Newsletter Sign-up

By MATTHEW BRODSKY, senior editor/Web editor of Risk & Insurance®

Listening to some loss-estimate reports from insurance companies, one might get the impression that catastrophe models did not get Hurricane Ike right. One might be right too.

And while Lloyd's announced in October that losses from Ike and Gustav were as expected, one of its London insurers, Advent, announced in January that its Ike loss estimate was nearly tripling--from £15 million to £42 million. The specialty insurer also announced that it would no longer write Gulf Coast offshore energy business with windstorm exposure.

In their announcements, the insurers explained these increases by saying that Hurricane Ike caused more destruction than initially expected out of a Category-2 storm. Particularly, it was more damaging to offshore energy facilities as well as inland property--far inland.

And where do insurers come up with these initial loss estimates, which in this case appear to have been incorrect? Typically, catastrophe models help provide an understanding of the range of industry exposure to a particular event. We have all seen the press releases blasted out right before and after a hurricane landfall from risk modelers like AIR Worldwide Corp., Risk Management Solutions Inc. and Eqecat Inc., predicting what likely damages a storm will cause for the industry. Carriers can use these industry loss estimates to gauge their own. Then when claims start coming in, they can update their estimates (as was the case with Partner Re, Paris Re, etc.)

THE MODELERS' ESTIMATES

"Most of our clients actually depend on us to provide a range of reasonable scenarios," said Peter Dailey, Ph.D., director of atmospheric science at AIR Worldwide. The modelers' job in making these loss ranges during a hurricane, he explained, is to estimate what the impact will be six months to a year later when all claims are tallied.

In real time in September, AIR had released estimates that Ike would cause $8 billion to $12 billion in industry-wide onshore coastal property losses. Then, when the remnants of Ike mingled with another storm inland leading to high winds and flooding over the Midwest, according to Dailey, AIR used its extratropical cyclone model to come up an additional loss range of $2 billion to $3 billion.

Added up, that led to a total onshore loss estimate of $10 billion to $15 billion provided during the storm--a range that AIR is sticking with now. The firm also announced that the storm would cause losses of $1 billion to $2 billion in offshore energy hits.

"We're still very confident in our estimates," said Dailey, adding that his firm made an effort to get Ike right. "We treated it very carefully and spent a great deal of time on it."

One of AIR's competitors, Eqecat, released an initial loss range of $8 billion to $18 billion. Nearly a week after Ike's landfall in Texas on Sept. 13, after receiving reports from its reconnaissance teams and additional data, EQE came out with an updated loss estimate of $8 billion to $12 billion. It also added a loss estimate range for offshore energy losses: property damage of $1.5 billion to $2 billion and lost production losses of $2.5 billion to $4 billion.

Another modeler, RMS, reported initial onshore loss estimates of $6 billion to $16 billion, then updated the lower figure to $7 billion a few days later for both onshore and offshore losses. And then--still following us?--on Oct. 24, RMS put out another press release updating its estimates further: total industry losses went to between $13 billion to $21 billion: $10 billion to 15 billion from wind and storm surge in Texas and Louisiana, $1 billion to $3 billion for offshore losses, and an additional--previously excluded--range of $2 billion to $3 billion for inland wind and flood losses from the aforementioned meteorological phenomenon over the Midwest.

"The reason that the update took place, there were a few things going on," explained Robert Muir-Wood, chief research officer at RMS. "This was not simple wind damage on its own."

How so? For starters, Ike was a huge storm with tremendous surge for having only Cat-2 winds that basically made a direct hit into the third largest concentration of coastal property in the United States, according to Muir-Wood. Add to that footprint power outages, and you have a very different--though not unforeseen--storm.

Offshore, losses were not so clear-cut either. "They were not at all well-established as to what had happened," said Muir-Wood, explaining that losses offshore were higher than expected for a Cat-2 storm, in part because wave heights were higher than one would anticipate.

And as explained previously, Ike got even hairier as it went inland, which RMS initially did not model for. Muir-Wood would even argue that, meteorologically, the Ike on the coast and the Ike in the Midwest were not the same phenomenon.

WHO'S WRONG?

So do all these numbers, updates and qualifications mean that AIR had it right to begin with, and RMS and EQE did not?

Yet ISO's Property Claim Service estimated that Hurricane Ike cost property insurers $10.7 billion, which would fall in the range of all three modelers' estimates for onshore property losses. So then, all the modelers could be right, despite discrepancies.

Or how about, because of the discrepancies among the three loss estimates, modeling overall was wrong again in 2008 at least with Ike, as it was in 2005 with Katrina?

"Ike was certainly more expensive and damaging than people were expecting," said Howard Mills, the former NYS insurance commissioner and now chief advisor of the global insurance industry practice at Deloitte & Touche LLP. Mills pointed as well to the scale of inland damage, to the scope of power outages and business interruption.

"In a holistic way, it was not very well handled," said Steve Smith, Ph.D, president of Property Solutions for ReAdvisory, a branch of reinsurance intermediary Carvill.

Sure, the models tried. Such as how AIR switched to its extratropical cyclone model after Ike went too far inland. But Smith doesn't necessarily think adding another model's output on top of their hurricane model's is the best approach; these storm models do not too good a job with flooding, he explained, which was key here, when Ike remnants dropped 10 inches of rain on Chicago, Smith's hometown. It is with this inland flooding, the atmospheric scientist said, that modelers could learn to do better.

"They can try as well as they can," Smith said about models, "but with such a unique event, it would be very hard to come up with a very credible number."

Overall, though, the models did perform relatively OK in 2008, according to Smith. They handled Hurricane Gustav, for example, which was a "fairly standard storm."

Last year could be "another one of those years where we'll take a few things away that aren't huge, but incremental improvements," said Smith.

January 21, 2009

Copyright 2009© LRP Publications