Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Health Insurance Adverse Selection and Market Segmentation

Abstract. Health insurance selection is a joint outcome of heterogeneity in medical expenditure risk, heterogeneous preferences, the information available to households and insurers, the prices and contractual attributes of coverage, and the institutions governing entry and transfers among risk pools. This monograph examines how individuals and groups select into insurance and among plans, how insurers may shape membership through product design, and how these processes alter expenditure, welfare, access, and distribution. It derives the basic demand–average-cost distinction, separates adverse from advantageous selection, evaluates empirical identification strategies, and interprets evidence from employer-sponsored plans, Massachusetts reform, U.S. exchanges, Medicare Advantage, Medigap, Australia, and international pooling institutions. It assesses risk adjustment as an incentive mechanism rather than simply a prediction exercise; identifies the interaction between benefit design, formularies, enrollment frictions, subsidies, and competitive responses; and specifies how comparative findings should be transported across jurisdictions. The central conclusion is that selection is a multi-margin allocation process: the risk distribution of a plan is determined jointly by membership, benefit-specific expectations, switching, access, and the revenue rule. Policy evaluation must therefore examine residual plan profitability by risk type, welfare for both participants and nonparticipants, and the distribution of medically necessary services. All empirical results are attributed to their original settings. This study offers an original analytical synthesis without proposing a new theory or claiming an independently reproduced causal estimate.

Keywords: health insurance; adverse selection; advantageous selection; asymmetric information; market segmentation; risk pooling; enrollment; endogenous plan choice; health-plan competition; risk adjustment; Medicare Advantage; Medigap; Affordable Care Act; consumer inertia; welfare economics; insurance regulation; international health financing

Introduction

Health insurance joins uncertain future need for medical services to prepayment arrangements operating across households, employers, insurers, providers, and public institutions. Individuals arrive at an enrollment decision with different expectations of illness, differences in liquidity and risk tolerance, distinct clinical requirements, and unequal ability to understand products. Insurers face a different information set and compete through premiums, networks, covered services, formularies, customer service, and the financial attractiveness of the enrollees those features draw. The resulting allocation cannot be inferred from the stated premium alone.

The organizing research question is how this interaction determines the membership and welfare consequences of health insurance markets. A secondary question concerns institutional scale: a voluntary supplementary market, a mandatory standardized primary-coverage market, an employment-based menu, and a publicly financed entitlement may all have insurers and medical claims, yet the location at which selection operates differs. The OECD’s 2025 coverage classification distinguishes primary, complementary, supplementary, and duplicate forms of health insurance. That classification matters because a comparison between insurance products is meaningful only after the underlying entitlement is specified.

A third question concerns inference. A high-cost plan may attract patients who expect greater medical needs; coverage may also increase utilization or alter the provider price paid. Observing larger claims after enrollment identifies neither private information nor moral hazard by itself. Cohen and Siegelman (2010) examine the pitfalls of tests based solely on a positive correlation between coverage and realized risk. Einav and Finkelstein (2011) describe the welfare geometry of selection and clarify why demand and claims cost must be estimated jointly.

The analysis proceeds from contractual and mathematical foundations to enrollment, private information, welfare, research design, and international institutions. Each quantitative statement refers to the population and identification method of its source. The principal analytical contribution is an integrated reading of heterogeneous risk, demand-side choice, supply-side screening, and revenue equalization, while preserving the differences between structural models, quasi-experimental evidence, administrative reports, and legal requirements.

Conceptual and Historical Foundations

Information asymmetry and the insurance allocation

The early information-economics foundation is Akerlof’s (1970) account of quality uncertainty, which demonstrates how the distribution of unobservable product quality can disrupt exchange. Its specific automobile-market construction is not a calibrated health-insurance model; its enduring relevance here is the possibility that privately held information affects which transactions occur. Rothschild and Stiglitz (1976) subsequently analyzed competitive insurance contracts under imperfect information. Their setting includes risk types, premiums, coverage levels, and contracts that may reveal type through self-selection. Under the paper’s assumptions, a separating outcome may arise, and a competitive equilibrium need not exist. These are conditional model findings, not a prediction that every real health insurance market must collapse.

Information asymmetry has direction and content. An applicant may know symptoms, family history, expected care use, a treatment preference, or willingness to tolerate restricted networks that is unobserved by the insurer. An insurer may know more about negotiated provider prices, utilization-management processes, formulary changes, likely out-of-pocket expenditures, or future network stability than the applicant. In a voluntary product, the applicant’s private cost expectations influence demand; in a choice of standardized mandatory plans, detailed treatment preferences may determine sorting even when basic membership is compulsory. The same observed market share can therefore encode several distinct information relations.

For analytical clarity, define selection as a difference in the distribution of expected risk or otherwise revenue-relevant member characteristics across enrollment states or products that arises through entry, exit, choice, or insurer design. Adverse selection, from the standpoint of a given pooled premium or risk-adjusted revenue rule, occurs when enrollment disproportionately attracts members whose expected net claims burden is high relative to the revenue assigned to them. Advantageous selection occurs when the pool disproportionately attracts favorable expected net-cost types. These definitions specify the comparison and financing rule; they do not assign a permanent label to a person.

Distinguishing risk, cost, preference, and realized claims

Expected medical cost must be separated from a realized bill. Let C_i(j) denote medical spending under plan j for individual i. The potential outcome depends on health, access, utilization incentives, provider price schedules, and benefit design. Let m_i(j)=E[C_i(j)|I_i] be expected expenditure conditional on an explicitly defined information set I_i. For a premium p_j and benefit design b_j, the individual's utility comparison includes expected medical protection, risk exposure, product attributes, liquidity, and switching costs.

The insurer’s average claims cost is a conditional quantity, AC_j=E[C_i(j)|J_i=j], where J_i is realized plan choice. This average depends on the composition of enrollees and therefore generally changes when p_j or the set of plans changes. It is conceptually different from a fixed claims cost estimated on an arbitrary baseline sample. If coverage itself affects demand for care, its causal effect must be added to the composition effect. Einav, Finkelstein, and Cullen (2010) use price variation jointly with demand and costs precisely because movement along the insurance demand curve changes the composition of those insured.

Claims paid are also not equivalent to health need. Contract exclusions, deductibles, network boundaries, negotiated prices, delayed care, diagnostic coding, and unmet need all alter observed expenditure. A low claim total could reflect robust health or constrained access. A risk adjustment formula trained on realized spending may therefore reproduce existing patterns of service receipt while incompletely measuring clinical need. Consequently, a fair comparison of plan risk composition requires explicit cost definitions and appropriate access and outcome checks.

Institutional units and economic boundaries

There are several nested markets. An individual decides whether to be insured and which product to buy. An employer chooses a menu and contribution policy. A plan decides which providers, benefits, and administrative processes to offer. A public sponsor may define a basic benefit, premium rating limits, transfers between plans, and guaranteed acceptance. The insured population can be mandatory at one level but voluntary in its choice among competing insurers. Treating these margins as one undifferentiated decision obscures where selection actually occurs.

