Premium Justice: An Egalitarian Defense of Risk-Based Insurance Pricing

Introduction

One of the most difficult and interesting questions of insurance regulation is whether insurers should be allowed to charge different rates based on an insured’s level of risk. Because individuals differ both in the likelihood that they will suffer a loss and in the size of their potential loss, insurers try to estimate the amount that each insured expects to lose in a given period, and then charge a “risk-based price” that reflects that amount. Whether this practice is appropriate has broad policy implications. Most state insurance regulators, for example, have a mandate to ensure that insurance prices are not “unfairly discriminatory.”1 There has also been a steady stream of public controversies during the last half-century relating to the defensibility of risk-based underwriting—from the use of sex in determining employment-based annuity rates,2 to the use of characteristics that indicate a high risk of contracting HIV when selling life or health insurance,3 to the denial of health and life insurance coverage to battered women,4 to the push to forbid health insurers from discriminating on the basis of preexisting conditions5 and genetic information,6 to the propriety of charging higher homeowners’ insurance premiums to those exposed to wildfire or hurricane risk.7

On one side of this issue are those who believe risk-based insurance pricing is presumptively unjust, especially when it turns on characteristics over which insureds lack control. Proponents of this argument tend to advocate for “community rating”8—a regulatory prohibition against charging different premiums on the basis of characteristics like sex, credit score, or pre-existing health conditions.9 On the other side are those who assert that risk-based pricing promotes economic efficiency.10 They argue that preventing insurers from varying premiums based on an insured’s level of risk will threaten the stability and availability of insurance, and will distort policyholders’ incentives to avoid loss in a way that will lead to economic waste.11 As Kenneth Abraham put it in 1985, “[A]ttitudes toward insurance always seem to be pulling in two directions—one that highlights the risk assessment, or efficiency-promoting features of insurance classification, and the other that stresses insurance’s risk distributional function.”12 This standoff has, for the most part, persisted.13 Occasionally, one side or the other wins ground in the legislative or regulatory arena, but at an intellectual level, they continue to talk past one another.14

This Article aims to escape that deadlock by arguing that the most powerful argument for risk-based pricing is grounded not in efficiency, but rather in egalitarian distributive justice. The availability of an “egalitarian” argument for risk-based insurance pricing is counterintuitive; charging everybody an equal price would seem to be the more egalitarian thing to do.15 But, as will be explained in more detail below, traditional efficiency-based arguments for risk-based pricing, mostly based on the problem of adverse selection, cannot motivate a categorical opposition to community rating.16 This suggests that the better way to understand the persistent controversy over risk-based insurance pricing is not as a tradeoff between efficiency and fairness, but rather as a disagreement over which conception of egalitarianism ought to apply. It is, this Article argues, a disagreement among egalitarians about what should be equalized: the burdens of insurance coverage (i.e., premium costs), or the benefits of insurance coverage (i.e., the welfare gain that insureds get when they purchase a policy).

An example can help to illustrate the basic intuition of the egalitarian argument for risk-based insurance pricing. Imagine a person, Charles, who suffers from a genetic disease that causes him to need special health care. He expects to incur healthcare expenses that would cost him $5,000 this year, on average, if he were paying out-of-pocket. (This is an estimate based on his past medical expenses. The costs could turn out to be significantly higher, or lower, in any given year.). Charles, because he does not like the unpredictability of such expenses, is willing to pay up to 25% more than his expected loss, or $6,250, for insurance that would cover all of his healthcare costs.17 Now imagine a healthier person, David, who expects to incur only $3,500 of expenses this year, and who is willing (for the same reason) to pay $4,375 for coverage. A health insurance company with zero operating costs could feasibly charge both Charles and David the same price, $4,250. At that price, the insurer’s premiums would be equal to what it expects to pay out in claims ($5,000 to Charles plus $3,500 to David equals $8,500 in combined premiums). Both Charles and David would voluntarily purchase coverage at this price; no mandate would be required to ensure their enrollment. But one of them would nonetheless appear to get a “better deal” than the other. Charles is willing to pay $6,250 for something that costs him $4,250 for a benefit of $2,000, whereas David, by the same logic, only gets a benefit of $125. If, instead, each insured were charged a price that corresponded to his individual expected loss, then there would be less of a difference in the benefit of coverage: if Charles were charged $5,000 and David were charged $3,500, then they would receive $1,250 and $875, respectively.18

Many who read this example will be inclined to argue that charging both Charles and David the same price is appropriate. As a matter of distributive justice, it would be improper to allow an insurer to discriminate against Charles because of his unfortunate genetics. Charging a flat premium, therefore, is morally correct because it creates a cross-subsidy from David to Charles that counteracts an undeserved inequality of genetic endowments.

This Article presents an alternative perspective. From this view, charging both individuals the risk-based price is the more just policy. The insurance arrangement “grew the economic pie” by creating $2,125 of benefit. Neither of them has a greater claim to that benefit than the other because insurance only works if there are multiple participants, joining their risks together into a pool. If one dollar of this benefit is worth the same to Charles as it is to David,19 then an equal division of the economic benefit between insureds would be the most just way to divide it, and risk-based pricing comes closer to achieving that goal than community rating.

This Article lays out an argument in support of the latter position, and shows that its logic applies not just to health insurance, but to all other types of insurance as well. It also shows why and how one could hold this position and still accept that redistribution is warranted to compensate unlucky individuals like Charles, even while maintaining that insurance regulation is not the proper way to do so.

The availability of an egalitarian argument for risk-based pricing can help to explain the persistence, and broad appeal, of the claim that risk-based pricing is not only efficient, but also fair. Such a claim is frequently used in public policy debates to argue that insurance should not be redistributive in any broad sense—that it should not, for example, be used simply to effect transfers from the rich to the poor, from the healthy to the chronically ill, or from those who are protected from climate risks to those who are exposed to them.20 This Article provides an analytically clear articulation of what, I suggest, is the motivating intuition for many who oppose such efforts in good faith. Shedding light on its logic and underlying assumptions will help us better understand risk-based pricing from a moral perspective.21 Even (and especially) those who are inclined to reject risk-based pricing can benefit from such an exploration because it will allow them to oppose it more effectively.

The Article proceeds as follows. Part I provides a brief background on how insurance markets work, and the current state of insurance antidiscrimination law in the United States. Part II makes the egalitarian argument for risk-based insurance pricing. It also responds to two important counterarguments: first, that risk-based pricing is morally objectionable because it perpetuates or exacerbates unfair inequality; and second, that the egalitarian argument is implausible because it would seem to require that all commercial gains be divided fairly between the parties involved. Part III reviews alternative arguments that have been made in support of risk-based insurance pricing, and argues that they fail to justify the practice whereas the egalitarian argument succeeds. Part IV concludes.

I. A Primer on the Theory and Practice of Insurance Pricing

In competitive insurance markets, insurers face pressure to lower premiums as much as possible while maintaining their solvency. The primary method for achieving this is to align each insured’s premium with her “expected loss”22—the anticipated claims she will generate during the coverage period—while adding a margin for administrative costs.23 Insurers employ two principal approaches to estimate expected losses.24 The first, “feature rating,” relies on observable characteristics that correlate with risk.25 Examples include calculating fire insurance premiums based on a property’s value or its fire suppression systems, or varying health and life insurance premiums based on preexisting conditions, sex, or smoking habits. The second approach, “experience rating,” uses the insured’s claims history to project future losses.26 This manifests in practices like increasing auto insurance premiums after accidents or adjusting workers’ compensation rates based on prior years’ claims experience.

Competition drives insurers to identify increasingly granular differences in insureds’ expected losses—a practice called, “risk classification.”27 An insurer that develops more precise classifications can profitably “skim” lower-risk customers away from competitors by offering reduced premiums that better reflect those customers’ true risk profiles while still covering the insurer’s expected payouts.28 If a competitor fails to adopt comparable classification schemes, its risk pool will deteriorate as lower-risk insureds depart for a better deal elsewhere, forcing it to raise premiums on the insureds that remain in the pool. In extreme cases, this process can trigger a “death spiral” whereby insurers that are unable to keep pace with evolving underwriting techniques face mounting losses and eventually experience financial collapse.29 Competition among insurers therefore pushes insurers to charge each insured a premium that approximates her true expected loss.30 A theoretically perfect risk-based pricing regime—one that charges each insured an amount that exactly reflects her expected loss—is sometimes called “actuarially fair” pricing.31 Of course, risk classification has its own costs. Developing the data and methods of distinguishing different levels of risk can be expensive. Eventually, the competitive benefits to an insurer of further improving its classification methods will not exceed these costs.32 This is one reason why insurance prices, even in a market with essentially no regulation of underwriting practices, never achieve perfect actuarial fairness.33

Competition between insurance companies thus explains why they have a tendency to continuously refine their risk classifications. But it does not justify the practice. When insurers charge higher premiums to insureds with greater expected losses, they discriminate against those insureds. Discrimination, as a general matter, is not considered just, especially when it turns on characteristics for which individuals are not responsible such as race, sex, age, or genetics.34 This suggests that the morally appropriate way to regulate insurance markets is to force a truce between competing insurers by forbidding them from making classifications that we deem morally suspect, even if those classifications are actuarially sound. Insurance, on this view, ought to be subject to antidiscrimination rules that require insurers to ignore characteristics like race, sex, age, genetics, and potentially many others.35

Puzzlingly, actual insurance regulation in the United States does not even come close to implementing such rules. Federal law provides only limited protections against insurance discrimination, primarily in the health insurance context,36 and no federal law prevents insurers from discriminating based on characteristics like race, religion, sex, or national origin.37 For historical reasons, most of the responsibility of regulating insurers has been delegated to the state governments.38 One might expect, therefore, to find state insurance laws that have filled this regulatory gap.39 But when three of the nation’s leading insurance law scholars conducted a comprehensive survey of state insurance regulations in 2014, they were surprised to find that “affirmative bans of insurer discrimination on the basis of potentially suspect policyholder traits are quite rare.”40 For the most part, insurers are not legally barred from using even the most suspect characteristics or traits in setting premiums. Insurance thus stands in stark contrast with other important social activities—such as employment, banking, education, housing, and public accommodation—that are subject to broadly-applicable antidiscrimination rules.41 This is morally troubling because even if relatively few insurers explicitly take those characteristics into account, it opens the door to discrimination on bases that are straightforward proxies for suspect characteristics, like zip codes, credit scores, education levels, and marital status.42

American insurance markets thus appear to be characterized by a surprising deference to insurers’ determinations of which classifications are appropriate.43 Insurers have, for example, generally succeeded in preserving their ability to charge different life and auto insurance rates to men and women.44 There have been several discrete efforts over the last half-century to push back against particularly controversial underwriting practices—such as when health insurers were discovered to be systematically discriminating against battered women,45 individuals who might be exposed to HIV,46 or those with pre-existing health conditions.47 Or when homeowners insurance companies faced scrutiny under disparate impact regulations promulgated by the Department of Housing and Urban Development (HUD), which sought to hold them accountable for discriminatory effects in their underwriting practices.48 But these moments of resistance have been the exception, rather than the norm. This raises a question: Is there any justification for our surprising unwillingness, as a country, to subject insurance to comprehensive antidiscrimination rules?49 The argument articulated in Part II, and explored in Part III, provides one.

II. The Egalitarian Argument for Risk-Based Insurance Pricing

This Part develops the egalitarian argument for risk-based insurance pricing in four steps. Section II.A presents the core argument: that risk-based pricing distributes the consumer surplus generated by insurance more equally than does community rating. Section II.B articulates and defends the conception of egalitarian distributive justice that underlies this argument. Section II.C then addresses three objections to risk-based pricing: first, that certain forms of insurance provide essential goods that should be universally accessible, regardless of individual risk; second, that risk classifications based on unchosen characteristics perpetuate unjust inequalities; and third, that certain classifications inflict dignitary harms and create relations of domination. Section II.D then responds to a separate challenge: if consumer surplus should be divided equally in insurance, why not in all commercial transactions? Together, these four Sections argue that although departures from risk-based pricing may be warranted in specific contexts, the egalitarian case for actuarial fairness in insurance markets remains generally compelling.

A. Distributing Surplus

The egalitarian argument for risk-based pricing focuses not on the price of insurance, but rather on the distribution of what economists call “consumer surplus,” the net benefit derived from purchasing coverage. Consumer surplus is a standard economic measure of the welfare gain an individual realizes when the price of a good is lower than the maximum amount she would be willing to pay.50 When, for example, a consumer is willing to pay $15 for a movie ticket but the ticket costs only $10, she gets $5 of consumer surplus.

The egalitarian argument for risk-based pricing states that, in a standard competitive insurance market where insureds are identical except for their levels of expected loss, the distribution of consumer surplus between any two insureds will always be more equal under risk-based pricing than under community rating. A more rigorous statement of this proposition is laid out in Appendix A. But the straightforward intuition behind it is captured in the introductory example involving Charles and David.

The Charles and David example illustrates a key assumption upon which the egalitarian argument for risk-based pricing depends. It posited that the amounts Charles and David were willing to spend for insurance were not only greater than their respective levels of expected loss (making them both “risk averse”51), but also positively correlated with their expected loss. In the example, both Charles and David were willing to pay 25% more than the amount they expected to lose, in any given year, in exchange for health insurance coverage. This phenomenon—where insurance consumers’ willingness-to-pay for coverage is positively correlated to their expected loss—creates what economists call “adverse selection.”52 Under such conditions, those insureds who expect to lose the most—and who are thus the costliest to insure—will also be willing to pay the most for coverage.53 The egalitarian argument for risk-based pricing only has force when this is the case.

It is important to emphasize that this argument for risk-based pricing is not grounded in the pursuit of efficiency.54 Nor is it the same as the banal insight that the most-risky insureds are charged less, and the least-risky are charged more, when their costs are pooled, and the members of the pool are charged a flat amount. This is instead a point about the relative distribution of surplus produced by different methods of pricing insurance. Analyzing the distributive consequences of insurance from this perspective requires examining not just the price of insurance coverage (i.e., the premium insureds are charged) but also the value of coverage (i.e., the maximum amount each insured is willing to spend, given all the other things on which she might spend her money). By considering these two things together, it is possible to see why risk-based pricing should be justifiable to both the low-risk and the high-risk insured, on the grounds that risk-based pricing distributes surplus among all insureds more equally than does community rating.

B. Defining a Conception of Egalitarian Distributive Justice

This Section articulates the conception of egalitarianism that motivates this Article’s argument in favor of risk-based insurance pricing.

