The spate of insured losses attributed to natural catastrophes the past two years has spurred a broad reassessment of how the insurance industry uses catastrophe models.

Bryon Ehrhart, Chicago-based chief strategy officer for Aon Benfield and chairman of Aon Benfield Analytics, said insurance companies are now striving to “own the risk” by achieving a more thorough understanding of the assumptions that underlie the models.

“The users of these models are now taking ownership of the models,” he said. “There's still something of a 'black box' effect; it's just that the companies are better understanding the assumptions and demanding more transparency.”

John DeMartini, New York-based leader of Towers Watson & Co.'s catastrophe risk management practice, said pressure from rating agencies and reinsurers are primary factors why insurers are looking to better own the risk in their portfolios. To leave behind this more passive view of catastrophe modeling, insurers must redefine how they use the output gleaned from models created by catastrophe modeling firms, such as Oakland, Calif.-based Eqecat Inc., Boston-based AIR Worldwide Corp., and Newark, Calif.-based Risk Management Solutions Inc., Mr. DeMartini said.

Historically, many insurers used the output from a single model to determine the probable maximum loss for a given book of business. More recently, insurers have begun to use multiple models from separate catastrophe modeling firms, often blending the results, Mr. DeMartini said. For example, if one model determined a probable maximum loss of $5 million and a second model estimated losses at $10 million, an insurer could average the two and assign a probable maximum loss of $7.5 million.

%%BREAK%%

Yet the manner in which the models are being blended is now changing, he said.

“Averaging the two model outputs does not constitute owning your own view of catastrophe risk, unless you understand the differences in the models and their core components and biases,” he said. “There's no way insurers can make an informed decision about how to blend models without going into the core components in each model.”

Mr. DeMartini cites the frequency and severity of hurricane risk as an example of the modeling firms using divergent methods to assign risk and the need for insurers to delve deep into the assumptions that populate a model.

“RMS prefers to look at a near-term view of hurricane risk, while AIR prefers to use long-term, historical averages,” he said. “That's a huge difference between the two models that companies have to take a position on.”

To make more informed decisions about the assumptions underpinning models, insurance and reinsurance companies are investing in internal research and development, including hiring in-house climatologists, said Cliff Hope, Atlanta-based executive vice president and chief property underwriter for U.S. Property for Aspen Insurance Holdings Ltd.

“One of the things that has happened over the last couple of years is that insurers are developing research and development teams within their own companies to asses not only model output but the science behind the models,” he said, noting that Aspen has a team to validate or, in some cases, invalidate the science behind the models. “Having that expertise in-house is invaluable because underwriters typically don't have the understanding of all the science that goes into the models.”

Mr. Hope said the need for insurers to own their risk necessitates greater interaction with the modeling firms. “The models companies are supportive of this,” he said. “They understand that companies want their own view of risk.”

%%BREAK%%

Jayanta Guin, Boston-based senior vice president at AIR Worldwide, said his company worked to make its models more transparent. “We are encouraging our clients to question the model, and dig in to the details and understand the uncertainty in the models,” he said.

This more collaborative approach to modeling ultimately will yield better results, Mr. Hope said. “My opinion is that in the next three to five years, you are going to see model output that has become better because of greater cooperation between insurers and the modeling firms,” he said.

Nonetheless, Mr. Hope said insurers need to guard against being too reliant on models at the expense of traditional underwriting techniques.

“To me, as an underwriter, the more important thing is understanding the risk,” he said. “Underwriters need to understand that models are a factor in your assessment of the risk and not the answer to whether you want to assume a risk or not.”

When it comes to commercial insurance, Mr. Hope said underwriters need to pay greater attention to the physical characteristics of risk being insured, as well as the business aspects of what they are insuring, especially business continuity and business interruption policies. “You have to understand the business that you are insuring and not just the model data,” he said.

Mr. DeMartini agreed that quality data about factors such as construction type and occupancy are vital to a more accurate view of risk.

“If you are really going to own your view of catastrophe risk, you have to be extremely confident that you have a complete and accurate set of data for analytical purposes,” he said.