Panel suggests ways to fine-tune “black box” insurance models
In a twist on the old saying “garbage in, garbage out,” a panel of experts predicts that property insurance rates could change dramatically in South Carolina if insurers feed more-accurate data into their secret computer models.
The panel’s finding is part of a preliminary report submitted recently to the state Department of Insurance. The department released a partially redacted summary late Friday.
The department hired the experts in the wake of The Post and Courier’s series, “Storm of Money,” which showed how insurers use controversial computer models to calculate potential losses in a hurricane. Based on those calculations, insurers have sought higher rates in South Carolina and other parts of the country.
Companies that produce the models guard them carefully to prevent competitors from stealing their algorithms, which also makes it difficult for regulators to determine whether insurers are gaming them to boost profits at the expense of ratepayers.
Ray Farmer, director of the agency, said the department’s goal was to make sure insurance companies and their models are assessing potential risks as accurately as possible. He said he didn’t see any major surprises in the panel’s preliminary report but added that, “it’s still a work in process.” He said the panel is doing additional work to finalize the study. “But I wanted to get this out to be as transparent as possible,” he said. He said a public hearing about the report’s findings will be held Oct. 9.
The panel included Martin Simons, a noted actuary, Jennie Evans, a meteorologist, and Masoud Zadeh, a structural engineer. All three were members of a group that examined Florida’s catastrophe models, in what is considered the most intensive look yet at how models affect rates.
In the highly technical report, the panel took great pains to say their analysis was not about whether rates were too high or low. Rather, their goal was to help regulators better understand the inner workings of catastrophe models so they can make better decisions for ratepayers and insurers alike.
The panel did find potential flaws, recommending that insurers avoid plugging in data about tropical storms and depressions, which might skew results. The panel also was critical of insurers who use “short-term” models, which are designed to predict losses in the next five years. Critics have said these models overstate potential losses and that insurers have wrongly used them to jack up rates. The panel also recommended that insurers feed into their models better data about whether buildings have been built to stricter engineering codes. Putting more accurate information into the models, “would result in substantial and systematic variations of insurance rates across South Carolina,” the report said.