Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Defaults and Active Choice in Insurance Enrollment

Abstract. An enrollment system establishes the consequences of inaction before consumers assess an insurance contract. Opt-in requires affirmative enrollment; opt-out makes coverage the initial allocation, conditional on eligibility and the applicable payment rules; preselection proposes a specific product; automatic assignment allocates an eligible person to an insurer or plan; and active choice requires an expressed selection or refusal. These procedures change participation, composition of risk pools, plan fit, administrative expenditure, and the meaning of consumer authorization. This monograph studies the causal and normative implications of initial enrollment defaults in private and public insurance, while using renewal and switching evidence to identify durable downstream effects. It combines original empirical research on Massachusetts health insurance enrollment, Medicare Part D, employer-sponsored plans, experimental information interventions, and retirement-plan active decisions; U.S. administrative rules; European insurance-distribution law; and international insurance-conduct standards. A central empirical result is that removing automatic enrollment in a Massachusetts exchange reduced enrollment by 33 percent, with disproportionate exclusion of younger, healthier and economically disadvantaged eligible people. Other studies show that active decisions can overcome inertia in a savings context, while informed plan choice can remain poor even when formal opportunities to select a plan are abundant. The monograph develops an explicit analytical separation among coverage take-up, insurer or plan allocation, choice quality, risk-pool selection, consent, and total welfare. Counterfactual equations show why improving the decisions of active shoppers can have ambiguous effects on market-wide welfare when premiums and risk composition adjust. The original contribution is a source-critical synthesis of competing mechanisms, precise estimands, identification strategies, and practical assessment criteria, with independently checked illustrative calculations. The synthesis establishes no new named theory, proprietary behavioral construct, causal estimate, or empirical discovery.

Keywords: Insurance enrollment; default options; opt-in; opt-out; automatic enrollment; automatic assignment; active choice; preselected plans; consumer inertia; administrative burden; enrollment ordeals; health insurance; Medicare Part D; Medicaid managed care; risk selection; adverse selection; health plan literacy; comparison friction; informed consent; insurance distribution; consumer protection; causal inference; randomized field experiments; welfare economics; insurance market design.

Introduction

At the moment a person becomes eligible for insurance, the institution has already selected an enrollment procedure. An eligible person may remain uninsured unless an application is completed; receive coverage unless a refusal is lodged; be assigned to one of several plans; be offered a preselected option; or be required to record a yes-or-no decision. The procedure changes what happens when the individual is busy, uncertain, unable to compare contracts, or unaware of the opportunity. It also changes the insurer's distribution of risks and the costs borne by the program, intermediary and consumer.

The research object of this dossier is the initial passage from eligibility to insurance coverage and plan allocation. Continuing a contract, renewing coverage, changing an insurer, and responding to altered premiums are distinct later events. Evidence about subsequent switching helps explain persistence of initial allocations, but a renewal intervention is not counted as an experiment on first-time enrollment. The difference is economically important because someone who does not complete a first-time enrollment may have no protection at all, whereas someone who does not switch an existing plan generally retains some form of coverage, subject to contract and payment rules.

The institutional stakes are evident in the Massachusetts auto-enrollment study by Shepard and Wagner (2025). The removal of a small administrative shortcut—automatic plan assignment—reduced enrollment by 33 percent. The people screened out were disproportionately younger, healthier and poorer. That result challenges any simple claim that an extra enrollment step principally excludes consumers with little to gain from protection. By contrast, Carroll and colleagues (2009) found in their active-decision study that requiring a response increased initial 401(k) participation relative to a conventional opt-in process. The different institutional results demand a joint account of available options, the response required, payment and eligibility consequences, rather than a single verdict on defaults.

The study uses three evidentiary levels. Published research with identifiable designs and results supports narrowly specified empirical claims. Regulatory and supervisory documents describe the obligations and available procedures of their own jurisdictions or frameworks. Analytical counterexamples and equations clarify the implications of assumptions but are explicitly constructed here and do not masquerade as measured market outcomes. The evidence ledger distinguishes full available text from publisher abstracts, agency guidance, and legal provisions.

The central questions concern how enrollment architecture affects take-up, which people are enrolled, which products they receive, whether the choice accords with their interests and understanding, and who bears its costs. Because insurance pools risks, the sum of individual switching benefits can diverge from a market-wide welfare result. Handel's (2013) analysis of adverse selection and inertia demonstrates the importance of this market-level perspective in a large employer setting. The focus here remains the first enrollment decision, while later inertia is treated as a consequence with its own empirical identification.

Enrollment and Choice Regimes

Defining the enrollment outcome

Insurance enrollment contains multiple legal and administrative events. Eligibility establishes whether an individual may be covered. Application records the requested transaction or consent. Assignment determines the designated insurer, network or product. Effectuation may depend on premium payment, eligibility verification, or contractual commencement. Actual coverage requires the relevant legal and financial conditions to be satisfied. A reported number of applications therefore does not automatically measure the number of people with effective protection.

Let E denote eligibility, A an application or recorded response, J an assigned or selected plan, P a qualifying payment where one is required, and C the state of effective coverage. In a simple contributory contract, E = 1 and A = 1 can coexist with C = 0 if an initial premium is unpaid. In a publicly financed plan with zero required consumer premium, an eligibility determination and valid automatic assignment may generate C = 1 without a separate payment event. The relationship among these variables depends on the governing program, not on the vocabulary of the enrollment screen.

The U.S. Marketplace illustrates the importance of effectuation. Its renewal and payment guidance explains that automatic re-enrollment of an existing Marketplace enrollee may occur after inaction, yet coverage in the new plan still depends on payment of the first monthly premium. This is a renewal boundary case and does not establish a universal rule about initial insurance enrollment. Outcome studies must state whether they count selection, assignment, premium payment, continuous enrollment or realized claim eligibility.

