By MATTHEW BRODSKY, senior editor/Web editor of Risk & Insurance®
Isn't it ironic that one of the entrepreneurs credited with launching the catastrophe modeling business back in the 1980s is now a critic? Karen Clark, who was a founder of Boston-based AIR Worldwide in 1987, left the company after 20 years to launch a catastrophe risk consulting businesses. Now AIR and the other two big modeling firms, Eqecat Inc. (EQE) and Risk Management Solutions Inc. (RMS), cringe whenever Karen Clark & Co. releases a report.
In particular, see her 2010 review of near-term hurricane model performance. The results are not impressive.
The report compared actual numbers from the past four seasons--on named storms, landfalls and insured losses--versus what the models project. Over those seasons, four U.S. landfalls occurred in total. The near-term models forecast anywhere from 8.4 landfalls (AIR) to 9.4 (RMS) landfalls. Actual insured losses during the period totaled $13.3 billion. The models forecast $48.8 billion (AIR), $54.5 billion (EQE) and $54.6 billion (RMS).
Considering this discrepancy, Clark's report stated, "Near-term projections do not have sufficient credibility to be used for important insurance applications such as product pricing ..."
That leaves open the implication that insurers that do use these models for pricing property-catastrophe insurance do so as an excuse to charge higher prices for an upcoming season.
Of course, the modelers themselves protest Clark's conclusions. RMS and EQE have issued statements that the report is based on misunderstanding of how catastrophe models work. AIR stated that its heightened-activity product isn't even a near-term model and is not designed to "forecast activity or losses for upcoming seasons, as near-term catalogs attempt to do."
After talking with folks who work with models in practice, it's apparent that near-term models are not relied upon to determine price.
"Karen makes it sound like they take this model output and that's it," said Carl Hedde, head of risk accumulation at reinsurer Munich Re America. "It's a lot more complicated than that."
Munich Re America prices risk based on many factors, Hedde explained, including the long-term view and short-term view of hurricane risk and analysis from in-house scientific resources.
Yet models "definitely drive the pricing," according to Al Tobin, national property leader for Aon Risk Services. Still, Tobin concurred that many factors go into it, one being that insurers usually don't just use one model.
"With all the models out now, these guys are using a wide view," agreed Ravi Singhvi, vice president of catastrophe risk modeling for wholesale broker NAPCO LLC. "I like how the insurance companies view it. They don't put all their eggs in one basket."
Also, pricing can be based on relationships between risk managers and carriers, how much money a carrier has in the bank, or how much CAT exposure it already has in a given region, added Tobin. Other factors are also entirely in risk managers' hands, such as the quality of their property data--the old garbage-in-garbage-out lesson.
Said Tobin about a risk manager who doesn't have detailed data in his property schedule to give to his insurer, "His risk is going to go to the bottom of the pile."
Hurricane catastrophe models are designed to help insurers, brokers and risk managers understand the extent of their exposure, and figure out how much they stand to lose should, say, a 100-year hurricane hit their properties. These are probabilistic tools, meaning the models simulate thousands of hurricane seasons to come up with this output.
Near-term models were introduced after Hurricane Katrina in 2005 to take into account the scientific consensus that the Atlantic basin is in a heightened state of hurricane activity. In effect, when running simulations on these versions, the outputs end up higher because of the assumption of increased storm frequency and landfalls.
January 26, 2010
Copyright 2010© LRP Publications