First, the argument presumes that it is possible not only to measure individual well-being,55 but also to make interpersonal comparisons of differences in individual well-being.56 This assumption is commonplace in much policy work (especially cost-benefit analysis) that attempts to analyze the well-being of individuals in society and to recommend policies that both maximize well-being and distribute it fairly. It is also consonant with many of our most basic moral intuitions and common sense.57

Second, the argument requires one to accept that interpersonal differences in consumer surplus are a plausible proxy for interpersonal differences in individual well-being.58 So long as egalitarianism aims for something like an equal or fair distribution of individual well-being, it will be necessary to decide upon at least one such proxy, because well-being is not directly observable.59 Microeconomics tends to focus on income as a proxy for well-being.60 The egalitarian argument for risk-based pricing makes a similar move, by treating consumer surplus dollars as income, and then assuming that each new dollar of income increases one person’s well-being as much as it does another’s.61

Third, this Article’s argument for risk-based pricing assumes that a system that aims narrowly to equalize the distribution of well-being produced by a voluntary cooperative activity (e.g., private insurance) among the participants in that activity, and that ignores everything else, can meaningfully be characterized as “egalitarian.”62 By focusing solely on the distribution of insurance-created surplus, rather than on the distribution of resources in society as a whole, the egalitarian argument for risk-based pricing satisfies a practical need to evaluate the fairness of insurance prices without simultaneously evaluating the fairness of every other aspect of modern life as well.63 It also treats insurance markets as markets—that is, enterprises in which individuals participate willingly, for their own private reasons.

Fourth, this Article’s argument for risk-based insurance pricing assumes that the proper way to distribute the well-being produced by cooperative enterprises is to spread it equally. That is, each participant in a cooperative venture ought to get the same boost to his or her well-being as every other, rather than a pro rata share based on some other metric like proportional contribution.64 The moral plausibility of this assumption is supported by experimental psychology studies that utilize the “ultimatum game.”65 In the game, two players divide a sum of money. One player, the proposer, offers a portion of the total to the other player, the responder. If the responder accepts the offer, both players receive their respective shares as proposed. If the responder rejects the offer, neither player receives anything. Traditional economic conceptions of individual rationality suggest the responder should always accept a non-zero offer.66 Empirical studies show, however, that responders frequently reject offers perceived as unfair. In the United States, participants tend to opt for a 50/50 split of the money.67 This suggests a shared preference for an equal division of cooperative surplus—which, as discussed above, is a reasonable proxy for well-being. While alternative frameworks may provide more nuanced mechanisms for dividing cooperative surplus,68 the simplicity and intuitiveness of equal division make it a good baseline for evaluating distributions for fairness.

C. Objections to Risk-Based Pricing

There are several objections to risk-based insurance pricing that will come naturally to many readers. This Section articulates three of them, and argues that although they can justify community rating in specific settings, they do not yield a generalized rejection of the egalitarian argument for risk-based pricing.

1.   The Need for Essential Coverage

First, one might assert that risk-based pricing is morally objectionable in the context of certain types of insurance, like health insurance, where the security that insurance provides is a basic human right or need.69 Allowing health insurance premiums to depend on an applicant’s health, for example, may render it unaffordable to the very people who need it most. The Affordable Care Act’s prohibitions on health insurance premium variations based on pre-existing conditions70 and the Genetic Information Nondiscrimination Act’s prohibition on health insurers using genetic information in underwriting71 can be interpreted as evidence of the force of this argument. How, then, does the egalitarian argument for risk-based pricing hold up to the allegation that it is morally objectionable because it fails to ensure that everybody is kept above a minimum humanitarian floor?

In answering this question, it is helpful to notice that the prohibitions against health insurance pricing based on preexisting conditions or disease-linked genes can be reframed in terms that are entirely consistent with the egalitarian argument for risk-based pricing. One could argue that insurance against the risk of being unable to afford basic healthcare ought to “bind” before one is even born.72 In such a position, all “insureds” would be identical in their endowments, levels of risk, and in their preferences, and so community-rated insurance would give each insured the exact same consumer surplus.73

When framed this way, it becomes evident that the argument against risk-based pricing in health insurance is not really an objection to risk-based pricing, but rather to the use of private insurance as a mechanism for ensuring that basic needs are met. This is because any real-world “insurance” arrangement that compensates individuals who have already suffered the misfortune of bad health, bad genes, or being unable to afford risk-rated coverage will require pure redistributive transfers from haves to have-nots.74 Unless individuals are simply altruistic, such transfers cannot emerge as a result of private contracts, which are only rational when there is at least the possibility of an improvement for all parties involved.75 Private insurance creates such a possibility when the amount that an individual is willing to pay for coverage is less than what the insurer must charge her to remain solvent, as illustrated in the example with Charles and David. But such an interaction requires the pooling of prospective expected losses, not known losses. No non-altruistic insured will rationally choose to pay premiums that are already earmarked for covering others’ already-materialized misfortune. As Joseph Heath put it, “There may be a moral case to be made for such behavior, but to propose such principles as a basis for the legal regulation of the marketplace is essentially to argue that there should not be a marketplace.”76 The party that is better-suited to ensure that all individuals’ basic needs are met is, of course, the government. The government’s power to tax and spend—its ability to force transfers that would not occur voluntarily in markets—is perhaps its defining feature.

Of course, it is not always politically feasible to fund a social safety net through taxes.77 Tax-and-transfer programs require legislative action, and legislatures may be gridlocked due to industry capture, supermajority requirements, or other institutional barriers.78 These frictions can block redistributive policies that might otherwise get democratic support.79 Moreover, legislators may find it politically easier to fund a social safety net through insurance regulation rather than explicit transfers, because antidiscrimination rules may appear less overtly redistributive.80 Insurance regulation also has institutional advantages: regulators can implement antidiscrimination rules through administrative action, and courts can enforce equal protection principles on insurance pricing methodologies, bypassing legislatures. In this way, insurance regulation presents itself as an alternative pathway for funding a social safety net when direct tax-and-transfer mechanisms are unavailable.

Under such nonideal circumstances, one might ask whether insurance markets truly are less well-suited than governments to effecting the sort of redistributive transfers that are necessary to ensure all basic needs are met. If regulation of private insurance markets is the most (or only) viable way of achieving that goal, then doesn’t that make such markets relatively well-suited for the purpose?

The “narrow” conception of egalitarianism described above answers this question in the negative. Tolerance for risk-based pricing, even when it raises the price of essential coverage, can be motivated by an intuition that there is something dangerous about shoehorning public goals into private markets—especially when the urge to do so arises out of the political difficulty of effecting resource transfers directly. Instrumentalizing markets in this way undermines the democratic legitimacy of such transfers. It also undermines the integrity of markets. This is illustrated by the fact that, under extreme circumstances, an insurance regulatory regime that aims to effect pure redistributive transfers from haves to have-nots will need to impose an insurance “mandate” to prevent the haves from opting out by foregoing coverage. But requiring individuals to buy something undermines the quintessentially voluntary nature of the contracts that constitute markets.

When markets are undermined in this way, their many salutary effects for society are jeopardized. Well-functioning markets produce what Daniel Markovits has called “transaction benefits”: various positive externalities of contractual exchange that help members of a pluralistic society to cooperate and coexist, even when they lack a shared conception of the good.81 Those who place special value on such benefits82 will be inclined to protect the integrity of markets, and therefore will find appealing the argument against using private insurance regulation as a way to create a social safety net. They may still recognize a moral imperative to ensure that all individuals stay above a minimum humanitarian floor, while nonetheless preferring that this be accomplished through other means.

This tension between the need to provide a social safety net and the need to protect the integrity of private markets is not easily resolved. Where, for example, private insurance regulation truly is the only politically viable way to provide a minimum humanitarian floor, the argument against risk-based pricing will be strong. This may explain why we see community rating in health insurance, but not everywhere else. Regardless, the argument against risk-based pricing in such limited contexts does not yield a principled objection to the practice, generally. Where the relevant type of insurance does not constitute a basic human need, or where that need can be met through non-market mechanisms, the argument for risk-based pricing will be stronger.

2.   The Problem of Unchosen Disadvantage

A second category of fairness-based objections to risk-based pricing arises from the theory of distributive justice known as luck egalitarianism.83 This framework, developed by theorists like Ronald Dworkin,84 Richard Arneson,85 and G.A. Cohen,86 rests on the moral intuition that inequalities resulting from people’s voluntary choices are acceptable, but inequalities that arise from circumstances beyond their control are not. If, for example, one person becomes poor after choosing to gamble away her savings while another remains wealthy by living frugally, most would view this inequality as acceptable. But if one person is born with a genetic disease requiring expensive treatment while another enjoys perfect health, this inequality seems unfair.

From the luck-egalitarian perspective, a just society should neutralize disadvantages arising from “brute luck”—those circumstances individuals cannot control—while permitting inequalities that flow from personal choice.87 This principle implies that resources should flow from those blessed with unearned advantages to those burdened by unchosen disadvantages. Because such transfers rarely occur unprompted, government intervention becomes necessary to realize the luck-egalitarian vision of distributive justice.

Applied to insurance markets, luck egalitarianism appears to condemn risk-based pricing that depends on morally-arbitrary characteristics. When unchosen traits such as race, sex, or genetics correlate with higher expected losses, permitting insurers to price based on these correlations perpetuates and exacerbates unjust inequalities, whereas prohibiting such classifications helps to address them. Thus, luck egalitarianism would seem to demand comprehensive antidiscrimination rules barring insurance classifications based on any characteristic determined by brute luck. But implementing such a comprehensive prohibition would have unpredictable effects as a matter of luck-egalitarian distributive justice.88 Correlations between morally-arbitrary disadvantages and expected losses can run in either direction—positive or negative—often simultaneously within the same risk pool. Consider the following example: A low-income neighborhood consists of modest homes adjacent to fire-prone wilderness, while an affluent neighborhood of expensive mansions is protected from wildfire risk. Assume that limited opportunity caused the poor families to live where they do, and that the rich families inherited their properties. A blanket prohibition on insurance classifications based on unchosen characteristics would prevent insurers from charging poor families higher premiums based on their involuntary exposure to wildfire risk. But it would also prevent insurers from charging these families lower premiums based on their homes’ modest values. The net distributional impact thus becomes chaotic and unpredictable.

This structural indeterminacy reveals that luck egalitarianism cannot coherently motivate a categorical prohibition on all insurance classifications based on brute luck. Unchosen disadvantage might increase or decrease expected losses, and thus risk-based premiums. Prohibiting insurers from classifying based on morally-arbitrary characteristics will not systematically produce the transfers from the lucky to the unlucky that luck egalitarianism demands.

But what if insurance antidiscrimination rules were applied in a targeted manner, such that they prohibited risk classifications only when doing so would have the effect of transferring resources from the lucky to the unlucky? Returning to our example, this approach would ban classifications based on wildfire exposure (creating cross-subsidies from the protected wealthy to the exposed poor) while permitting classifications based on home value (preserving lower premiums for modest homes).

The conventional economic objection to such targeted redistribution through private law rules is well-established: using insurance regulation as a redistributive mechanism will distort market incentives and reduce aggregate social welfare.89 Orthodox economic theory therefore suggests instead that competitive markets should be allowed to operate efficiently, maximizing total surplus, and that the government should then employ tax-and-transfer systems to achieve distributional outcomes.90

Though this logic is taken by many to dispose of the argument in favor of achieving redistribution through targeted insurance antidiscrimination rules, Kyle Logue and Ronen Avraham have shown that such rules can, under specific conditions, achieve redistributive goals more efficiently than traditional tax-and-transfer mechanisms.91 Consider the case of genetic variants that substantially increase disease risk and reduce life expectancy (for example, Huntington’s Disease and BRCA-linked cancers). The luck-egalitarian case for transfers from the genetically-fortunate to the carriers of these variants is compelling, given the profound and unchosen nature of the disadvantage.

Logue and Avraham identify the surprising efficiency of using insurance regulation to effect luck-egalitarian redistribution in such settings. Insurance antidiscrimination rules would harness what Logue and Avraham call the “invisible redistributive hand.”92 Insurance markets subject to such rules would automatically calibrate the magnitude of cross-subsidies to reflect actual cost differences attributable to genetic variation, without requiring explicit government determination of compensation levels. The result would be “something approximating the right amount of redistributive transfer”93 without the need for centralized bureaucratic intervention. A tax-and-transfer system, by contrast, would require the government to identify individuals with the relevant genes and to quantify the compensation they are due for reduced life expectancy, elevated medical costs, and diminished quality of life—all of which are administratively complex, expensive, and potentially controversial tasks.

Logue and Avraham admit that the insurance-regulation-based approach has some drawbacks. First, if potential insureds know whether or not they have disease-related genes, then those who lack them might forego insurance rather than pay a cross-subsidy.94 In extreme cases, this dynamic can cause premiums to rise for those remaining in the pool, leading the lowest-risk insureds progressively to drop out. The inefficiency that results when low-risk individuals cannot purchase insurance that reflects their low-risk status is called “deadweight loss.”95 This must be weighed against any efficiencies gained by the “invisible redistributive hand” of insurance antidiscrimination rules.

Second, insurance-based redistribution can generate new kinds of arbitrary and inequitable misallocations.96 If certain genetic variants concentrate in particular ethnic or geographic communities, and thus within specific insurance pools, the burden of cross-subsidization will fall disproportionately on members of those pools, regardless of their individual circumstances or ability to pay.97 This geographic and demographic lottery will, at least in theory, be more arbitrary than allocating those extra costs across the entire tax base.

The first of these two problems could be addressed by mandating that all individuals purchase community-rated insurance.98 The second could be addressed by joining all risk pools together—creating a single monopolistic insurer. Whether these problems are likely to be serious enough, as a practical matter, to require such drastic measures are empirical questions about which reasonable people might disagree.99

Logue and Avraham’s analysis thus provides a luck-egalitarian justification for community rating in specific settings where unchosen disadvantages are severe and also more costly to address through conventional redistribution. This may justify targeted legislative interventions like the Affordable Care Act’s prohibition on health status underwriting and various state restrictions on credit score classifications (potentially proxying for race-based disadvantage). Nevertheless, their argument does not make a comprehensive luck-egalitarian case against risk-based pricing generally. The efficiency costs of broader antidiscrimination rules, the indeterminacy of the distributive effects of such rules, and the principled concerns about instrumentalizing private markets for public purposes all counsel against wholesale abandonment of risk-based pricing.

3.   Dignitary Harms and Non-Domination Concerns

Even when actuarially sound, risk classifications that track suspect categories may devalue classes of individuals and undermine their equal citizenship. These objections provide arguments against risk classification that do not sound directly in egalitarian distributive justice.

Deborah Hellman’s work on discrimination explains why certain insurance classifications may be impermissible regardless of their actuarial accuracy.100 Hellman argues that discrimination is wrongful when it demeans—when it expresses the inferior moral status of individuals or denies their equal worth. If insurers were to charge higher premiums based on race, for example, they would perpetuate practices that mark racial minorities as inferior or less deserving of concern.101 Even if actuarially justified, such practices would communicate that these communities were less valuable and less entitled to economic opportunity.

The work of Hanoch Dagan and Avihay Dorfman can be used to make a similar argument.102 Their concept of “relational justice” emphasizes the moral quality of institutions that is captured not by how they distribute resources, but by the character of the relationships they structure and sustain.103 Dagan and Dorfman argue that relational justice requires evaluating institutions according to whether they enable participants to interact as equals—not merely in the sense of receiving equal shares of some good, but in the deeper sense of relating to one another with mutual respect and standing.104 Applied to insurance markets, this framework could suggest that risk-based pricing will become problematic when it expresses and reinforces hierarchical relationships among insureds. Such relationships are especially likely to result from risk classifications based on suspect categories or immutable characteristics.