Opt-in, opt-out, preselection and automatic assignment

Under opt-in, inaction leaves the individual outside the proposed coverage arrangement. Under opt-out, inaction preserves the coverage arrangement selected by the institution, subject to applicable eligibility and consumer-protection conditions. Preselection places a particular plan or coverage level before the consumer as a proposed choice; it becomes an opt-out default only if failing to change or reject it creates the relevant contract or allocation. Automatic assignment determines a plan or administrator for eligible persons who have not actively selected one. Automatic assignment is compatible with mandatory coverage, an opt-out entitlement and certain public benefits, and it should not be equated with consumer agreement to every product term.

Active choice requires an individual to express an option, a refusal, or a plan selection rather than leaving silence to do the work. A pure active-decision design with no valid inaction path can delay or prevent coverage if a response is difficult to obtain. A structured active-choice design can include a safe fallback at the response deadline; then it combines required attention with an explicitly disclosed default. These combinations must be coded separately in research because a nominally identical 'active enrollment' label may describe substantively different exposure to uninsurance.

Enrollment rules may be represented by a mapping from eligibility, responses, deadlines and payment status to plan and coverage outcomes. For a set of available plans J, the mapping g(E, A, t, P) -> (J, C) makes the treatment operational: two websites with identical visual presentation may implement different rules when the response deadline passes. This treatment definition is necessary before causal comparisons or legal evaluation.

Active decisions and insurance-specific institutions

In employer-sponsored coverage, the employer's enrollment period and benefits platform can offer opt-in, default coverage or an explicit response requirement, but eligibility and payroll contribution terms differ from those of Medicare, Medicaid and individual-market insurance. In Medicare Part D, the state can intervene in the absence of a consumer plan selection: U.S. regulation addresses automatic enrollment for subsidy-eligible individuals, including random assignment among qualifying basic-benefit prescription drug plans when more than one meets the relevant subsidy limit. CMS separately explains auto- and facilitated enrollment for eligible categories of beneficiaries. A consumer's initial plan may therefore be produced through program rules even when a personal choice was available.

In Medicaid managed care, the state may organize a choice period or passive assignment. The provisions of 42 CFR § 438.54 distinguish voluntary and mandatory enrollment systems and prescribe notices, processes for default assignment, and criteria relating to capacity and existing provider relationships. This creates a sharp analytical distinction between default coverage and default plan allocation: for a mandatory managed-care population, the policy question may be which qualified organization provides care rather than whether the person should remain uninsured.

Historical origins and limits of analogy

Madrian and Shea (2001) found in their automatic 401(k) enrollment analysis that changing the default raised participation and that many workers retained the default contribution rate and investment allocation. The study established a major empirical reference for behavioral choice architecture, but its measured outcome was retirement saving in one employer context. A 401(k) contribution trades current consumption for accumulated assets; an insurance premium finances contingent protection and participates in a pool of heterogeneous risks. The analogy therefore identifies plausible human responses to defaults while leaving the insurance-specific implications for adverse selection, contract fit and payment to separate research.

Carroll and colleagues (2009) formalized a choice between default enrollment and requiring active decisions. Their empirical evidence establishes that the requirement to respond can increase participation when the alternative is a standard opt-in process and procrastination is salient. It does not imply that replacing automatic health-insurance enrollment with mandatory active selection will improve coverage. The relevant counterfactual is always the actual baseline institution. Three comparisons are different: active choice versus opt-in inaction, active choice versus opt-out protection, and personalized choice versus an automatic plan assignment in an already mandatory program.

These distinctions also resolve a frequent conceptual ambiguity. A default's apparent success measured by coverage take-up may coexist with poor plan fit, while an active choice's apparent success measured by decisions recorded may coexist with lost coverage. No enrollment architecture is adequately assessed through its response rate alone.

Default Effects and Mechanisms

Attention, inertia and the action threshold

The immediate behavioral effect of an enrollment rule can be described using a threshold. Let v_i be consumer i's expected subjective value of coverage, p_i the perceived premium, h_i the effort and time cost of completing the action, and m_i any additional perceived complexity, uncertainty or attention burden. Under a basic opt-in arrangement, the individual enrolls when v_i - p_i exceeds h_i + m_i. Under opt-out, the consumer receives the default allocation unless the perceived gain from rejecting or replacing it exceeds the costs of the required exit action. These expressions illustrate decision rules rather than estimate them from observed enrollment alone.

Inertia is a label for persistent action or inaction, not a unique psychological diagnosis. Inaction can arise from procrastination, limited attention, time pressure, high comparison costs, uncertainty about preferences, confidence in institutional recommendations, perceived procedural risk, or a genuine preference for the current outcome. The same observable refusal to change a default can reflect very different motivations. Empirical identification requires measures or experimental manipulations that distinguish mechanisms.

The retirement evidence of Madrian and Shea (2001) supports both inertia and a potential interpretation of the preset allocation as advice. In insurance, automatic assignment can similarly be understood as a recommendation, especially when consumers believe a public authority, employer or intermediary has evaluated suitability. Yet the rule may instead be administrative, based on capacity, rotation or cost constraints. The communicative meaning attributed to the default may exceed the information actually used to choose it.

Defaults as information and recommendations

A default can communicate a claim of institutional competence without explicitly stating one. Informed choice therefore requires an explanation of who selected the default, what information informed its selection, which alternatives remain available, the financial and contractual consequences, and how to reject or modify the assignment. Clear disclosure is especially important where the default is selected by an interested seller or intermediary whose incentives differ from those of the consumer.

Preselection has two distinct effects that should be tested separately. First, it changes the cognitive starting point by reducing the search effort needed to act. Second, it can suggest that a particular product is normal, favored or adequate. If the preselected contract is costly, excludes relevant risks or uses an unsuitable provider network, the same psychological mechanisms that promote take-up may entrench unfavorable allocations. Product heterogeneity makes the welfare quality of the default an empirical and actuarial matter, not merely a user-interface property.

