Cat model angst grows after storms.

Catastrophe models that heavily influence insurers' underwriting and pricing decisions are being scrutinized closely in the wake of recent Gulf coast storms.

Commercial insurers are questioning why these sophisticated models failed to forecast a devastating event such as Hurricane Katrina, and asking whether it's time to rethink the tools used to create realistic disaster scenarios.

Thomas F. Motamed, vice chairman and chief operating officer of Chubb Corp., which sustained a loss of $511 million as a result of Katrina, said the magnitude of the losses experienced by both the industry and Chubb "significantly exceeded" what was suggested under existing catastrophe models.

"We're now seeking to answer a number of questions: Why did the models prove so much more inaccurate with Katrina than with other hurricanes, such as in Florida during 2004?" Motamed said during an earnings conference call with investors. "It's clear we need to understand better whether there were flaws in the model we used or in its underlying assumptions."

But Ryan Ogaard, a practice leader at Marsh & McLennan Cos.' Guy Carpenter & Co., said squabbling over whether the models were right or wrong in forecasting Katrina oversimplifies their use, because in any single event there are complicating factors such as business interruption and flood-related losses, and a number of wild-card factors such as political risk that aren't represented in the data.

"No one who knows about these models thought they considered all of that; and anybody who did think that the models were going to be an all-encompassing view of any given event, just doesn't know how these things are used," said Ogaard, whose division runs cat models for clients on a daily basis. Guy Carpenter licenses data from all three of the major modelers--Eqecat Inc., AIR Worldwide Corp. and Risk Management Solutions Inc.--and helps clients understand how to use them efficiently.

Ogaard said it is necessary, however, in light of the string of stormrelated disasters, to look at how better to gauge the frequency of hurricanes. "It seems to boil down to this: Are we in some multidecadal pattern? There is some evidence in the last few years that something is happening. So, the big question is, is the 1-in-100-year event, really a 1-in-50, or a 1-in-20?"