Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Distributional Evaluation of Insurance Policies

Abstract. Insurance policy changes the distribution of premiums, claims, risk, consumption, fiscal costs, and economic opportunity. Measuring this distribution requires identifying who receives coverage, who ultimately finances it, how contractual protection responds to realized shocks, and how households value reductions in exposure to uncertain losses. This monograph develops a source-grounded evaluation of insurance reforms, mandates, premium subsidies, public insurance expansion, and risk-based pricing. It distinguishes ex ante insurance value from realized claims, statutory from economic incidence, population averages from heterogeneous causal effects, and fiscal expenditure from social benefit. A transparent accounting decomposition links household resources to premiums, taxes, subsidies, claims, provider payments, risk reduction, and behavioral responses. Its interpretation is conditional on contract terms, risk selection, market adjustment, and financing. Evidence from randomized Medicaid access in Oregon, the introduction of Medicare, state Medicaid expansions, agricultural index insurance experiments, U.S. flood insurance reform, and international financial-protection monitoring demonstrates the value of studying insurance across multiple dimensions. The empirical literature supports meaningful financial-risk protection in certain settings, but also reveals incomplete enrollment, heterogeneous gains, delayed or unmeasured outcomes, nontransparent subsidies, and limits to cross-context extrapolation. The monograph sets out identification requirements for credible group-specific incidence estimates, practical welfare comparison under alternative social weights, a fully labeled illustrative numerical calculation, and an evidence-based framework for comparing policy options without asserting a new formal theory. The central conclusion is that distributional evaluation must measure the entire policy-and-financing arrangement, including nonparticipants and future taxpayers, rather than equating a coverage increase or paid claim with its net distributive effect.

Keywords: Insurance policy; distributional incidence; fiscal incidence; risk pooling; premium subsidies; insurance mandates; insurance coverage; household welfare; health insurance; agricultural index insurance; flood insurance; adverse selection; causal inference; heterogeneous treatment effects; financial protection; social welfare weights.

Introduction

Distributional evaluation of insurance policies investigates the allocation of economic consequences across persons, households, firms, geographical areas, and generations when insurance arrangements change. The policy object may be a new coverage entitlement, an altered deductible, a subsidy or mandate, a tax-financed public pool, a rating reform, or an adjustment in claims eligibility. Its outcomes must be measured both before and after uncertain states of the world are realized. A household that receives no claim may nevertheless obtain substantial risk protection; a household that collects a large indemnity may still remain poorer after a disaster; a government that funds a subsidy may displace an employer contribution or change wages. Each comparison consequently needs an explicit counterfactual.

The intellectual foundation combines choice under uncertainty, insurance equilibrium, and public-finance incidence. Mossin's analysis of rational insurance purchasing shows why pricing and risk aversion affect desired protection. Rothschild and Stiglitz's competitive insurance market analysis makes information and sorting central to contract equilibrium. Arrow's welfare economics of medical uncertainty explains why the institutional conditions of health services depart from frictionless commodity exchange. Later, Einav, Finkelstein, and Levin (2010) reviewed empirical models of insurance markets, emphasizing joint information on demand, insurer costs, and selection.

The object of this dossier differs from a general account of solidarity or redistribution. It concerns the empirical and analytical work required to estimate distributional consequences of particular policies, across income and risk groups and across financing channels. A classification of a program as universal, contributory, private, or targeted supplies institutional context; it does not itself establish progressivity, social value, or causal effectiveness. An evaluation must specify the population, intervention, time horizon, counterfactual policy, welfare metric, exposure to uncertainty, and boundary of the government budget.

The monograph uses four forms of evidence: original economic analyses of insurance choice and fiscal incidence; experimental or quasi-experimental estimates for particular reforms; administrative assessments of insurance affordability; and internationally standardized monitoring of financial protection. The examples deliberately span health, agriculture, and property catastrophe. They support comparisons of mechanisms while preserving the limitations of their distinct study populations. The work conducts no new household-level econometric estimation and therefore reports source-specific findings rather than manufacturing pooled international effects.

For precision, “benefit” in the chapters denotes a defined object: a cash transfer, an expected indemnity, a reduction in loss variance or downside exposure, access to a service, or a money-metric welfare gain. These objects are not interchangeable. “Incidence” means the final allocation of costs or benefits after prices, wages, taxes, participation, and market responses adjust. “Equity” is evaluated against stated normative criteria and the selected population rather than presumed from an institutional label.