Coverage, plan allocation and financial commitment

A two-stage structure clarifies the result. At stage one, the consumer is either covered or uncovered. At stage two, among covered people, a plan j from an eligible menu J_i determines benefits, premiums, deductibles, networks and service terms. An institutional assignment can alter both stages or only the second. A benchmark of successful design therefore requires at least effective coverage probability, the distribution of selected plans, expected net financial burden and the frequency of unsuitable matches.

The dominated-plan evidence of Bhargava, Loewenstein and Sydnor (2017) illustrates why a recorded election is not a sufficient measure of choice quality. Among 23,894 employees at a U.S. firm, a majority selected financially dominated plans under the studied menu, with excess expenditure corresponding to 24 percent of chosen premiums. Their experiments pointed toward limited understanding of insurance rather than solely the menu's size. A consumer may have actively chosen and still made a consequential financial error. A default chosen from an evidence-based benefit and network evaluation could improve some of those decisions, provided it preserves relevant personal preferences and consent.

Financial commitment adds another mechanism. A default that initiates a genuine premium obligation differs from an initial allocation to a zero-premium public benefit. Even when economic coverage value is positive, cash-flow timing and premium payment can prevent an intended enrollment from becoming effective. Policies that automate plan assignment must account for billing, affordability, eligibility recertification and the policyholder's ability to recognize a new obligation.

Heterogeneity, risk pools and adverse selection

Insurance markets transform the behavioral response into a selection problem. Suppose people have expected insured claims cost c_i, and enrollment under regime r is denoted C_i(r). Ignoring benefit heterogeneity for the moment, the expected average cost among enrolled persons is

cˉr=E[ciCi(r)]E[Ci(r)],\bar c_r = \frac{\mathbb E[c_i C_i(r)]}{\mathbb E[C_i(r)]},

provided the denominator is positive. A default changes both the number of enrolled persons and which people enter the average. If the additional participants have below-average expected costs, the pool's mean claims burden can fall. If they have above-average expected costs, it can rise. Neither direction is entailed by the word 'default'.

Shepard and Wagner (2025) provide an insurance-specific example of the first possibility: requiring action when automatic enrollment had previously been available disproportionately excluded younger and healthier people. Their study of enrollment ordeals finds that the extra step screened out groups with lower insurance costs alongside economic disadvantage. The important inference is about the joint distribution of coverage value, medical risk and administrative capacity. An administrative hurdle may screen on ability to navigate bureaucracy more strongly than on the social efficiency of providing protection.

The equilibrium response can be more complicated than the direct enrollment change. When the selected risk pool becomes more costly, premiums, subsidies, benefit design, insurer exit and public finances may adjust. Handel's (2013) work shows that a policy reducing inertia among existing health-plan enrollees can increase adverse selection sufficiently to offset individual gains. Because that paper addresses switching, it supplies evidence of a downstream market mechanism, not an estimated first-enrollment default effect. The shared implication is that enrollment design cannot be evaluated solely at the consumer who changes plans.

Active Choice and Forced Attention

The active-decision design

An active-decision intervention makes doing nothing procedurally difficult or impossible within a specified decision window. Its causal effect depends on the baseline: if the comparison group would otherwise fail to enroll in any plan, a prompt may promote coverage; if the comparison group would otherwise receive a good default plan without action, the same prompt may introduce an enrollment barrier. The procedure's actual deadline, the consequences of missed responses, the user interface, employer assistance and the ability to obtain personal advice are parts of the treatment.

The design can be organized into distinct institutional variants. A neutral active choice asks whether to enroll. A structured active choice additionally provides accessible and personalized cost or benefit information. A guided active choice presents a suggested plan but requires explicit assent. A protected active choice retains a disclosed fallback when the individual does not respond. The latter is a hybrid choice architecture, since silence still has consequences. These designs should not be pooled under a single estimate without verifying that their action requirements and fallback conditions coincide.

Requiring a decision can resolve procrastination when the relevant information is comprehensible and the decision can be made quickly. It can also convert scarce time, poor language access, inaccessible technology or incomplete documentation into a loss of coverage. The difference between a response obligation and a meaningful ability to respond is especially consequential for consumers with low literacy, disabilities, fluctuating work schedules, unstable housing or limited access to internet-enabled devices. The empirical issue is which constraints a procedure actually removes or introduces.

What the retirement experiment identifies

Carroll, Choi, Laibson, Madrian and Metrick (2009) developed an account of when active decisions outperform a passive opt-in arrangement, combining formal analysis with workplace savings evidence. In their original study of active decisions, requiring new employees to make an express 401(k) enrollment decision increased initial participation by 28 percentage points relative to the standard opt-in system. Three months after employment began, the distribution of savings resembled what the standard procedure would attain only after approximately 30 months. The result is strong evidence that requiring action can change the timing and incidence of participation.

The institutional counterfactual limits the lesson. The standard procedure required employees to initiate savings; an active prompt overcame their tendency to postpone an elective action. A health insurance program already providing automatic protection has a different baseline and potentially different costs of a missed response. It would be a category mistake to place the 28-percentage-point retirement estimate next to the Massachusetts 33-percent relative reduction as if they were estimates of the same treatment or outcome. One measures the effect of adding an explicit response to a conventional opt-in savings regime; the other measures the effect of removing automatic health coverage and requiring an additional enrollment step.

A further limitation concerns interpretation of participation. Retirement saving is often judged against projected wealth accumulation and consumption preferences. Insurance enrollment must also consider the distribution of medical or other insurable risks, benefit exclusions, risk pooling, premiums paid, and the value of protection against a low-probability but costly loss. A mechanical increase in participation is therefore neither a proof of improved risk protection nor a direct welfare estimate.

Health insurance literacy and comparison friction

Active decisions are useful only to the extent that consumers can compare the available policies. Handel and Kolstad (2015) combined claims, choices and survey information in their health insurance information-frictions analysis. Their evidence shows that misunderstanding and hassle costs are important determinants of observed plan choices, and that ignoring them distorts estimates of underlying risk preferences. This matters for default design because the observed plan selected under a nominally active system is not automatically an unambiguous revelation of informed preference.

