This is a tooltip for the edit command button


In recent discussions about Big Data, artificial intelligence, and the use of predictive modelling – most notably at the meetings of the National Association of Insurance Commissioners – the terms “unfair discrimination” and “disparate impact” have been used almost interchangeably.  Each is a term of art particular to context.  Under insurance law, “unfair discrimination” occurs when similar risks are treated differently.  “Disparate impact” is not an insurance term, but rather, it is a method of proving unlawful discrimination against members of a protected class, but only under certain circumstances. 

Insurance regulators have authority to prohibit unfair discrimination by making sure similar risks are treated the same, and they have authority to prohibit discrimination on the basis of certain factors.  Whether regulators have authority to rely upon a disparate-impact theory to prove discrimination based on such factors is uncertain given the language of state insurance laws and the long line of Supreme Court decisions on disparate impact.  This article explains the differences and explores whether insurance regulators have the requisite statutory authority to use a disparate impact theory of liability in enforcing insurance laws.  For simplicity’s sake, the analysis is limited to private passenger automobile insurance.[1]

What is the Difference?

The term “unfair discrimination” has a special meaning in state insurance laws and is unrelated to “disparate impact”, a legal theory of liability first adopted by the U.S. Supreme Court in 1971 in an employment discrimination case.  Disparate impact is a method of proving discrimination when there is no overt discrimination against members of a protected class.

Under state insurance laws, “unfair discrimination” occurs when similarly situated risks are treated differently.  Unfair discrimination laws are distinct from other insurance laws that prohibit consideration of certain factors, such as race, national origin, and religion.  Some have urged insurance regulators to adopt a disparate impact theory to enforce such insurance laws, but that may not be as simple as it sounds.   

Unfair Discrimination in Insurance

Insurance laws in every state prohibit “unfair discrimination” in rates, coverages, benefits, terms and conditions of insurance policies.  Under those laws, unfair discrimination occurs when similar risks are treated differently.  Unfair discrimination is defined in the Arizona insurance code.

As to kinds of insurance other than life and disability, a person shall not make or permit any unfair discrimination in favor of particular persons or between insureds or subjects of insurance having substantially like insuring, risk and exposure factors, or expense elements, in the terms or conditions of any insurance contract, or in the rate or amount of premium charged.[2]

Ohio prohibits similar conduct.

Making or permitting any unfair discrimination between individuals of the same class and of essentially the same hazard in the amount of premium, policy fees, or rates charged for any policy or contract of insurance, other than life insurance, or in the benefits payable thereunder, or in underwriting standards and practices or eligibility requirements, or in any of the terms or conditions of such contract, or in any other manner whatever.[3]

Unfair discrimination in insurance has nothing to do with protected classes or prohibited factors.[4]  Unlike federal and state anti-discrimination laws, unfair discrimination laws do not identify specific prohibited factors or protected classes of persons.  The term “unfair discrimination” is often described as an actuarial principle, and Principle 4 of the Casualty Actuarial Society’s Statement of Principles Regarding Property and Casualty Insurance Ratemaking provides:

Principle 4: A rate is reasonable and not excessive, inadequate, or unfairly discriminatory if it is an actuarially sound estimate of the expected value of all future costs associated with an individual risk transfer. 

In other words, since premiums should be tied to the expected risk, similar risks should pay the same.  This actuarial principle tracks statutory standards for insurance rates – that rates not be excessive, inadequate, or unfairly discriminatory.  Under the insurance laws, what is “unfair” is treating similar risks differently.   

Prohibited Discrimination under State Insurance Laws

Insurance is inherently discriminatory by nature.  With the permission of their regulators, insurance companies discriminate among different risk profiles.  They may not, however, discriminate on the basis of certain specified factors.  Depending on the state, factors such as race, national origin, and religion may not be used or considered in establishing risk classifications or in underwriting decisions.[5]  Even if such factors are predictive of loss, it is against the public policy of the state for insurance companies to consider them.

The following are examples of insurance laws that prohibit discrimination on the basis of certain factors. 


The declination, cancellation or nonrenewal of a policy for private passenger nonfleet automobile insurance is prohibited if the declination, cancellation or nonrenewal is based: (1) On the race, religion, nationality or ethnicity of the applicant or named insured; …[6]


(a) A person may not refuse to insure or provide coverage to an individual, refuse to continue to insure or provide coverage to an individual, limit the amount, extent, or kind of coverage available for an individual, or charge an individual a rate that is different from the rate charged to other individuals for the same coverage because of the individual's:

(1) race, color, religion, or national origin…[7]

North Carolina

No insurer shall refuse to insure or refuse to continue to insure an individual; limit the amount, extent, or kind of coverage available to an individual; or charge an individual a different rate for the same coverage, because of the race, color, or national or ethnic origin of that individual.[8]

The concern is about underwriting and rating practices that do not use any prohibited factors but which might result in an adverse impact on a protected class of persons.  A key question in the debate about the use of Big Data, artificial intelligence, and predictive modelling that is often ignored is whether insurance regulators have legal authority to take action against an insurance company based on a showing of a disparate impact on a protected class.