These types of dignitary and relational harms caused by risk-based pricing can be particularly acute, given insurers’ role as a gatekeeper to full economic and social participation. As Tom Baker has shown, insurance can act as a form of “private governance,” with insurers effectively determining access to housing, transportation, healthcare, and business opportunities.105 When insurers classify based on suspect characteristics, they exercise quasi-governmental power without the corresponding democratic accountability or constitutional constraints.

These objections are strong, but, again, they do not defeat the egalitarian case for risk-based pricing in all or even most contexts. Rather, they identify competing moral considerations that must be weighed against efficiency and conceptions of distributive justice. In some cases, the efficiency and distributive justice benefits of risk-based pricing may outweigh costs to dignity, formal equality, or autonomy. Geographic rating for flood insurance, for example, has little demeaning effect and generates significant fairness and efficiency benefits in the form of loss-prevention. But where classifications track suspect classifications, perpetuate stereotypes, or reinforce patterns of domination, these harms may be dispositive. As Logue and Avraham put it, “[w]hatever the statistical insurance data reveal, it may well be that allowing insurers to discriminate on the basis of such characteristics perpetuates or reinforces invidious stereotypes and, on that basis alone, should be prohibited.”106

D. The Distinctiveness of Insurance Markets

This Section responds to the following objection: If one accepts that the surplus produced by insurance transactions ought to be divided equally among insureds, what distinguishes insurance from all other commercial transactions? What prevents the argument for risk-based pricing from implicitly suggesting that even the most routine commercial interactions must also be subjected to judicial or regulatory assessment of whether the resulting surplus has been allocated fairly among the parties involved?

This critique has weight because the everyday operation of our economy shows that we don’t do this. Markets routinely tolerate price uniformity despite different consumer valuations. When merchants charge identical prices to consumers with different willingnesses-to-pay, we typically view the resulting surplus allocation not as something requiring correction, but as the natural and legitimate consequence of voluntary exchange. The common law doctrine of consideration makes this clear: when determining whether a contract is enforceable, courts ask whether an exchange has occurred, not whether the exchange distributed value equitably.107 How, then, does the argument for risk-based pricing respond to the allegation that it appears to contradict both the general morality of markets and basic doctrines of contract law?

To answer this question, it is helpful to begin by understanding why market economies generally eschew any requirement that there be an equitable division of surplus. Any attempt to do so would require sellers to charge different prices based on each customer’s willingness to pay. When sellers can charge each consumer her precise reservation price (a practice known as “perfect” or “first-degree price discrimination”), all consumers are left with the same amount of surplus: none. This outcome, or something approaching it, may seem appropriate as a matter of distributive justice, at least as between consumers.108

But price discrimination is only possible when markets are not fully competitive. In competitive markets, price discrimination is unsustainable because rival sellers will find it profitable to undercut each other to capture market share. Such competition drives prices towards an equilibrium at which the marginal cost of producing an additional unit equals the marginal benefit to the consumer who values the good least among all actual purchasers, thereby ensuring that all mutually-beneficial trades are completed.109

There are many benefits to competitive markets—they allocate resources efficiently, for example.110 These benefits tend to weigh in favor of promoting the competitiveness of markets and against any attempt to achieve distributive fairness through price discrimination. There are exceptions to this general rule, of course. When it comes to markets that are subject to a natural monopoly—e.g., public utilities, telecommunications infrastructure, and regional transportation networks—it may be possible to promote price discrimination in service of distributive fairness without sacrificing the benefits of competition, precisely because competition is structurally impossible or inefficient. But where such conditions are not present, modern economic theory suggests that the better course is to let competitive markets flourish, thereby maximizing total social surplus, and then use ex post tax-and-transfer mechanisms to reallocate that surplus equitably among all involved.111

Interestingly, insurance markets differ from most other markets in a fundamental way that aligns competition with allocative fairness. The distinctive feature of insurance markets is identified particularly clearly in the work of economists Liran Einav and Amy Finkelstein, who make a distinction between what they call “selection markets” and “conventional product markets.”112 In selection markets—of which insurance represents the paradigmatic case—there is a structural link between supply and demand that is absent in markets for conventional goods like toaster ovens, cars, and houses, where supply and demand are independent.113

In conventional product markets, higher prices make production more profitable, attracting new entrants and incentivizing existing producers to expand their operations. This positive relationship between price and quantity-supplied creates the familiar “upward-sloping” supply curve of introductory economics textbooks (as seen in Appendix A, Figure 1b). The curve reflects increasing marginal costs as production scales up due to factors like resource scarcity, capacity constraints, and diminishing returns to capital. Different producers face different marginal costs of production, and those who do not find it profitable to produce goods at the equilibrium price will direct their resources elsewhere. The demand curve is determined by entirely separate factors, such as consumer preferences, income levels, prices of substitute goods, and demographic trends. It slopes downward because consumers’ marginal utility diminishes with additional consumption and because lower prices make the good appealing to consumers who value it less.

In insurance markets, by contrast, the cost of providing coverage depends not on the insurer’s production technology, but on the risk profile (i.e., expected loss) of those who buy it (i.e., those who are “selected” into the risk pool).114 As Peter Siegelman concisely put it, “[in insurance markets,] customers are simultaneously purchasers and costs to the seller.”115 This means that, when the higher-risk insureds are also willing to pay more for coverage, insurers who lower prices to attract more customers will also draw in individuals who are progressively cheaper to cover. The marginal cost of coverage therefore decreases as the quantity of insurance supplied increases—creating an unusual “downward-sloping” supply curve, illustrated in Appendix A, Figure 1a.116 The demand curve in insurance markets slopes downward in the usual way, reflecting the standard principle that lower prices induce more individuals to purchase a good. Einav and Finkelstein’s revolutionary insight is that, in selection markets, the slopes of these two curves are intrinsically linked through the composition of the pool of insureds.

This means that, in competitive insurance markets with negligible transaction and administrative costs, all insurers have the same marginal-cost functions because they all compete for the same heterogeneous population of potential insureds. Competition also drives insurers to charge risk-based prices, rather than a single uniform price for all consumers. Such risk-based pricing produces the same sort of outcome that price discrimination in conventional markets might seek to achieve—namely, an equitable distribution of surplus among consumers.

This distinctive feature of insurance makes it possible to advocate for fair allocation of consumer surplus in insurance, without extending the same requirement to commercial transactions generally. Whereas implementing distributive fairness in conventional markets typically requires sacrificing competition and thus efficiency—a tradeoff that efficiency usually wins, in our market-oriented economy—insurance markets present an extraordinary situation where fairness and efficiency point in the same direction.

III. The Power of the Egalitarian Argument

This Part begins, in Section III.A, by focusing on the classic efficiency-based argument that is most commonly made in favor of risk-based insurance prices: that adverse selection will cause deadweight loss unless insurers are allowed to classify risks as they see fit.117 It suggests that this argument is not as strong as it may seem, and thus that it may well be possible to subject insurance markets to more rigorous antidiscrimination regulation without sacrificing efficiency. This Part then reviews three alternative justifications that are also frequently given for risk-based insurance pricing:118 (i) that departures from risk-based pricing are morally problematic when insureds have control over the relevant risk, (ii) that risk-based pricing is justified by the principle that everybody ought to “carry their own weight,” and (iii) that libertarian political theory justifies the practice. Sections III.B, III.C, and III.D review each of those arguments and argue that none can provide a morally plausible and generalized defense of risk-based pricing in insurance. Section III.E then shows how the egalitarian argument for risk-based pricing succeeds where the others fail.

A. Adverse-Selection Arguments Prove Too Little

Where adverse selection is present, it is often argued that limiting insurers’ ability to classify risks will cause low-risk individuals to forego coverage, creating deadweight loss.119 Under extreme circumstances, such dynamics can cause the average risk-level of the remaining pool of insureds to rise, necessitating further premium increases, which then accelerates the exodus of the lowest-risk insureds. When entire insurance markets unravel in this way, the loss to social welfare is even greater: not only the lowest risk, but also the highest risk insureds are deprived of the opportunity to make a welfare-enhancing purchase. Thus, it is often argued, any restriction on insurers’ ability to charge actuarially-fair prices will require a sacrifice, perhaps a great sacrifice, to efficiency.

But there is good reason to think—both as a theoretical matter and based on empirical studies of insurance markets—that these efficiency-based adverse selection concerns are overblown. Economic theory suggests that the inefficiencies described above can generally be eliminated by rough risk classifications—i.e., without achieving anything near actuarially-fair prices. So long as a low-risk insurance applicant remains risk-averse, then she will still choose to purchase insurance that is “overpriced,” as compared to her individual risk level.120 Such risk-aversion creates a sort of cushion, allowing insurers to price coverage based on relatively crude distinctions between insureds, without losing any of the lowest-risk insureds in the pool. The Einav Finkelstein Framework makes that possibility especially easy to illustrate, as shown in Appendix B.

This means that the traditional efficiency concerns related to adverse selection can motivate an argument for crude risk classifications, but not one for actuarially-fair pricing. There may, moreover, be a relatively large set of crude classification schemes available in any given insurance market that would prevent deadweight loss.121 Thus, for example, it might be that deadweight loss caused by adverse selection in the auto insurance market could be eliminated by classifications based solely on the color of the insured’s car, the gender of the driver, the driver’s past record, or some other simplistic distinction. And although competition will drive all insurers towards a classification scheme that is within the set that does not cause progressive exit of low-risk insureds,122 it is difficult to say where within that set the market will land. The economic arguments for actuarial fairness therefore cannot provide normative guidance as to (a) how to select from among multiple economically-feasible classification schemes, or (b) why to continue pushing towards increasingly-accurate classifications, once the efficiency concerns related to adverse selection have been addressed. It appears, moreover, that most of the debate about the propriety and necessity of risk-based pricing takes place when the actual risk-classification schemes at issue are within the set of schemes that do not create deadweight loss. This is evidenced by the fact that it is surprisingly difficult to observe deadweight loss and death spirals in the real world.123

Various kinds of “frictions” also may have the effect of making adverse selection less likely to cause deadweight loss than traditional economic theory would suggest.124 Cognitive limitations, for instance, may lead consumers to misperceive their own risk levels or fail to optimize their insurance purchasing decisions in ways that would exploit informational asymmetries. When low-risk individuals incorrectly assess themselves as higher-risk, or when computational complexity prevents them from recognizing that pooled premiums exceed their true reservation price, they may remain in the insurance pool despite having economic reasons to exit. Behavioral inertia represents another powerful friction that can maintain market stability even in the presence of adverse selection. The well-documented tendency of consumers to maintain status quo insurance choices, even when market conditions change substantially, means that low-risk individuals who might theoretically benefit from exiting the pool often remain enrolled through passive decision-making.125 This “stickiness” in insurance markets—reinforced by default options, automatic renewal provisions, and the psychological costs of active decision-making—can preserve risk pooling that would collapse under frictionless conditions. Such frictions, operating simultaneously and often reinforcing one another, can create a market environment in which the stark predictions of adverse selection theory fail to materialize.

If crude risk classifications, in conjunction with various demand-side frictions, are likely to prevent most of the economic downsides of adverse selection, then adverse selection fails to justify anything more than rough risk classifications. This suggests that antidiscrimination rules may have far fewer efficiency costs than generally believed.

B. Arguments Based on the Insured’s Control Beg the Question

Perhaps the next-most-commonly made argument for risk-based insurance pricing is based on the concern that community rating is improper when the underlying risk is within the insured’s control. There is an efficiency-based and a fairness-based version of this argument.

The efficiency-based version focuses on the fact that classifications based on variables within an insured’s control126 can incentivize insureds to invest in efficient amounts of loss-prevention.127 Thus, for example, where homeowners insurance premiums vary based on exposure to climate risks like hurricanes or wildfires, there will be less demand for homes in areas exposed to those risks, and more investment in risk mitigation by the owners who remain.128 Where insurance premiums are not permitted to vary on such bases, insureds will have less incentive to prevent and avoid loss. This is called “moral hazard.”129 By allowing insureds to “externalize” the cost of their risky behavior onto the pool, insurance that fails to classify risks appropriately causes a wasteful increase in the aggregate amount of losses.130

The fairness-based version of this argument suggests that, while it would be unjust to allow insurers to charge more based on unchosen characteristics like genetics, it is entirely proper to raise prices on insureds whose choices increase their expected loss.131 Doing so prevents insureds who make such choices from unfairly shifting responsibility for the costs of their choices onto others.132 Any insurance regulation that prohibits insurers from doing so therefore interferes with conceptions of autonomy and responsibility that are central to luck-egalitarian distributive justice.

The problem with both arguments is not that they are incorrect. Surely, there are certain things that are within an insured’s control, on the basis of which it would be both efficient and fair for an insurer to raise premiums. Choosing high-risk hobbies like skydiving, or failing to replace the batteries on one’s smoke detectors, come to mind. The problem, rather, is that disagreements about the propriety of risk classifications generally arise where the extent of insureds’ control over the relevant trait is deeply contested.133 As Ronen Avraham, Kyle Logue, and Dan Schwarcz have put it, in the context of health insurance,

[I]t is theoretically possible to hold people responsible only for health features that involve choice, such as smoking, eating, and working out. But even in these domains, it is difficult to determine what choice means. Much behavior that seems voluntary may actually be the result of habits adopted in young age or addiction and, in any event, is highly correlated with numerous social factors, such as growing up in poverty or in a particular cultural setting.134

The same arguments can be made about differences in the life expectancy of men versus women,135 or the responsibility that homeowners should bear for living in an area exposed to climate risks.136 Thus, this type of argument for risk-based insurance pricing is only helpful where it isn’t necessary—i.e., where efficiency and fairness uncontroversially point in the same direction.

C. “Carry One’s Own Weight” Arguments Prove Too Much

Another argument often made in favor of actuarial fairness is grounded in the proposition that each insured ought to “carry his own weight.” Spencer Kimball, for example, posits that “the goal of the ratemaking process is to discriminate fairly—to measure as accurately as is practicable the burden shifted to the insurance fund by the policyholder and to charge exactly for it, no more and no less.”137

The problem with this argument is not just that it is conclusory (Why is it “fair” for every insured to pay an amount equivalent to her expected cost of coverage? Which costs are “hers” in the relevant sense? Etc.). The more important problem is that, when the principle is framed this way, there is nothing about it that suggests it ought only to be applied to insurance transactions rather than to commerce generally. But, as noted above, such a moral principle is conspicuously absent in market economies—i.e., there is no general norm that each purchaser is entitled, as a matter of justice, to a price equivalent to the marginal cost of producing the unit of goods she purchased.138 Though something like this may sometimes happen, such as when delivery services charge more based on the distance to be traveled, there are also many situations in which producers charge “flat-rate” prices that smooth differences in the costs of serving different customers.139 Indeed, most markets do not supply a framework for assigning the “weight” of production costs to various consumers.140 Instead, traditional welfare economics simply asks whether the transactions that produce uniform market-clearing prices are beneficial for all parties involved—i.e., whether the market equilibrium is Pareto efficient.141 If so, then no participant has any legitimate claim to a price corresponding to his “own weight.”142 The “carry one’s own weight” principle therefore fails because it proves too much.

