To Our Clients and Friends,

In our last monthly edition of the Insurance Industry Corporate Governance Newsletter we focused on the growing role of shareholder activism as a force in the insurance sector. 

In this edition, we look at an aspect of diversity, equity and inclusion (“DEI”) in insurance—specifically, how the insurance industry and insurance regulators are focusing on unfair discrimination (including proxy discrimination) in insurance underwriting and rating.

Unfair Discrimination and Insurance Consumers—Restricted or Prohibited Factors

Laws against unfair discrimination prohibit disadvantaging people on the basis of certain protected characteristics, and “proxy discrimination” generally refers to the use of an otherwise non-prohibited, facially neutral variable as a “proxy”—or stand-in—for a protected class characteristic. Regulators have been more closely scrutinizing apparently neutral attributes. In our February edition of this newsletter here, we focused on how the growing use of artificial intelligence (“AI”) and machine learning (“ML”) by insurers is bringing such issues into sharper focus and raising new issues.

Relatedly, we have discussed and predicted regulatory developments and trends regarding insurance, Big Data, and AI, including in our webcasts here and here.  In our recent blog post here, we discuss the highly significant Notice by the Connecticut Insurance Department titled “The Usage of Big Data and Avoidance of Discriminatory Practices” that requires regulated entities to certify compliance by September 1, 2022 and annually thereafter -- potentially positioning Connecticut as the new forerunner in AI insurance regulation in the United States.

Much of the recent DEI discussion is centered on race, but the work by regulators and industry also relates to other historically underrepresented groups, including by seeking to improve insurance access and affordability to underserved (both uninsured and underinsured) populations.  

We noted previously that credit scores and credit-based insurance scores have drawn particular scrutiny as potential proxies for protected characteristics, including race. Many states have longstanding laws restricting the use of credit scores, primarily in underwriting or rating private passenger automobile and homeowners insurance.  Examples of such restrictions include whether an insurer’s decision may be made solely on the basis of credit, whether it may cancel or nonrenew a policy based on credit, how the absence of credit may be viewed, what disclosures must be made and rules regarding dispute resolution. Some states have enacted such laws more recently or are revisiting their rules.  

Due to the COVID-19 pandemic’s dramatic impact on people’s livelihoods, regulators’ concerns with the use of credit have increased. Some states enacted emergency measures in an attempt to stop the chain reaction of negative impacts, such as moratoria against evictions and banning negative reporting to credit agencies. There were similar efforts in New York following Superstorm Sandy.

Specific to insurance, as we noted in our client alert here, the New York Department of Financial Services (“NYDFS”) is investigating the use of credit by authorized insurers writing private passenger automobile, commercial automobile and homeowners insurance in the State of New York, seeking extensive and detailed information about their use of personal credit for underwriting and ratings. NYDFS requires companies to submit information regarding their credit scoring models, initial tier placement algorithms, consumer notifications and interactions, data analysis, and general views and observations regarding the use of personal credit scores as a predictive tool.  

When presenting at the Insurance Federation of New York Annual Award Luncheon on March 31, 2022, New York Superintendent of Financial Services Adrienne A. Harris mentioned the investigation and how NYDFS engaged with advocates to help frame the questions and data gathering process. Superintendent Harris emphasized that she has zero tolerance for discriminatory practices and said that regulation or partnering with the New York legislature for new laws may be necessary to address this issue. 

This is reminiscent of NYDFS’s investigation years ago into private passenger auto insurers using policyholders’ education or occupation in placing the policyholders in particular tiers in multi-tier programs affecting the rates they paid. That resulted in NYDFS promulgating a regulation effectively prohibiting an insurer from using a policyholder’s education or occupation unless it demonstrates that doing so would not result in a rate that is excessive, inadequate or unfairly discriminatory.

What Companies Should Do Now

The issue of unfair discrimination, including proxy discrimination, in underwriting and rating factors is highly likely to continue heating up. Certain regulators have indicated that they view existing laws as providing a sufficient basis for enforcement actions if they find unfair discrimination, including if it results from the use of AI, ML or algorithmic models.  

Using certain criteria in underwriting and rating appears riskier, however, even if actuarially predictive of loss costs. An insurer showing a nexus between a factor and the risk (such as a reasonable explanation of causality) is generally helpful to justify its use of the factor. When that connection is less obvious, the criterion is more likely to be considered suspect by regulators, particularly if it results in any disparate impact. Therefore, we see a trend to shift the burden onto insurers to disprove, or otherwise be able to demonstrate, that any given attribute used in the underwriting or claims process (especially a novel one) is not a proxy for an otherwise prohibited characteristic.  

Both credit scores and credit-based insurance scores have drawn scrutiny as potential proxies for protected characteristics, with potentially disparate negative effects on underrepresented groups. Further restricting or prohibiting the use of credit as a factor in underwriting and rating therefore seems like a high priority target for legislative or regulatory action.

As a matter of good corporate governance, companies should take this emerging risk into consideration as they evaluate their existing business models or consider entering into new areas. Further, as we have highlighted previously, using an input for an AI or ML model that is later found to be problematic can be very difficult and costly to unwind.  

Conclusion

Industry, regulators and legislators are working to prevent unfair discrimination in insurance. There is a lot of work to be done in the area, and significant differences of opinion (e.g., on the definition of “proxy discrimination”), but regulators might be viewing some changes as lower-hanging fruit.

Certain underwriting and rating criteria that have been historically used are being questioned, particularly those that show less (or no) apparent causal link to the risks. Further, certain criteria (e.g., education, occupation and credit) have significant disparities such that using them in underwriting and rating potentially disproportionately negatively affects historically disadvantaged groups. This likely increases the burdens on insurers to disprove that any unfair discrimination results to defend their use; otherwise, their use may be restricted or prohibited by legislation, regulation or in states’ rate review processes. 

One might argue that using variables over which people have control (e.g., driving behavior as tracked by telematics or rate reductions for engaging in healthy behaviors) results in greater fairness and influences insureds toward better behavior to reduce risks (e.g., auto accidents or reducing morbidity and mortality for life or health insurance).

No matter which point of view one favors, there is no doubt that traditional attributes and actuarial approaches to underwriting have been placed under serious scrutiny, creating risk to traditional business models and meriting attention by those responsible for overseeing insurance businesses.