Policy Incidence and Heterogeneous Households

The unit of analysis and the counterfactual

Suppose a population contains households indexed by i, with pretreatment income y_i, assets a_i, risk characteristics r_i, and an exposure-specific loss L_i(s) in state s. A policy k changes the eligible contract set, premium P_i(k), subsidy S_i(k), tax T_i(k), loss indemnity I_i(s,k), and perhaps labor income, service use, or loss prevention. Define disposable resources in state s as

Ci(s,k)=yi(k)+ai−Ti(k)−Pi(k)+Si(k)−Li(s,k)+Ii(s,k)−Oi(s,k),C_i(s,k)=y_i(k)+a_i-T_i(k)-P_i(k)+S_i(k)-L_i(s,k)+I_i(s,k)-O_i(s,k),

where O is an additional direct payment for services or losses not already embedded in L. The accounting convention must be applied consistently: if L includes all direct medical expenditure, O is zero; if L measures health-related income loss, O may record out-of-pocket medical spending. Double subtraction would corrupt incidence estimates.

An ex post net financial change for state s is the difference C_i(s,1)-C_i(s,0). The expected financial change integrates this difference over relevant states with specified probabilities. A policy can also change probabilities of loss through prevention, consumption, treatment, and investment. Its total welfare consequence therefore generally exceeds the expectation of cash transfers. Ex ante value under a known utility function is

ΔVi=∑spi(s∣k=1)ui[Ci(s,1)]−∑spi(s∣k=0)ui[Ci(s,0)].\Delta V_i=\sum_s p_i(s\mid k=1)u_i[C_i(s,1)]-\sum_s p_i(s\mid k=0)u_i[C_i(s,0)].

If state probabilities differ between policies because exposure or health behavior changes, that difference belongs in the comparison. If the exercise is a mechanical incidence calculation with no behavioral response, probabilities should be held fixed and the result labeled accordingly. Money-metric compensating or equivalent variation additionally requires specifying preferences and the reference price or resource state. Subtracting mean claims from premiums alone ignores the value of insurance against severe downside outcomes.

Multiple dimensions of group identity

Income quantiles are an indispensable axis, but they are insufficient. The same income can accompany different wealth, debt exposure, expected loss, vulnerability, access to providers, family structure, geography, and eligibility. Group definitions can be written g(i) = (income class, risk stratum, location, tenure, coverage status). Analysis should report separately the overlap and relative size of strata. Cross-tabulation prevents an apparent benefit to low-income participants from hiding exclusion of low-income nonparticipants. Distributional estimates should identify whether income is measured before a shock, after a shock, or after insurance transfers: sorting households by post-policy income can mechanically turn an insured loss into apparent targeting success.

The denominator matters. A subsidy per enrolled household answers a different question from subsidy spending per eligible household, per resident, per covered loss dollar, or per dollar of baseline income. The extensive margin is the decision to take up insurance; the intensive margin is the amount and terms of coverage. A program concentrated among low-income enrollees may still allocate most resources to higher-income persons if eligible low-income households face liquidity, information, or documentation barriers.

Contract and equilibrium response

Insurance is purchased in markets with information frictions. The Rothschild–Stiglitz selection framework links risk types to equilibrium contract design. An evaluation of a flat premium subsidy that holds insurer prices constant can overstate the consumer share if supply is inelastic, administrative costs rise, or risk composition changes. An identical subsidy can improve protection for the marginal enrollee while lowering the pool's average risk if lower-risk buyers enter; in another design it may attract higher-risk households and raise average costs. The direction is an empirical question.

Einav, Finkelstein, and Cullen (2010) demonstrate that demand and cost curves jointly identify welfare in an employer insurance setting when relevant price variation is available. Their particular application found adverse selection but modest measured welfare effects from inefficient pricing. That study is a reminder that the presence of an information problem is not itself a quantitative policy evaluation. Distributional analysis needs a demand response and insurer-cost response for each substantively relevant group, including people priced out or never observed as customers.