WHO guidance on pooling and fragmentation favors larger and more diverse compulsory pools because fragmented voluntary enrollment limits redistributive capacity. The WHO principle concerns financing architecture; it should not be mistaken for a claim that every compulsory scheme uses an identical insurer model. OECD comparative coverage definitions likewise show that the Netherlands and Switzerland achieve near-universal core coverage through strongly regulated compulsory private insurance, whereas other systems rely principally on national health services or social insurance. A supplementary private product in one country thus is not directly comparable with a regulated core-benefit policy in another.

Historical progression of the research object

The development of selection research can be described in four substantive stages. Foundational models establish potential failures of pooling under asymmetric information. Employer-plan investigations study how contribution and price changes move employees among competing plans. Individual-level administrative and survey research develops tests for private information and decision frictions. Regulated-market research then evaluates mandates, risk adjustment, and plan-design incentives. Cutler and Reber’s Harvard study is a key example of the second stage, whereas Geruso and Layton’s 2017 review articulates the policy instruments of the fourth.

This progression has an important analytical implication. A static equilibrium model supplies mechanisms, while a policy reform supplies a setting in which one or more mechanisms can be tested. Extrapolating a model's conditional possibilities as empirical prevalence, or interpreting a single employer's result as a universal parameter, exceeds the evidence. A mature comparative account must preserve both forms of knowledge.

Risk Composition and Enrollment Choices

Entry into coverage and sorting among plans

Selection has an extensive margin and an intensive margin. The extensive margin concerns whether a person enters a particular coverage pool at all. The intensive margin concerns which plan, benefit level, deductible, provider network, or additional insurance layer that individual selects. These margins may move differently. Subsidized mandatory core insurance can stabilize entry while preserving strong sorting among insurers; employer financing can stabilize participation while inducing migration between a generous and a restrictive plan.

For an enrollment population N, let q_j(p_j,x) denote the probability of choosing plan j under price p_j and observed variables x. Total enrollment is Q_j=Σ_i q_j. Its expected gross cost is K_j=Σ_i q_j m_i(j), with average expected cost AC_j=K_j/Q_j for Q_j>0. Even holding every m_i(j) constant, a price change alters AC_j when the response of enrollment varies by risk. If changes in b_j, provider prices, or treatment also change m_i(j), the composition and treatment effects must be estimated separately.

The conditional average-risk response can be written as dAC_j/dp_j=(1/Q_j)Σ_i (dq_j/dp_j)[m_i(j)−AC_j] when potential expenditures are held fixed. This identity has a straightforward interpretation: changes in enrollment matter for average costs according to the departing or entering individual’s cost relative to the incumbent mean. It is an algebraic description of selection, not an estimated behavioral law. The formula assumes differentiability and a fixed eligible population; real discontinuous enrollment windows and heterogeneous household contracts require discrete versions.

Premium contributions, employer menus, and price salience

A plan premium is not necessarily the price the household faces. Employer contributions can vary by plan; income-related subsidies change marginal out-of-pocket premium costs; tax provisions alter effective prices; and family enrollment can bundle choices. The object of estimation should be the premium increment that changes the actual decision, together with deductibles, copayments, and nonprice characteristics.

Cutler and Reber (1998) document the consequences of a Harvard University contribution reform that changed the relative attractiveness of competing plans. Their published analysis estimated selection-related welfare losses of 2–4 percent of baseline spending and premium reductions associated with competition of 5–8 percent in that setting. It also showed why a premium decrease paid for by reducing insurer profits does not automatically offset a social welfare loss from selection: transfers among purchasers, employers, and insurers must be distinguished from changes in total surplus.

Bundorf, Levin, and Mahoney (2012) use small-employer plan choice data to estimate pricing-related losses, reporting estimates equivalent to 2–11 percent of coverage costs relative to a benchmark allowing risk rating. The benchmark is a useful economic comparison but does not mean individualized risk rating is a complete or legally desirable social policy. Such rating changes the distribution of reclassification risk and financial protection across healthy and ill households.

Health status, anticipated use, and contract attributes

Two applicants with identical diagnoses can have different willingness to buy a product because they expect different care, trust different providers, or place different values on protection against tail expenditure. A person with an expensive prescription may value a specific formulary more than a generically generous actuarial value. A family may choose a plan because an important specialist is in network. Contract characteristics can simultaneously increase legitimate treatment value and alter a plan's selection incentive.

Geruso, Layton, and Prinz (2019) find evidence in U.S. ACA exchanges that formulary design can screen consumers, despite risk adjustment and reinsurance offsetting much of the expected cost differences. Their analysis identifies drug-class-specific residual incentives that can make particular users systematically unprofitable. This is evidence about the examined exchange context and formularies, not proof that any restricted formulary is deliberately designed to select risk. Inferential caution requires combining product change, predictable profitability, and credible behavioral or contractual identification.

Plan networks operate similarly. A narrow network may be an efficient contracting arrangement that lowers the price of services; it may also reduce the attractiveness of a plan to patients with particular clinical needs. Evaluating selection through networks requires comparing provider availability and treatment quality as well as cost. Claims-only analysis risks treating reduced access as lower morbidity.

Time, enrollment windows, and switching

Selection is dynamic because health expectations, employment, income, family composition, and plan familiarity change. Open enrollment, special enrollment periods, renewal defaults, and lock-in rules determine when individuals can respond. An annual plan population includes new entrants, persistent members, and switchers with potentially different risk profiles. A model based only on new sales can misstate the full pool.

Handel’s 2013 inertia study shows that easing choice frictions in an employer-based setting may worsen plan-level adverse selection and increase total welfare loss, even when the direct decisions of individuals improve according to the modeled choice criterion. This result clarifies a general mechanism: improving individual choice need not improve an equilibrium with endogenous premiums. Its estimated magnitude and institutional structure remain setting-specific.

Handel and Kolstad (2015) combine administrative choices and claims with survey information about plan understanding, highlighting how information frictions and hassle costs influence decisions and the estimation of risk preferences. Choice-model residuals should not automatically be interpreted as a stable taste for insurance. They may capture misunderstanding, inertia, unobserved provider relationships, or household constraints.

Segmentation beyond simple actuarial risk classes

Market segmentation has at least four observable dimensions: product segmentation through benefits and cost sharing; institutional segmentation through employers, income, residence, eligibility, and public programs; clinical segmentation through expected services and networks; and behavioral segmentation through search, literacy, switching, and the perceived value of protection. These dimensions interact. A policy that standardizes premiums may leave product segmentation; a plan that receives diagnosis-based transfers may still prefer enrollees whose service needs are poorly measured.

One can distinguish sorting that improves matching from sorting that imposes a welfare cost. A household choosing a plan with a suitable provider may gain real value. A contract that attracts a healthier pool while excluding otherwise valuable care may impose spillovers or foreclose a beneficial match. The test concerns counterfactual access, total costs, and welfare under the institutional funding rules. Segmentation is thus an empirical description requiring outcome evaluation, rather than a verdict by itself.