Insurance laws prohibiting discrimination on the basis of certain factors may be less prohibitive than is commonly assumed.  Regulators and the industry take it for granted that race, for example, is a prohibited factor that may not be used in personal lines underwriting and rating.  While that may be true with respect to risk classifications, when it comes to decisions about whether to issue or continue a policy the laws of 15 states only prohibit race from being the sole factor.[9]  Even more startling, perhaps, is that the insurance laws of four states do not contain restrictions on the use of race as an underwriting or rating factor for personal automobile insurance.[10]  This is not to suggest that regulators permit the use of race, and there may be non-insurance laws that apply.  The point is simply to demonstrate that anti-discrimination provisions of state insurance laws are not necessarily as robust as one may believe and, as discussed below, may not support a disparate impact theory of liability.

Disparate Impact

The United States Supreme Court has recognized two ways to establish illegal discrimination against protected classes.  The first is “disparate treatment” - which is established by showing that an actor intends to treat a protected class of persons differently from non-protected classes.  The second way to establish illegal discrimination is to show that a business practice has a “disparate impact” or what the Supreme Court has called a “disproportionately adverse effect” on a protected class.[11]

The Civil Rights Act of 1964 (“Civil Rights Act”) established “race, color, religion, sex, and national origin” as protected classes.  It was an employment case brought under Title VII of the Civil Rights Act that resulted in the Supreme Court establishing disparate impact as a theory of liability when employment practices are facially neutral but disproportionately impact a protected class.  The relevant portion of Title VII provides:

(a) Employer practices

It shall be an unlawful employment practice for an employer -

(1) to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual's race, color, religion, sex, or national origin; or

(2) to limit, segregate, or classify his employees or applicants for employment in any way which would deprive or tend to deprive any individual of employment opportunities or otherwise adversely affect his status as an employee, because of such individual's race, color, religion, sex, or national origin.

In Griggs v. Duke, widely referred to as the first disparate impact case, an employer required a high school diploma or a skills test for certain jobs.  It was shown that the requirement was applied equally to all races and that there was no racial purpose or invidious intent.  The Supreme Court determined that the requirement was not related to job performance and that the requirement disproportionately made minorities ineligible for such jobs.  The court said that “good intent or the absence of discriminatory intent” was not enough to save requirements that operate as “built-in headwinds for minority groups and are unrelated to measuring job capability.”[12] The court held that, in enacting Title VII, Congress was focused on the consequences of employment practices, not just an employer’s motivation, and that Congress proscribed “not only overt discrimination but also practices that are fair in form, but discriminatory in operation.”[13]

Since the Griggs case, the Supreme Court has explored the appropriateness of “disparate impact” under other federal laws, and has held that a disparate-impact theory of liability is available under the Age Discrimination in Employment Act (“ADEA”) as well as the Fair Housing Act (“FHA”).[14]  

Today, there is a significant body of case law on disparate impact and much can be said about how disparate impact cases are proven, including how statistics are used and what constitutes an actionable disparate impact, but it is important to understand that it is not enough to merely show a disparate impact on a protected class.  The plaintiff must prove the challenged business practice caused the disparity and, when the defendant establishes a business justification for the practice, the plaintiff must prove there is an alternative practice that serves the business’ needs with less disparate impact.[15]  In 2015, the Supreme Court discussed the limitations on disparate-impact liability and reaffirmed the causality requirement of causality.[16] 

Statutory Language Matters

The Supreme Court has made it abundantly clear that the language of a statute is fundamental to determining whether a disparate-impact theory of liability is permissible.  

Together, Griggs holds and the plurality in Smith instructs that antidiscrimination laws must be construed to encompass disparate-impact claims when their text refers to the consequences of actions and not just to the mindset of actors, and where that interpretation is consistent with statutory purpose. These cases also teach that disparate-impact liability must be limited so employers and other regulated entities are able to make the practical business choices and profit-related decisions that sustain a vibrant and dynamic free-enterprise system. And before rejecting a business justification—or, in the case of a governmental entity, an analogous public interest—a court must determine that a plaintiff has shown that there is “an available alternative . . . practice that has less disparate impact and serves the [entity’s] legitimate needs.”[17] (emphasis added)

The FHA contains language similar to Title VII, making it unlawful to “refuse to sell or rent…or otherwise make unavailable or deny, a dwelling to a person because of race…”[18]  In Texas Dept. of Housing & Community Affairs v. Inclusive Communities, the Supreme Court focused on the phrase “or otherwise make unavailable” in comparing the FHA, the ADEA, and Title VII, finding that such provisions in each of those laws refer to the consequences of an action rather than the actor’s intent.

