How Can We Uncover Bias and Discrimination in Insurance Underwriting, Rating and Claims Practices?
Our world is familiar with the cries for a more diverse and inclusive society. In turn, the commercial insurance and risk management space has long heard these cries directed at their doors from without and within.
Efforts to foster a more inclusive, equitable environment seem sincere, but despite these efforts, the industry has been criticized for allowing systemic racism to seep into the everyday practices of underwriting, rating and claims. The proof of that disparity, critics say, are in premium rates in low-income areas that are higher than in more well-off zip codes.
The question at hand, then: How can risk professionals not only become more aware of underwriting bias but also play a vital role in dispelling it?
This was the question pondered, discussed and hopefully answered at the 2022 RIMS annual conference session “Defining Bias and Discrimination in Insurance: What Risk Professionals Need to Know.”
The session, held April 12, featured speakers Mallika Bender, DE&I staff actuary, Casualty Actuarial Society; and Claire Howard, senior vice president of general counsel and corporate secretary for the American Property Casualty Insurance Association.
Utilizing data and findings from Casualty Actuarial Society’s research “Race and Insurance Pricing,” the session looked at various terms used regularly when determining rating plans for risk portfolios, how the different forms of bias are impacting consumers and how to make this confusing landscape much clearer.
Understanding the Terms
The speakers said the insurance industry has some catching up to do when it comes to ensuring fair and adequate rate pricing.
But to even begin this conversation, what first has to be acknowledged is the use of varying language regarding bias and discrimination and how confusing that can be.
One term that took center stage throughout the session was “proxy discrimination,” which Bender said sees much debate about its true impact on the industry.
Proxy discrimination is linked closely to another term used in this discussion — disproportionate impact.
From the American Academy of Actuaries, “disproportionate impact” relates to “when a rating tool results in higher or lower rates on average for a protected class,” which is a “group of people who share a common characteristic for whom federal and state laws have created protections that prohibit against discrimination because of that trait,” according to Bender.
These types of characteristics could include race, religion and natural origin.
So while disproportionate impact looks at how this type of incident could occur, proxy discrimination intends to look at why it occurred.
Bender described proxy discrimination as “using potentially neutral characteristics to stand in for other variables.”
“If one characteristic is correlated with another, [perhaps] you can use this [trait] as a sort of proxy for the other [trait], for the purposes of prejudicing for a certain group,” Bender said.
While this definition seems easy enough to understand, there are several utilizations of the term by multiple stakeholders, which creates confusion and deters alignment on the issue at hand, which is bias and discrimination.
Bender said three factors play into why it’s so difficult to align on a central definition of proxy discrimination: intent, proxies and enforceability.
One particular and well-known form of proxy discrimination is redlining, which is “the practice of determining mortgage loan eligibility based on characteristics like the property, location, and characteristics of the borrower,” per Bender.
Another term discussed in the session was “disparate impact,” which is a practice that will impose a discriminatory impact on a protected class.
In sorting through these terms and definitions, taking action on underwriting inequity can get lost in the shuffle.
Bender said, “When we’re all talking about whether or not proxy discrimination is happening, but we can’t agree on what it actually is, we don’t move forward.”
It’s Confusing, But It Doesn’t Have to Be
Following Bender’s presentation of the various definitions used in this context, Howard addressed the confusion around how bias manifests and what to do about it.
As she said, “The landscape is confusing, but it doesn’t have to be.”
Howard went on to say that these terms ultimately fall into two categories: Those that have legal definitions and those that have been given new definitions by stakeholders. That lack of uniformity creates issues when attempting to pinpoint instances of bias and discrimination.
From a legal perspective, Howard said the use of proxy discrimination was so concerning, because it could be used to cover up “intent to exclude a protected class.”
Using a hypothetical example of evaluating the risk factors of where a consumer lives, works and drives, Howard laid out questions that regulators and actuaries could ponder to ensure that rates were not being unfairly priced. The questions included:
- Do minority policyholders living, working and driving in densely populated urban areas produce more adverse outcomes, or higher rates, than majority policyholders living, working and driving in the same area?
- Is there a valid interest being served by that risk factor?
- Is there an equally effective alternative to using where a policyholder lives, works and drives as a risk factor?
While this was a specific hypothetical situation, it painted a picture and, perhaps, a detailed framework for how regulators and actuaries could ensure the risk factors they are using to determine rate prices are not intentional or unintentional discrimination.
Moving forward, as every U.S. state has its own insurance regulations and codes, Howard said, “The question becomes, ‘Can adverse outcomes for protected classes be addressed within the existing statutory writing?’ ”
If the answer is no; regulation might have to return to the rate pricing drawing board. &