Insurers' crystal ball
Computer models, the tools that provide the statistical data upon which rates are determined, are in the center of Florida's latest storm.
By TOM ZUCCO
Published November 5, 2006
Most homeowners have gotten the news. It probably came quietly in the mail from the insurance company.
Property insurance premiums in Florida have risen as much as 400 percent in the last year, causing chaos in the housing market, panic among business owners and the beginning of an exodus out of the state. People can't afford insurance, can't find it or both.
The key question, of course, is why?
Insurers say the eight hurricanes that struck the state in 2004-05 caused about $40-billion in insured losses. The storms drove up the cost of reinsurance, which insurance companies buy to protect themselves from those same catastrophic losses.
But another factor is at play. Something that state regulators have little control over and some see as a license for insurance companies to raise rates as high as they want.
There is an ongoing cat-and-mouse game that insurance companies play with state regulators, and the models are the mouse hideouts.
It works like this:
Every year since the mid 1980s, insurance companies gather mountains of climate and construction data to determine the likelihood that they'll have to cover claims in the future and how much that could cost. They feed that data into computer catastrophe models they rent from about a half dozen private companies.
Based on 100-year or sometimes even five-year projections, the models then predict the storm damage. The insurance companies use that prediction to determine the rates they'll need to protect themselves.
But most consumer advocates think this is a severely flawed arrangement, and their argument is this: if an insurance company pays for the data it uses to make money, why would it accept any data that would prevent it from doing that?
In a letter to state insurance commissioners this year, the Consumer Federation of America noted that Risk Management Solutions, one of the modeling companies many major insurers use, had just changed its procedures for estimating future risk, paving the way for double-digit rate increases.
"RMS is the vehicle for collusive pricing by insurers and that is not only unfair to consumers, it is anti-competitive," the letter reads.
"RMS claims this massive rate increase is necessary for scientific reasons, but they appear to be bowing to pressure from insurers, who provide a great deal of financial support to them."
In fact, no one other than the insurance and modeling companies knows exactly how the numbers are used. That's a closely guarded trade secret, and the reason why the models are termed "black boxes" by those who monitor the industry.
And because the insurance companies hire the modeling companies to do the work, there is at least the potential that the modeling companies will tell the insurers what they want to hear, if for no other reason than to retain their business.
Officials from RMS, based in Newark, Calif., did not respond to requests for comments. But in an interview last month, company CEO Hemant Shah acknowledged problems after record losses from 2005.
"Models had some gaps," Shah said. "We didn't anticipate secondary effects from a catastrophe."
And he added this: "The model didn't tell you insurers to write the business."
This year, Florida decided it shouldn't rely on private models used by the insurers and created its own. The public model was put together by a team led by researchers at Florida International University at a cost to taxpayers of $2.7-million.
It was touted by state officials as providing a "transparent model independent of the insurance industry and state regulatory agencies."
But even that has been called into question.
Steve Burgess, the state's insurance consumer advocate, used the example of rate filings made this year by State Farm, the state's second-largest insurer, and top-ranked Citizens Property Insurance, the state-run insurer of last resort that must charge rates higher than the private market.
When the state model was used, Burgess said, it showed State Farm could have asked for nearly twice the 80 percent increase the company was seeking, while a Citizens' rate request was significantly lower than what it had sought.
"We're trying to figure out how the two biggest insurers in the state had very divergent numbers," Burgess said. "I'm troubled by the fact that State Farm was rejected for a higher amount."
State Farm eventually agreed to a 53 percent average statewide increase - down from the 85 percent increase it sought. But that was because of other factors, Burgess said, most notably a reduction in State Farm's underwriting profit.
Burgess has recently begun examining the public model to make sure it is performing as expected.
"It's extremely important to the credibility of the process," he said.
"We need the ability to actually run the public model ourselves. It's not so much that I'm bringing anybody's judgment into question.
"Let's make this process much more open."
Officials from the Florida Office of Insurance Regulation, or OIR, whose actuaries approve or deny rate hike requests based largely on the public model, say huge differences between public and private models don't occur often, and the public model, although not perfect, is credible.
"The best gauge we have is Hurricane Wilma," said OIR spokesman Bob Lotane, "and the state model performed remarkably well, within 2 points of the actual Wilma losses. It was markedly better than the private model.
"We have no problem with the public model, but we're still in the infant stages of modeling."
As a backstop, Lotane said, regulators (and the insurance companies) rely on their own actuaries to interpret the data.
"You still need to depend on human beings," Lotane said. "We know how we got there."
Not so, he said, with the private models.
"They (private insurers) never want to submit their black box information," Lotane said, "so we have no way of knowing how they arrived at their numbers."
Just last Thursday, the uncertainty of one of those private models was at the center of a heated rate hearing involving state regulators and Allstate Floridian.
"The current models are susceptible to problems because the insurance companies pay to use them," said Bill Newton, executive director of the Florida Consumer Action Network. "If you're a modeling company, you'll tell the insurance companies what they want to hear. There's a built-in prejudice or bias. ... Just the way it's set up is an open invitation to problems."
Even if the models are correct, Newton argues, they should be adjusted because the risk of hurricanes went down.
"Why didn't our rates come down when the risk was changed?" Newton asked. "Where was the midyear adjustment?"
Instead, every major property insurer in the state has asked for at least a 10 percent increase this year.
The insurance companies say they'd reveal the black box secrets if they could.
"It's not us who wants to keep that secret," said State Farm spokesman Chris Neal. "The modeling companies that developed them don't want the information out."
Still, if the state model is used to verify every rate increase request that's made, why would the insurance companies bother with their own models?
"We have a lot of questions to be asked before we trust billions of dollars in exposure to that one model," Neal said. "I don't think we'd be opposed to using their model, but the models we use are the best tool that the insurance industry has come up with to set rates and try to predict what losses will be in the future.
"Nobody knows what hurricane insurance in Florida should cost. We lost every dime after (Hurricane) Andrew because we grossly underestimated the risk.
Neal also argues that if Burgess is right and the private models are rigged to favor the insurers, then why are companies like Allstate quietly pulling out of the state?
"They would be flocking to Florida by the boatload," Neal said, "and we know that the exact opposite is true."
What will make rates stabilize, Neal said, is something far outside the box.
"If we can string enough of these calm years together, we'll have money to build reserves. Now, we're operating on borrowed money."
While many policyholders are operating on borrowed time.
Tom Zucco can be reached at firstname.lastname@example.org or (727) 893-8247.