Scope of incidence

A complete accounting boundary includes consumers, employers, insurers, reinsurers, providers when relevant, governments, and persons exposed to public debt or future taxes. In a small open insurance market, reinsurance and international capital can spread losses outside the jurisdiction; the incidence of reinsurance premiums and returns remains part of the economic analysis. In a compulsory public insurance system, payroll or general tax financing affects workers, employers, consumers, and asset owners depending on their market conditions. Initial legal payment assignments do not settle the eventual burden.

Policy evaluation also requires a time dimension. A temporary premium reduction can increase enrollment immediately, while a later insurer price adjustment or public borrowing reallocates costs into future periods. Transitional groups, such as owners who purchased property under prior pricing rules, can have a different incidence from future buyers facing capitalized risk information. These intertemporal channels should be reported before a static table is interpreted as a lifetime distribution.

Income and Risk-Group Effects

Expected loss, ability to pay, and ability to insure

Let q_i denote the probability of a covered loss of size d_i. The actuarially expected insured loss is q_i d_i under full indemnity and fixed probabilities. If a contract charges m_i q_i d_i, where m_i exceeds one because of expenses or capital costs, identical loadings can produce very different premium-to-income ratios. A high-risk household with low liquid wealth can face the largest insurance need and the weakest purchasing capacity. By itself, a lower premium for low-risk households need not demonstrate distributional harm or fairness: the relevant evaluation reports risk-adjusted price, protection, affordability, and the explicit fairness rule.

A subsidy equal to a common percentage of premium expenditure allocates more dollars to policies with higher premiums. Whether that is progressive depends on who holds these contracts, how the subsidy changes take-up, and whether the risk is socially concentrated. A fixed-dollar subsidy favors lower-premium policies in percentage terms but may be insufficient for households with substantial exposure. An income-tested subsidy directs eligibility to lower-income groups yet may create phaseout incentives, administrative exclusion, or implicit tax rates. A risk-tested subsidy supports high exposure but can capitalize into property values and can discourage mitigation unless designed carefully. These are conditional implications of contract arithmetic, not universal empirical rankings.

Insurance purchases under constrained liquidity

Risk aversion alone does not predict actual coverage. In randomized field experiments in rural India, Cole and coauthors (2013) found that rainfall insurance demand responded to price, while trust, liquidity, and salience barriers also constrained take-up. The source addresses a particular index insurance product and rural population; extrapolation to commercial health insurance or urban flood coverage requires new evidence. Nevertheless, it supports a general identification warning: low enrollment does not identify low valuation when financial and institutional constraints are present.

A premium subsidy paid at the moment of enrollment can have a different effect from an equal reimbursement paid after purchase when purchasers lack cash. Similarly, renewal reminders and comprehensible claims terms can have distributional effects without changing the actuarial expected indemnity. These differences become economically important when a poor household's marginal utility of liquidity is high or when credible claim settlement is uncertain.

Agricultural risk and productive investment

Karlan, Osei, Osei-Akoto, and Udry (2014) conducted experiments in northern Ghana that varied cash grants and rainfall index coverage. Their agricultural risk-constraint experiments found higher farm investment and riskier production choices associated with insurance provision. These results support treating the value of coverage as partly dynamic: by reducing exposure to catastrophic loss, insurance can change productive decisions before any claim is filed. They also emphasize payout experience, trust, and basis risk. A program whose index fails to pay during a particular farmer's loss may deliver weaker risk protection than its headline sum insured suggests.

Distributional evaluation of index insurance should therefore estimate joint distributions of actual household damage and index-triggered claims, not simply aggregate payouts. A high average payout ratio can coexist with poor protection for a subgroup whose location or crops are weakly correlated with the index. The relevant outcomes include post-shock consumption, debt, agricultural investment, crop diversification, and next-season participation. Attribution must distinguish the insurance treatment from accompanying grants, information, and contract availability in experimental designs.

Health risk and income gradients

Health insurance affects the prices households face when obtaining care and the financial consequences of illness. The Oregon lottery first-year study used random selection for the opportunity to apply for Medicaid among a low-income uninsured population. Selected adults were approximately 25 percentage points more likely to have insurance; the researchers reported greater medical utilization, lower out-of-pocket spending and medical debt, and better self-reported outcomes in the first year. The estimate is tied to a specific eligibility mechanism, population, and horizon. It does not imply a universal 25-point enrollment effect of Medicaid reforms.

