Quantifying Catastrophe: Inside Cat Bonds

The scene in New York City was surreal, bordering on cinematic: streets nearly empty, subway trains sitting passenger-less in their tunnels, the concourse at Grand Central Terminal silent but for the wind outside. Hurricane Irene, tracking north along the Atlantic Coast last August, was menacing its way toward New York, turning the city of 8 million into a virtual ghost town. Yet in the hours leading up to Irene’s landfall, a few people were as active as they’d ever been.

Insurance-linked securities (ILS) were conceived for events like Irene. For the busy investment managers specializing in “non-life” products like catastrophe bonds—or cat bonds—a major storm in New York City was their time to shine.

In his 2008 book The Ascent of Money, historian Niall Ferguson called catastrophe bonds and other weather-influenced securities “the most interesting kind of derivative” offered today. Equally enticing, though, is the evolution of catastrophe modeling technologies that has successfully backed the market’s steady rise.

From Alternative to Fixed Income
Among ILS products, catastrophe bonds are probably the most well-known. Simply put, they are securities that allow insurance and reinsurance firms to finance risk through the capital markets as those firms find themselves subject to Solvency II capital requirements and already cash-strapped after 2011’s record payouts. Besides their relatively high yield, the securities are made unique by their terms of default, which are related to different types of “peak-peril” disasters, their specific characteristics, and often, the property damage that is caused.

With performance uncorrelated to the market, the bonds are steadily moving from a niche alternative asset to a fixed-income product in their own right, attracting pension funds, sovereign wealth funds, and hedge funds, among others. Issuances in 2012 are outpacing previous highs reached in 2007, and include the largest single transaction ever, a $750 million tranche called Everglades Re.

Widening volume has allowed partnerships to arise providing access to institutional investors—among them a stand-alone Star Cat Bond fund from $50 billion diversified asset management firm GAM, managed by Westport, Conn.-based Fermat Capital Management.

Joe Gieger, managing director for GAM, says the evolution of cat bonds are following a model similar to that of the so-called “junk,” or high-yield, corporate bonds in the 1980s.

“Given the state of where fixed-income yields are in the market, at the 2 to 4 percent range, to have a bond that’s offering a rate of return that is higher than any of the actuarial assumptions the plans have, typically at least 7.5 percent plus Libor, that is consistent, BB-rated and secured—it becomes very attractive,” Gieger says.

Measurable Characteristics
Assessing those incentives requires hard science. Tom Larsen, senior product architect at modeling firm Eqecat, says “a key reason that this industry has grown so strongly is the ability to provide contracts that have clear and equitable settlement based upon measurable catastrophe characteristics that have minimal moral hazard.”

Those characteristics are three, according to Pete Vloedman, CEO at Wayne, Pa.-based Anchor Risk Advisors, which specializes in ILS investment management, and include the hazard itself, the structural vulnerability—“think shake-tables simulating earthquakes,” he says—and the financial loss involved. The models can be provided as an advisory service or, increasingly as with Eqecat’s RQE platform, through an XML-based application programming interface (API) that directly integrates with firms’ portfolio management systems.

Assessing the stochastic, or random, probability involved requires a diverse set of tools. Many are physics based, including the “heat-pump” convection model that uses sea surface and air temperature in the lower stratosphere to predict hurricane strength, notes Peter Nakada, director of risk markets at Risk Management Solutions (RMS).

On the engineering side, taking account of Donald Rumsfeld’s famous “unknown unknowns” is also required, says Larsen. After the 1994 Northridge Earthquake in California, for example, a previously unobserved flaw was discovered in the construction practices for steel frame buildings—leaving a steel backer bar in behind a weld, stiffening the beam-column connection—that led to significant losses to insurers.

Multiple Models
Today, demand for this information is also driven by capital markets participants no longer viewing the details of risks offered in an issuer’s prospectus as sufficient.

“The sponsor will generally use the ‘cheapest’ model with lowest expected loss, which gives a negative bias to the investment—low-balling the loss estimates. “Since RMS upgraded modeling last year, for example, their models have predicted appreciably higher expected loss. Logically, then, none of the issuances around US wind used their models,” says Serge Chiaramonte, director of ILS at Credit Suisse.

Credit Suisse and Fermat, and many mature cat bond participants, use two or three different modeling services when considering a particular event, capturing the significant deviation between the models while using historical data to determine which comes closest. “In the case of the US, after RMS’ 2011 update, you can have certain impacts that can be double the expected loss for RMS than for [another model provider] AIR—so that leaves quite a large gap between modeling outputs for the same event,” explains Chiaramonte.

