Insurers and reinsurers are coming to increasingly depend on catastrophe modeling technology to evaluate risks, but at least one expert thinks a lack of detailed data and improper use of such models might render their conclusions inaccurate.

“Companies really have no choice but to use these models–if not, we're back to the Stone Age,” said Tom Conway, principal of Ernst & Young's Insurance and Actuarial Advisory Services practice in Chicago. “But some inaccuracies in the models over the last couple of years have been a result of the way models were used by insurers, rather than flaws in the models themselves.”

Mr. Conway also blamed poor quality data for inaccuracies, contending this has been “documented in a lot of cases.” Data details are not always captured properly, he explained–a multilocation commercial property, for example, might only include data on the main location and not for other satellite locations.

“Models are geared off replacement value,” he added, “so estimating that value is another key thing that may not have been done well [because the] tools available and time given were inadequate.”

The key issue in catastrophe modeling today, according to Mr. Conway, is the recent increase in the number of catastrophic events, and whether that trend will continue.

“Historically, models take a long-term view of an overall average, and that is generally fine,” he said, noting there was a much higher frequency of major hurricanes in the 1930s and 1940s, but that number tended to decrease in the following decades. “From the 1970s to 1990s, it was pretty much at an all-time low, but in 2001, the frequency was back up to where it was in the 1940s.”

Will that higher frequency continue? “It seems it is pushing that way,” said Mr. Conway, citing possible factors that include global warming. “If I was running an insurance company, I would assume that it is [going to continue] and I would build my plans around it. I would not assume that what comes out of these hurricane models is the final answer.”

According to Mr. Conway, “until the last two years, companies tended to look at models as black boxes, and they trusted the answers they got, but the problem is that this is a key risk management issue for the insurance industry. This is a make-or-break issue. We can't afford to just trust the models.”

Mr. Conway asserted that insurers must develop expertise with the models, just as reinsurers have. “They need to know how the models work and what the flaws are,” he said, likening the knowledge gap to the difference between a professor and a student.

“Some of that expertise is only gained when using the models over and over and looking at why some models give different results than other,” he pointed out.

He said reinsurance brokers will often have access to multiple models, adding that “it would make sense” for insurers to have more than one model–even to the point of “coming up with their own fixes.”

How can models be improved? “I would like to see more of an ability to do scenario testing,” he suggested. “Companies need to be able to vary assumptions in the model, instead of focusing on some result as the be-all and end-all.”

Models are “constantly evolving,” he said, “but if you're aware of the flaws, you can make quantitative and qualitative adjustments. Don't wait for the software company. It's your job to get it done right.”

Special Reports /

Cat Models Only As Good As Those Insurers Using Them

Related Stories

Recommended For You