D. Libertarian Arguments Are Morally Unappealing

It is also sometimes argued143 that the most coherent non-economic argument in favor of risk-based insurance pricing is grounded in a type of libertarian political theory often associated with the work of Robert Nozick.144 Nozick proposes a theory of property rights grounded in a principle of self-ownership, which provides that even if someone’s “natural talents and abilities” are morally arbitrary in the sense that she did nothing to deserve them, she nonetheless has a right to the benefits of those attributes.145 Nozick then couples that with a separate principle that individuals have a right to transact with others, and that the terms of those transactions are necessarily just. The result is that the talented, the healthy, the good-looking, etc., are able to benefit more than the rest from their participation in the marketplace, and that this is acceptable as a matter of distributive justice.146 This view suggests that individuals with low levels of risk have a liberty interest in being able freely to associate with others of similarly low risk, so they can benefit from insurance coverage at low rates.147 As Tom Baker put it, this conception of “[a]ctuarial fairness saddles people with all the consequences of their high risk status, whether deserved or not. Conversely, it entitles other people to all the benefits of their low risk status, also whether deserved or not.”148

Such a theory is straightforwardly anti-egalitarian.149 Political theorists like Rawls, Dworkin, and Cohen have, often in explicit opposition to Nozick, grounded their theories of justice in the foundational proposition that the outcome of the natural lottery is not something that individuals deserve, and therefore that people have no moral claim to all the fruits of their natural endowments.150 For readers who share this intuition, the libertarian argument for actuarial fairness will be unappealing. It suggests not just that insurance coverage ought to be priced according to individual risk levels, but also that society generally ought to be structured in a way that explicitly underwrites any manner of inequality that arises out of voluntary transactions. Such a proposition, it feels safe to say, is not in line with the contemporary American political morality that generally supports at least some degree of progressive redistribution through taxes.151

E. Comparing the Egalitarian Argument to Alternatives

The egalitarian argument for risk-based pricing succeeds where the arguments outlined above fail.

Whereas the efficiency-based argument focused on adverse selection cannot motivate a call for anything more than rough risk-based pricing, thus rendering it useless in the context of many of the debates about risk classifications, where deadweight loss is not practically a concern, the egalitarian argument can make the case for perfect actuarial fairness.

Whereas the arguments based on moral hazard and luck egalitarianism can only motivate a call for risk-based pricing in the context of relatively uncontroversial classifications, the egalitarian argument can make the case for risk-based pricing even when such classifications are based on characteristics over which the insured has no control, or where the extent of an insured’s control and responsibility for a given characteristic is a matter of debate.

Whereas arguments grounded in the principle that everybody “ought to carry his own weight” cannot explain why that principle should apply only to insurance rather than to all commercial transactions, the egalitarian argument relies on an essential distinction between insurance markets and other markets that makes it possible to argue for “fair” prices in insurance without entailing a generally-applicable theory of just prices.

Finally, whereas the libertarian argument can only motivate a call for risk-based pricing by relying on controversial and implausible propositions about the deservedness of unearned differences between individuals, the egalitarian argument for actuarial fairness does not rely on such notions. It focuses instead on the distribution of the surplus that is created through insurance. Because that surplus is the product of the cooperative enterprise that is made possible by all participants in an insurance arrangement, it is not deserved by one participant any more than by another. One can thus find the egalitarian argument for risk-based pricing convincing, while remaining committed to the proposition that broader redistribution is warranted to address undeserved inequality.

For these reasons, the egalitarian argument distinguishes itself as the most powerful motivation for risk-based pricing.

Tom Baker has pointed out that, even if there are moral arguments that “would obligate insurers to do more than classify applicants accurately within the context of their ordinary business arrangements,” the main proponents of risk-based pricing tend not to go so far.152 This is because the main proponents of risk-based pricing are not individual low-risk insureds claiming a moral right to a lower rate, but insurance companies seeking to skim low-risk insureds away from their competitors.153 Such insurers do not argue for permission to engage in more-accurate risk classifications than are profitable to them, but rather merely to apply the risk-classifications that suit their business interests.154 Nor do we vest low-risk insureds with a legal right to a price that accurately reflects their risk.155 This, Baker argues, suggests that “[i]f there is a fundamental moral principle at work here, it is . . . utility,” and therefore that there is no good faith moral argument in favor of risk-classification at work.156

This, I believe, is mistaken. The fact that insurance companies serve as the most visible defenders of risk-based insurance pricing does not mean that their reasons (i.e., profit) are the only reasons to support the practice. The fact that insurance companies do not classify insureds more accurately than is profitable, or the fact that we don’t vest insureds with a legal right to actuarially-fair prices, does not undermine the moral force of the argument for doing so. Instead, it simply shows that, especially in a world with meaningful transaction costs, market competition among insurers might not produce risk-classifications that exceed a certain level of accuracy.157 The alignment between efficient insurance prices and fairness that the egalitarian argument achieves breaks down, beyond this point, under such non-ideal conditions. How one weighs the efficiency costs of continuing to pursue ever-more-accurate risk classifications, against the egalitarian argument for doing so is a difficult policy question involving tradeoffs between efficiency and (“actuarial”) fairness. That we tend, in practice, to defer to insurers when determining how close we should try to get to actuarially fair prices shows that we prioritize efficiency over fairness when the two come into conflict. It does not show that there are no arguments for doing otherwise.

IV. Conclusion

This Article articulates a previously-untheorized argument against regulating insurance prices on antidiscrimination grounds—one that sounds not in efficiency, but in fairness and equality. The availability of this argument, even if it is not made with analytical clarity in everyday policy debates, helps to explain why previous analyses of the insurance classification controversy seem to miss something, namely, that many advocates for risk-based pricing appear to be motivated not only by efficiency concerns but also by a good-faith intuition that risk-based pricing is actually more fair than a system where customers are charged prices that do not depend on their likelihood of making a claim. By articulating this intuition, identifying its assumptions, and describing its strengths, this Article aims to reframe the relevant controversy not as a conflict between efficiency and fairness, but rather as a disagreement over the meaning of egalitarianism. Doing so can help to sharpen serious debates about the propriety of insurance antidiscrimination rules, and to render disagreements in this field more tractable to policymakers charged with ensuring that insurance prices are not “unfairly discriminatory.”

But, some might ask, why bother rendering any argument against antidiscrimination with analytical clarity? My answer to this question is as follows.

On the one hand, I think doing so valuably sheds light on an important and under-appreciated perspective on insurance: that its defining feature as a social practice is that it allows for a mode of redistribution that is, in a sense, conservative (in the sense that it only recommends Pareto-improving interactions rather than more aggressive redistribution to which the “haves” might not consent) and narrow (in the sense that it only aims to redistribute the gains from such interactions among those who participate in them), while still being substantively egalitarian (in a way that normal markets are not). There are virtues to such a minimal modality of redistribution. It is broadly palatable in diverse societies where more radical forms of redistribution will be difficult to enact through democratic means. Such minimal redistribution is, one might say, “better than nothing” from an egalitarian perspective. It also generates a type of social solidarity that is scarce in such diverse communities. It can bind strangers together, even in the absence of altruistic motives.158

On the other hand, a rigorous application of the “egalitarian” argument for risk-based pricing would be difficult to stomach. Allowing insurance companies to charge higher prices based on race, sex, genetics, or any other chance characteristic, would feel cruel. This is not just because insurance has become an essential good in contemporary society, and so raising its price can render a necessity unobtainable to those of modest means. It is also because such premium-price discrimination can look like an insult—indeed, in the case of highly-individualized insurance classifications, a personal one. Suggesting that unbridled insurance risk classification can be rendered morally acceptable by effecting more direct tax-and-transfer redistribution is not a satisfying response to either of these concerns. And in an era where legislatures seem chronically unable or unwilling to legislate, it is difficult to take the suggestion seriously. Moreover, the dignitary harms caused by, say, race-based insurance premiums are unlikely to be remedied by any taxpayer-funded redistribution designed to benefit a specific racial group.

In the end, I take this tension to be evidence of the inevitability of conflict between, on the one hand, a claim that insurance’s limitedness ought to be preserved in order to protect its ability to provide distinctive social goods, including not just consumer surplus but also a special type of social solidarity; and on the other hand, an intuition that treating individuals differently on the basis of underserved characteristics can infringe on their autonomy and dignity. The reconciliation of conflicting moral propositions like these is, inevitably, imperfect and messy. Conscience may lead reasonable people to weigh them differently, especially in different insurance contexts. It is my belief, however, that understanding these conflicts as conflicts is preferable to attempting to fit them within a single unified moral framework. The latter project will eventually and inevitably fail. It will also be perceived, in the meantime, as an academic attempt to gaslight those who hold reasonable but different moral commitments into believing that they simply do not grasp an idealized overarching theory capable of answering all moral questions. So long as humans remain imperfectly rational creatures, it is important to recognize the possibility of different moral perspectives, and to support democratic processes and competitive markets that aim, however imperfectly, to produce agreement rather than truth.

Appendix A

This Appendix contains a more technical statement of the egalitarian argument for risk-based pricing introduced in Section II.A. It relies on an innovative framework for modeling insurance markets, published in 2011 by Liran Einav and Amy Finkelstein,159 and adopted by many other economists since.160 Though the “EF Framework”161 caused a “sea change” in the economic study of insurance,162 it has had little impact on the legal academy’s discourse on insurance, nor has it been used to analyze the distributive consequences of insurance antidiscrimination rules.163

The EF Framework begins by making several simplifying assumptions about insurance markets.

Supply-side assumptions. The Framework assumes a marketplace of multiple insurers, indistinguishable from the perspective of prospective insureds, each offering a single, identical, insurance policy form. This means that the only choice prospective policyholders make is whether to purchase coverage;164 they do not have any reason to choose one insurer over another or one policy over another. The model also assumes that the insurers’ only cost of doing business is paying claims, and thus that there are no transaction costs or operational costs.165 Moreover, the model assumes there is perfect competition within the insurance industry, which requires that the market is not dominated by a single monopolist insurer with the power to set prices, and that the insurers earn essentially no profit.

Demand-side assumptions. As to insureds, the EF Framework assumes that the expected cost of paying an insured’s claims is the only source of heterogeneity among them, and that the insureds are otherwise identical.166 The model also allows for the assumption that insureds can accurately estimate their individual expected loss, and that they are risk averse.167

Figure 1a presents the EF Framework, a graphical representation of an insurance market characterized by these assumptions. The horizontal axis depicts the “quantity” of insurance that is purchased, with Qmax representing a market where all individuals purchase insurance. The vertical axis depicts the expected cost of providing coverage (to the insurer), as well as the cost of purchasing coverage (to the insured). Each point along the Demand curve represents an individual’s maximum willingness to pay for insurance.168 The Marginal Cost curve represents the expected cost of coverage for each individual insured, or put differently, the amount of losses that would be expected to be covered for each purchaser of insurance. The “Average Cost” curve denotes the average cost of insuring any quantity, Q, of insureds (i.e., all insureds to the left of any given point, Q, on the horizontal axis).

Because the model assumes that all individuals are risk-averse, the Demand Curve lies above the Marginal Cost curve. Because the model assumes no heterogeneity in levels of risk-aversion, the individuals willing to purchase insurance at a higher price, who are located on the left side of the horizontal axis, are also those who have a higher level of expected loss. Prospective insureds are thus sorted along the horizontal axis according to their amount of expected loss. These high-loss consumers are the ones with the highest willingness to pay for insurance, and vice-versa. (More formally, one can say that for any insured i and j, D(i) > D(j) if and only if MC(i) > MC(j).169) As the price falls, consumers with successively lower willingness-to-pay find it appealing to buy. The Marginal Cost thus slopes downwards because insureds with the higher expected losses are always to the left of insureds with lower expected losses. This positive correlation between willingness-to-pay and expected loss is what causes adverse selection.170

In a competitive market where all insureds are charged the same premium—perhaps because there is an antidiscrimination rule requiring community rating, or because insurers lack the information they would need to distinguish insureds based on their expected losses—the “equilibrium” price will be the point at which the Demand curve intersects the Average Price curve. In Figure 1a, this is denoted by Peqm. This price satisfies the condition that insurers make no profit; they charge an amount sufficient to cover the cost of paying claims, and no more.171 At price Peqm, all individuals to the left of Qeqm will purchase coverage (because the price of insurance will be less than what they are willing to pay), while individuals to the right of Qeqm will not. This is true even though the latter group would like to purchase insurance if they could buy it at a price that reflected their lower level of expected loss. This illustrates how adverse selection, coupled with undifferentiated prices, can create deadweight loss, denoted by the shaded region to the right of Qeqm.172

To leverage the EF Framework to model the egalitarian argument for risk-based pricing, it is helpful to revise its assumptions temporarily, by assuming all individuals are very risk-averse, such that the Demand curve shifts upwards enough that it lies everywhere above the Average Cost curve. This is illustrated in Figures 2a and 2b.173 Under these conditions, if insurance is community-rated, then the equilibrium price Peqm will be equivalent to the average cost of insuring the entire population of insureds. This is denoted by the point where the Average Cost curve intersects the vertical axis on the right. Under these conditions, all individuals will choose to purchase coverage at the equilibrium price regardless of their risk profile, thereby eliminating the deadweight loss caused by adverse selection.

We can now measure the consumer surplus174 for any given insured. Figure 2a illustrates the consumer surplus that two insureds will receive, respectively, under conditions of community rating. An individual insured, i, will be willing to pay D(i) for coverage, but will only be charged the equilibrium price, Peqm, resulting in a surplus for her of D(i) - Peqm. By contrast, a less-risky individual insured, j, located towards the right of i on the horizontal axis, will be willing to pay D(j) for coverage, and will be charged the uniform price Peqm, resulting in a surplus for him of D(j) - Peqm. The relative size of these two surpluses is illustrated by the solid vertical line segments. As is visible on the Figure 2a, i’s surplus far exceeds j’s surplus.

Figure 2b illustrates the scenario in which, while keeping costs and demand the same, actuarially-fair prices are charged. Under these conditions, insured i is charged a price equivalent to the cost of providing her coverage, MC(i), which is equal to her expected loss, resulting in a surplus to her of D(i) - MC(i). Insured j, by the same logic, receives a surplus of D(j) - MC(j). As Figure 2b shows, although i’s surplus is still greater than j’s, the difference between them is smaller than it was under community rating. This narrowing of the differences in consumer surplus is a general result that does not depend on any other assumptions not specified above.

The egalitarian argument for risk-based pricing thus states that, for any two insureds i and j, where D(i) > D(j), the following inequality will be true:

(D(i) – Peqm) – (D(j) – Peqm) > (D(i) – MC(i)) – (D(j) – MC(j)).

The left side of this inequality measures the difference in consumer surplus enjoyed by the two different insureds under community rating, whereas the right side of the inequality measures the difference in consumer surplus enjoyed under risk-based pricing. The inequality reduces to MC(i) > MC(j), which is true by assumption if D(i) > D(j).