Bhargava, Loewenstein and Sydnor (2017) offer a particularly informative benchmark through dominated alternatives. Where one plan is more expensive across the relevant claim states without compensating benefits, selection of it cannot be justified through ordinary financial risk aversion or private forecasts of medical spending. Their employee health-plan study makes the distinction between freedom to select and capability to evaluate concrete. The experimental component supports an important role for health insurance comprehension. It does not establish that every default would outperform every active choice: a poorly selected default can also be dominated, and individuals value networks, access and nonfinancial characteristics that may not be observed in a restricted menu.

Kling, Mullainathan, Shafir, Vermeulen and Wrobel (2012) identify a practical response to comparison friction. In a randomized Medicare Part D information experiment, personalized comparative cost information was proactively mailed to some consumers. Plan switching reached 28 percent in the treatment group versus 17 percent in the comparison group, and predicted consumer expenditure fell by approximately $100 a year for recipients. Information had already been publicly accessible, but actively delivering a focused comparison mattered. The intervention did more than present neutral data: it indicated the consumer's lowest predicted cost plan and the potential savings. Hence the causal contrast combines salience, accessibility, a comparison aid and an identifiable recommended option.

These experiments imply that the relevant choice architecture includes not just the default state but the quantity and usability of information supplied at the moment of decision. A valid design comparison may therefore involve a matrix: opt-in versus protected automatic allocation crossed with standard disclosure versus personalized, comprehensible plan comparison. Such a factorial test can distinguish improvements caused by coverage protection from those caused by better plan evaluation.

Experimental design for initial insurance enrollment

A field study should define the eligible population before randomization, then allocate participants among permissible procedures while preserving statutory rights and continuity-of-care safeguards. Where law permits, a study might compare a conventional opt-in process, an active choice with assistance, and automatic assignment with an easy and informed opportunity to opt out or substitute a plan. Any compulsory or otherwise legally required coverage is a boundary condition, not a treatment option that researchers may disregard.

The prespecified primary outcome should be effective coverage at a fixed date, verified through policy or program records rather than self-reported intention. Important secondary outcomes include days without protection, the selected insurer, expected and realized out-of-pocket cost, plan quality, consumer understanding, opt-out rates, complaints, payment failures, claims service experience and continuity of care. Baseline measures of anticipated claims, income, age, provider relationships and digital access permit heterogeneity analysis without assuming that differences in take-up reveal differences in welfare.

The policy estimand should normally be intention-to-treat: assignment to the enrollment design rather than compliance with every screen. An analysis limited to those who responded to the active-choice message selects on the very effort and attention affected by treatment, generating post-treatment bias. Compliance, nonresponse, crossover and assistance take-up belong in secondary analyses with an explicit identification argument. In design and interpretation, a randomized comparison of an entire enrollment procedure estimates the procedure's combined effect; isolating a psychological mechanism requires a separate experimental contrast.

Insurance and Health Plan Enrollment

Massachusetts: automatic enrollment and eligibility transitions

The strongest directly relevant causal evidence in the reviewed literature concerns subsidized insurance in Massachusetts. Shepard and Wagner's (2025) American Economic Review investigation examines an institutional shift that removed an automatic health-plan enrollment policy and imposed an additional active step. Their estimate is a 33-percent reduction in enrollment, expressed as a relative change rather than 33 percentage points. The effect was disproportionately concentrated among younger, healthier and economically disadvantaged persons. The empirical design exploits an actual program change; it is not presented as a randomized trial of all conceivable defaults.

The composition result is essential. A common argument for administrative application burdens is that a requirement to act screens out people who value coverage less. Insurance creates a different incentive structure because healthier people can have lower private expected utilization and lower expected cost to the pool simultaneously. Excluding them need not improve fiscal targeting or the risk pool. The paper's selection-market analysis demonstrates why evaluating an ordeal solely by reduced administrative spending or reduced enrollment is incomplete. The estimate pertains to the Massachusetts institution studied and cannot be treated as a universal elasticity of enrollment to a default switch.

A related but analytically distinct Massachusetts policy concerns continuation of existing coverage. McIntyre, Shepard and Wagner (2021) study automatic retention of existing enrollees when a low-income beneficiary otherwise would be disenrolled for failure to pay a premium: an available free plan was used as a fallback. Their published study reports that the policy switched roughly 14 percent of consumers annually and differentially retained lower-cost, healthier individuals. This supplies evidence about continuity and risk-pool effects at a later eligibility or payment transition. It is not an estimate of initial enrollment and remains outside the primary research object.

Together the two Massachusetts studies illustrate an institutional principle: the danger of a passive path depends on the legal outcome of inaction. Under a baseline requiring application, doing nothing can mean uninsurance. Under an automatic allocation baseline, doing nothing can mean continuous protection. The interpretation of an 'active choice intervention' therefore reverses if the institution silently changes the inaction outcome.

Medicare Part D: automatic assignment and plan fit

Medicare Part D offers a setting where public assignment rules interact with private competing plans. The U.S. regulation on low-income subsidy enrollment specifies automatic assignment to qualifying prescription drug plans for eligible individuals who do not enroll, with random assignment among qualifying plans in specified circumstances. CMS provides operational enrollment guidance, including beneficiary notices. The state-created default addresses access to a covered prescription benefit, while the specific assigned plan may be more or less suited to an individual's medicines and pharmacy preferences.

The distinction between allocation and matching can be measured directly through Part D choice research. Abaluck and Gruber (2011), analyzing beneficiary prescription data and available plans in their study of Part D choice inconsistencies, found that consumers weighted premiums more heavily than out-of-pocket spending and often made choices inconsistent with the efficient use of information on their drugs. Their partial-equilibrium analysis estimated substantial possible welfare improvement from better plan choices. That is evidence of imperfect plan matching under existing choice opportunities; it is not an evaluation of a new opt-out enrollment mandate.

