Insurance companies are in the business of discrimination. Insurers attempt to classify insureds into separate risk pools based on their differences in risk profiles, first, so that they can charge different premiums to the different groups based on their risk and, second, to incentivize risk reduction by insureds. This is why we let insurers discriminate. There are, however, limits to how much discrimination, or the types of discrimination, we will allow even insurers to engage in. But what exactly are those limits, and how are they justified? To answer this question, rather than starting with top-down grand normative theories, this Article takes a bottom-up approach and explores the state laws that govern this question in the first place. The Article makes use of a unique hand-collected dataset of laws regulating insurer risk classification in all 51 jurisdictions. Among our findings are that state insurance anti-discrimination laws vary a great deal, in substance and in the intensity of regulation, across lines of insurance, across policyholder characteristics, and across states. The Article also finds that, contrary to expectations, a surprising number of jurisdictions do not have any laws restricting insurers’ ability to discriminate on the basis of race, ethnicity, or religion. In addition, the Article finds that much of the cross-line variation in state insurance regulations can be explained in terms of adverse selection. Specifically, we show that in the lines of insurance where concerns about adverse selection are strong, states allow insurers more leeway in discriminating among insureds. The Article’s primary contribution is descriptive; however, it has potential normative implications as well.The authors comment that
This article takes a different approach to exploring insurance discrimination. Rather than starting with top-down grand normative theories, it takes a bottom-up approach and explores the state laws that govern this question in the first place. This is much harder than it initially sounds, which may be why we are the first to undertake the task. States’ regulation of insurers’ risk classification practices is remarkably multi-faceted and variable. In fact, it varies along at least four core dimensions. First, risk classification regulation often varies across insurance lines. For instance, a state may prohibit automobile insurers from taking into account age in discriminating among policyholders, while explicitly permit life insurers to do so. Second, state laws may vary across different policyholder characteristics: whereas automobile insurers may be permitted to take into account gender in classifying policyholders, they may be prohibited from taking into account race or national origin. Third, the intensity of state risk classification regulation is quite heterogeneous: such laws may simply require an actuarial justification for discriminating on the basis of a characteristic, or it might categorically prohibit insurers from using that characteristic. Finally, risk classification laws differ dramatically by state. To capture these details, we use a unique hand-collected dataset of state laws on the regulation of insurer risk classification. Since the law in this area is determined primarily at the state level, we identified and analyzed the relevant state statutes and regulations in all 50 states (and Washington DC) and coded those laws for five different lines of insurance and for nine different characteristics. Once we had completed a thorough search and coding of the existing state statutory and regulatory law, we went back to every state and looked for any court decision or administrative ruling that interpreted the relevant statutory or regulatory language to determine whether the initial code assigned to that law, for a given line or characteristic, should be adjusted. The result is the first ever database of insurance anti-discrimination laws in the U.S.
We then investigate the laws. Our empirical results are summarized below through numerous graphs and pictures that are hard to distill in just few words. But they show, for example, that state laws regulating risk-classification practices in the auto and property/casualty insurance lines are the most restrictive. By contrast, the least restrictive state regulations apply to disability andlife insurance, which means that they permit the highest amount of discrimination or risk- classification for those lines of insurance. We also document more specific cross-line variations. For example, forty-eight of the fifty-one jurisdictions (as well as the federal government) completely prohibit the use of genetic endowment for health insurance, even though in the other four lines of insurance genetic endowment is among the least restricted characteristics. And while race, ethnicity, and religion are the most intensely restricted characteristics in every line of insurance, we find, contrary to conventional wisdom, that a surprising number of jurisdictions do not, in fact, have any laws restricting insurers’ capacity to discriminate on the basis of these characteristics.
Having mapped out these complexities, we analyze the results based on the various explanatory factors that are emphasized in the extant theoretical literature. Our goal in doing so is to provide a descriptive, bottom-up understanding, of the reasons and possible justification for the pattern of state laws that we observe. By doing so, we are able to solve several puzzles regarding variations in anti-discrimination insurance laws which look initially as unsolvable. Why is gender more strongly restricted in auto, disability, and property/casualty insurance than it is in health and life insurance? Or, as was just mentioned above, why do states prohibit the use of genetic endowment for health insurance, but permit it in the other four lines of insurance?
Surprisingly, we find that concerns about adverse selection can explain and justify a large part of the cross-line variations that are embedded in state insurance regulation. In particular, we show that in lines where concerns about adverse selection are strong, states allow insurers to discriminate (in fact, to classify) among the insureds. By contrast, where concerns about adverse selection are weak, insurers are more like to be prohibited from engaging in such discrimination. A more diverse set of factors is needed to explain cross-characteristic variation in state laws.
While the article’s contribution is primarily descriptive, it also has various potential normative implications as well. Among other things, it reveals potentially troubling cross-state variation in the regulation of insurance discrimination. For instance, twenty-one jurisdictions permit using gender in life insurance compared with twenty jurisdictions that strongly limit it; one state, North Carolina, goes as far as to prohibit it. And, as was mentioned above, state laws are also inconsistent about seemingly settled questions, such as the use of race, national origin, and religion. Yet, one would expect to see all states following roughly the same normative commitment. Discrimination is such a loaded and litigated concept in the U.S., as reflected in various landmark Supreme Court cases and federal antidiscrimination laws, that one would expect to see all states striking the balance between permissive classification and prohibited discrimination at roughly the same point. But they don’t. Perhaps this suggests that a federal legal response may in some cases be called for. Or, alternatively, perhaps it suggests that norms are doing some of the work that state insurance anti-discrimination statutes are not—especially in such areas as race, ethnicity, and religion.
The Article proceeds as follows. Part I describes the various factors we extracted from the voluminous literature on discrimination. This literature review is the first to systematically integrate scholarship published in insurance economics journals with the insights of legal academics. It distills from these literatures twelve different factors that may shape the normative case for laws restricting insurers’ capacity to discriminate among different policyholders. Part II presents our empirical results, documenting fifty-one anti-discrimination insurance law regimes with respect to nine different characteristics (race, ethnicity, religion, gender, age, credit score, sexual orientation, genetic endowment and zip code) in five difference insurance lines (life, health, homeowners, automobile and disability). Finally, Part III attempts to explain cross-line and cross-characteristic variations using the factors we identified in Part I.