This helps to demonstrate why the argument for risk-based pricing is not motivated by the traditional concern with deadweight loss caused by adverse selection (see Section III.A). Because we temporarily assumed that all insureds are very risk averse, the move from community rating to risk-based pricing does not eliminate any deadweight loss. The shaded areas in Figures 3a and 3b, which denote the sum of all the consumer surplus enjoyed by the entire population of insureds, are equal in size.175 They are, however, differently shaped, which shows that the different pricing regimes distribute the surplus “pie” differently. The argument outlined above says, in geometrical terms, that the shaded area in Figure 3b will always be “less wedge shaped” than the shaded area in Figure 3a, and thus a more equal distribution. That difference in shape is the core of the egalitarian argument for risk-based insurance pricing. If the assumption about extreme risk-aversion were relaxed, the same statement about equality of distribution would still be true, but risk-based pricing would also eliminate deadweight loss. This would produce a separate, efficiency-based, argument in favor of risk-based pricing. The egalitarian argument might coincide with this argument, but it does not depend on it.

Appendix B

Figure 1b (reproduced below as Figure 4a) illustrates the way that community rating can create deadweight loss under adverse selection. Figure 4b shows an insurance market where the insurer is able to separate insureds into two groups based on their levels of expected loss. Charging the relevant community rates within these two groups eliminates deadweight loss as well as the possibility of death spirals. It is, however, a far cry from actuarially-fair pricing.

• See, e.g., Tex. Ins. Code Ann. § 560.002(b)(2)(C) (West 2009) (prohibiting insurance rates that are “unfairly discriminatory”); Conn. Gen. Stat. § 38a-686(a) (2019) (same); Cal. Ins. Code § 1861.05 (West 1988) (same); Fla. Stat. Ann. § 627.062(1) (West 2024) (same). See also Daniel Schwarcz, Towards A Civil Rights Approach To Insurance Anti-Discrimination Law, 69 DePaul L. Rev. 657, 665–68 (2020) (discussing state insurance regulations containing prohibitions on “unfairly discriminatory” rates); Kenneth S. Abraham, Efficiency and Fairness in Insurance Risk Classification, 71 Va. L. Rev. 403, 406 (1985) (same).
• A long-running academic and public policy debate followed the Supreme Court’s decisions in City of Los Angeles, Department of Water and Power v. Manhart, 435 U.S. 702 (1978), and Arizona Governing Committee v. Norris, 463 U.S. 1073 (1983), which each dealt with the question of whether federal employment discrimination law banned sex classifications for employer-sponsored benefits. See Spencer L. Kimball, Reverse Sex Discrimination: Manhart, 4 Am. Bar Found. Rsch. J. 83 (1979); Lea Brilmayer, Richard W. Hekeler, Douglas Laycock & Teresa A. Sullivan, Sex Discrimination in Employer-Sponsored Insurance Plans: A Legal and Demographic Analysis, 47 U. Chi. L. Rev. 505 (1980); George J. Benston, The Economics of Gender Discrimination in Employee Fringe Benefits: Manhart Revisited, 49 U. Chi. L. Rev. 489 (1982); Lea Brilmayer, Douglas Laycock & Teresa A. Sullivan, The Efficient Use of Group Averages as Nondiscrimination: A Rejoinder to Professor Benston, 50 U. Chi. L. Rev. 222 (1983); George J. Benston, Discrimination and Economic Efficiency in Employee Fringe Benefits: A Clarification of Issues and a Response to Professors Brilmayer, Laycock, and Sullivan, 50 U. Chi. L. Rev. 250 (1983). For a more thorough review of this debate and the many other contributions to the dialogue, see Leah Wortham, Insurance Classification: Too Important to Be Left to the Actuaries, 19 U. Mich. J.L. Reform 349, 356–58 (1985).
• See Katy Chi-Wen Li, The Private Insurance Industry’s Tactics Against Suspected Homosexuals: Redlining Based on Occupation, Residence and Marital Status, 22 Am. J.L. & Med. 477, 477 (1996); Norman Daniels, Insurability and the HIV Epidemic: Ethical Issues in Underwriting, 68 Milbank Q. 497 (1990); Joyce Nixson Hoffman & Elizabeth Zieser Kincaid, AIDS: The Challenge to Life and Health Insurers’ Freedom of Contract, 35 Drake L. Rev. 709, 710–11 (1987). [Cited in Hellman 1997 at 355].
• See Deborah S. Hellman, Is Actuarially Fair Insurance Pricing Actually Fair?: A Case Study in Insuring Battered Women, 32 Harv. C.R.-C.L. L. Rev. 355, 355–57 (1997).
• The Health Insurance Portability and Accountability Act of 1996 (HIPAA) requires group insurers to meet specific exceptions to deny health coverage due to an insured’s preexisting condition. See 29 U.S.C. § 1181 (2018). The Patient Protection and Affordable Care Act of 2010 (ACA) prohibits all health insurers from denying coverage on the basis of preexisting conditions. See 42 U.S.C. §§ 300gg-3 to -4. For a history of the debate over this issue, see Sallie Thieme Sanford, The Struggle to Bury Pre-Existing Condition Consideration, 7 St. Louis U.J. Health L. & Pol’y 405 (2013) (detailing the arguments for and against the ACA’s mandate to prohibit insurer discrimination based on preexisting conditions); Jennifer M. Franco, Undermining the Protection of Health Insurance: The Preexisting Condition Clause, 30 New Eng. L. Rev. 883 (1995).
• The Genetic Information Nondiscrimination Act of 2008 protects individuals from genetic discrimination in health insurance and employment. Pub. L. No. 110-233, 122 Stat. 881 (2008) (codified as amended at scattered sections of 26, 29, & 42 U.S.C.). For a discussion of genetic discrimination in insurance, see Michael Hoy & Michael Ruse, Regulating Genetic Information in Insurance Markets, 8 Risk Mgmt. & Ins. Rev. 211 (2005); Robert Lowe, Genetic Testing and Insurance: Apocalypse Now, 40 Drake L. Rev. 507 (1991).See Avraham et al 2014 at 198.
• See Kenneth S. Abraham, About the California Fires, Harv. L. Rev. Blog (Jan. 27, 2025), https://harvardlawreview.org/blog/2025/01/about-the-california-fires/; Congressional Budget Office, Climate Change, Disaster Risk, and Homeowner’s Insurance (Aug. 2024), https://www.cbo.gov/publication/60674; Lloyd Dixon, Allocating Costs for California Wildfires, Rand Corp. (July 24, 2019), https://www.rand.org/pubs/commentary/
  2019/07/allocating-costs-for-california-wildfires.html.
• On the origin of the term “community rating,” see Mark V. Pauly, The Welfare Economics of Community Rating, 37 J. Risk & Ins. 407, 407–08 (1970).
• See, e.g., Eric Mills Holmes, Solving the Insurance/Genetic Fair/Unfair Discrimination Dilemma in Light of the Human Genome Project, 85 Ky. L.J. 503, 563 (1996) (“Since one cannot choose one’s genetic make-up, arguably there should be no duty to pay more for insurance because of a poor genetic make-up”); Hellman, supra note 9, at 410–11; Deborah Stone, Beyond Moral Hazard: Insurance as a Moral Opportunity, in Embracing Risk: The Changing Culture Of Insurance And Responsibility (Tom Baker & Jonathan Simon eds., 2002); Regina Austin, The Insurance Classification Controversy, 131 U. Pa. L. Rev. 517 (1983).
• See, e.g., Michael J. Miller, Disparate Impact and Unfairly Discriminatory Insurance Rates, Cas. Actuarial Soc’y E-Forum, Winter 2009, at 276 (arguing that risk-based pricing is efficient); Colin S. Diver & Jane Maslow Cohen, Genophobia: What Is Wrong with Genetic Discrimination?, 149 U. Pa. L. Rev. 1439, 1445 (2001) (“[T]he effort to ban genetic discrimination within insurance and employment markets would, if successful, cause significant welfare losses due to the distortion of allocative efficiency.”); Richard A. Epstein, The Legal Regulation of Genetic Discrimination: Old Responses to New Technology, 74 Bos. U.L. Rev. 1, 21 (1994) (“[T]he anti-discrimination approach often leads to the adoption of general and neutral rules that are inefficient for the firm and society at large.”); George L. Priest, The Current Insurance Crisis and Modern Tort Law, 96 Yale L.J. 1521, 1545 (1987) (“It is a well-established implication of this approach that the more narrowly risk pools can be defined, the more broadly insurance can be offered in the society.”); Abraham, supra note 1, at 420–28 (reviewing efficiency-based arguments); Jan Mossin, Aspects of Rational Insurance Purchasing, 76 J. Pol. Econ. 553, 556 (1968).
• See Abraham, supra note 1, at 420-28 (reviewing these arguments).
• Id. at 404.
• See, e.g., Ronen Avraham, Kyle D. Logue & Daniel Schwarcz, Understanding Insurance Antidiscrimination Law, 87 S. Cal. L. Rev. 195, 201–202 (2013) (framing the issue as a “tradeoff[] between the ‘efficiency’ costs of regulation and the ‘fairness’ benefits”); Hoy & Ruse, supra note 13, at 212 (casting the problem as “a classic efficiency-equity trade-off”).
• See Kyle Logue & Ronen Avraham, Redistributing Optimally: Of Tax Rules, Legal Rules, and Insurance, 56 Tax L. Rev. 157, 204 (2003) (noting “[t]he historical disconnect” between the “fairness” and “efficiency” literatures).
• To my knowledge, there has been no sustained expression or exploration of what this Article calls “the egalitarian argument for risk-based insurance pricing.” The closest to a similar proposal of which I am aware is contained in a passing comment in Tom Baker, Kyle D. Logue & Chaim Saiman, Insurance Law and Policy: Cases and Materials 258 (5th ed. 2023) (“Note that the argument for the use of actuarially fair insurance premiums is very similar to the argument for the use of ‘benefit taxation’ among public finance scholars. Under a benefit tax, individual taxpayers are taxed according to the benefit they receive from the public expenditure being funded by the tax.”).
• See infra Section III.A.
• Note that this hypothetical health insurance policy, for the sake of simplicity, has no deductibles, copays, or co-insurance.
• For reasons that will be explained in more detail below, under normal circumstances, risk-based pricing will necessarily result in a more equal distribution of consumer surplus than community rating. See infra Appendix A.
• See infra Section II.B for further discussion of this assumption.
• See, e.g., John H. Cochrane, Health-Status Insurance: How Markets Can Provide Health Security, Cato Inst. Policy Analysis No. 633 (Feb. 18, 2009), https://papers.ssrn.com/abstract=1462376 (arguing that health insurance should be structured around individual risk-based pricing rather than forced transfers); Greg Ip, Why It Matters How We Define ‘Insurance’, Wall St. J. (Mar. 24, 2017), http://blogs.wsj.com/economics/2017/03/24/why-it-matters-how-we-define-insurance/; Dixon, supra note 7; Tom Baker, Containing the Promise of Insurance: Adverse Selection and Risk Classification, 9 Conn. Ins. L.J. 371, 383–84 (2003) (reviewing the history of fairness arguments against cross-subsidies).
• Like Tom Baker, in his seminal article on adverse selection, “[m]y goal . . . is . . . to focus attention on the morality that is implicit in arguments for and against risk classification.” Baker, supra note 20, at 375. Baker concluded that there was no intelligible egalitarian argument in favor of risk classification. Seeid. at 396 (describing the idea of actuarial fairness as “a watered-down form of liberalism that privileges individual interests over the common good and that privileges, above all, the interests of insurance institutions organized on its terms.”). This Article argues otherwise. See infra Section III.E.
• See Abraham, supra note 1, at 408 (defining expected loss as “the predicted probability that an insured will suffer a loss multiplied by the predicted severity of the loss”).
• Seeid. at 421 (“Other things equal, insurers strive to charge insureds in accord with expected costs, which equal their expected losses plus a portion of the other costs of providing coverage.”).
• Seeid. at 413–14.
• Seeid.
• Seeid. at 414.
• For a definition of the term “risk classification,” see id. at 403. In this Article, I use the term “risk classification” to describe the practice of classifying insureds based on differences in their levels of expected loss. Risk classification is necessary for the practices of “risk-based pricing,” and “risk-based underwriting,” i.e., setting prices that vary depending on risk classifications. See also Baker, supra note 20, at 376 (“Insurance risk classification is the process of sorting insurance applicants into categories believed to correspond to differences in expected risk. Common examples include sorting life insurance applicants by age, health insurance applicants by health status, workers compensation insurance applicants by type of industry, and property insurance applicants by the nature of the construction of the property to be insured (e.g. wood versus brick).”).
• For a canonical statement of this competitive dynamic in insurance, see Mark V. Pauly, Is Cream-Skimming a Problem for the Competitive Medical Market?, 3 J. Health Econ. 87 (1984). Seealso Logue & Avraham, supra note 14, at 211 (“Insurers have an economic incentive to classify or segregate individuals into relatively narrow risk pools. Put differently, it is in the nature of a competitive insurance market that insurance companies will be induced to charge each individual insured a premium that approximates that insured’s expected costs. Economists sometimes (and insurance companies frequently) refer to such premiums as ‘actuarially fair’ premiums. If an insurer fails to price its policies actuarially fairly, other insurers will compete away the first insurer’s business.”); Baker, supra note 20, at 377 (“An insurer that discovers a new way to identify and exclude high risks improves its competitive position in two ways: it lowers its average risk and, assuming the people it rejects go elsewhere, it increases the average risk of its competitors. This competitive power of risk classification produces a classification ‘arms race,’ in which insurers either maintain their classification edge or face the loss of low risks to the competition and the migration of the high risks to their insurance rolls.”); Abraham, supra note 1, at 408 (“The more refined (and accurate) an insurer’s risk classifications, the more capable it is of “skimming” good risks away from insurers whose classifications are less refined. If other insurers do not respond, either by refining their own classifications or by raising prices and catering mainly to high risks, their “book” of risks will contain a higher mixture of poor risks who are still being charged premiums calculated for average risks.”); Leah Wortham, The Economics of Insurance Classification: The Sound of One Invisible Hand Clapping, 47 Oh. State L.J. 835, 860 (1986) (“In the insurance market . . . competition pressed prices to the actuarially fair premium, that is, expected loss.”).

  Note that empirical evidence shows that consumer inertia—when policyholders stick to their old insurance choice even when they could switch to a better alternative—can hinder such cream-skimming efforts. See Benjamin R. Handel, Adverse Selection and Inertia in Health Insurance Markets: When Nudging Hurts, 103 Am. Econ. Rev. 2643 (2013) (finding that policyholders tend not to switch to better insurance alternatives); Alma Cohen & Peter Siegelman, Testing for Adverse Selection in Insurance Markets, 77 J. Risk and Ins. 39, 65–66 (2010) (discussing reasons why policyholders might fail to adjust their purchasing decisions based on their risk profiles).

