He added that users need to ensure data is appropriately characterized and complete. “Circumstances like this call for a certain degree of humility–from all of us,” Mr. Shah said of the recent round of U.S. catastrophes.

Mr. Shah gave the audience preliminary indications of the magnitude of loss cost changes that will flow from a model revision–Version 6–scheduled for April 2006 release. For example, he noted that frequency assumptions will likely increase by about one-third next year.

For hurricanes, the issue of frequency fueled by climate change “needs to be put front and center,” he said, noting that while the long-term average for severe hurricanes is 0.6 per year, the average has been 3.5 for the last two years.

This needs to be reflected not just in models, but in insurance company management thinking going forward, he said, noting that science suggests elevated levels of frequency and increased storm intensity for the next five-to-10 years.

Newark, Calif.-based RMS is “now buying into” the messages from scientists, he said, explaining why frequency assumptions in the RMS model will rise. RMS will also undertake an annual process of reviewing forward-looking frequencies on a five-to-10-year time horizon, he said.

As for the increase in storm intensity that's occurring once storms are formed, he said that will “play out differently in different parts of the country.” In some areas, he said, it might increase expected losses by 30 percent or more. “These are only preliminary thoughts,” he cautioned, adding that more insights on model changes would be forthcoming in February 2006.

Earlier in the session, Mr. Shah also noted that Hurricane Katrina had “shed some very provocative light on the idea of 'super-catastrophes'” and the “post-lost amplification factors” that drive loss amounts skyward once they reach a certain severity level.

“We're not just talking about traditional demand surge,” he said, referring to increased costs of labor and building materials. Putting aside the failure of models to predict the collapse of the levee system in New Orleans, he said, “correlations and feedback loops” need to be better understood.

Near the close of the session, Steven Dreyer, managing director and North American practice leader for New York-based Standard & Poor's (one of the event's sponsors) tried to pin Mr. Shah down for more information on how high model indications might rise next year.

Noting the intense conversation around questions of how revisions in rating agency capital models will impact the reinsurance market in 2006, Mr. Dreyer said “we're not requiring more capital, but there's an expectation that [catastrophe] model results will be different,” which may filter through rating agency analyses, he added.

Mr. Dreyer noted there are “a dozen or so companies being formed” in Bermuda, where people have to make decisions about this “today,” explaining the urgent need for such indications.

Mr. Shah replied, “I can't give a figure on the impact because it's portfolio- and line-dependent,” referring instead to the benchmarks he had already provided. The impact will be different for a homeowners writer in Florida and a commercial specialty writer in Texas, he noted.

Besides S&P, the conference was sponsored by PricewaterhouseCoopers, Black Diamond Group, Morgan Stanley and LeBeouf Lamb.

“Circumstances like this call for a certain degree of humility–from all of us,” says RMS.