Adverse and Advantageous Selection

The classical coverage–risk relationship

Under a simple voluntary premium fixed across otherwise similar individuals, higher expected health expenditure can increase willingness to obtain insurance. If greater risk is privately known and compensated only imperfectly by revenue, the average expected claim cost among purchasers may exceed that among nonpurchasers or among less generous plans. Rothschild–Stiglitz competitive-contract analysis formalizes why offers with different coverage and prices may induce self-selection and why existence of equilibrium depends on the contractual environment.

This expected direction is conditional. Risk-averse people may have lower medical expenditure risk because they undertake preventive behavior. Higher-income households may be healthier and more likely to afford supplementary coverage. People with greater cognitive ability may make better purchase decisions or plan further ahead. Provider access itself may produce higher observed spending. Consequently the sign of the coverage–claims correlation is a compound result, not a direct readout of the sign of private information.

Cohen and Siegelman’s empirical review finds that empirical coverage–risk correlations vary by insurance market and dataset. They discuss how risk preferences, heterogeneous beliefs, moral hazard, and learning complicate straightforward tests. The appropriate conceptual outcome is not a universal positive-correlation hypothesis but a family of hypotheses tied to a particular information set and decision environment.

Advantageous selection and multidimensional private information

Fang, Keane, and Silverman (2008) report advantageous selection in the U.S. Medigap market when conditioning on the study's price controls: Medigap purchasers had lower average medical spending than nonpurchasers. After conditioning on health, the spending comparison reversed. Their analysis connects the observed sorting to income, education, financial-planning horizon, longevity expectations, and cognitive ability. It therefore illustrates how an apparently favorable aggregate comparison can coexist with privately known risks that favor purchasing insurance in more comparable subgroups.

The adjacent evidence on long-term care is analytically important as a boundary case. Finkelstein and McGarry (2006) show that two private-information dimensions—higher expected risk and stronger preference for insurance—can offset in aggregate data. Their sample concerns long-term care insurance, which has distinct eligibility, benefit, and underwriting institutions. Its relevance to health insurance is methodological: the absence of aggregate positive correlation cannot exclude economically important private information.

Buchmueller and colleagues (2013) study preference heterogeneity and selection in Australian private health insurance. This setting helps examine choices where public entitlement and voluntary private protection coexist. Findings about the relative roles of private insurance, preferences, and public coverage must be interpreted under Australia's financing and regulatory setting; they do not identify the same margin as the mandatory standardized basic insurance in the Netherlands.

The measurement of plan-level selection

Suppose a regulated insurer receives premium P_i and risk-adjustment transfer T_i for member i and expects claims C_i and administration A_i. Its expected member contribution to operating margin is M_i=E[P_i+T_i−C_i−A_i|information]. A plan can have a high measured morbidity score and still attract profitable members if the payment formula overstates their counterfactual cost, just as a relatively low-score patient can be unprofitable when clinically important needs are omitted.

This distinction separates selection on gross medical risk from selection on residual profitability. Under regulated rating and transfers, observed risk-score averages may improve while incentives to select persist. Brown and colleagues’ Medicare Advantage study found that changes in risk adjustment were associated with higher coded scores and lower observed costs conditional on scores, with no evidence of a net reduction in the overpayment problem they investigated. The Medicare Advantage payment setting and their historical reform period define the conclusion.

ACA formulary-screening evidence provides a second illustration. A benefit can be designed to be less attractive to a particular group for whom expected costs exceed associated revenues. If high-cost services remain medically necessary, the potential social loss concerns unmet or displaced care as well as cost reallocation. Selection metrics based solely on the plan's average incurred claims could reward such outcomes mechanically.

Adverse selection, moral hazard, and endogenous care

Selection and moral hazard have different objects. Selection concerns the composition of people who take coverage or choose a plan. Moral hazard, in the conventional insurance-economics use, concerns changes in behavior or service use induced by coverage or reimbursement. Both can affect realized claims. A broader source of endogeneity arises when providers respond to payment incentives, networks limit options, or coding systems reward documentation intensity.

An elementary decomposition helps. Let Y be realized spending, D enrollment, and R pre-enrollment expected risk. Differences in E[Y|D=1] and E[Y|D=0] contain differences in R, any causal effect of D on utilization and unit prices, and unmeasured confounding. Conditioning on diagnoses recorded after enrollment may create bias because coverage influences diagnostic contact and coding. A well-designed study establishes temporality and states which counterfactual quantity its estimator identifies.

Einav and coauthors’ 2010 estimation framework uses price-induced marginal enrollment changes to learn both willingness to pay and marginal costs. It thus addresses a question more informative for welfare than whether insured individuals happen to have larger average bills. The 2011 selection review makes the same distinction accessible through demand and average-cost curves.

Segmentation, cream skimming, and risk adjustment as imperfect responses

An insurer may select indirectly by contracting with providers, positioning a product, designing formularies, adjusting marketing, or affecting service quality. The incentives are clearest when revenue depends weakly on unobserved costs. Glazer and McGuire (2000) show, in a formal managed-care framework, how appropriately designed risk adjustment can respond to insurers' incentives to compete on risk selection rather than efficiently serve different needs.

The selection equilibrium must therefore include the insurer's action set. A conventional risk-adjustment comparison that assumes a fixed set of products measures only part of the problem. If a transfer rule changes, insurers can modify benefits, networks, enrollment efforts, and coding investments. A policy may reduce one form of risk sorting and intensify another. These possibilities demand empirical verification but follow directly from the behavior represented in the models.

Causal caution at the boundary of related insurance markets

Hendren’s 2013 study of private information explains rejected applications and missing markets in long-term care, disability, and life insurance; it does not estimate rejections within contemporary ACA-compliant U.S. individual major-medical plans. The study is relevant for economic mechanisms concerning information and trade, yet applying its specific rejection mechanism to a guaranteed-issue market would be a category error.

Similarly, supplemental coverage, employer group plans, and public-program plan choice cannot be pooled statistically without recording eligibility and benefit entitlements. An insured individual's measured expenditures are always conditional on the coverage and service environment. International analysis begins with this institutional measurement rather than ends with it.

Coverage and Welfare Implications

A graphical welfare account in words and equations

Let p be the full incremental premium for a given contract. Define inverse demand D(q) as the marginal purchaser's willingness to pay and average cost AC(q) as expected insurer cost among the first q enrolled individuals, sorted by willingness to pay. Total expected claims expenditure is TC(q)=q×AC(q), and marginal cost is MC(q)=dTC(q)/dq where smoothness permits. A competitive premium tied to the enrollment pool may cross demand at D(q)=AC(q), whereas an efficient quantity under the simplified single-product, complete-surplus benchmark satisfies D(q)=MC(q).