Title VII’s and the ADEA’s “otherwise adversely affect” language is equivalent in function and purpose to the FHA’s “otherwise make unavailable” language. In these three statutes the operative text looks to results. The relevant statutory phrases, moreover, play an identical role in the structure common to all three statutes: Located at the end of lengthy sentences that begin with prohibitions on disparate treatment, they serve as catchall phrases looking to consequences, not intent. And all three statutes use the word “otherwise” to introduce the results-oriented phrase. “Otherwise” means “in a different way or manner,” thus signaling a shift in emphasis from an actor’s intent to the consequences of his actions...[19]

The potential problem for insurance regulators is that insurance laws that ban the use of particular factors, such as race, national origin, and religion do not have the “otherwise adversely affect” or “otherwise to discriminate” language that has been the basis of the Supreme Court’s willingness to allow liability based on disparate impact.  Instead, state insurance laws use language such as “based wholly or partially on” or “because of” and do not contain the word “otherwise.”  It is not at all clear that a “disparate impact” theory would be available to insurance regulators.  


As the discussions about Big Data, artificial intelligence, and predictive modelling continue, the insurance industry and regulators should be careful to use precise language and avoid referring to unfair discrimination when the concern is about illegal discrimination.  Language matters.  Insurance regulators have clear authority to prohibit unfair discrimination, but whether they may use a disparate-impact theory to prove illegal discrimination is a completely different issue.


[1] State laws, other than insurance laws, were not reviewed for potential applicability to insurance underwriting and rating practices.

[2] Ariz. Stat. § 20-448 (C)

[3] Ohio Rev. Code § 3901.21 (M)

[4] In addition to having general unfair discrimination laws, a few states also have laws that provide that rates based on certain factors, such as race and sex are, by definition, unfairly discriminatory.  See e.g., Tenn. Code § 56-5-302, Wyo. Stat. § 26-14-103

[5] See e.g., NAIC Property and Casualty Model Rating Law (#1775, 1776 and 1780) (prohibiting risk classifications on basis of race, creed, national origin or religion); NAIC Automobile Insurance Declination, Termination, and Disclosure Model Act (#725) (prohibiting declinations and terminations based upon race, religion, nationality or ethnic group); NAIC Unfair Trade Practices Model Act (#880) (prohibiting insurers from refusing to insure, refusing to continue to insure, or limiting the amount of coverage available to an individual because of the sex, marital status, race, religion or national origin of the individual).

[6] Conn. Gen. Stat. § 38a-358

[7] Tex. Ins. Code § 544.0002

[8] N.C. Gen. Stat. § 58-3-25

[9] Ariz. Rev.Stat.  § 20-1631; Ark. Code § 23-66-206; Colo. Rev. Stat. § 10-4-626; Fla. Stat. § 634.282; Haw. Rev. Stat. § 431:10C-111; Iowa Code § 515F.6; Ky. Rev. Stat. § 304.20-040; La. Rev. Stat.§ 22:35; Mo. Rev. Stat. § 379.114; Nev. Rev. Stat. § 687B.390; N.H. Rev. Stat. § 417-A:3; N.Y. Ins. Law § 2606 (b); Ohio. Rev. Code § 3937.39; S.D. Codified Laws § 58-11-55; Va. Code § 38.2-2213 

[10] Alaska, Idaho, Indiana, Vermont.  In some of these states, race is a prohibited factor in lines other than personal automobile insurance. Non-insurance laws that might apply in these states were not examined

[11] Ricci v. DeStefano, 557 U.S. 557, 577 (2009)

[12] Griggs v. Duke Power Co., 401 U.S. 424, 432 (1971)

[13] Griggs, at 431, 432

[14] Tex. Dep’t of Hous. & Cmty. Affairs v. Inclusive Cmtys. Project, Inc. 135 S. Ct. 2507 (2015); Smith v. City of Jackson, 544 U.S. 228 (2005)

[15] Tex. Dep’t of Hous. & Cmty Affairs, at 2518

[16] Tex. Dep’t of Hous. & Cmty. Affairs, at 2523-24.  In 1991, Title VII was amended to establish a shifting burden of proof under a disparate impact theory of liability.  See, 42 U.S.C. § 2000e-2

[17] Tex. Dep’t of Hous. & Cmty. Affairs, at 2518

[18] 42 U.S.C. § 3604

[19] Tex. Dep’t of Hous. & Cmty. Affairs, at 534