Property data emerges as key element of California wildfire risk model debates

di-google-drone-042319.jpg
An Alphabet Inc. Google X Project Wing delivery drone flies during a demonstration in Blacksburg, Virginia. Aerial photography is one source of data for property intelligence used by insurers.
Charles Mostoller/Bloomberg

Property intelligence technologies have become a critical element of the risk models being debated by carriers and consumer advocates for California Department of Insurance (CDI) consideration of how to regulate the handling of wildfire risk.

Birny Birnbaum - LI credit.jpg
Birny Birnbaum, director of the Center for Economic Justice.
LinkedIn

The property intelligence used in insurers' catastrophe and risk models at issue in California and throughout the U.S. can come from aerial photography or granular terrain maps, and generates a "tremendous amount of data," said Birny Birnbaum, director of the Center for Economic Justice.

"You can see virtually everything on the property," he said. "You can get detailed information about the slope, and risks that are close to the property or alleged risks that are close to the property."

StephanieDalwin_Aite-Novarica.jpg
Stephanie Dalwin, advisor, Datos Insights.

Property intelligence is often evaluated using AI, according to Stephanie Dalwin, advisor at Datos Insights, a consultancy which has been researching the subject. "A lot of these platforms are obviously incorporating AI to quantify the risk or potential claim severity associated with different properties and property characteristics," she said. "Maybe AI could help understand these unforeseen risks."

The basics of property intelligence for wildfire risk include data points such as the distance between a property and brush or overhanging trees, and where the "wildlife-urban interface actually starts compared to your property line," said Dan Dick, executive managing director and global head of property analytics at Aon, a consulting firm that specializes in risk mitigation products.

Dan Dick - Aon.jpg
 Dan Dick, executive managing director and global head of property analytics at Aon.

CDI has allowed some insurers to set wildfire hazard scores for underwriting, according to Dick. These scores, produced in part with AI and captured imagery of properties, are shared with homeowners and insureds. 

"It's a way that an underwriter can get their hands on information that might not otherwise be available or might not be readily available through even a property application," he said. 

Property intelligence data and evaluations for underwriting coverage for wildfire damage — and related technological advances — could be used to mitigate damage before it occurs, he added. In California, the federal and state governments may be the first to look at this application to mitigate wildfire damage, because they both have significant land holdings of their own in the state, Dick believes.

Dalwin raised the "black box" issue. Insurers should keep in mind that AI systems cannot be used just as "some sort of weird ether that all your data is getting thrown into, and you can't explain why decisions are being made," she said. "There's going to be some sort of auditability and explainability coming down the pipeline."

For reprint and licensing requests for this article, click here.
Artificial intelligence Machine learning Big data Homeowners insurance California Predictive modeling Data modeling
MORE FROM DIGITAL INSURANCE