RMS Withdraws Five-Year Hurricane Model from Florida

May 18, 2007

Risk Management Solutions has withdrawn a short-term, five-year hurricane forecasting model submitted for approval to Florida officials after being told the model did not meet required standards.

Mitch Sattler, RMS vice president for public policy, said representatives of the Florida Commission on Hurricane Loss Projection Methodology concluded that the short term model did not meet some of the standards normally associated with previously accepted long term models.

“There was fundamental concern of the professional team based on the number of standards RMS did not meet,” said FCHLPM member Howard Eagelfeld. He said RMS had the option of either requesting a second visit by the professional team or proceeding to a presentation in front of the commission.

But RMS has opted not to proceed with the short-term model and will stick with its more traditional long-term version. “We decided not to request the second visit,” Sattler said. “We will use the long term historical averages and resubmit the model at a June 20 meeting.”

The models are used by property insurers to gauge future expected losses in hurricane-prone areas and help them determine what rates to charge.

Sattler defended RMS’s use of the short term or five-year model, saying that models – no mater what the length of projection – are always forward-looking. He said the new model represents a five-year moving timeframe based on different parts of the historical record with concentrated consideration on the current cycle of hurricane activity.

While Sattler claimed that RMS “did not unilaterally make a decision on the five-year model,” Eagelfeld said the team “had uncertainties” with regard to the meteorologists on the RMS panel.

Specifically of concern were the RMS meteorologist selection process and certain questions asked of them, according to Eagelfeld. Another concern is that RMS used warm sea surface temperature years in their model calculations, he added.

“Concerning their elicitation methodology process, the team couldn’t verify that it was an acceptable technique,” Eagelfeld said. “The mission did not have an absolute process that rejects the five-year model, but it requires scientifically sound practices.”

Consumer groups have criticized the “near term” hurricane prediction models based on five-year forecasts. They have urged state regulators to reject them as a basis for hiking coastal property insurance rates. Georgia and now Florida are the only states to reject the models for use by insurers.

The Consumer Federation of America and the Center for Economic Justice in March called on state insurance regulators to reject what they say are “severely flawed hurricane projections” used by insurers to “sharply” increase rates on property insurance policies in states along the Atlantic and Gulf coasts.

“We informed the NAIC a year ago that modeling changes made by RMS would lead to unjustified rate increases for consumers, but the NAIC and every state with the exception of Florida and Georgia failed to act to protect consumers,” said J. Robert Hunter, CFA’s director of insurance. “Since that time, rates have risen sharply in coastal areas and impartial scientists have strongly criticized the use of ‘near term’ projections by RMS and other firms that have increased estimated loss costs by up to 90 percent in some areas. Even one of the firms that markets catastrophic risk models, AIR Worldwide Corp. has criticized the practice.”

Birny Birnbaum, executive director of CEJ, has maintained that the models are not as scientific as their creators claim. “It is a sham for RMS to claim that its catastrophe models are scientifically sound when they make an ad hoc adjustment at the end of the process that doubles loss projections,” he argued.

He, too, urged state insurance regulators to take up the issue. “The NAIC claims the primary job of state insurance regulators is consumer protection, but it has done nothing to protect consumers from massive and unjustified rate hikes. It is a sad commentary on state insurance regulation that consumer groups have to repeatedly demand that regulators take action to stop these dramatic and unfair increases,” Birnbaum added.

RMS responded to those criticisms in March. “Our model reflects the widespread agreement among researchers that hurricane activity in the North Atlantic has increased since 1995 and that this period of elevated activity will last for at least another 10 years,” said Sattler at the time. “The long-term historical average significantly underestimates the hazard posed by hurricanes for the foreseeable future.”

Randy McKee, meteorologist in charge for the National Weather Service’s Mobile, Ala. office, told Insurance Journal that in terms of weather trends, a five-year sample is meaningless. “For any kind of study to be valid, sometimes 100 years isn’t even enough,” McKee said. “Climactic variations may last hundreds of years. How are you going to come to conclusions with five-year predictions?”

McKee said the hurricane activity witnessed over the past several years is part of a “multi-decadal trend” that began in the mid 1990s: “We’re in a pattern where we expect hurricane activity to be above normal,” he said. “These patterns usually last one to two decades.”

The models have historically been based on over a hundred years of historical data, but last year, RMS announced that it would use a forecast of only five years because hurricane activity over the next few years will be above the historical average.

But the CFA and the CEJ said scientists and insurance experts have increasingly questioned the scientific legitimacy of the modeling changes. “The wheels are coming off of the ‘science’ that RMS said it employed,” said Hunter.

The decision by RMS to withdraw its short-term model is a result of Florida continuing to engage in “wholesale denial over hurricanes” in general, according to Robert Hartwig, chief economist for the industry’s Insurance Information Institute.

“RMS has concluded that in Florida the environment is hostile and it may be counter productive to push the short term model at this time,” Hartwig said. “There are numerous models out there but no model would bring actuarial justification to the current environment in Florida. Florida’s recent legislation assumes there won’t be anything more than an average storm in the future.”

RMS said it had been in discussion with the FCHLPM over the past year “regarding use of a new method of estimating future hurricane activity over the next five years, drawing upon the expert opinion of the hurricane research community, rather than relying on a simplistic long-term historical average which does not distinguish between periods of higher and lower hurricane frequency.”

Sattler said RMS had been optimistic that the certification process would accommodate a more robust approach, so it was disappointed that the FCHLPM was “unable to verify” that the company had met certain model standards relating to the use of long-term data for land-falling hurricanes since 1900.

Sattler noted and Eagelfeld confirmed that a hurricane model alternative, the Florida Public Hurricane Model, also has not been certified as presented this year. Funded with nearly $2.7 million in public money, this project was contracted out to the Florida University system with several institutions participating including Florida State University, Florida Institute of Technology, the University of Florida and the University of Miami.

“The FIU model shows losses higher than all the other models out there,” according to Sattler.