Learning may moderate initial errors. Ketcham, Lucarelli, Miravete and Roebuck (2012) analyzed nonpoor beneficiaries in 2006 and 2007 and found in their Medicare learning study a mean reduction in overspending of $298, with improvements among 81 percent of the sample. Their evidence warns against treating initial observed mistakes as permanently fixed; experience, market feedback and switching can improve decisions. Yet the benefit of learning depends on access to feedback, an opportunity to correct a plan and the consequences of mistakes during the interim.

The Part D default can also shape future market behavior. Ericson (2014) used enrollment data and a regression-discontinuity design in his analysis of initial defaults and firm pricing. He found persistent default-related inertia and reported older plans with premiums roughly 10 percent above comparable new plans. This finding connects initial assignment to later insurer price incentives. The analysis belongs to the dossier as a documented dynamic consequence, not as an estimate of first-time coverage take-up. CMS's separate plan reassignment guidance confirms that authorities sometimes reassign subsidized beneficiaries when a plan exceeds the applicable benchmark or terminates, with specific exceptions for those who actively chose their plan.

Medicaid managed care: required choice and assignment rules

The legal design of Medicaid managed-care assignment demonstrates that 'default' can refer to a plan rather than an insurance decision. Under federal managed-care enrollment rules, voluntary programs may provide an active-choice period or passive enrollment with an opportunity to make a delivery-system selection. Mandatory managed-care arrangements require a default procedure for individuals who do not select a qualified managed-care entity. Notice requirements address the implications of action and inaction, available alternatives, and choice or disenrollment procedures. The provisions also address continuity of provider relationships, capacity and related assignment criteria.

These rules make normative goals concrete. Default design can protect uninterrupted access to care while honoring existing treatment relationships. It can also impose a network or administrative burden if the assignment disregards known medical needs. A comparison of default algorithms would need to distinguish random or equitable distribution, prior provider continuity, access measures, available plan capacity, and beneficiary preference. The precise regulatory requirements depend on the state's managed-care program and the applicable version of federal rules; no universal assertion of a right to a particular plan follows from the mere existence of default enrollment.

The rule is also a caution against aggregating all Medicaid default studies into a single opt-in/opt-out effect. Within mandatory coverage, every eligible person may be assigned a plan if no choice is made, so the outcome is matching quality and administrative continuity rather than a binary difference between coverage and uninsurance. Within voluntary arrangements, coverage or delivery system choice can remain a substantive decision. Institutional classification must precede estimation.

Employer-sponsored insurance and exchange architecture

Employer insurance requires attention to the characteristics of the offered menu and its pricing rules. In their analysis of information frictions, Handel and Kolstad (2015) show that health-plan choice cannot be reduced to stable risk preferences without measuring consumers' information and practical costs. Bhargava and colleagues (2017) demonstrate dominated plan selection in a particular employer menu. These results support testing guided choice and low-friction comparisons, but neither finding assigns a unique optimal default for all employers. Payroll contributions, taxable benefits, family eligibility, network access, underlying risk and regulatory protections materially affect the correct policy choice.

A restricted employer menu can reduce complexity while limiting customization. Dafny, Ho and Varela (2013) use a large insured-lives panel and a demand model in their study of expanded plan choice; under their stated equivalent-subsidy and price assumptions, they estimate meaningful potential welfare gains from a broader exchange choice set. That result balances the dominated-choice evidence: expanding alternatives can benefit heterogeneous consumers if they can evaluate and access relevant plans. The welfare assessment should therefore distinguish the economic value of variety from the cognitive costs of comparison and should assess both empirically.

Design changes beyond defaults also matter. Ericson and Starc (2016) studied Massachusetts Health Insurance Exchange standardization and found in their product-standardization analysis that changed cost-sharing displays and product availability altered plan generosity and brand choices without making consumers more price-sensitive. A policy that simultaneously changes the available menu, its presentation and the default cannot identify a pure default effect unless the research design or data allow the components to be distinguished.

Legal and supervisory boundaries

A useful default must operate within the law governing the contract and consumer. Under Article 20 of the European Union's Insurance Distribution Directive, distributors must ascertain a customer's insurance demands and needs from obtained information and provide objective, comprehensible product information before contracting. The proposed contract must be consistent with those demands and needs. This provision is not an instruction to implement a particular digital default, and national implementation and the product category remain relevant; it establishes an important standard against which a preselected private product must be assessed.

At the international supervisory level, the International Association of Insurance Supervisors' 2025 fair-treatment application paper supports Insurance Core Principle 19 and emphasizes removing barriers and treating diverse consumers fairly across the product lifecycle. It is supervisory guidance, not a single globally self-executing enrollment statute. It directs attention to accessibility, comprehensibility, institutional conduct, and the cumulative effects of product design and distribution. An insurer's default should therefore be assessed for both the fairness of entry and the suitability of the resulting contract.

Causal Tests and Selection

Causal estimands and counterfactuals

The primary causal question concerns how the enrollment arrangement would change outcomes for the same eligible population. Let R_i indicate the institutional regime assigned to person i, with R_i = 1 representing a specified alternative and R_i = 0 its specified baseline. Let C_i(r) be effective coverage at a prespecified date under regime r. The average causal effect on coverage is

τC=E[Ci(1)−Ci(0)].\tau_C = \mathbb E[C_i(1) - C_i(0)].

This definition requires the researcher to describe each regime in full. 'Automatic' could mean that a qualified person enters a zero-premium default plan without action; it could also mean that a prior coverage relationship is maintained subject to payment. 'Active' could mean a required yes-or-no response, a mandatory plan selection, or the ability to accept a suggested plan with one click. Without these procedural facts, the estimand has no stable interpretation.

