After All These Years, States Still Don’t Understand Credit Scoring
You may have missed it, as some bigger events have been dominating the headlines in recent weeks, but the West Virginia Legislature and Maryland General Assembly both recently adjourned after having considered bills (SB 659 and SB 17, respectively) that would have significantly restricted the use of credit in personal lines insurance. Neither bill passed.
Similar legislation has been introduced in the Oklahoma State Legislature, which technically remains in session until May 29, though the state Senate was forced to go on lockdown March 17 after a member tested positive for the COVID-19 virus. And, of course, in the U.S. House, Rep. Rashida Tlaib (D-Mich.) has introduced H.R. 1756, which would effectively bad credit-based insurance scoring nationwide.
The trend is not new. Bills to prohibit insurers from using credit in underwriting and rate-setting, placing insureds in operating companies with preferred rates or awarding discounts for good credit have been considered in most states over the past 20 years.
How credit is used in property and casualty insurance is not widely understood—even by insurance agents—so restricting its use can seem like a reasonable public policy. In fact, in 2006, Bob Hunter, director of insurance for the Consumer Federation of America and the former Texas insurance commissioner, claimed: “We are beginning to see a turn in the direction of getting rid of (insurance credit scoring), or at least putting more controls on it.”
What sold insurers on using these black box scoring formulas was that they really produced good loss predictability, particularly when married up with driving records in states that kept good records. This became the new gold standard in auto-insurance underwriting. But it’s useful to remember that part of what interested the companies is that they were also, in some sense, the fairest systems devised to date. The formulas that were developed didn’t depend on income, ethnicity, where you live, cultural background, gender, age, sex or any of the factors whose use people resent because they are considered “unfairly discriminatory.”
Instead of being a proxy for sorting drivers and homeowners into classes or conditions they couldn’t change, it was argued that these scoring formulas were only a proxy for responsibility. Most of the state laws that were passed to restrict their use focused on areas where responsibility was compromised by things one couldn’t control, like crushing medical bills.
Auto and property insurance prices are always just sophisticated guesses, because the actual cost of the product is never known when the price is calculated. Companies that sell these products spend years working on their rating engines, because guessing low could bankrupt them and guessing high would push sales out to their competition. So, they work diligently to match risk profiles with the prices they charge. With these formulas, the odds that a relationship between the credit score and relative loss ratios does not exist for a given random sample of policyholders is usually in the range of 500-to-1, 1,000-to-1 or even 10,000-to-1, according to industry information provided to regulators.
A common misconception is that credit scoring by insurers is analogous to credit reporting used in banking, where the product is an assessment of capacity to pay back a loan. Instead, by using up to 50 different elements of how people manage their financial affairs, insurers and the vendors who craft these models figured out that aggressive use of credit often indicates aggressive use or even overutilization of an insurance policy, compared to other customers. Extensive use of no interest for first-year purchases and dozens of other indications point irrefutably to customers who are more likely to want a new carpet because of dripped candle wax or who refuse to accept remanufactured wheels for their Porsche which they raced into a ditch. “More likely” is the key phrase here, but if it’s 1,000 times more likely, the insurer has enough predictability to figure out charges with some confidence.
Before any state government, much less the federal government, enacts any new prohibitions, it might be useful to ask the companies what they will do instead to predict the eventual losses, and whether that might be preferred by most of their customers.