• See Peter Siegelman, Adverse Selection in Insurance Markets: An Exaggerated Threat, 113 Yale L.J. 1223, 1254–63 (2004) (reviewing the literature on insurance and evidence for insurance death spirals). Note, there is a subtle distinction between different types of insurance “death spirals.” One is the phenomenon, caused by adverse selection, whereby all insurers in a given market will enter a death spiral, as low risks choose to self-insure. See Liran Einav & Amy Finkelstein, Selection in Insurance Markets: Theory and Empirics in Pictures, 25 J. Econ. Persps. 115, 120 (2011). Another is the phenomenon whereby an individual insurer fails to keep up with risk-based underwriting standards and experiences a death spiral as its policyholders leave to other, more competitive, insurers. See Baker, supra note 20, at 378 (describing this dynamic as “insurer-side adverse selection”); Casper H. de Jong, Risk Classification and the Balance of Information in Insurance; an Alternative Interpretation of the Evidence, 24 Risk Mgmt. & Ins. Rev. 445, 457 (2021).
• See Keith J. Crocker & Arthur Snow, The Theory Of Risk‐Classification, in Handbook of Insurance 281, 281 (Georges Dionne ed., 2nd ed. 2013); J. David Cummins et al., Risk Classification in Life Insurance 30–35 (1983); DeJong, supra note 29, at 449 (reviewing the economics literature on this point, and proposing that “[i]n a competitive environment and regulation permitting there will be a tendency towards ever more refined risk classification.”); Bertrand Villeneuve, Competition between Insurers with Superior Information, 49 Eur. Econ. Rev. 321, 323 (2005) (“[T]he common view is that competition should result in full revelation of available information (the consumer’s type) and first-best efficiency (full insurance at an actuarially fair rate).”). See also David A. Cather, Cream Skimming: Innovations in Insurance Risk Classification and Adverse Selection, 21 Risk Mgmt. & Ins. Rev. 335, 335 (2018) (noting instances in the auto insurance context where cream skimming occurred); Am. Acad. of Actuaries, On Risk Classification (Nov. 2011), available at https://www.actuary.org/sites/default/files/files/publications/RCWG_RiskMonograph_Nov2011.pdf (noting that the introduction of smoking as a risk factor in US life insurance forced all companies to offer different rates on that basis).
• Scholars focusing on this problem have tended to conclude that the term “actuarial fairness” has a question-begging, conclusory quality. In what sense are perfectly risk-based prices “fair”? Why should we care whether something appears fair from an “actuarial” standpoint? Etc. The general consensus has been that “actuarial fairness” in insurance premium pricing has no intrinsic connection to more common notions of fairness, but rather is merely another way of claiming that risk-based prices are efficient. See, e.g., Avraham et al., supra note 13, at 203 (“[S]upporters of risk classification have co-opted the term ‘actuarially fair’ insurance, which has long been used by economists to describe insurance that is priced at expected cost. Despite these complications, [we] summarize the extant normative literature on risk classification and insurance discrimination through the ‘efficiency vs. fairness’ lens.”); Baker, supra note 20, at 396 (“The particular moral vision is that of ‘actuarial fairness’—-a watered down form of liberalism that privileges individual interests over the common good and that privileges, above all, the interests of insurance institutions organized on its terms.”). This Article aims to explain the sense in which actuarially fair prices truly are “fair” according to a reasonable conception of distributive justice.

  For what may be the earliest use of the term, “actuarial fairness,” see Mark V. Pauly, The Economics of Moral Hazard: Comment, 58 Am. Econ. Rev. 531, 532 (1968).

• See Abraham, supra note 1, at 422.
• See id. at 423. Note that another reason why insurers cannot feasibly charge perfectly-actuarially-fair rates is because, in reality, they have operating costs which must be recovered out of premiums (or, possibly, from income gained investing the premiums they hold while awaiting claims). Such costs tend to require insurers to charge an amount that exceeds the actuarially fair rate. Seeid. at 426 (discussing the practice of premium “loading” to allow insurers to cover operating expenses).
• See generally Sharona Hoffman, The Importance of Immutability in Employment Discrimination Law, 52 Wm. & Mary L. Rev. 1483 (2011) (arguing that federal antidiscrimination laws, including those covering race, sex, age, and genetics, are unified by the principle of protecting individuals from discrimination based on immutable traits for which they are not responsible); see, e.g.,Frontiero v. Richardson, 411 U.S. 677, 686 (1973) (describing sex as “an immutable characteristic determined solely by the accident of birth” for which the individual carries no responsibility).
• See Hellman, supra note 4, at 402 (1997); Holmes, supra note 9, at 563; Stone, supra note 9, at 287; Daniels, supra note 3, at 497 (1990). In 2011, the Court of Justice of the European Union adopted this perspective when it determined that insurers violated European Union antidiscrimination law when they used gender as a factor to calculate premiums and benefits. Case C-236/09, Association Belge des Consommateurs Test-Achats v. Conseil des Ministres, 2 C.M.L.R. 38 (2011). See Geert De Baere & Eveline Goessens, Gender Differentiation in Insurance Contracts after the Judgment in Case C-236/09, Association Belge Des Consommaeurs Test-Achats ASBL v. Conseil Des Ministres Case Law, 18 Colum. J. Eur. L. 339 (2011) (discussing the Test-Achats decision).
• The Affordable Care Act and HIPAA prohibit consideration of preexisting conditions and gender. See Patient Protection and Affordable Care Act, Pub. L. No. 111-148, § 1201, 124 Stat. 119, 154-55 (2010) (codified at 42 U.S.C. § 300gg-3) (prohibiting the exclusion of preexisting conditions); id. § 1201 (codified at 42 U.S.C. § 300gg-4) (prohibiting discrimination based on health status); id. § 2701(a)(1)A) (codified at 42 U.S.C. § 300gg) (omitting gender as a permissible factor for health insurers to consider). The Genetic Information Nondiscrimination Act bars genetic discrimination in health insurance. Genetic Information Nondiscrimination Act of 2008, Pub. L. No. 110-233, 122 Stat. 881. In housing, Department of Housing and Urban Development (HUD) regulations and the Fair Housing Act prohibit homeowners insurance practices with disparate racial impacts. 24 C.F.R. § 100.70(d)(4) (implementing Fair Housing Act). For a review of the federal laws prohibiting insurance discrimination, see Avraham et al., supra note 12, at 198–99.
• See Avraham et al., supra note 14, at 199 (“Besides [the ACA, HIPAA, GINA, and HUD regulations], there are no federal laws expressly forbidding insurers from engaging in any form of discrimination in the underwriting process.”).
• See Jonathan R. Macey & Geoffrey P. Miller, The McCarran-Ferguson Act of 1945: Reconceiving the Federal Role in Insurance Regulation, 68 N.Y.U. L. Rev. 13, 14–17 (1993).
• Under the McCarran-Ferguson Act, passed in 1945, federal laws that affect insurance are “reverse preempted” by any conflicting state law, unless the federal law expressly provides that it is meant to apply to insurance. See 15 U.S.C. § 1012(b) (2012). See generally Jonathan R. Macey & Geoffrey P. Miller, The McCarran-Ferguson Act of 1945: Reconceiving the Federal Role in Insurance Regulation, 68 N.Y.U. L. Rev. 13 (1993) (providing a history of the McCarran-Ferguson Act).
• Avraham et al., supra note 14, at 201. See alsoid. at 201 (“[M]ore than half the jurisdictions do not ban the use of race in life, health, and disability insurance, twenty-three states do not ban its use in auto insurance, and seventeen do not ban its use for property/casualty insurance, which includes homeowners insurance. Similar statements can be made about national origin and religion. We also found similar gaps in state laws for other policyholder characteristics: only fifteen states ban the use of sexual orientation in health insurance and only nine states ban the use of gender in auto insurance. As all this suggests, affirmative bans of insurer discrimination on the basis of potentially suspect policyholder traits are quite rare.”).
• See, e.g., Title VII of the Civil Rights Act of 1964, 42 U.S.C. §§ 2000e to 2000e-17; Age Discrimination in Employment Act of 1967 (ADEA), 29 U.S.C. §§ 621–34; Americans with Disabilities Act of 1990 (ADA), 42 U.S.C. §§ 12101–03; Equal Pay Act of 1963, 29 U.S.C. § 206(d); Pregnancy Discrimination Act of 1978, 42 U.S.C. § 2000e(k). See also U.S. Const. amend. XIV, § 1 (guaranteeing equal protection for all persons).
• See, e.g., Robert K. Yass, Homeowner’s Insurance and Credit Score: A Critical Race Theory Perspective, 27 Conn. Ins. L.J. 286 (2020); Gregory D. Squires & Charis E. Kubrin, Racial Profiling, Insurance Style, 24 J. Ins. Regul. 33 (2006) (discussing the use of variables as a proxy for rate in insurance underwriting); Li, supra note 3, at 477 (1996) (discussing insurer discrimination during the HIV epidemic against male hairdressers).
• Indeed, state laws prohibiting “unfair discrimination” in insurance markets have generally been interpreted to require only that there is an actuarial justification for a discriminatory practice. See Schwarcz, supra note 1, at 659.
• See Avraham et al., supra note 14, at 267 (reflecting, at the conclusion of a comprehensive survey of U.S. insurance discrimination rules, on “the stunted development of state insurance law and regulation on the topic”).
• See Hellman, supra note 4, at 354–57 (reviewing the outcry against insurers discriminating against battered women).
• See Daniels, supra note 3.
• See Sanford, supra note 5, at 405 (detailing the arguments for and against the ACA’s mandate to prohibit insurer discrimination based on preexisting conditions).
• See U.S. Dep’t of Housing & Urban Development, Discriminatory Effects Final Rule Fact Sheet (2023), https://www.hud.gov/sites/dfiles/FHEO/documents
  /DE_Final_Rule_Fact_Sheet.pdf.
• Cf. Baker, supra note 20, at 382 (noting the need for an explanation for the “social decision to leave decisions about insurance risk classification in the hands of insurance companies”).
• For a definition and explanation of the concept of consumer surplus, see Varian, infra note 166, at 247–58; Mankiw, supra note 58, at 134–39.
• See Ronen Avraham, TheEconomics of Insurance Law: A Primer, 19 Conn. Ins. L.J. 29, 37 (2012) (defining risk aversion as “the preference for certainty over uncertainty with regard to future losses,” and providing a history of the concept).
• See Siegelman, supra note 29, at 1223 (“The phrase ‘adverse selection’ was originally coined by insurers to describe the process by which insureds utilize private knowledge of their own riskiness when deciding to buy or forgo insurance”); Liran Einav & Amy Finkelstein, Empirical Analysis of Selection and Welfare in Insurance Markets: A Self-Indulgent Survey, 48 Geneva Risk Ins. Rev. 167, 170 (2023) (describing “the well-known adverse selection property of insurance markets: the individuals who have the highest willingness to pay for insurance are those with the highest expected costs”); Avraham, supra note 51, at 44.
• Though the opposite correlation (i.e., where the least-risky insureds are also willing to pay the most for coverage, a phenomenon sometimes called “propitious selection”) is possible, adverse selection is generally accepted as a standard and reasonable assumption to make about consumer preferences in insurance markets. See Siegelman supra note 29, at 1264. It also makes intuitive sense. So long as other differences between insureds are not introduced (e.g., differences in their levels of wealth that might cause the most-risky insureds to be the least-able to spend money on things like insurance), it is reasonable to expect that insureds who expect to lose more will place a higher value on protection.
• See Appendix A.
• For a discussion of different conceptions of well-being, see Matthew D. Adler, Measuring Social Welfare 10–11 (2019) (defining “individual well-being” as “how well an individual’s life goes” and discussing four ways of interpreting what this means).
• See Logue & Avraham, supra note 14, at 160–61 (reviewing economists and welfare scholars for whom “the ultimate target for a redistribution-minded policymaker [is] differences in overall well-being”). See also Adler, supra note 55, at 10–15 (specifying four different types of well-being comparisons).
• See Matthew D. Adler & Eric A. Posner, New Foundations of Cost-Benefit Analysis 40 (2006) (“For at least seventy years . . . the orthodox view in welfare economics has been that overall welfare is a meaningless notion [when comparing two outcomes that are ‘Pareto-noncomparable’”); id. at 41–42 (“[A] little reflection will show that the orthodox view is deeply counterintuitive. In one world, Felix feels fine but thousands die painful, premature deaths; in another world, the deaths and pain are averted but Felix has a headache. In one world, many workers lose their jobs, and suffer depression, family strife, or other ill effects of protracted unemployment; in another world, the unemployment doesn’t occur but a small number of top executives can now only afford to buy BMWs rather than Lamborghinis. Although the two worlds in these scenarios are Pareto-noncomparable, the second world in each scenario is (intuitively) much better for overall welfare than the first.”); id. at 42 (“The orthodox economic view that denies the possibility of interpersonal welfare comparisons thus leads to an ‘error theory’ of ordinary practice. It means seeing our pervasive practices, individual and governmental, of weighing the welfare gains to some persons against the welfare losses to others, as mistaken.”). Cf. Daniel Markovits & Gabriel Rauterberg, Contracts: Law, Theory, and Practice 21 n.6 (2018) (“Pareto efficiency has no conception of treating different people’s flourishing as equally good, because it makes no interpersonal comparisons of flourishing at all.”).
• This is a common assumption in the field of microeconomics, especially in the field of Industrial Organization. See, e.g., N. Gregory Mankiw, Principles of Microeconomics 138 (8th ed. 2017) (“Consumer surplus . . . measures the benefit that buyers receive from a good as the buyers themselves perceive it. Thus, consumer surplus is a good measure of economic well-being if policymakers want to satisfy the preferences of buyers.”); Edward E. Schlee, Expected Consumer’s Surplus as an Approximate Welfare Measure, 34 Econ. Theory 127, 127 (2008) (“Expected consumer’s surplus remains a popular measure of consumer welfare in applied microeconomic models with uncertainty.”); R.D. Willig, Consumer’s Surplus Without Apology. 66 Am. Econ. Rev. 589, 589–97 (1976) (arguing that the change in consumer’s surplus often approximates the willingness to pay for a price change). The assumption is not without critics, however. See, e.g., Mark Glick & Gabriel Lozada, The Erroneous Foundations of Law and Economics 84 (Inst. for New Econ. Thinking, Working Paper No. 149, 2021), https://papers-ssrn-com.ezproxy.law.uconn.edu/abstract=3812839 (arguing that consumer surplus is “not a credible measure of welfare”).
• See Logue & Avraham, supra note 14, at 204; Chris William Sanchirico, Deconstructing the New Efficiency Rationale, 86 Cornell L. Rev. 1003, 1020–22 (2001); Adler & Posner, supra note 57, at 160.
• For discussions of this tendency to use income as a proxy for well-being, see Logue & Avraham, supra note 14, 160–61; Daniel N. Shaviro, Inequality, Wealth, and Endowment Symposium on Wealth Taxes—Part II: Commentary, 53 Tax L. Rev. 397 (1999). See also, e.g., Louis Kaplow & Steven Shavell, Fairness versus Welfare 28–29 n.25 (2009) (collapsing the distinction between welfare and income “for convenience”).
• For discussion of such a move, see Mark Glick, The Unsound Theory Behind the Consumer (and Total) Welfare Goal in Antitrust, 63 Antitrust Bulletin 455 (2018); Yoon-Ho Alex Lee, The Efficiency Criterion for Securities Regulation: Investor Welfare or Total Surplus, 57 Ariz. L. Rev. 85 (2015); Barbara H. Fried, Fairness and the Consumption Tax, 44 Stan. L. Rev. 961, 983 (1991).