If higher-risk customers disproportionately have high willingness to pay, expanding coverage to marginal customers can lower average claims cost, making MC(q) lower than AC(q) over the relevant range. The average-cost pricing equilibrium then tends toward underinsurance in the usual adverse-selection diagram. Advantageous selection reverses the relevant sorting and can generate overinsurance under analogous assumptions. Einav and Finkelstein’s graphical synthesis and their empirical welfare approach with Cullen establish why these distinctions matter. The formulas are analytical benchmarks, not claims that all real premiums equal a single cost curve.

The quantity-based surplus comparison assumes that premiums are transfers between households and insurers except for resource and administrative costs, that outside options and benefit value are specified, and that care supplied does not create unmodeled externalities. Public subsidies impose financing and distributional effects. If a policy changes network quality, treatment choices, or health outcomes, a demand–claims graph alone is insufficient.

The welfare consequences of employer-plan sorting

Harvard plan-choice evidence illustrates an identifiable distributional and efficiency problem. The contribution policy improved some dimensions of premium competition while increasing selection-driven misallocation. The authors distinguish savings from reduced insurer margins from the aggregate welfare cost of selection; a plan purchaser's price gain, an insurer's loss of markup, and a resource saving have different welfare accounting.

Bundorf and colleagues’ welfare estimates offer another benchmark based on small employers and plan pricing. The particular benchmark permits risk-dependent incremental premiums, which can support matching but expose people to future reclassification and medical-cost risk. To assess a real reform, one needs a social welfare criterion including ex ante financial protection and the treatment of individuals with serious illness, rather than mechanically adopting the lowest measured static allocation loss.

Mandates, enrollment incentives, and stable pools

An enrollment mandate can reduce the option to wait until illness becomes known before buying insurance, thereby altering the extensive margin. Its economic effect depends on enforcement, exceptions, subsidies, the price schedule, and opportunities for switching. Hackmann, Kolstad, and Kowalski (2015) analyze Massachusetts's 2006 reform and estimate substantial improvements in individual-market selection. Their model-based estimate of reduced adverse-selection losses implies a welfare gain of 4.1 percent per person, or $51.1 million annually, within the study's Massachusetts market. Those are the paper's modeled estimates, not universally applicable fiscal returns from mandates.

Voluntary enrollment periods and late-enrollment penalties may pursue similar timing objectives, but they also create coverage discontinuities if households cannot afford premiums or face administrative barriers. Mandates and subsidies should therefore be evaluated jointly. A change in the observed claims mean after a mandate can indicate healthier entrants, altered utilization, or simultaneous market regulation; credible causal analysis must isolate the mechanisms.

Competitive price effects and selection effects

More insurers may create stronger price competition, more product diversity, and more opportunities for risk sorting. Dafny, Gruber, and Ody (2015) analyze initial ACA marketplace pricing and insurer participation. The research supports a competition–premium channel in that early market configuration; it is distinct from estimating a net welfare effect attributable to selection alone. The marginal impact of an additional insurer can depend on market structure, subsidy formulas, provider contracting, and the risk composition of demand.

Likewise, larger public transfers to plans need not translate entirely into cheaper prices. Cabral, Geruso, and Mahoney (2018) estimate that 45 percent of increased Medicare Advantage payments passed through to lower premiums and another 9 percent to more generous benefits in their design, with important variation by competition. The outcome illustrates how insurer market power and subsidy incidence interact with selection, rather than attributing every consumer-price difference to adverse selection.

Consumer information and the equilibrium externality

Policies facilitating choice can improve consumer understanding and yet alter the risk distribution across plans. Handel's inertia counterfactual demonstrates this possibility in an employer menu: reducing inertia increased modeled selection-induced welfare losses enough to outweigh direct improvements in individual decisions in the evaluated scenario. Handel and Kolstad’s information study adds that a model treating all observed behavior as rational risk preference can misvalue coverage and interventions.

The appropriate objective is to evaluate consumer welfare under the new equilibrium, not just the predicted quality of each isolated plan choice. Information provision can affect search, plan choice, plan pricing, network design, and the number of available plans. An evaluation that fixes insurer behavior can miss those mediated effects. These observations argue for causal designs or structural counterfactuals that incorporate credible supply responses.

Distribution, clinical access, and financial protection

Average total surplus is one criterion; distribution across risk types, income groups, and clinically vulnerable people is another. A reform may reduce aggregate premium expenditure while shifting expensive chronic patients into narrower access networks. Medical expenditure risk is particularly sensitive to income, since the same out-of-pocket shock has a different welfare impact at different wealth levels. The protection provided by a predictable premium is part of the value of insurance even if total expected claims remain unchanged.

WHO’s pooling guidance makes redistribution across people with diverse health needs a central financing objective. OECD's population-coverage methodology explicitly cautions that nominal coverage of core services does not demonstrate actual access. A defensible welfare analysis should therefore report coverage, financial protection, consumer surplus, medically appropriate utilization, and distribution by health and income. This is a demand for multidimensional evaluation within existing health-economics methods, not a claim to a newly invented welfare theory.

Data, Identification, and Model Validation

Defining the research population and time

A scientifically interpretable dataset begins with the eligible population, enrollment unit, and time frame. The unit may be person-month, family-year, employer-plan-year, insurer-state-market-year, or clinical episode; these are not interchangeable. Eligibility rules, premiums, subsidies, employer contributions, enrollment timing, and plan choice sets must be observed or reconstructable. A health insurer’s claims extract alone generally excludes people who decided not to enroll and will not identify the extensive-margin demand response.

For every major comparison, record the jurisdiction, coverage type, population denominator, exposure period, inclusion restrictions, administrative source, and outcome definition. Cost means permitted charges, insurer-paid claims, total allowed spending, out-of-pocket payments, or prospective expected medical need according to the research question. Using an unqualified word such as cost can conceal differences that reverse a selection result.

A natural observational table might contain anonymized enrollment identifier; contract and benefit year; eligibility and socioeconomic variables where lawfully available; quoted and paid premiums; subsidy and employer contribution; offered alternatives; formulary and network characteristics; prospective risk score based on prechoice information; realized allowed and paid claims; and change in enrollment status. Each variable requires an information timestamp. Predictors recorded after policy choice can contaminate a model of selection known at choice.

The positive-correlation test and its limits

A conventional test compares risk among people holding more and less coverage, conditional on contract price and observables. A positive coverage–risk association can be consistent with adverse selection, but also with utilization induced by coverage or omitted determinants of both choice and cost. A zero or negative association does not rule out offsetting private-information dimensions. These are central conclusions of Cohen and Siegelman (2010) and the related Finkelstein–McGarry evidence.

Tests should distinguish expected risk from ex post realized loss. Where available, pre-enrollment diagnoses, historical utilization, predicted expenses, and subjective beliefs are informative; none is an unproblematic gold standard. Claims history may be unavailable to the researcher, selective through access to care, or subject to coding incentives. Survey expectations can be noisy and measured after the decision. Replication requires the exact timing of each covariate and the handling of left-truncated coverage histories.

