Insurers aren’t pricing the real risk of wildfires in California, study finds

A home burns as the Camp Fire rages through Paradise, Calif on Nov. 8, 2018
A home burns as the Camp Fire rages through Paradise, Calif on Nov. 8, 2018. (AP Photo/Noah Berger, File)

The historic drought and warmer temperatures are stoking fears of an early wildfire season. Areas in the state scorched by wildfires increased fivefold from 1979 to 2019. The following year, the burn area more than doubled.

And a new United Nations report warns of a “global wildfire crisis” and predicts that extreme wildfires could increase by 57% by the end of the century.

With fire comes the threat of property loss for homeowners.

A team of Berkeley Haas researchers warn that despite this clear upward trajectory, the risks posed by wildfires are worse than we—or, at least, the insurance and mortgage markets— are willing to account for.

An analysis by Professors Nancy Wallace and Richard Stanton, along with two PhD alumni, suggests the financial firms whose insurance products help protect homeowners from the financial devastation of fires could soon find their own businesses financially wrecked, unless their risk models and pricing (and the government regulations overseeing both) undergo dramatic changes.

“We have a problem, and the political will to deal with this is not even trying to catch up with the speed of the rapidly changing risks,” says Wallace, the Lisle and Roslyn Payne Chair in Real Estate and Capital Markets, chair of the Berkeley Haas Real Estate Group, and co-chair of the Fisher Center for Real Estate and Urban Economics.

Wallace and Stanton, along with Paulo Issler, PhD 13, MBA 98, director of the Haas Real Estate & Financial Markets Lab, and Carles Vergara-Alert, MFE 04, PhD 08, now a professor at Spain’s IESE Business School, teamed up with physicists from the Lawrence Berkeley National Laboratory (LBNL) who study the fluid dynamics of fire. In a unique collaboration, they linked the physicists’ sophisticated measurement models—using satellite and radar data—to the  comprehensive Fisher Center for Real Estate and Urban Economics’ real estate and mortgage-record data. This allowed them to forecast the risks posed by wildfires to lenders and insurers in California.

The researchers found that their site-specific estimates of wildfire risk in the state were quite different from the risk maps developed by the California Department of Insurance. This difference was most pronounced in the zones marked “zero-risk” on the state’s maps, because the researchers found that there was some level of risk in many of those places.

Separately, the researchers found that neighborhoods damaged by wildfires tend to return as gentrified—populated by larger and more expensive homes and residents with higher wealth than those outside of the burn area. That’s because the insurance industry incentivizes bigger and more expensive rebuilds, they noted.

But the system is not sustainable.

Mapping dynamic risks

As economists focused on the mortgage market, Wallace says the researchers are accustomed to working with static maps, which are “imperfect measures of the physical space because they’re so infrequently updated.”

The granular, digitized measurements over a 1.5-by-1.5 kilometer grid developed by the LBNL physicists are, on the other hand, based on dynamic hourly data. They incorporate information about the locations of thousands of California wildfires between 2000 and 2015, as well as meteorological factors including direction and speed of wind, humidity levels, and temperatures. The Haas team also considered a location’s slope, elevation, and vegetative density.

The Haas team mapped the owner-occupied, single-family residences included in a database that covers properties in California between 2000 and 2018. Their dataset included information about each property’s size and price, as well as household characteristics like wealth, income, and mortgage performance.

Building back bigger

The models they developed suggest that wildfire risk in California is more dispersed than the state’s static maps suggest. In fact, they found that important risks to housing stock falls into what the state labels as “zero-risk” zones which contain large numbers of single-family homes. In contrast, their predictive models estimate there is some risk of wildfire in most of these areas.

They also show that five years after a fire, the rebuilt houses are larger and more expensive than those outside of the burn area. Five years later, the houses in the post-fire areas have 3.4% higher prices and are larger than those outside the burn area. This is partly because new homes are required to be built to code, which often means improvements in safety and quality over the previous housing stock. Insurance payouts enable owners to invest heavily in a rebuild, and because everyone in the area is doing the same thing, the whole area increases in value.

Of course, these incentives rely on the existence of a healthy property insurance industry. And that, the researchers say, may be threatened by the patterns highlighted by their predictive models.

From deterministic to probabilistic

Insurers are canceling policies at a growing clip. In the past three years, there’s been a 31% increase in policy cancellations, Stanton says. In 2020 and 2021, he adds, California insurers lost nearly two years of premia. “They can’t sustain providing insurance in this state unless there’s a policy response,” he says.

The problem, the researchers argue, is that the insurers are relying on deterministic models of wildfire risk, based on where fires have happened, rather than probabilistic models that predict incidence of fires. The insurers have no choice—the California Department of Insurance (CDI) requires them to price based on deterministic maps. The researchers argue that the CDI policy needs to change so insurers can base their risk and pricing on probabilistic models.

California regulators also prohibit insurers from using reinsurance margins in the rate structure, which is insurance to cover tail risks—in other words, extreme events. In recent years insurers that offer financial protection from hurricanes and earthquakes have been relying on the reinsurance market. Introducing reinsurance would likely raise customers’ premia, so the researchers propose that the solution could involve subsidies for people who can’t afford the price hikes. The new structure should also shift the current incentives.

“If insurance products really reflected the risk, it would be much more costly, and homebuyers would have a decision to make,” Wallace says. “‘Do I want this home enough to pay these premia, and to take this risk with my life?’ Right now, the real risk isn’t priced in accurately enough for people to understand what their exposures are.”

Wallace has personal experience with those trade-offs: She was one of thousands of people who fled the 1991 Oakland Hills Firestorm 30 years ago, which killed 25 people, injured about 150, and burned 2,900 homes.

As climate change continues to take its toll, the structure of the insurance market will have to change. That may mean a shift from today’s indemnified insurance, which carefully covers losses, to what’s called parametric insurance, which pays out based on the occurrence and magnitude of a specific weather event.