A follow-up analysis by Baicker and colleagues (2013) documented clinical and financial outcomes after two years. Among measured clinical endpoints, statistically significant improvements were not detected for average blood pressure, cholesterol, or glycated hemoglobin; use of preventive services and diabetes detection increased, depression screening outcomes improved, and catastrophic out-of-pocket expenditures were nearly eliminated. An accurate distributional account can report substantial financial-risk protection while recognizing that the study did not establish short-run improvements in every physical-health measure.

Group averages and the most exposed tail

Mean expenditure hides severe risk. If the 90th percentile of uninsured out-of-pocket payments is high, reducing it may be important even when average spending changes modestly. Conversely, high claims spending on behalf of a severely ill person is not identical to a welfare gain equal to the provider reimbursement. The household's valuation, health outcomes, and the public financing cost all matter. Quantiles, upper-tail probabilities, poverty transitions, and debt collections should accompany means when protection from catastrophic exposure is a stated objective.

Fiscal Transfers and Tax Incidence

Visible subsidies and underlying financing

Define government net spending attributable to policy k as G(k) = program subsidies + public claims + administration + debt-service effects - dedicated contributions - applicable tax receipts, with accounting conventions reported for each item. A claim that program spending is progressive requires identifying the people whose benefits rise and the people who bear the taxes or opportunity costs. A national outlay is neither a household benefit estimate nor a measure of progressivity.

The U.S. Congressional Budget Office and Joint Committee on Taxation's 2026–2036 subsidy projections provide an instructive fiscal baseline: projected subsidies total $33.6 trillion for the period, allocated across Medicare, Medicaid and CHIP, employment-based coverage, premium tax credits, and other benefits. These are estimates under a particular policy baseline, including legislative and administrative assumptions; they are not a study assigning that spending to final household beneficiaries or taxpayers. A full distributional evaluation would combine income-specific enrollment, tax incidence, medical risk, insurer transfers, and general-equilibrium responses rather than read an income gradient directly from expenditure totals.

Payroll and employer mandates

Employer-paid premiums may be shifted into wages. Gruber's (1994) study of mandated maternity benefits found substantial shifting of group-specific insurance mandate costs to wages for women of childbearing age in the particular historical U.S. policy setting. This is evidence that formal employer liability need not equal ultimate incidence. The numerical magnitude is not transferable without evidence about bargaining, labor supply, wage rigidities, union contracts, legal constraints, and employer behavior. The appropriate outcome is total compensation and household welfare, not only the employer's recorded insurance bill.

Tax instruments may also be perceived differently by households. Chetty, Looney, and Kroft (2009) show in research on tax salience that the response to taxes depends partly on how taxes are presented. The study concerns consumer taxation rather than a direct randomized test of insurance subsidies; its relevance here is methodological. If people respond differently to an embedded premium credit and a later tax refund, standard assumptions of complete price salience require assessment.

A transparent illustrative calculation

Consider two equal-sized illustrative household groups, A and B, each containing 100 households. All dollar figures are invented exclusively to demonstrate accounting; they are not estimates from any jurisdiction. Group A has an annual income of $20,000 per household and Group B $80,000. Without policy, the expected uninsured loss is $600 per A household and $1,000 per B household. A policy provides an expected indemnity of $500 to A and $900 to B. Gross premiums are $700 and $1,100, respectively. Government subsidies are $500 and $200, leaving premiums paid by households of $200 and $900. Assume the same probabilities and losses under policy and no policy, no behavioral response, no administrative costs outside gross premiums, and annual tax financing of $300 per A household and $400 per B household.

Annual component, per household Group A Group B
Pretreatment income $20,000 $80,000
Gross premium $700 $1,100
Subsidy $500 $200
Household premium $200 $900
Expected indemnity $500 $900
Attributed financing tax $300 $400
Expected cash-incidence change: indemnity less premium and tax $0 -$400
Expected indemnity divided by income 2.5% 1.125%

The expected cash-incidence calculation is $500 - $200 - $300 = $0 for A and $900 - $900 - $400 = -$400 for B. The illustrative total subsidy budget is $70,000, financed by $70,000 in illustrative attributed taxes. Insurer receipts are $180,000 gross and expected indemnities $140,000, leaving $40,000 in assumed insurer loading or other costs; this is part of the policy's resource cost and must not be counted as a benefit to households. With concave utility, A and B may value the reduction in loss exposure, so expected cash incidence is not an estimate of their welfare changes. Even the groups' relative welfare ranking cannot be determined without loss distributions, utility or willingness-to-pay evidence, and the use of any remaining insurer revenues. This example exposes a frequent accounting error: treating the full subsidy and the full indemnity as independent positive transfers while ignoring premiums, financing, and the way subsidies flow through contracts.