Price variation and marginal cost

An alternative strategy estimates demand from exogenous variation in the effective price of coverage and then measures the expected cost of the marginal individuals induced to enroll or switch. Einav, Finkelstein, and Cullen (2010) use this logic for welfare identification. A valid instrument or reform must shift the effective choice price without independently affecting unobserved health risk, expected care use, or product quality in ways that violate the design. Employer contribution reforms may help but can alter menus and information at the same time.

The demand–average-cost method has a particular attraction: it measures the cost of marginal enrollees rather than inferring their risk from baseline diagnoses alone. Its limitation is equally important: a price reform in one employer group may identify local responses that do not generalize to older public-program beneficiaries, subsidized low-income households, or jurisdictions with stronger enrollment mandates. Structural assumptions are needed to extrapolate beyond observed prices and choices.

Structural estimation of selection and information frictions

Structural models typically represent household choice as a function of premiums, cost-sharing distributions, risk aversion, plan quality, and unobserved preferences. These models permit counterfactual premiums and welfare calculations beyond the observed support, subject to maintained functional forms. Handel's choice-and-inertia analysis and Handel and Kolstad’s survey-linked model show why switching cost and imperfect information can be central parameters rather than harmless residuals.

A responsible estimation report should identify which parameters are pinned down by premium variation, claim history, choice-set changes, surveys, or auxiliary assumptions. Standard errors and confidence intervals alone do not capture sensitivity to information sets, functional forms, demand substitution restrictions, and equilibrium price determination. Report alternative specifications for high-cost tails and heterogeneous risk preferences; compare predictions in held-out enrollment periods where permitted.

Dynamic selection and actuarial model validation

Actuarial morbidity prediction and economic selection analysis answer related but distinct questions. A highly accurate prediction of next year's cost does not identify whether people self-select because they privately anticipate those costs. A poorly calibrated prediction can nonetheless create real financial incentives for insurers if payment relies on it. The relevant validation questions include discrimination, calibration by subgroup, sensitivity to diagnoses and missing care, net-margin residuals after transfers, and prospective stability under switching.

Layton, McGuire, and van Kleef (2018) demonstrate why a risk-adjustment regression optimized for predicting total expenditure can differ from weights designed to reduce inefficient service-level selection incentives. Their analysis supplies an economic motivation for evaluating margin-specific distortions. Glazer and McGuire's optimal adjustment model similarly focuses on incentives created by payment, rather than treating a smaller prediction error as the unique policy objective.

Consider a residual profitability diagnostic r_i=P_i+T_i−E[C_i+A_i|X_i]. A mean residual near zero across all members is insufficient if a clinically identifiable subgroup has strongly negative margins and insurers can discourage its enrollment. Evaluate residuals across diagnosis groups, drug-use groups, regions, network needs, and the joint distribution of age and income, while controlling for legitimate benefit and provider-cost differences. For rare high-cost events, credible inference requires careful uncertainty intervals and disclosure of small-cell privacy protections.

Design of a reproducible comparative study

A well-documented project specifies a preregistered or clearly dated analysis plan, including the outcome, target population, causal estimand, comparison group, and institutional reform date. A before-and-after study should compare concurrent trends where possible, inspect market entry and exit, and report whether payment formulas, coding incentives, and enrollment windows changed in the same period. Difference-in-differences and discontinuity designs need explicit identifying assumptions and sensitivity checks. Simulation studies must mark all parameter values as assumed or empirically estimated.

Suggested sensitivity dimensions are: (a) insurer-paid versus allowed costs; (b) diagnoses recorded before versus after plan choice; (c) inclusion of zero-claim members; (d) treatment of switching and mortality; (e) alternative premium subsidy incidence; (f) high-cost outlier thresholds; (g) alternative risk-adjustment versions; (h) market area definitions; and (i) network adequacy and unmet-need measures. The purpose is to learn which conclusions survive changes that would plausibly matter for policy.

Provenance, health data, and source hierarchy

Health data are sensitive and often restricted. Under the European Union's General Data Protection Regulation, Article 9, processing data concerning health engages the special-category regime and its specified exceptions and safeguards. This is a jurisdiction-specific legal fact, not a statement that identical rules apply globally. The present monograph examines published aggregate findings and primary literature and does not access individual-level health records.

Evidence is ranked according to the question. A final legal instrument establishes what regulation says within its jurisdiction and effective period. Official statistical releases establish reported aggregates within their definitions. Peer-reviewed empirical articles provide estimates conditional on data and identification design. A theoretical paper proves implications under stated assumptions. Reviews synthesize bodies of evidence and identify conflicts. None of these source classes substitutes for another, and this work does not represent its literature synthesis as a fresh observational study.

Integrated International Synthesis

Competitive choice under mandatory basic coverage

In a compulsory core-coverage system, the opportunity for people to remain outside the basic pool may be sharply reduced, yet selection among insurers and services persists. The OECD’s 2025 comparative account identifies the Netherlands and Switzerland as systems achieving near-universal core coverage through compulsory private insurance supported by strong rules and subsidies. Their institutional identity differs from voluntary supplementary plans and from employment-based health insurance in the United States.

The central comparative implication is that compulsory participation addresses one boundary of selection while creating a continuing task for the interinsurer transfer system. Plan competition may occur through service quality, access, supplemental benefits, administration, or insurer performance. If revenue equalization fails to reflect clinically important needs, insurers may still face incentives to attract profitable members or undersupply particular services. A fully comparative evaluation must obtain the applicable national scheme rules and current transfer formula; the OECD classification establishes system type but does not by itself measure whether an individual insurer has adverse selection.

U.S. employer-based choice and individual-market reforms

The U.S. literature has unusually detailed employer-plan and individual-market quasi-experiments. Cutler and Reber (1998), Bundorf and coauthors (2012), Handel (2013), and Handel and Kolstad (2015) jointly show the importance of contribution design, risk sorting, inertia, and plan understanding. Their study populations differ materially. Combining their published welfare estimates into a single pooled number would erase the relevant price variation, eligibility rules, and welfare criteria.

The Massachusetts mandate analysis provides evidence about an extensive-margin intervention in a historical individual market. The ACA's subsequent regulated-market framework adds standardized benefits, rating restrictions, financial assistance, enrollment windows, and risk transfers. The 2027 federal payment-parameters final rule describes the current HHS-operated risk-adjustment methodology and data validation framework for the relevant U.S. markets. Its 2027 benefit-year provisions should be read as law and administration for that period, rather than retrospective proof that earlier research estimates remain numerically stable.

A detail of the 2027 framework is especially relevant: the federal methodology predicts plan liability using factors including age, sex, diagnoses and, as applicable, prescription-drug and interaction information. Transfers are based on risk-adjustment rules and market-level comparisons. This helps counter plan-specific differences in expected enrollees but creates the need to test residual margins and coding incentives. Geruso–Layton–Prinz formulary evidence documents how benefit-level screening can survive broad risk compensation in an earlier ACA exchange setting.