Much of that uncertainty, of course, has to do with the severity of the event, and many investors might be surprised by which kind of disaster is easiest to model. “Conventional thinking might be that the more experience you have [over time] the less the uncertainty. But that does not apply directly to catastrophe modeling, which is just as much about placement risk as severity risk. And that placement risk is determined by scenario analysis, where the placement risk and the severity risk of the scenario are separately informed by experience. The end result is that a one-in-100-year event is often actually easier to model than a one-in-10-year event,” says Fermat’s CEO, John Seo.

One example of the difference is the recent trend in property configurations that involves concentrating greater—and better insured—commercial property in a more compact area, thereby increasing the potential cost of a tornado, which is otherwise a relatively low-severity catastrophe event. “The related bonds might be higher yield, but don’t always move up in line with the increased risk—and recognizing that is tricky,” Seo says.

Understanding those subparts becomes even more critical for firms active in the secondary and real-time, or “live-cat” markets, led today by secondary market-makers Swiss Re, Aon, and Rochdale Securities. RMS’ Nakada says when the firm added a “live-cat” function, the value-add is not the meteorology but rather granularity of property detail—down to roof design, the length of the nails used and types of shingles—and what that could mean for damages. Meanwhile, to hedge against delays to external model­ing updates, Chiaramonte says Credit Suisse’s London-based ILS team maintains real-time access to meteorological data as hurricanes form in the Atlantic.

Catalogs of bond issuance history have also become necessary, a new demand that has some firms considering white-labeling. Sandro Kriesch, a partner at Twelve Capital in Zurich, says, for example, that rival investment management firms in the space have inquired about Twelve Analytics, which the firm first developed when it was still known as Horizon21. “We seriously considered it, but for the near future, determined it is just too valuable for our clients,” he says.

Complex Valuation
With real progress on the hazard and vulnerability ends, modeling firms say the next step for cat models is to take up the challenge of integrating valuation.

RMS’ Nakada says, for instance, that the firm’s separate Miu Pricing module made recently available to investors and based on an options platform, recognizes that fixed-income participants may well require a pricing engine in-house, while leaving the heavy-duty science to the modelers. The module calculates the Z-spread, the spread you would earn if there were no cat risk at all, or essentially yield to worst; the cat adjusted spread, or the risk premium after accounting for expected cat losses; and the difference, which is the spread you expect to lose due to cat risk, called “catastrophe cost.”

Where things get more complicated for the models, says Vloedman at Anchor Risk, is in dealing with the details of the bonds’ unique triggers on the financial side of the equation.

“Thinking about it in the context of a pyramid, one could start with parametric triggers on the bottom, modeled loss next, then industry loss, and finally indemnity on top, as you get to the most complex,” he explains. Whereas parametric bonds are triggered with an event simply occurring, with no bearing on actual damage, those at the top deal with a whole host of intricacies specific to the insurance industry—e.g., whether claims adjusters are employed in-house or contracted out—and, in the case of indemnity, the operations of single companies. “Those all have an impact on valuation,” says Vloedman.

Similarly complicating matters is that triggers can be based on a single event, or aggregate multiple disasters over time, as was the case with Avalon Re, a bond issued in 2005 that covered both Hurricane Katrina and the Buncefield oil depot explosion in the UK within its first year.

According to a report from reinsurer Swiss Re, bonds with those more complex triggers have jumped from 30 to 60 percent of the market since 2007. While this is not necessarily a bad thing, firms must recognize the increased variance of the models when considering those bonds. “It is simply down to the old Clint Eastwood line, ‘a man has got to know his limitations,’” Vloedman concludes.

A Kind of Purity
Even while valuation remains a differentiator, continued developments in the cat bond market are certain to rely on the modelers, as the geography and liquidity in the markets incrementally expand to an estimated $20 billion by the end of this year. “As the market grows and continues to offer these attractive yields, like the Everglades’ T-bill plus 17 percent, we expect more institutional clients, including funds of funds, to enter,” GAM’s Gieger says.

The asset class will also become more mainstream once an index of pricing information is published—Nakada at RMS, for example, says a project is in the works with Bloomberg. And bonds for non-natural catastrophes have potential as well, including terrorism and even, as suggested by George Washington University law professor Lawrence Cunningham, spreading the risks auditors face under Sarbanes-Oxley.

Yet for veterans of ILS, there remains something unique to the science of measuring nature’s own risks, and the scientists who model them. “People in the asset class love cat modelers because of their willingness to change and evolve. Cat modeling firms were born out of thin air as entrepreneurial ventures with a relative purity in their approach to the market—a contrast to most financial models, such as those backing mortgage-backed securities, which were sponsored by broker-dealers who were trying to sell products, and where the pressure to leave the models alone was therefore tremendous. With a cat modeling firm, once there is a scientific basis to evolve the view, to not change it would be to lose all credibility,” says Fermat’s Seo.

Indeed, with the markets hanging on every word and folly of politics, consumer confidence, and human behavior, there is something strangely rational—if knowingly imperfect—about sizing up a disaster, and betting against it.