National Underwriter: What are thecompetitive advantage of AIR Worldwide's U.S. Hurricane Modelvis-à-vis the competition?

Ming Lee: What sets our model apart is its greater stabilityover time. It has long been considered the most consistent,realistic and comprehensive view of hurricane risk. From revisionto revision, the model gets more refined and more granular, but arobust model, which ours is, should not require dramatic updates.Also, there is a tremendous amount of transparency around ourmodel: We have literally thousands of pages of documents explaining[our methodology], and our scientists are available to answerquestions.

NU: How has the model'sview changed regarding the extent and severity of inland lossesthat hurricanes can cause?

Lee: We've always been capturing the risks frominland losses, and the potential for these losses is not new. Since1900, over 60 hurricanes have caused significant inland losses.From the inception of the AIR model, we covered Arkansas, Kentucky,Ohio and other inland states; and in 2010, we added Illinois,Indiana and Missouri, so the model now covers a total of 29 states.There has been a lot of fuss over how hurricane models changedbecause of [Hurricane] Ike in 2008 causing a new view of inlandrisk. That's just a lot of fuss about nothing—Ike was not the firststorm to cause inland losses.

NU: What impact have the recent changes toRisk Management Solution's cat model (RMS Version 11.0 U.S.Hurricane Model) had on AIR's business?

Lee: There has been growing interest in AIRover last few years, period. The impact of the RMS Version 11release acceleratef that interest—it has prompted companies to justhave the conversation sooner. What we hear from these clients isthe unexplainable nature of the changes in the RMS model—not thechange itself, but the fact the changes couldn't be explained verywell.

Another point I'd want to make related to this: Severe stormshappen all the time, and some insurers think they can anticipatetheir losses based on historical-claims information. Thecatastrophe experience of 2011 has a lot of companies realizingthat rather than viewing modeling from a historical-claimsperspective, they should take a more scientific approach to theirlevel of risk.

NU: What advice would you give to a companyconsidering changing models or using one for the first time? Whatquestions should they ask?

Lee: Ask the modeler what its approach is tovalidation: Do the components of the model make sense overall—inother words, how do they make sure the output makes sense? Acompany should hear that all components of the model are validatedto have scientific merit.

A company should investigate how the model has performedhistorically. Run storms that occurred in the past with thecompany's book of business and see what the model says the losseswould be if that historical storm occurred today.

Also, take a look at how the model performs in real time: Whatdid the model say that the losses would be as a storm occurred, andhow does that reflect the reality of what actually happened?

Finally, ask why you should trust the modeler's experts—and theanswer should be the extent the model has been peer reviewed byoutside people.

NU: Your European wind model has also seensome recent developments. Tell us about them.

Lee: In 2011, we expanded the model to includeadditional countries, and the latest version of the AIRExtratropical Cyclone Model for Europe now captures the financialimpact of European wind storms on 18 countries. And the 2010release reflects a lot more realism in our creation of thesimulated storms, such as explicit modeling of storm clustering, animportant phenomenon in Europe that must be accounted for whenmanaging wind-storm risk. We also added support of additionalconstruction/occupancy combinations, business-interruptioncoverage, and auto and forestry lines of business.

NU: And what's the latest on theterrorism-modeling front?

Lee: We reconvened our team of experts withexperience in military and government, and they considered theability of a full range of terrorist groups to plan and accomplisha wide variety of attacks. And based on that analysis, the revisedmodel reflects a less frequent and lower severity risk for U.S.terrorism.

NU: Final thoughts?

Lee: The most important concept to remember isthat a cat model's output is a currency by which risks aretransferred—like money. And like any currency, a cat model hasvalue only to the extent there is confidence in that output. Andthat confidence should rest on sound science—science that ensuresthe numbers make sense.