Medicare Advantage: selection within public purchasing

Medicare Advantage links publicly financed coverage with competing private plans and risk-adjusted payments. Newhouse and McGuire (2014) assess its performance, including evidence that historical risk selection was mitigated under earlier policy changes. Brown and coauthors (2014) focus more narrowly on how selection responded to a payment-formula change, finding that plans could enroll individuals whose costs were favorable conditional on their risk scores. These findings address distinct estimands and can coexist: reducing one measured selection margin does not imply eliminating all residual incentives.

Cabral, Geruso, and Mahoney (2018) add the role of market power and incomplete subsidy pass-through. A public payment change can be absorbed as lower premiums, more generous benefits, greater insurer margins, or altered contract design. Risk adjustment and competitive pass-through are therefore complementary questions. The correct welfare accounting includes tax financing, distribution of benefits, consumer choice, insurer behavior, and access to care.

Medigap and supplemental private choice

The Medigap advantageous-selection study demonstrates that private supplemental coverage can attract comparatively healthier and more financially sophisticated people, depending on adjustment and the relevant comparison. Its findings challenge any automatic inference that greater coverage must imply worse baseline morbidity. Yet the study's data, pricing regime, and Medicare supplement context cannot be mechanically transferred to ACA individual markets, or to countries where private insurance serves as the compulsory basic entitlement.

Supplemental insurance decisions may be affected by the underlying public benefit, the timing of first eligibility, tax or employer support, and whether the product covers cost sharing or offers duplicate access. OECD classification of secondary coverage therefore provides an essential first step. The denominator should be eligible people for the particular secondary product rather than the whole population whenever eligibility is selective.

Australia and the public–private boundary

Australia provides a distinct market configuration involving public entitlement and voluntary private insurance. Buchmueller and colleagues’ Australian study examines preference heterogeneity and selection in this context. The current regulator also publishes insurer-level equalization accounts: APRA's March 2026 statistics report the financial-year 2024–25 net results based on regulatory returns. These are official accounting observations, not a causal estimate that equalization eliminated selection.

This distinction is essential for cross-country synthesis. A transfer total can demonstrate that money moved among insurers according to a statutory mechanism. It cannot alone establish whether consumers selected differently because of the mechanism or whether care quality changed. Comparative causal inference would need institutional timing, insurer-level risk and claims composition, and a credible counterfactual.

Fragmentation and the worldwide financing problem

For much of the world, the central risk-pooling challenge lies above the product-choice margin. Insurance may be fragmented by occupation, formal employment, region, income, or institutional membership, while many households face large direct payments. WHO's guidance on reducing pooling fragmentation recommends larger, more diverse, and compulsory pools. The implication for adverse-selection analysis is that equalization among competing private insurers is one specific instrument within a broader financing architecture.

OECD coverage and access distinctions establish why participation figures should be accompanied by access and financial-protection metrics. Voluntary participation can be shaped by ability to pay as well as health risk; compulsory membership can coexist with exclusions or unmet need. Markets differ in the identity of the payer, the distribution of risk across financing funds, the existence of risk equalization, and the scope of medical entitlement. An international monograph must compare those institutions before comparing regression coefficients.

Operational risk compensation in the Netherlands, Switzerland, and Germany

The term risk equalization refers to different legal funding and accounting mechanisms, and those differences matter for interpreting market segmentation. The Netherlands, Switzerland, and Germany each organize compulsory or statutory coverage with risk-adjusted interinsurer or interfund allocation, but the legal payer, variables, and premium rules differ. Their current official descriptions provide a more precise comparison than a general claim that a country has risk adjustment.

In the Netherlands, the 2027 framework described by Zorginstituut Nederland's Zvw 2027 documentation allocates risk-equalization contributions using characteristics including age, sex, income, prescription-drug use, and hospital-care patterns. Each member is assigned to applicable risk classes with prescribed amounts, and insurer payments are determined from the corresponding membership distribution, accounting for nominal premiums. This is a year-specific operational framework for the mandatory Dutch basic insurance, not a universal description of every voluntary supplemental policy. The 2027 rules were published during 2026; an analysis of 2025 plan outcomes would require the earlier payment specification, not retroactive substitution of the 2027 one.

Switzerland combines community-type premiums within the same insurer, canton or premium region, and model with a compensating transfer mechanism. The Swiss Federal Office of Public Health identifies age, sex, hospital or nursing-home stays in the preceding year, and pharmaceutical cost groups as risk-compensation inputs. It describes a zero-sum national transfer system, with adjustments applied prospectively from prior-year indicators. In this setting, changes in expected patient risk can affect the insurer's financing position even though the compulsory benefit entitlement is regulated. The office's description distinguishes a compensation for expected risk from reimbursement of actual expenditure. That distinction makes the incentive to improve efficiency conceptually distinct from the incentive to avoid predictable losses from specific patient groups.

Germany's statutory health insurance uses a different institutional accounting chain. The Federal Office for Social Security's 2026 notice specifies allocations from the Gesundheitsfonds, including a basic per-member amount, risk-adjusted supplements and deductions, and other legally specified expenditures. It cites Section 266 of the German Social Code, Book V, and the applicable risk-structure-adjustment ordinance. The notice is an official determination of allocation adjustments as of its September 2026 issuance, rather than an outcome study of whether all risk-based selection incentives were eliminated.

These cases illustrate three questions an international comparison must ask. Which legal entity collects revenue and makes transfers? Which observed characteristics determine prospective payment, and which forms of care need remain unmeasured? How do premium restrictions and the ability to differentiate benefits affect the financial consequence of a high-cost member? Neither the presence of a compensation fund nor the precision of its annual coefficients answers the outcome question. Evaluators should compute residual profits after the correct jurisdiction-specific payment rule, with due attention to provider cost and the actual benefit entitlement, and compare relevant clinical access measures. An official regulation or payment notice verifies the institutional mechanism; it provides no substitute for a causal estimate of its welfare impact.

Worked selection arithmetic under explicit assumptions

A numerical illustration clarifies why average claims costs respond mechanically to selection. Consider a hypothetical population of one hundred people consisting of twenty members whose expected annual spending is $10,000 and eighty whose expected spending is $1,000. Assume constant individual expected spending, a single identical benefit, no administrative cost, no moral-hazard response, no price negotiation, and a common actuarially balanced premium. Total expected spending is $280,000 and the initial premium is $2,800.

Suppose, solely for illustration, that forty low-cost members stop enrolling while all twenty high-cost members remain. The remaining sixty members generate $240,000 of expected spending, raising average cost and the balanced premium to $4,000. If an additional twenty low-cost members leave, the surviving forty members generate $220,000 of expected spending, and the balanced premium rises to $5,500. This is a mathematical description of compositional change. It does not predict that people will actually respond to these premium levels: that requires estimated willingness to pay, subsidy exposure, switching costs, and available outside coverage. It also does not imply insurer insolvency when premiums or transfers can adjust.

