Journalism won’t die if you donate. Support Voice of San Diego today!
When insurance companies deny coverage or raise prices in areas at risk of wildfire, they’re often relying on secret formulas designed to predict what homes are likely to burn, according to a Voice of San Diego review of thousands of pages of regulatory filings by eight of California’s largest home insurers.
The reason for cutting policies or hiking prices seems obvious. Destructive fires have burned California again and again, destroying thousands of homes in recent years and costing billions of dollars.
Less obvious is how companies decide who gets insurance and how much to charge.
Companies use risk predictions to draw hard lines around certain homes, making some properties uninsurable by name brand companies like State Farm, Farmers Insurance and AAA.
The most popular wildfire risk model, known as Fireline, uses a 0 to 30 scale, with a score of 30 assigned to the very riskiest homes.
Northern California’s AAA auto club insures about 470,000 homes. Of those, 355,000 have a Fireline score of 0. The company has insured only one home in the past several years with a score of 30, according to regulatory filings.
Several researchers say the insurance industry’s models don’t use all available fire science.
Fireline, for instance, typically considers three things: topography, the amount of burnable vegetation around a home and how easily fire departments can reach a home.
Those factors certainly help predict fire. But scientists say all kinds of other things make a home more or less likely to burn, including whether the home is new and built with fire-resistant features. Fireline does not account for how a home is built.
Officials at California’s Department of Insurance have similar concerns about the wildfire models but still allow companies to use them.
In fact, regulators label the models an objective measure of wildfire risk. Yet models by different insurance companies or industry consultants make different predictions about how likely homes are to burn. If nothing else, that shows some models are more right than others and suggests some models could simply be wrong.
When it comes down to it, industry spokespeople said it’s not surprising that they are trying to figure out how at risk homes are of burning. After all, insurance is all about betting on risk.
If a company takes on too much risk or doesn’t charge enough for insurance, it could go out of business.
“A fundamental and core function of any insurance company is the selection and ongoing evaluation of risk,” said Sevag Sarkissian, a spokesman for State Farm, California’s largest home insurance company. “Insurers will find some risks unacceptable.”
The insurance industry does not talk much to the public about the models or how they are used. But some homeowners turned away by insurance companies are generally aware that they’ve been deemed too risky to insure. Some homeowners have also learned about the risk models through their insurance agent or letters from their insurer explaining why a company is not renewing an insurance policy. Homeowners insurance policies typically last a year. People with mortgages almost always have to have insurance.
While it’s no secret that the models exist or what factors some companies use to assess risk, it’s not clear how the factors are weighed, and state regulators say they have not validated all of the data being used in the models. In June, the state Commission on Catastrophic Wildfire Cost and Recovery recommended that the models be vetted publicly.
For some industry watchdogs, the models run afoul of Proposition 103, the 1988 ballot measure that made insurance companies publicly justify what they plan to charge customers.
“You don’t get to avoid Proposition’s 103 mandate of disclosure by hiring some company in another state that claims its model is, quote, proprietary,” said Harvey Rosenfield, the author of Proposition 103.
The insurance industry says it is disclosing all it has to the state Department of Insurance.
For example, according to regulatory filings:
- Starting in 2011, State Farm stopped offering insurance to new renters in dozens of ZIP codes across California with “high” or “very high” wildfire risk. That prohibition was gradually expanded across the whole state, though such strict limits still only apply to new renters and condo owners looking for insurance. The company is more lenient toward existing customers and homeowners.
- The United Services Automobile Association, known as USAA, may decline to issue new policies to homes “where it is determined there is a significantly higher than average risk of brushfire losses.”
- Farmers, the state’s second largest insurer, requires some homeowners to get a second policy that covers fire insurance because the companies will only cover other risks – like burglaries. People with high Fireline scores who want fire insurance from Farmers are sent to the California FAIR Plan, the state’s industry-backed insurer of last resort.
- Northern California’s AAA, known as CSAA, is proposing to lower rates on most of its low-risk homes. It is also asking to double rates on several thousand high-risk homes. At least 10 homeowners in fire-prone Santa Cruz, Monterey and Mendocino Counties could see their insurance costs nearly triple if they stick with the company.
- Liberty Mutual and Mercury Insurance require customers in risky areas to get special approval before they can get insurance.
It’s unsurprising that insurance companies are reacting this way – 10 of California’s 20 most destructive fires have occurred in the past five years.
“Now there’s a whole bunch of companies after two historically bad years trying to figure out how can we not screw our companies up,” said Rex Frazier, the head of the Personal Insurance Federation of California, an industry trade group.
There is no question that the wildfire models highlight areas likely to burn, and there’s no question destructive fires are not going away anytime soon. But how effective the models are remains little understood.
Whether the models are being used properly is also debatable. Some group homes by ZIP code, but in the real world, fires don’t follow mail routes.
“ZIP code is a terrible unit for assessing risk,” said Bob Frady, CEO of HazardHub, an upstart wildfire risk modeling company. “It’s for delivery of the mail, it’s not for assessing risk.”
HazardHub is trying to enter a market dominated by two consulting firms, Verisk and CoreLogic. Verisk owns the Fireline formula.
In February, Verisk said about 58 percent of the homes destroyed in last November’s Camp and Woolsey fires were in areas Fireline deemed to have high or extreme wildfire risk.
Max Moritz, a wildfire specialist at the University of California’s cooperative extension, said the company’s apparent predictive success may come from considering a lot of the state riskier than it is.
“They err on the side of painting with a high-hazard brush,” he said. “And you know it’s not hard after the fact to show, ‘Look how well we did in predicting losses’ when you painted the whole landscape red.”
Right now, Verisk’s Fireline formula estimates that over 2 million California housing units are in areas with high or extreme fire risk, including a quarter-million homes in San Diego.
HazardHub sees things different. Frady said his company’s formula would consider very risky some homes that Fireline considers less risky.
It’s not always clear what risk model insurers are using. Some companies may use one model, a combination of models from different consulting firms or they may even have their own.
Officials at the California Department of Insurance, who agreed to be interviewed on the condition they not be named, said they found the models have a satisfactory link to risk. But they have not totally evaluated all of the models and their components.
In a 2018 report, the department also found issues with the models because they don’t take into account work by homeowners to clear brush around their home or the fire-resistance of new homes built according to new building codes.
Steve Clarke, a spokesman for Verisk, said Fireline isn’t yet sure how to take those things into account.
“One of the challenges we’ve had with mitigation is a lack of good sound data upon which to do a meaningful analysis of many of these factors,” he said.
Tammy Nichols Schwartz, an insurance industry consultant who used to work for the California FAIR Plan, said the models help insurers figure out what different customers should pay. The industry knows that everyone shouldn’t be paying the same rate.
But she argues that the models shouldn’t be used to deny coverage to people in wildfire areas. Instead, companies should charge people based on the risks they face – even if that’s a lot more.
“I think that the answer is that the Fireline isn’t the right score but it’s the best score we have and it’s better than no score and it’s getting better,” she said.