Coverage take-up and plan quality are different dependent variables. For a person with effective coverage, J_i(r) identifies the plan. A plan-match evaluation might compare expected net financial burden or an independently specified measure of adequacy, denoted M_i(J_i(r)), while allowing an uncovered state to have its own expected consequences. If M is only measured among covered consumers, comparing mean M across regimes can be misleading: the regimes themselves determine who enters the sample. The appropriate population-level quantity is an outcome defined for everyone eligible, including those who are uncovered, supplemented by clear analyses of outcomes conditional on coverage.

A default can also change expected financial expenditure through insurer pricing, subsidies or cost pooling. Let p_j(r) denote the equilibrium premium of plan j under institutional regime r, and c_i(j) the expected insured cost associated with person i's enrollment in that plan. A consumer's response to a proposed plan depends on p_j(r), so an intervention that changes the risk composition and later premiums may change the behavior of people who were never directly prompted. In that case, the stable unit treatment value assumption used in simple individual-level experiments requires refinement. The estimand is a regime-level or market-level policy effect, and an individually randomized trial may identify only its partial-equilibrium component.

Identification, threats and transportability

Randomized assignment to enrollment procedures, implemented before the affected decision and without differential eligibility, can identify an intention-to-treat contrast in a defined population. Practical hazards include differential delivery of notices, changes in eligibility verification, baseline coverage misclassification, bundled financial inducements, unexpected changes in plan choice menus, selective consent to research participation, and outcome records that capture application but not coverage effectuation. Attrition is especially concerning when consumers with weak access to the system are both less likely to enroll and less likely to respond to an evaluation survey.

A natural experiment relies on a different identifying structure. When a jurisdiction removes automatic enrollment, a before-and-after comparison may be contaminated by simultaneous changes in eligibility, subsidy amounts, economic conditions or administrative capacity. Credible difference-in-differences designs require examination of pre-intervention trends and plausible comparability; discontinuity designs require attention to manipulation, sorting and continuity around the assignment rule. The design of Shepard and Wagner (2025) is informative precisely because it exploits an identifiable policy change and examines composition, rather than attributing every fluctuation in Massachusetts coverage to consumer psychology. Its published result should be transported only after considering eligibility rules, available zero-premium products and enrollment infrastructure.

Kling and colleagues' (2012) randomized comparison-information experiment identifies the effect of receiving a particular letter and comparison package during a Part D choice period. It does not isolate the effect of a bare default, because the treatment also gives personalized price projections and highlights an alternative. Ericson and Starc's (2016) standardization setting likewise altered multiple elements of the environment. Different research designs answer different questions even when each is described colloquially as 'choice architecture.'

Evidence is also subject to transportability limits. Employer-sponsored health insurance can have payroll deductions and restricted networks; Medicaid may have zero-premium public coverage; Medicare Part D has subsidy and formulary rules; catastrophe and property policies often require underwriting; commercial insurance may involve brokers and negotiated contracts. Consumers differ in professional assistance, literacy, experience and risk exposure. An intervention effective where all assigned plans meet a regulated minimum may perform poorly where defaults could involve materially different exclusions or financial commitments.

Source criticism is part of identification. A published abstract can substantiate a paper's stated sample, design and principal conclusion but may not support exact subgroup effect sizes, standard errors, sensitivity analyses or causal mechanism claims beyond those published. The accompanying evidence ledger records whether full open text, publisher abstracts or official regulatory text were available. No invented confidence interval, pooled meta-analytic estimate or novel empirical calibration appears in this monograph.

Selection and dynamic market responses

For a pool with premiums determined by expected cost, suppose regime r yields coverage share q_r and selected mean cost c_r. Expected claims per eligible person are q_r c_r; expected claims per enrolled person are c_r. A policy that increases q may lower c, raise total insured claims and expand insurance value at the same time. These three directions are compatible: more people are covered, the average enrolled person is healthier, and the insurance system pays more aggregate claims because its insured population is larger. Reports must distinguish these denominators.

Consider two groups with expected costs c_L and c_H, and enrollment probabilities a_L(r) and a_H(r) for regime r. If their population shares are pi_L and pi_H, then

qr=πLaL(r)+πHaH(r),q_r = \pi_L a_L(r) + \pi_H a_H(r),

and

cr=πLaL(r)cL+πHaH(r)cHqr.c_r = \frac{\pi_L a_L(r)c_L+\pi_H a_H(r)c_H}{q_r}.

The sign of the premium effect cannot be inferred from the overall rise in q. It depends on whether the marginal entrants differ from the people already insured and on benefit design, cost sharing, risk adjustment and the applicable premium rules. When a new default attracts relatively healthy persons, it may reduce the average cost of the pool while increasing expenditures on their coverage. If the default attracts high-cost people not previously insured, the mean may rise while expanding socially valuable financial protection. A complete analysis must assess who gains coverage and how the underlying financing system responds.

Dynamic responses are equally important. A default assignment may initially distribute enrollees among plans but create persistent customer relationships. Ericson's (2014) Medicare plan-pricing evidence indicates that insurers can price differently when consumers exhibit inertia; the reported approximately 10-percent premium difference between older and comparable newer plans pertains to his study's market and period. It supports monitoring later markups and the ease of moving to better plans. It does not justify a universal ten-percent adjustment to any defaulted insurance premium.

Handel's (2013) equilibrium welfare analysis establishes another structural warning: better individual plan decisions can redistribute risks in a way that destabilizes pooled pricing. This finding does not prove that increasing initial take-up through defaults harms the market; to the contrary, adding healthy low-cost people can strengthen it. The sign depends on the particular change in selection. The legitimate synthesis is that choice architecture is partly market architecture whenever insured risks are pooled.

Worked analytical examples

The following numbers are constructed solely to check equations and demonstrate interpretation. They are not insurance data and do not represent a policy simulation validated against any jurisdiction.