Now consider a hypothetical balanced risk transfer at the original full-population premium of $2,800. A transfer of negative $1,800 for each low-cost member and positive $7,200 for each high-cost member would precisely equalize expected gross claims costs for these two observable groups: $2,800 minus $1,800 equals $1,000, while $2,800 plus $7,200 equals $10,000. Aggregate levies of 80 × $1,800 equal aggregate high-risk payments of 20 × $7,200, or $144,000 each. This demonstrates how risk-based revenue can reduce the incentive to enroll one observable class rather than another.

Actual risk-adjustment systems use far more detailed rules and do not perfectly recover every person's counterfactual clinical needs. Imagine that a subgroup of high-risk patients costs $13,000 but receives a payment calculated for an expected $10,000. The resulting negative $3,000 contribution, before administrative and other adjustments, persists despite perfect balance for the broad two-class example. If a plan can make itself less attractive to this subgroup through a drug formulary or network, an incentive to select can survive a zero-sum transfer system. The observed source-based motivation is Glazer and McGuire's risk-adjustment analysis and the empirical ACA formulary-screening evidence. The chosen hypothetical dollar amounts are teaching parameters and are not taken from those studies.

A second distinction concerns reclassification. If an initially low-cost person becomes chronically ill next year, the social value of stable community financing arises from protection before knowing one's future health status. An administrative system that makes current-period costs equal across insurers may still need prospective risk sharing, predictable benefits, and protection from loss of access over the life cycle. These goals require a welfare analysis of individuals before and after health transitions, rather than a single-period ledger.

Institutional interpretation of the comparison

The Netherlands and Switzerland connect compulsory basic coverage to insurer competition and explicit risk compensation. Germany coordinates statutory sickness funds through a national health fund and risk-structured allocations. Australia publicly reports private health insurance equalization transfers in a mixed financing environment, while the United States operates distinct employer, individual exchange, and Medicare Advantage financing rules. Such systems may share a general problem of asymmetric information while placing the policy response at different administrative levels.

The comparison also shows why an international table of insurer loss ratios is an unstable ranking device. A high insurer loss ratio may reflect generous benefit design, higher morbidity, contractual provider prices, limited administrative loading, or inadequate revenue compensation. A low loss ratio may reflect favorable selection, efficient purchasing, restricted access, or an accounting difference in what is included as claims. A risk-adjustment transfer may be posted in a separate account. Comparability demands matched benefits, risk pools, time windows, and the exact definitions of medical expenditure, premium revenue, and transfers.

Finally, legal equality of premium within a specified category is distinct from equality of effective price after subsidies and from equality of service access. Risk compensation helps regulate the insurer's expected financial consequences, while household subsidy policy determines affordability and provider contracting determines much of the actual care experience. These are connected but separately measurable dimensions of health insurance market performance.

A consolidated interpretation of mechanisms

The evidence supports six interdependent relationships. First, selection arises at enrollment and between plans; an intervention can affect one margin without closing the other. Second, the meaning of adverse selection depends on the revenue rule and on the information available before choice. Third, a positive coverage–cost correlation is neither a necessary nor sufficient standalone demonstration of privately informed risk sorting. Fourth, payment adjustment alters insurer incentives but remains vulnerable to residual profitability differences and responsive product design. Fifth, consumer understanding and inertia modify selection equilibria as well as individual choice quality. Sixth, welfare requires a counterfactual accounting of access, medical expenditure, risk protection, subsidies, and competitive conduct.

These are analytical conclusions drawn from the reviewed literature, not new named propositions. Their value is comparative precision: they specify what an empirical claim must measure before it can inform an institutional decision.

Disagreements and transportability

Studies may appear to disagree because their periods, populations, products, and estimands differ. A Medigap result about supplemental purchasing need not have the same sign as an employer-plan result about switching. A Medicare Advantage estimate of conditional margin selection is not contradicted by evidence of an improvement in aggregate selection. A study estimating consumer surplus under an assumed pricing equilibrium answers a different question from an official report measuring insurer transfers.

Transportability therefore requires at least five matching conditions: eligibility and compulsory participation; price and subsidy exposure; plan benefit and provider access; observation of prechoice risk and realized outcomes; and the insurer revenue formula. When these conditions fail, an estimate can still motivate a hypothesis, but it is not an identified quantitative forecast for the destination market. This is a central editorial discipline for the international encyclopedia.

Conclusion

Health insurance adverse selection and market segmentation are best understood as the equilibrium outcome of privately informed expectations, heterogeneous preferences, contractual design, effective premiums, regulated transfers, and population eligibility. The analysis yields several grounded conclusions. An insurer's average claims cost is endogenous to enrollment and product choice; the marginal enrollee's expected cost is essential to welfare estimation. Private information can produce adverse or advantageous aggregate sorting because financial circumstances, cognition, preferences, and clinical needs are multidimensional. Plan-level incentives depend on expected net profitability after premiums and transfers, so a good risk-score prediction may coexist with undesirable service-specific selection.

Empirical research establishes that the effects of insurer competition, mandates, inertia, and risk adjustment cannot be judged on premiums or claim means alone. Employer-plan reforms identify contribution-sensitive sorting; the Massachusetts reform documents a context-specific improvement from broader enrollment; Medigap research establishes advantageous selection under its study design; U.S. exchange studies reveal benefit-level screening incentives; Medicare Advantage research documents the interaction of payment adjustment, residual selection, and competition. International WHO and OECD materials demonstrate that the relevant institutional level may be the national financing pool rather than an individual insurance product. APRA's published equalization accounts exemplify the distinction between an observed transfer mechanism and a causal evaluation of its effects.

The strongest comparative inference is methodological and substantive: policy analysis should jointly measure the risk composition of enrollment, the price faced by households, the clinical value of contracts, residual insurer profitability, treatment access, and ex ante distributional welfare. Static plan comparisons are incomplete when eligibility, price, benefits, or revenue formulas change. Well-designed pooling and equalization arrangements can address powerful selection incentives, while their actual success remains a matter for source-specific evaluation.

The research agenda follows directly from established evidence. It requires prospective data with prechoice risk measures; credible exogenous variation in net premiums or enrollment rules; plan-level measurement of service-specific residual profitability; evaluation of network and formulary consequences; and internationally comparable descriptions of eligibility and financing institutions. The completed synthesis advances the interpretation of the field through connected source criticism and economic reasoning without asserting a novel theory, a new welfare formula, or an original causal replication.

Methodological and Editorial Notes

This work is an original narrative and analytical synthesis of checked scholarly publications, international institutional documents, official statistical material, and jurisdiction-specific regulation. Mathematical expressions are standard identities or schematic representations with their assumptions stated nearby. No patient-level records were obtained, no insurer data were re-estimated, and no meta-analytic pooling was performed. Reported study findings are attributed to the exact populations, years, and design described in the cited publications. The 2027 U.S. regulatory source is identified by its final-rule date and benefit year; it is not substituted for current law in other jurisdictions.