  This assumption has some interesting and perhaps surprising implications if one also adopts the separate assumption that individuals experience diminishing marginal utility from income and wealth. See Sarah B. Lawsky, On the Edge: Declining Marginal Utility and Tax Policy, 95 Minn. L. Rev. 904, 915–19 (2010) (defining declining marginal utility); Thomas D. Griffith, Progressive Taxation and Happiness, 45 B.C. L. Rev. 1363, 1395–98 (2004). Under such circumstances, the amounts of well-being that different individuals will gain from additional surplus dollars will be influenced by differences in their other sources of wealth. See John Bronsteen, Christopher Buccafusco & Jonathan S. Masur, Well-Being Analysis vs. Cost-Benefit Analysis Forty-Third Annual Administrative Law Symposium: A Happiness Approach to Cost-Benefit Analysis, 62 Duke L.J. 1603, 1627 (2012) (“Because of the diminishing marginal value of money, two individuals with differing levels of personal wealth can obtain vastly different amounts of welfare from the same gain (or loss) of income.”); Robert H. Frank, The Frame of Reference as a Public Good, 107 Econ. J. 1832, 1834–35 (1997) (discussing variation in the significance of income’s role in satisfaction across income levels). Under some circumstances, such as when the highest-risk insureds are also the most wealthy (i.e., where expected loss correlates positively with wealth), the egalitarian argument for risk-based pricing is weakened. A full exploration of the implications of wealth heterogeneity is beyond the scope of this Article. For the moment, however, it is possible avoid complicating the mapping of surplus dollars onto well-being by assuming that, aside from differences in expected loss, all individuals are otherwise identical. This is a standard assumption in economic models of insurance markets. See Appendix A.

