A new flashpoint is the use of algorithmic models for predicting catastrophe losses. Insurers have long used these models to project losses from natural disasters but insurers, homeowners and regulators disagree on how they should be applied to assess California’s wildfire risk.
Insurers are pushing for broader permission from the state to use catastrophe models when setting rates. But some consumers say those models are opaque and fail to adequately capture wildfire risk.
The debate over catastrophe models is part of a broader disagreement between insurers and consumer advocates over how to respond to two years of record-breaking wildfire damage in the Golden State. Many residents of fire-prone areas are either paying more for insurance or having problems securing it, and the state has warned that the situation is likely to get worse.
California wildfires cost insurers more than $24 billion in 2017 and 2018. One insurer, Merced Property & Casualty Co., was pushed to insolvency and liquidated by the state following the 2018 Camp Fire.
“Insurers are acting as you would expect where the risk is going up. They’re adjusting their pricing and deciding in some cases that the risk is too high,” said Dave Jones, senior director at the Nature Conservancy and a former California insurance commissioner. But without drastic steps to address climate change, he said, “we’re marching steadily towards a future where the wildfire risk is uninsurable.”
Catastrophe models are sold to insurers by vendors including RMS, a unit of Daily Mail & General Trust PLC, and AIR Worldwide, a unit of Verisk Analytics Inc. Insurers also use products like Verisk’s FireLine, which assigns properties a score of zero to 30 based on their wildfire risk.
In California, insurers cannot use catastrophe models when setting statewide average rates for standard home-insurance policies, which include wildfire coverage. State regulations require insurers to rely on historical-loss data when calculating expected future losses.
California’s insurance regulations are more stringent than those of most states. Insurers need to get approval before they can change the rates they charge, and rate filings are made public.
Though an insurer’s average California home-insurance rate has to be based on historical data, the insurer can use catastrophe models to determine how to price policies for certain regions or segments of the market. Insurers also can use these models when deciding whether or not to insure a specific property.
But the insurance industry says its pricing would be more accurate if it could use the forward-looking models when setting overall rates. Catastrophe models run tens of thousands of simulations to project a range of possible scenarios.
Insurers say models provide a more comprehensive view of risk than historical data and are able to take climate change into account.
“We’re told there’s a new normal” due to climate change, which regulators and scientists have said increases wildfire risk, said Rex Frazier, president of the Personal Insurance Federation of California. But under current regulations, “we cannot use probabilistic modeling that attempts to build a rate based upon where models suggest losses will go.”
Consumer advocates say the models don’t give homeowners and communities enough credit for taking steps to mitigate their wildfire exposure through brush-clearing and other efforts.
“They want to have a black-box model and not let the public or advocates see what’s in it,” said Jamie Court, president of Consumer Watchdog.
Patty Swenson of Montecito, Calif., was told earlier this year that the insurance policy for her five-bedroom home wouldn’t be renewed. Her insurance agent said the insurer had cited the house’s proximity to a forest as a reason, Ms. Swenson said.
But the forest near Ms. Swenson’s home burned in the 2017 Thomas Fire, she said, which she thinks makes it less likely to burn again. Ms. Swenson added that her property is cleared of flammable brush and that the Montecito community has invested in safety measures to protect it from fire-related debris flow.
Ms. Swenson was able to find another insurer, but her new coverage limit is lower, leaving her underinsured, she said.
“Everything we’ve done has further mitigated the risk of a future disaster,” Ms. Swenson said. “That should count for something.”
One concern for regulators is that until recently, wildfire models were less sophisticated than hurricane and earthquake models. Many insurers were shocked by the extent of the damage in 2017 and 2018 and said the previous models didn’t predict those losses.
Both AIR and RMS have released new wildfire models in the past two years that they say capture a wider range of factors and are more predictive than past models.
“We’re hopeful that the regulators are open to hearing about new modeling technologies,” said Michael Young, vice president of product management at RMS.
But different models might still reach different conclusions about the risk facing a certain property or book of business. “Any model that is used in a rate filing would have to be vetted through a public process,” said Joel Laucher, special consultant for the California Department of Insurance.
The state’s insurance regulators are “balancing the risk that insurers will nonrenew customers against the risk that customers will be asked to pay more than they can afford,” he said.