Suppose a cohort contains 1,000 eligible people: 600 low-risk persons with expected annual insured claims of $1,000, and 400 high-risk persons with expected annual insured claims of $5,000. Under an opt-in baseline, 40 percent of low-risk persons and 80 percent of high-risk persons enroll. The covered population equals 240 + 320 = 560, or 56 percent of those eligible. Expected insured claims equal $240,000 + $1,600,000 = $1,840,000, corresponding to an average of $3,285.71 per enrollee.

Under a protected enrollment default, suppose 80 percent of each group enrolls. Coverage becomes 480 + 320 = 800, or 80 percent of eligible persons. Expected insured claims are $480,000 + $1,600,000 = $2,080,000, and the mean cost falls to $2,600 per enrollee. The coverage gain is 24 percentage points, average selected cost falls by $685.71, and total insured expected claims rise by $240,000. An actuarially balanced community-rated premium without expenses or subsidies would reflect the different average costs; real rates depend on regulation, risk adjustment, reserves, loading and other variables not included here.

Now reverse the marginal composition while preserving the same original pool. Suppose the default leaves low-risk enrollment at 40 percent but raises high-risk enrollment from 80 percent to 100 percent. Coverage increases from 560 to 640, yet expected claims become $240,000 + $2,000,000 = $2,240,000 and the mean rises to $3,500. The same phrase 'a default increases enrollment' is compatible with lower or higher mean claims cost. The effect of a default is therefore not a scalar independent of risk heterogeneity.

A separate decision-quality example illustrates the second stage. Imagine two plans with equal covered benefits, provider access and contractual obligations across all relevant states, but one costs $120 more annually for the same person. Under these stipulated assumptions, assigning the person to the more expensive plan generates a measurable $120 loss in consumer resources relative to the cheaper plan. If provider networks, convenience, covered services or risk protection differ, the $120 premium difference no longer establishes dominance by itself. In real studies, the researcher must show that an apparent dominated choice remains dominated after all relevant plan attributes and consumer-specific costs have been accounted for.

Finally, interpretation of percentage changes must be precise. If enrollment falls from 75 percent to 50.25 percent, the decline is 24.75 percentage points and 33 percent relative to the initial rate. A statement that a policy 'reduces enrollment by 33 percent' cannot be rewritten as a 33-percentage-point effect. The distinction is especially important when comparing Shepard and Wagner (2025) with the 28-percentage-point result in Carroll and colleagues (2009).

Welfare and Consumer Protection

The components of social and consumer welfare

An insurance default can be judged on at least five outcomes: protection obtained; fit between coverage and the person's risk, needs and providers; financial cost and affordability; administrative effort and accessibility; and system-wide consequences for the insurer, pool or public budget. These outcomes are linked but are not interchangeable. For a person who cannot navigate an application, avoiding the ordeal can be valuable even when a short survey would report indifference to a plan. For a consumer assigned an unsuitable network, easy enrollment may be insufficient.

Let v_i(j) denote the money-equivalent social value of the protection and relevant service characteristics of plan j to eligible individual i, c_i(j) its real expected claims and resource cost, a_i(j,r) the real administrative and decision burden under regime r, and k_i(j,r) any external resource cost. A simplified allocation-based social surplus is

W(r)=∑iE[Ci(r){vi(Ji(r))−ci(Ji(r))−ai(Ji(r),r)−ki(Ji(r),r)}],W(r) = \sum_i \mathbb E[ C_i(r) \{v_i(J_i(r)) - c_i(J_i(r)) - a_i(J_i(r),r)-k_i(J_i(r),r)\} ],

where the uncovered state can be assigned a separate welfare value or loss if needed. Premium payments and subsidies are primarily financial transfers within a sufficiently broad accounting boundary, although their distribution, financing and behavioral incentive effects remain essential. This expression is not a calibrated structural model. It forces the analyst to declare which benefits, real costs and externalities enter the accounting and warns against treating gross enrollment as a comprehensive social objective.

Private welfare and social surplus can differ. Consumers may value indemnification, security and service access; insurers face claims cost, capital requirements, adverse selection and administrative expenses; public programs face budget constraints; and other insured persons may bear changing pooled premiums. A welfare conclusion should specify whose perspective is assessed, which transfers are counted, what opportunity costs are included and whether the analysis assumes fixed prices or permits insurer responses.

Consent, transparency and appropriate defaults

Opt-out design changes the consequences of silence and should therefore be accompanied by a clear account of whether silence creates legal coverage, monetary obligations, ongoing renewal commitments or merely a preliminary plan assignment. The distinction is central to consent. A public authority may assign an eligible person to a fully financed program under applicable law, while a private intermediary that uses an inconspicuous prechecked insurance purchase may create consumer-protection concerns. The appropriate legal assessment depends on the relevant jurisdiction, product type, demands-and-needs rules and contract formation requirements.

In the European Union, Article 20 of the Insurance Distribution Directive requires comprehensible objective information and a proposal consistent with the customer's stated demands and needs. This makes the quality and basis of a suggested default a distribution issue as well as a behavioral one. The IAIS consumer fair-treatment guidance broadens the institutional lens to diverse consumer characteristics and barriers throughout the insurance lifecycle. Appropriate evaluation includes the accessibility of refusal, human help, language, dispute resolution and downstream servicing.

An effective notice should identify the precise default plan, its meaningful exclusions and benefits, the premium or absence of consumer premium, applicable deductibles, relevant networks and limits, action deadlines, the consequences of doing nothing, and a functional method of changing or rejecting the assignment. The prominence of the notice and its comprehension should be tested, not assumed from a legal disclosure document's existence. Different consumer groups may require different communication channels and assistance even while substantive eligibility and rights remain equal.

Default selection itself requires governance. For public assignment programs, criteria such as network adequacy, accessible benefits, provider continuity, price and quality may be ethically relevant. In Medicaid managed-care enrollment rules, federal provisions explicitly attend to provider relationships, qualified plans and notice obligations. For employer or distributor recommendations, conflicts of interest and compensation arrangements require appropriate scrutiny. If a default reflects the administrator's convenience or commercial incentive, its descriptive popularity cannot establish that it serves the consumer.