• This narrow conception of egalitarianism is grounded in the influential strain of “social contract” theory that includes the work of John Rawls and David Gauthier. See David Gauthier, Morals by Agreement (1986); John Rawls, Justice as Fairness: A Restatement 75–76 (2001); Joseph Heath, On the Scope of Egalitarian Justice, 1 Les ateliers de l’éthique 21, 21–31 (2006). Such theories take the central question of distributive justice to be how to divide up the benefits and burdens of cooperation. There are multiple ways to split the surplus that cooperation creates, including dramatically unequal divisions that would nonetheless make all participants better-off than they would have been on their own. The question thus becomes, “What constitutes a fair distribution of cooperative surplus?” Social contract theorists argue that cooperative surplus ought to be distributed according to principles of justice that individuals would voluntarily choose before they enter the cooperative enterprise. This perspective has some appealing features. Because it requires individuals to consent to any principled method of distributing surplus, it allows one to argue that such principles are just, even if they are not efficient. And because it focuses narrowly on dividing surplus produced by cooperation—i.e., activities that produce gains which no single individual could achieve on her own—it can justify requiring an individual to split the fruits of cooperation, rather than claim all of them for herself. Cf. Joseph Heath, Morality, Competition, and the Firm 145 (2014); Rawls, supra note 62, at 178–79. It suggests the possibility of principles of distributive justice that are broadly acceptable and thus broadly applicable, even in pluralist societies. Such “contractarian“ theories of distributive justice would seem to be good models for any method of evaluating the fairness of institutions that are actually constituted through contract—including private insurance markets.
• See Heath, supra note 62,  at 147 (“[I]ndividuals . . . expect their particular interactions and private associations to be fair, above and beyond whatever contribution the outcome may make to the fairness of the entire society (so that, for instance, even in the context of an unjust society, particular institutions or domains of interaction might nevertheless be just).”). If it were only possible to evaluate the fairness of a specific cooperative enterprise—such as a commercial merger, a family dynamic, or a local zoning board vote—by first evaluating the fairness of society writ large, it would be difficult ever to gain confidence that one had a “complete enough picture” to evaluate the justice of any particular interaction. Seeid. at 31 (“Wide-scope egalitarianism runs contrary to the widely held intuition that it is possible for a particular group of individuals, dealing with a particular problem, to resolve it in a way that respects the principle of equality, independent of how every other domain of society is organized.”).
• Cf. Gauthier, supra note 62, at 153 (“The co-operative surplus is in the fullest sense the joint product of the co-operators No one may reasonably or fairly expect more, and no one should reasonably expect or fairly accept less, than an equal share of the co-operative surplus, where equal shares may be determined.”). Note that Gauthier’s method of dividing surplus is not strictly equally, but rather according to his more complicated rule of “minimax relative concession.” Seeid. at 113–56. See also Russell Hardin, Bargaining for Justice, 5 Soc. Phil. & Pol’y 65 (1988) (discussing Gauthier’s theory).
• See Werner Güth, Rolf Schmittberger & Bernd Schwarze, An Experimental Analysis of Ultimatum Bargaining, 3 J. Econ. Behav. & Org. 367 (1982) (the first ultimatum game experiment).
• See Richard H. Thaler, Anomalies: The Ultimatum Game, 2 J. Econ. Persps. 195, 196 (1988).
• See, e.g., Alvin E. Roth, Vesna Prasnikar, Masahiro Okuno-Fujiwara & Shmuel Zamir, Bargaining and Market Behavior in Jerusalem, Ljubljana, Pittsburgh, and Tokyo: An Experimental Study, 81 Am. Econ. Rev. 1068, 1082 (1991) (“In the United States . . . the modal proposal is 500 [out of 1000.]”). But see Joseph Henrich et al., In Search of Homo Economicus: Behavioral Experiments in 15 Small-Scale Societies, 91 Am. Econ. Rev. 73 (2001).
• David Gauthier’s minimax relative concession theory is a prime example. See Gauthier, supra note 62, at 113–56.
• See, e.g., Norman Daniels, Just Health Care 36–58 (1985) (arguing that health care is special because disease and disability restrict the range of opportunities open to individuals); see also Deborah A. Stone, The Struggle for the Soul of Health Insurance, 18 J. Health Pol. Pol’y & L. 287, 290 (1993) (discussing the conception of health insurance as social insurance rather than actuarial insurance).
• See 42 U.S.C. § 300gg (2018) (prohibiting discrimination based on health status); 42 U.S.C. § 300gg-4(a) (prohibiting exclusions for preexisting conditions).
• See Genetic Information Nondiscrimination Act of 2008, 42 U.S.C. § 300gg-53 (prohibiting health insurers from using genetic information for underwriting purposes).
• Cf. John H. Cochrane, Time-Consistent Health Insurance, 103 J. Pol. Econ. 445 (1995).
• Ronald Dworkin, What Is Equality? Part 2: Equality of Resources, 10 Phil. & Pub., Affs. 283, 293–304 (1981).
• Heath, supra note 62, at 358 (describing such transfers in the context of insurance cross-subsidies as “win-lose transformations”).
• Id.
• Id.
• Cf. Zachary Liscow, Is Efficiency Biased?, 85 U. Chicago L. Rev. 1649, 1664–66 (2018) (arguing that there is a plausible case that the distributional consequences of efficient legal rules will not always be offset by taxes and transfers).
• See Sarah A. Binder, Stalemate: Causes and Consequences of Legislative Gridlock (2004) (describing the causes and consequences of legislative gridlock in the United States).
• See Logue & Avraham, supra note 14, at 256 (raising a similar argument). Alternatively, one might argue that such “legislative transaction costs” are really indications that the proposed legislation merely lacks the necessary amount of democratic support—and thus that regulators or judges who pursue the same ends do so illegitimately. Seeid.
• See Kyle D. Logue, Reparations as Redistribution Symposium: The Jurisprudence of Slavery Reparations, 84 B.U. L. Rev. 1319, 1371–72 (2004) (“It is interesting to note that the expressive or symbolic objections to racial redistribution seem to diminish when the transfers are indirect (by geography, for example, rather than explicitly by race) or when they are structured as automatic cross-subsidization within insurance pools or lending markets. Why this is so is difficult to say. Perhaps voters do not understand that these sorts of programs implicitly involve racial redistribution . . . . Perhaps, [alternatively], people just view implicit transfers differently, as an empirical and even normative matter.”)
• See Daniel Markovits, Transactions Benefits, 38 Yale J. Reg. 633 (2021). Markovits outlines three types of benefits created by markets. First, markets generate “publicity benefits” when they produce market prices. In competitive markets, such prices establish a shared, public metric of what things are worth. This allows people who disagree fundamentally about values to agree on prices and thereby to coordinate economic activity. Second, contracts produce “legitimacy benefits” that sustain agreement even against backgrounds of injustice. “Contract law does not require setting the world right before contracts may be struck.” Id. at 650. Third, markets create “solidarity benefits.” Because parties are imperfectly rational, they must invest effort in understanding each other’s interests and intentions. This constitutes a form of mutual recognition. Id. at 654. As an illustration of this dynamic, Markovits points to Lisa Bernstein’s empirical work, in which she shows that extended negotiations and detailed contracting can deliberately build “trust-based social capital” as parties establish reciprocity norms that transcend mere self-interest. Id. at 655 (citing Lisa Bernstein, Beyond Relational Contracts:Social Capital and Network Governance in Procurement Contracts,7 J. Leg. Analysis 561, 589–90 n.98 (2015)).
• These benefits will be particularly important to anyone who is worried about increasing polarization, partisan manipulation of historically independent institutions, and political volatility that creates the conditions for escalating waves of reciprocating backlash.
• See Elizabeth Anderson, What is the Point of Equality?, 109 Ethics 287, 289 (1999) (coining the term “luck egalitarianism” to describe the conception of egalitarianism “which takes the fundamental injustice to be the natural inequality in the distribution of luck”); Samuel Scheffler, What Is Egalitarianism?, 31 Phil. & Pub. Aff. 5 (2003) (discussing luck egalitarianism).
• See Dworkin, supra note 73, at 293.
• See Richard J. Arneson, Equality and Equal Opportunity for Welfare, 56 Phil. Stud.: Int’l J. for Phil. Analytic Tradition 77 (1989).
• See G.A. Cohen, On The Currency of Egalitarian Justice, 99 Ethics 906 (1989).
• See Anderson, supra note 83, at 292.
• Cf. Logue & Avraham, supra note 14, at 182–85 (discussing “the problem of inaccuracy” that arises when legal rules are used to effect redistributive transfers, because the relevant categories of plaintiffs and defendants will not predictably correspond to the income classes that are the target of redistribution). See also Heath, supra note 14, at 148–49 (“[I]f one is to abandon the principle of actuarial fairness, one must do so across the board. And there are many cases in which doing so will not have progressive consequences.”).
• See Liscow, supra note 77, at 1653 (describing this position as a “mantra” among economic analysts).
• Id. at 1663 (“To lay observers, a more familiar example of this argument comes from trade policy. The longtime refrain from economists of (nearly) all stripes has been that countries should adopt free trade, notwithstanding potentially negative impacts on the poor, because trade increases the size of the economic pie, and those gains can be redistributed to the poor through taxes and transfers.”).
• Logue & Avraham, supra note 14, at 227; Logue, supra note 14, at 1369–70.
• Id. at 214.
• Id.See also Logue, supra note 80, at 1369 (“Simply forbidding the use of race—enforcing a norm of color blindness—automatically produces the desired level of cross-subsidization.”)
• Logue & Avraham, supra note 14, at 215.
• Deadweight loss is a cost to society created by market inefficiency, representing the loss of total social surplus that occurs when mutually beneficial trades are not made because the market is not operating at an equilibrium price and quantity. See, e.g., N. Gregory Mankiw, Principles of Microeconomics 77 156–62 (8th ed. 2017).
• Logue & Avraham, supra note 14, at 215.
• See Richard A. Epstein, The Legal Regulation of Genetic Discrimination: Old Responses to New Technology, 74 Bos. U.L. Rev. 1, 21 (1994).
• See Logue & Avraham, supra note 14, at 216.
• Seeid. at 215–17.
• See Deborah Hellman, When Is Discrimination Wrong? 23–27 (2008).
• Id.
• Hanoch Dagan & Avihay Dorfman, Justice in Private: Beyond the Rawlsian Framework, 37 Law & Philos. 171 (2018).
• Id. at 173–74.
• Id.
• See Tom Baker & Jonathan Simon, Embracing Risk, in Embracing Risk: The Changing Culture of Insurance and Responsibility (Tom Baker & Jonathan Simon eds. 2002).
• Logue & Avraham, supra note 14, 225 n.208. See also Abraham, supra note 1, at 441–45 (discussing suspect variables and the conditions under which they ought to be inadmissible in insurance).
• See Restatement (Second) of Contracts § 71 (Am L. Ins. 1981) (defining consideration without reference to equivalence of exchange); see alsoBatsakis v. Demotsis, 226 S.W.2d 673 (Tex. Civ. App. 1949) (enforcing promise to pay $2,000 in exchange for 500,000 drachmae worth approximately $25); Hamer v. Sidway, 27 N.E. 256, 257 (N.Y. 1891) (stating that courts “will not ask whether the thing which forms the consideration does in fact benefit the promisee or a third party, or is of any substantial value to any one.”).
• This ignores the obvious possibility of maldistribution of surplus as between consumers and producers, construed as classes. Whenever producers are able to engage in perfect price discrimination, they capture all of the social surplus. This may, rightly, be regarded as unfair. See Einer Elhauge, Tying, Bundled Discounts, and the Death of the Single Monopoly Profit Theory, 123 Harv. L. Rev. 397, 435 (2009) (explaining that perfect price discrimination transfers all consumer surplus to the seller and arguing that extracting individual consumer surplus is “unattractive on distributive grounds”).
• See Mankiw, supra note 95, at 144–45.
• See Markovits, supra note 81.
• See Liscow, supra note 77, at 1653.
• Liran Einav & Amy Finkelstein, Selection in Insurance Markets: Theory and Empirics in Pictures, 25 J. Econ. Persps. 115, 117–18 (2011).
• Id. at 118 (“The distinguishing feature of selection markets is that the demand and cost curves are tightly linked, because the individual’s risk type not only affects demand but also directly determines cost.”).
• See Liran Einav & Amy Finkelstein, Empirical Analysis of Selection and Welfare in Insurance Markets: A Self-Indulgent Survey, 48 Geneva Risk Ins. Rev. 167, 171 (2023) (“Marginal cost is driven by demand and thus varies with the selection of customers.”); Einav & Finkelstein, supra note 29, at 117–18 (“The link between the demand and cost curve is arguably the most important distinction of insurance markets . . . from traditional product markets. The shape of the cost curve is driven by the demand-side customer selection. In most other contexts, the demand curve and cost curve are independent objects; demand is determined by preferences and costs by production technology. The distinguishing feature of selection markets is that the demand and cost curves are tightly linked, because the individual’s risk type not only affects demand but also directly determines cost.”). See also Ronen Avraham, TheEconomics of Insurance Law: A Primer, 19 Conn. Ins. L.J. 29, 33 (2012) (making a similar point).
• Siegelman, Risky Business: A Review Essay on the Law & Economics of Selection in Insurance Markets, 30 Conn. Ins. L.J. 221, 253 (2024).
• See infra Appendix A, and specifically Figures 1a and 1b.
• See supra note 95, and accompanying text.
• Cf. Baker, supra note 20, at 383 (“The leading moral justifications for risk classification are the following: 1) without risk classification, low risks are unfairly forced to subsidize high risks; 2) risk classification promotes socially beneficial efforts to prevent loss; and 3) risk classification promotes individual responsibility.”).
• See supra note 95, and accompanying text. See, e.g., Priest, supra note 10, at 1541 (“Adverse selection is a problem central to every insurance context, and it dominates the insurance function.”). See also Cohen & Siegelman, supra note 28, at 42 (reviewing arguments made by economists, regulators, courts, and legal scholars, expressing concern about adverse selection); Siegelman, supra note 29, at 1226–34 (same).
• Cf. Baker, supra note 20, at 380–81 (“Faced with the choice between no insurance and expensive insurance, many or even most low risk applicants would not change their purchasing decision because of the higher price.”).
• See Avraham et al., supra note 13, at 204–09; Heath, supra note 62, at 370; Wortham, supra note 28, at 861. Note that this argument still assumes that individuals are reasonably risk averse, such that there is a relatively large “gap” between the Demand curve and the Marginal Cost curve. As risk-aversion decreases, and the gap between those two curves narrows, the classification schemes will need to approach perfectly-actuarially-fair classifications, in order to prevent efficiency losses.
• See supra note 28, and accompanying text.
• See Cohen & Siegelman, supra note 28, at 77–78 (finding mixed evidence of adverse selection, and finding very little documented evidence of insurance death spirals).
• See Johannes Spinnewijn, Heterogeneity, Demand for Insurance, and Adverse Selection, 9 Am. Econ. J.: Econ. Pol’y 308, 308–09 (2017) (reviewing various frictions that provide an explanation for empirical studies that show an individual’s risk type often plays only a limited role in explaining her demand for insurance, including limited cognitive ability, biased risk perceptions, inertia, and information frictions); Handel, supra note 28, at 2643; Justin Sydnor, (Over)Insuring Modest Risks, 2 Am. Econ. J.: Applied Econ. 177, 198 (2010). See also Pauly, supra note 28, at 93 (“One of the things that theory [suggests] is that only a little bit of adverse selection may cause market equilibrium to unravel. But then only a little bit of consumer inertia is needed to reinstate it.”).
• See Handel, supra note 28, at 26.
• See Abraham, supra note 1, at 413 (“In this context[,] ‘control’ encompasses not only the ability to conduct activities more safely, but also the capacity to vary levels of activity or production to reduce or prevent losses.”).
• See Baker, supra note 20, at 390 (listing “many instances in which risk classification is justified on the grounds of loss prevention”); see also Abraham, supra note 7, at 422 (“Because insurance is priced in accord with expected cost, insureds have the incentive to compare the cost of protecting against risk through insurance with the cost of reducing risk through loss prevention. Efficient classification discourages insureds from purchasing insurance when they can more cheaply protect against risk by investing in loss prevention. In contrast, inefficient classification may produce suboptimal loss prevention incentives. When coverage is priced below expected cost, for example, insureds may not take safety precautions that would otherwise be worthwhile. In this situation, they can obtain equivalent protection against risk by purchasing insurance at a lesser cost than the precautions.”).
• See Abraham, supra note 7; Howard C. Kunreuther & Erwann Michel-Kerjan, At War With The Weather: Managing Large-Scale Risks In A New Era Of Catastrophes 266 (2009).
• See Avraham, supra note 52, at 66–68; Tom Baker, On the Genealogy of Moral Hazard, 75 Tex. L. Rev. 237, 237–39 (1996).
• See, e.g., Avraham et al., supra note 52, at 208 (“Legal restrictions on insurers’ ability to classify risks can result in moral hazard, causing policyholders to take less than socially-optimal levels of care.”).
• See Logue & Avraham, supra note 14, at 212.
• Id.
• See Kenneth S. Abraham, Distributing Risk: Insurance, Legal Theory, and Public Policy 28 (1986) (stating, “controllability . . . [is a] normative conclusion[] about the risks for which individuals can properly be asked to bear insurance responsibility,” and citing Michael Sandel for the argument that these normative conclusions “are derived from premises about the personal characteristics that are properly considered individual assets and those that are properly considered the collective property of society.”); Abraham, supra note 1, at 438 (“The notion of ‘control’ is a normative conclusion about the characteristics for which individuals can properly be asked to bear insurance responsibility. This notion does not solve risk-distributional issues; rather, it directs one’s attention to them.”).
• Avraham et al., supra note 13, at 215.
• For instance, do women live longer because of genetics, or because they are more careful?
• See Abraham, supra note 133, at 28 (“Territorial classification of residential fire and automobile insurance, for example, might have to be tempered to account for the semidetermined character of people’s choices about where they live.”). Moreover, because moral hazard may be a goal of insurance (e.g., marine insurance encourages commercial risk-taking), there may be disagreements not over whether the insured actually exercises control, but rather over whether the increased risk is actually socially bad. See Logue & Avraham, supra note 14, at 210 (arguing that liability insurance for high-risk medical specialties like obstetrics is good because it creates moral hazard). This requires agreement about whether an increased risk is net good or net bad for society as a whole. Thus, to determine whether the tradeoff between insurance and loss prevention is efficient requires a broad view of all of the externalities of loss prevention, which is difficult to form, much less establish consensus about.
• Kimball, supra note 2, at 105. Many others have justified risk-based insurance pricing with reference to the same principle. See, e.g., Joseph Heath, Reasonable Restrictions on Underwriting, 7 Insurance Ethics for a More Ethical World 127, 132 (Patrick Flanagan et al. eds., 7th ed. 2007) (discussing “the principle . . . that the premium paid by an individual should be equal to that individual’s expected loss,” and describing it as “the natural thing”); id. at 150 (reframing the same principle in terms of cost-externalization); Richard A. Miller, Discrimination by Gender in Automobile Insurance: A Note on Hartford Accident and Indemnity Co. v. Insurance Commissioner, 23 Duq. L. Rev. 621, 624–25 (1984) (arguing that “costs [of production] should be reflected in the prices that consumers pay. . . . identical prices to different customers, despite different costs of supplying those customers, also involves discrimination” and interpreting the statutory term “unfairly discriminatory” to mean “rates not adequately related to costs of providing insurance”); Wortham, supra note 2, at Baker Logue casebook at 258 361 n.41 (reviewing similar arguments); Benston, supra note 2, at 498 (“The task, then, is to calculate how much each [insured] should contribute, so that each pays a ‘fair’ share—an amount that reflects only the cost of the risk each person imposes on the group.”); Herman T. Bailey, Theodore M. Hutchison & Gregg R. Narber, The Regulatory Challenge to Life Insurance Classification, 25 Drake L. Rev. 779, 781–82 (1975).
• Quite the opposite, in fact. Everyone pays the same price, based on the cost of producing the last unit of output—the unit consumed by the marginal consumer—not “their” unit. See Mankiw, supra note 58, at 139–40.
• Such a simplified pricing schedule may well be efficient, given the transaction costs involved in charging each consumer a different price.
• If the “carry one’s own weight” requirement were applied to market transactions generally and seriously, it would create a very thorny “what is the cost of what” problem. See Guido Calabresi, The Cost of Accidents: A Legal and Economic Analysis 198 (1970) (describing the “what-is-the-cost-of-what” problem).
• See Nicholas Barr, The Economics of the Welfare State 78 (1998) (discussing Pareto efficiency).
• Or any other conception of a “fair” or “just” price, for that matter. As Alasdair Macintyre put it, “What is necessarily absent in [ ] markets is any justice of desert. Concepts of a just wage and a just price necessarily have no application to transactions within those markets.” Alasdair Macintyre, Marxism & Christianity at x (1995). See also Heath, supra note 62, at 190; Friedrich A. Hayek, Competition as a Discovery Procedure, 5 Q.J. Austrian Econ. 9, 9–23 (Marcellus S. Snow trans.) (2002) (arguing that the central advantage of market economies over planned economies is that the latter, being subject to “conscious direction,” “would always have to aim for prices that are considered fair,” and yet “an economic system in which everyone received what others felt he deserved could not help but be a highly inefficient system”).
• See, e.g., Baker, supra note 20, at 393; Hellman, supra note 4, at 402.
• SeeRobert Nozick, Anarchy, State, and Utopia (1974).
• Id. at 213.
• See id. at 160–64 (discussing the example of Wilt Chamberlain).
• See, e.g., Daniels, supra note 2, at 504–05 (arguing that the argument for actuarially fair premiums rests upon a “controversial premise . . . that individuals should be free to pursue the economic advantage that derives from any of their individual traits, including their proneness to disease and disability”); Hellman, supra note 4, at 398–402.
• Baker, supra note 20, at 394.
• See John E. Roemer, Theories of Distributive Justice (1996) (describing Nozick’s Anarchy, State, and Utopia as “probably the most important antiegalitarian contribution to contemporary political philosophy.”).
• See, e.g., Rawls, supra note 62, at 74–75; Ronald Dworkin, Sovereign Virtue: The Theory and Practice of Equality 74–81 (2002); G.A. Cohen, On The Currency of Egalitarian Justice, 99 Ethics 906, 931 (1989) (“[A] large part of the fundamental egalitarian aim is to extinguish the influence of brute luck on distribution.”).
• Joseph Bankman & Thomas Griffith, Social Welfare and the Rate Structure: A New Look at Progressive Taxation, 75 Cal. L. Rev. 1905, 1906–09 (1987).
• Baker, supra note 20, at 394. See also, Abraham, supra note 1, (“The argument against inaccuracy in general seems therefore also to include the idea that there should be a right to purchase accurately priced coverage, even when purchasing is not required.”).
• Id. at 394 (“[T]he fairness argument has been mobilized in public policy debates, not to protect the rights of low risk individuals, but rather to promote the freedom of insurance organizations to classify insureds through any means they wish.”).
• Id. at 394.
• See id.
• Id. at 395. Cf. with Abraham, supra note 1, at 430 (“Critics of inaccura[te classifications] must be making [the argument that] when the amount of inaccuracy in the system is already economically optimal, . . . everyone should be charged more for coverage simply to assure greater accuracy in general. No one would favor such a system, however, because no one would benefit from it.”).
• For a statement of the assumptions underlying the EF Framework, see Appendix A.
• This stands in contrast to an appealing, but utopian, solution to the “classification controversy” proposed by Regina Austin. See Austin, supra note 9, at 581 (“The ultimate solution to the controversy must . . . lie in the achievement of real voluntary groups or communities whose members are engaged in important economic and social tasks relevant to the control of risks and the provision of support and resources for the victims of accidents and losses. These activities must be undertaken pursuant to a genuine consensus of values. . . . This solution requires far-reaching changes in the structure of economics, politics, kinship, and community.”).
• Einav & Finkelstein, supra note 29.
• See Liran Einav & Amy Finkelstein, Empirical Analysis of Selection and Welfare in Insurance Markets: A Self-Indulgent Survey, 48 Geneva Risk Ins. Rev. 167 (2023) (reviewing other uses of the EF Framework).
• See id. at 168 (acknowledging others who have referred to their framework as the “Einav-Finkelstein Framework” and abbreviating it to the “EF Framework”).
• Siegelman, supra note 115, at 232. See also Einav & Finkelstein, supra note 29. (reviewing subsequent studies that utilize their framework).
• But see Michael Geruso, Timothy Layton & Adam Leive, The Incidence of Adverse Selection: Theory and Evidence from Health Insurance Choices (Nat’l Bureau of Econ. Rsch., Working Paper No. 31435, 2023) (discussing a setting in which “adverse selection combines with income differences in demand for more generous insurance to create an equity-efficiency trade-off: reducing the efficiency losses from adverse selection involves a subsidy that disproportionately benefits higher-income consumers. Such a disparity in the incidence of subsidies used to correct selection was, to our knowledge, previously overlooked.”).
• As Einav and Finkelstein note, however, “[o]ne can think of the binary choice [represented in this chart] as one between no insurance and a given insurance product, or between lower-coverage and a higher-coverage insurance product.” Einav & Finkelstein, supra note 29, at 169.
• See id. at 169–70 (“We also assume that the firm costs are simply the insurance claims that they must pay out; in other words, we abstract from other firm costs, such as marketing their products or processing submitted claims.”); Siegelman, supra note 115, at 252 n.93 (describing this as “an entirely standard assumption in the economics of insurance.”). The EF Framework also assumes insurers are unable to earn revenue by investing the premiums they hold in anticipation of future losses. See Einav & Finkelstein, supra note 160, at 169. See also Priest, supra note 10, at 1528 (1987) (“Insurance company profits are the net of premium plus investment returns less losses (payouts and expenses).”)
• This is a standard assumption not just in insurance economics but in mainstream intermediate microeconomics. See, e.g., Hal R. Varian, Intermediate Microeconomics: A Modern Approach 215–33 (2006).
• Risk aversion means that each dollar that an individual loses from an unlucky loss is worth more to her than the previous dollar she lost. Intuitively, this makes sense: for each dollar that one loses, one moves further away from a merely inconvenient loss and closer to a loss that is a catastrophic disruption to one’s livelihood. Seeid. at 224–25 (defining risk aversion).
• See Einav & Finkelstein, supra note 160, at 169–70 (“The resulting demand curve is standard, and reflects (one minus) the cumulative distribution function of individuals’ willingness to pay for the insurance contract.”).
• See id. at 170–71 (“[i]n many settings, it seems natural to assume that individuals’ willingness to pay for insurance is” an increasing function of their expected costs.); Einav & Finkelstein, supra note 160, at 124. Intuitively, this means that the person with the lowest expected loss (located on the rightmost side of the horizontal axis) is willing to pay more for coverage than what she expects to lose in a given year, but that the overall amount she is willing to pay for coverage is less than the overall amount that the person with a greater expected loss (located to her left on the horizontal axis) is willing to pay. The same is true for any other two insureds located between those two points.
• See Einav & Finkelstein, supra note 160, at 170 (describing “the well-known adverse selection property of insurance markets: the individuals who have the highest willingness to pay for insurance are those with the highest expected costs”); Ronen Avraham, TheEconomics of Insurance Law: A Primer, 19 Conn. Ins. L.J. 29, 44 (2012).
• Note that the assumed competitiveness of the market, coupled with the other supply-side assumptions, reinforces this price from both above and below: any insurer that attempts to charge more than Peqm will be undercut by competitors and driven out of the market; any insurer that attempts to charge less than Peqm will have fewer premium dollars coming in than claims dollars going out, and will become insolvent.
• See Siegelman, supra note 115, at 253; Einav & Finkelstein, supra note 29, at 117. Moreover, if all insurers are subjected to the same community rating requirement, then the entire market can potentially enter a “death spiral.” See Siegelman, supra note 29, at 1254–58 (reviewing arguments about the possibility of such death spirals); Einav & Finkelstein, supra note 29, at 120.
• Cf. Einav & Finkelstein, supra note 29, at 119 (using a similar chart to illustrate the situation where “individuals’ risk aversion is high (that is, the demand curve lies well above the MC curve)”).
• For a definition and explanation of the concept of consumer surplus, see Varian, supra note 166, at 247–58; Mankiw, supra note 58, at 134–39. See also Alfred Marshall, Principles Of Economics 78, 241, 272 (1890) (for an early use of the term “surplus”). For further discussion of the ability of consumer surplus to serve as a proxy for well-being, see supra Section II.B.
• The equivalence of the size of the two shaded areas can easily be demonstrated geometrically.