Source hierarchy: primary journal articles for study claims, the Federal Register and EUR-Lex for legal provisions, APRA for regulator-reported financial data, WHO and OECD for their institutional classifications and policy guidance. Authors and years appear in the narrative, while short meaningful hyperlinks connect directly to each checked source. The final bibliography is complete for the sources entered in the research evidence ledger. This is internal scholarly and editorial work. External peer review has not been commissioned.

References

  • Akerlof, George A. 1970. “The Market for ‘Lemons’: Quality Uncertainty and the Market Mechanism.” The Quarterly Journal of Economics 84 (3): 488–500. DOI or official source

  • Rothschild, Michael, and Joseph Stiglitz. 1976. “Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information.” The Quarterly Journal of Economics 90 (4): 629–649. DOI or official source

  • Cutler, David M., and Sarah J. Reber. 1998. “Paying for Health Insurance: The Trade-Off between Competition and Adverse Selection.” The Quarterly Journal of Economics 113 (2): 433–466. DOI or official source

  • Einav, Liran, Amy Finkelstein, and Mark R. Cullen. 2010. “Estimating Welfare in Insurance Markets Using Variation in Prices.” The Quarterly Journal of Economics 125 (3): 877–921. DOI or official source

  • Einav, Liran, and Amy Finkelstein. 2011. “Selection in Insurance Markets: Theory and Empirics in Pictures.” Journal of Economic Perspectives 25 (1): 115–138. DOI or official source

  • Geruso, Michael, and Timothy J. Layton. 2017. “Selection in Health Insurance Markets and Its Policy Remedies.” Journal of Economic Perspectives 31 (4): 23–50. DOI or official source

  • Hackmann, Martin B., Jonathan T. Kolstad, and Amanda E. Kowalski. 2015. “Adverse Selection and an Individual Mandate: When Theory Meets Practice.” American Economic Review 105 (3): 1030–1066. DOI or official source

  • Handel, Benjamin R. 2013. “Adverse Selection and Inertia in Health Insurance Markets: When Nudging Hurts.” American Economic Review 103 (7): 2643–2682. DOI or official source

  • Handel, Benjamin R., and Jonathan T. Kolstad. 2015. “Health Insurance for ‘Humans’: Information Frictions, Plan Choice, and Consumer Welfare.” American Economic Review 105 (8): 2449–2500. DOI or official source

  • Fang, Hanming, Michael P. Keane, and Dan Silverman. 2008. “Sources of Advantageous Selection: Evidence from the Medigap Insurance Market.” Journal of Political Economy 116 (2): 303–350. DOI or official source

  • Finkelstein, Amy, and Kathleen McGarry. 2006. “Multiple Dimensions of Private Information: Evidence from the Long-Term Care Insurance Market.” American Economic Review 96 (4): 938–958. DOI or official source

  • Buchmueller, Thomas C., Denzil G. Fiebig, Glenn Jones, and Elizabeth Savage. 2013. “Preference Heterogeneity and Selection in Private Health Insurance: The Case of Australia.” Journal of Health Economics 32 (5): 757–767. DOI or official source

  • Bundorf, M. Kate, Jonathan Levin, and Neale Mahoney. 2012. “Pricing and Welfare in Health Plan Choice.” American Economic Review 102 (7): 3214–3248. DOI or official source

  • Geruso, Michael, Timothy Layton, and Daniel Prinz. 2019. “Screening in Contract Design: Evidence from the ACA Health Insurance Exchanges.” American Economic Journal: Economic Policy 11 (2): 64–107. DOI or official source

  • Brown, Jason, Mark Duggan, Ilyana Kuziemko, and William Woolston. 2014. “How Does Risk Selection Respond to Risk Adjustment? New Evidence from the Medicare Advantage Program.” American Economic Review 104 (10): 3335–3364. DOI or official source

  • Glazer, Jacob, and Thomas G. McGuire. 2000. “Optimal Risk Adjustment in Markets with Adverse Selection: An Application to Managed Care.” American Economic Review 90 (4): 1055–1071. DOI or official source

  • Layton, Timothy J., Thomas G. McGuire, and Richard C. van Kleef. 2018. “Deriving Risk Adjustment Payment Weights to Maximize Efficiency of Health Insurance Markets.” Journal of Health Economics 61: 93–110. DOI or official source

  • World Health Organization (WHO). n.d. “Pooling Revenues and Reducing Fragmentation.” Health Financing, WHO; accessed November 14, 2026. DOI or official source

  • Organisation for Economic Co-operation and Development (OECD). 2025. Health at a Glance 2025: OECD Indicators. Paris: OECD Publishing; section “Population Coverage for Healthcare.” DOI or official source

  • U.S. Department of Health and Human Services, Centers for Medicare & Medicaid Services (CMS). 2026. “Patient Protection and Affordable Care Act; HHS Notice of Benefit and Payment Parameters for 2027; and Basic Health Program; Final Rule.” Federal Register 91 (97), May 20: 29526–29877, FR Doc. 2026-10050. DOI or official source

  • Australian Prudential Regulation Authority (APRA). 2026. “Private Health Insurance Risk Equalisation Statistics.” Published March 31, 2026, including financial-year 2024–25 insurer returns. DOI or official source

  • European Parliament and Council. 2016. Regulation (EU) 2016/679 of 27 April 2016 (General Data Protection Regulation), especially Article 9 on special categories of personal data. Official Journal L119. DOI or official source

  • Cohen, Alma, and Peter Siegelman. 2010. “Testing for Adverse Selection in Insurance Markets.” Journal of Risk and Insurance 77 (1): 39–84. DOI or official source

  • Newhouse, Joseph P., and Thomas G. McGuire. 2014. “How Successful Is Medicare Advantage?” Milbank Quarterly 92 (2): 351–394. DOI or official source

  • Cabral, Marika, Michael Geruso, and Neale Mahoney. 2018. “Do Larger Health Insurance Subsidies Benefit Patients or Producers? Evidence from Medicare Advantage.” American Economic Review 108 (8): 2048–2087. DOI or official source

  • Dafny, Leemore, Jonathan Gruber, and Christopher Ody. 2015. “More Insurers Lower Premiums: Evidence from Initial Pricing in the Health Insurance Marketplaces.” American Journal of Health Economics 1 (1): 53–81. DOI or official source

  • Hendren, Nathaniel. 2013. “Private Information and Insurance Rejections.” Econometrica 81 (5): 1713–1762. DOI or official source

  • Zorginstituut Nederland. 2026. “Zvw 2027: Risicoverevening Zorgverzekeringswet.” Official 2027 risk equalization documentation and payment framework, accessed November 14, 2026. Official 2027 documentation

  • Swiss Federal Office of Public Health (FOPH). n.d. “Health Insurance: Risk Compensation.” Official compulsory health insurance risk-compensation description, accessed November 14, 2026. Risk compensation guidance

  • Bundesamt für Soziale Sicherung (BAS). 2026. “Bekanntmachung zum Gesundheitsfonds Nr. 13/2026.” September 30, 2026. Official health fund notice