Evaluation, audits and adaptive governance

A valid audit should report the following separately: number of eligible persons, number receiving communications, number recording a choice, number placed in a plan, number with effective coverage, number paying a required premium, and number whose coverage remains effective after a prespecified period. It should also report coverage gaps, financial burden, complaints, plan substitutions, medically important access disruptions and changes in risk distribution. The denominators must be declared for every percentage.

For choice quality, auditors can measure expected costs under the consumer's observable medical or other risk profile, compare plans with common coverage and adequate networks, and assess whether choices are demonstrably dominated under transparent assumptions. They should avoid attributing all apparent mismatches to cognitive error when private preferences, unobserved provider needs or volatility of future claims are plausible. Abaluck and Gruber (2011) provide one empirical strategy using claims and available plan menus; Handel and Kolstad (2015) demonstrate why measuring information and hassle costs is necessary for welfare interpretation.

For causal evaluation, the strongest practicable design is a prespecified randomized rollout or a credibly identified administrative policy change with clear eligibility and implementation records. A study should consider statistical power for low-frequency harms, protect against differential missingness, use intention-to-treat primary analyses, and measure treatment fidelity. Consumer comprehension tests, complaints and subgroup participation are complementary evidence rather than substitutes for enrollment records. A policy should not be declared successful solely because the click-through rate increased.

For market effects, monitoring must extend beyond the immediate enrollment window. The distribution of enrolled risk, insurer pricing, risk adjustment, default-plan quality, subsequent premium increases, switching opportunities and insurer responses may change after an intervention. Ericson's (2014) Part D evidence and Handel's (2013) health-insurance market model show why later behavior can matter for the original design. Assessment should separate short-run administrative benefits from long-run equilibrium effects, with uncertainty stated at the source-specific level.

For governance, an institution should retain an auditable record of the legal authority for automatic allocation, the default algorithm, any risk or quality criteria, notices, accessible alternatives, consumer complaints, significant changes in products, and reasons for subsequent reassignment. Independent review of algorithmic or administrative assignments should consider who lacks the resources to challenge an unsuitable default. This is a practical research and institutional-control agenda derived from existing evidence and norms, not a newly claimed doctrine or regulatory regime.

Implications across insurance systems

The strongest evidence base assembled here concerns U.S. health insurance and prescription drug coverage. The principles can motivate questions in life, disability, agricultural, property and commercial insurance, but their empirical effect sizes do not transfer by assumption. Private protection frequently has underwriting, exclusions, premium collections, anti-fraud screening and risk-specific pricing. A default that silently initiates a paid add-on policy differs from a public zero-premium safety-net allocation. Group insurance through employers, associations and microinsurance providers may combine collective eligibility with individually variable consent and household budgets.

For inclusive insurance, an enrollment default may overcome language, transport, documentation or digital barriers while raising demands for meaningful explanation and simple exit rights. For compulsory insurance, the choice problem is often coverage composition, insurer quality or compliance rather than whether to purchase any protection. For products requiring individual underwriting, the institution may be unable to guarantee immediate effectuation and should distinguish prequalification or preselection from actual coverage. These are analytical implications requiring product-specific empirical studies, not asserted findings from the U.S. health evidence.

The cross-jurisdictional conclusion follows from the separation among action rules, financial commitment and risk pooling. A system may legitimately seek widespread protection, informed selection, institutional solvency and consumer consent simultaneously. Those objectives must be measured separately and then evaluated jointly. Automatically assigning coverage can substantially reduce administrative losses, and requiring conscious comparison can improve product fit, but each depends on the consequences of nonresponse and the available institutional safeguards. The appropriate comparison is between precisely described policy packages, not between isolated words such as 'default' and 'choice'.

Conclusion

Initial insurance enrollment is an institutional allocation problem with behavioral, legal and actuarial consequences. Opt-in, opt-out, preselection, automatic assignment and active choice specify different consequences for inaction. They influence whether protection takes effect, where insured persons are allocated, what they understand, which risks enter a pool and how insurers or public programs later respond.

The empirical literature establishes several distinct findings. Shepard and Wagner (2025) demonstrate that a seemingly small removal of automatic health-insurance enrollment can substantially reduce take-up and selectively exclude younger, healthier and poorer eligible consumers. Carroll and colleagues (2009) demonstrate that an active decision requirement can raise participation relative to an opt-in savings baseline. Bhargava and colleagues (2017), Abaluck and Gruber (2011), and Handel and Kolstad (2015) show that formal plan choice may coexist with serious information and comparison problems. Kling and colleagues (2012) identify gains from the direct provision of personalized comparative information. Ericson (2014) and Handel (2013) connect initial allocations or later switching frictions to the economics of insurer pricing and risk selection. None of these results supplies a universal optimal default independent of product, coverage rules and the counterfactual inaction state.

The original analytical synthesis is a disciplined separation of five dimensions: actual coverage, plan allocation, choice quality, selected risk composition and social welfare. Explicit estimands and constructed arithmetic examples establish how participation can increase while either reducing or increasing average insured claims costs, and how a decision recorded as active can still be financially dominated. The relevant empirical object is the full enrollment procedure, including its fallback, information, payment and legal effect. Credible evaluation consequently needs coverage records, consumer-specific plan attributes, risk composition, accessible choice mechanisms and where necessary a market-level account of pricing feedback.

For insurance institutions, the substantive conclusion is operational: protect access to valuable coverage, establish comprehensible and legally valid choices, evaluate plan fit, and monitor the consequences for all insured persons and the financing system. The evidence supports careful, source-specific policy evaluation and a durable scholarly distinction between initial enrollment and renewal. No new theory, doctrine, experimental result or universal causal coefficient is claimed.

References