companies to game the system," says Fitch's Auden, but he couldn't see why they would consider it to be in their best interest.
He adds that, just in case, ratings agencies "ask questions" to uncover if companies are doing so. Best claims to do the same.
PROBABILITIES BE GONE
Some insurers are wondering whether probabilities of any kind are in their interest.
After all, great debate has erupted over how and why hurricane activity seems to be on an uptick. Science is not capable of delivering accurate return periods for individual earthquake faults.
And most anybody will tell you (except those modeling it) that the probability of terrorism is impossible to figure.
Dan Munson, founder of mapping solutions provider CDS Business Mapping LLC, says the dread of terrorism is driving folks to use other tools besides return periods to calculate their exposure.
"People have realized that the only way you could manage aggregates or make sure you didn't lose a lot of money is to make sure you didn't have too much concentrated in one place," he says.
"Well, if we don't want to lose more than $100 million," Munson says, "we better not have more than $100 million at risk."
Says Beckman, Carvill and others in the industry are hunting for a tool for insurers similar to total sums insured, which reinsurers use to add up their exposure to events.
"It's taking that frequency issue off the table," he says. "Let's not worry about how often it happens. If it does happen, is your company still going to be around to pay its claims afterward?"
Speaking of reinsurers, they have long been sophisticated at taking the probability out of the problem. Take how Lloyd's manages the CAT exposure of its syndicates. It uses deterministic scenarios--their so-called Realistic Disaster Scenarios.
"The question of return period is one that's vexed us somewhat over the years," says Paul Nunn, head of exposure management at Lloyd's.
So the venerable market of markets devised a way, implemented back in 1995, to require that its syndicates test against nine events in "key disaster areas" where Lloyd's has peak exposure. Another nine scenarios are required for syndicates that have exposure over a certain threshold for them.
"What we do know is that we have material exposure in all the areas where we've asked these questions, and that the loss levels that we're pitching the scenarios at would generate significant losses for the industry and for Lloyd's," he says.
Some examples of mandatory RDS include a New Madrid quake at both the $40 billion and $90 billion thresholds (depending on, what else, the return period of the event); two $65 billion California quakes; two different $100 billion Florida windstorms; and a $95 billion two-event scenario involving South Carolina and Northeast hurricanes.
For each, Lloyd's prescribes to the syndicates what lines would be affected, what size and location the event footprint would have, and whether storm surge or demand surge should be considered--and how to set these guidelines for each of the major modeling brands.
Munson says reinsurers are passing such thinking down to insurers, prohibiting insurers from writing more than a certain amount of limits within a geographical area--for example, not writing more than $100 million in aggregate of wind storm in Florida's Miami-Dade and Broward counties.
"I think that you are going to see more of that," he says, "where people (reinsurers) are going to set hard numbers and say, 'Don't go over this.' "
The largest insurers are also becoming masters of their long tail on their own initiative, managing their apocalyptic risk through exposure optimization and diversification, rate increases and alternative reinsurance mechanisms such as CAT bonds--one practice that seems to have significant potential.
"I think you'll see a number of organizations--I know a few out there, even on the primary commercial side--that have CAT bond facilities to address their exposures above that one-in-250-year event, to give them some security to be able to sleep at night," says DeRose from A.M. Best.
These lessons are even getting down to the smaller players in the CAT world. One such carrier is James River Insurance Co., an excess and surplus lines property insurer.
The company, says Brian Haney, vice president and chief actuary, learned after Katrina that the best way to control its CAT exposure was to cap its total limits to $50 million in a 25-mile radius, a strategy that, Haney says, has produced a "nice moderate band" of exposure in earthquake and hurricane targets.
Top to bottom, this old-school focus on aggregations and geography is important and good, say the experts. "How long that lasts, we'll have to see," says Auden.
MATTHEW BRODSKY is associate editor of
Risk & Insurance®.
May 1, 2007
Copyright 2007© LRP Publications