If households vary in take-up, the example must be extended to eligible nonbuyers. If the subsidy changes gross premiums, wages, provider prices, tax rates, or risk probabilities, those responses must enter the accounting before policy incidence can be inferred. If government spending is financed by public debt, one must identify a consistent intertemporal attribution convention rather than arbitrarily assigning zero current cost to all residents.

Insurance subsidy incidence and public budgets

A direct subsidy to an insurer, premium tax credit to a household, or discounted public premium may all lower the immediate price of coverage. Their incidence differs when market power, supply constraints, or selection alters premiums. Fiscal transparency matters because hidden premium discounts can distribute a public resource to owners whose property values are already high. The 2023 U.S. Government Accountability Office review of NFIP ratemaking recommended consideration of means-based assistance transparently reflected in the federal budget. Its analysis demonstrates the tension between full-risk pricing, affordability, and financial sustainability, while its historical figures should not be mistaken for current premiums in every location.

Distribution of Claims Protection

Claims as contingent performance

An insurance claim is a payment conditional on a covered event and contract compliance. Its distribution depends on risk occurrence, reporting, adjuster practice, exclusions, deductibles, coverage limits, waiting periods, insurer solvency, and timeliness of settlement. A claims-based distribution computed only among successful claimants omits denial, underinsurance, and the value of protection for households without realized losses. The appropriate data structure follows eligible households from enrollment through covered and uncovered shocks, attempted claims, settlement, and subsequent consumption.

For household i define a shortfall at threshold z as max{0, z - C_i(s,k)}. A policy's distributional protection can be described by changes in the probability C_i < z, average shortfall, lower-tail consumption quantiles, and high out-of-pocket expenditure incidence. When the benefit under consideration is a health service, treatment use and clinical outcomes should be recorded separately from the financial bill. This preserves the distinction between payment protection and improvements in health.

U.S. Medicare and Medicaid evidence

Finkelstein and McKnight (2008), studying the introduction of Medicare in 1965, report a substantial reduction in elderly out-of-pocket medical spending risk and limited detectable short-run mortality change over their analysis window. Their evidence shows why risk protection can be consequential even when measured mortality does not move. It does not imply that medical care or insurance never affects mortality, because the study evaluates a particular introduction and horizon.

The Oregon randomized design provides evidence on the margin of Medicaid expansion through an application lottery. In the first-year Oregon evidence, lower financial strain and medical debt accompany higher care utilization. The two-year clinical analysis establishes an important separation between the financial security gains documented in the study and its statistically non-significant average effects on selected physical-health biomarkers. A complete policy scorecard can reflect both without forcing one to stand in for the other.

Sommers, Baicker, and Epstein (2012) used state comparisons for earlier Medicaid expansions and reported associations with lower adult mortality, alongside improved access and coverage. Their quasi-experimental design differs from the Oregon lottery, so the studies answer different questions. An internal editorial synthesis should preserve the stronger randomization-based identification of lottery eligibility for compliers and the greater institutional scope but additional comparison assumptions of state expansion analyses.

International financial protection and measurement revisions

The World Health Organization's 2025 universal health coverage monitoring reports major remaining gaps in effective service coverage and financial protection. It estimates that 2.1 billion people experienced financial hardship from out-of-pocket health spending in 2022. The worldwide count describes direct payment hardship; it does not isolate the causal effect of an insurance mandate or identify an insurer's claims performance.

WHO's revised financial-hardship indicator defines SDG 3.8.2 in its 2025 specification with positive out-of-pocket spending exceeding 40% of the household discretionary budget and makes country-level disaggregation available. Comparisons across time or studies must distinguish this revised measure from the earlier 10%-of-total-expenditure version. Researchers should state the exact definition and source vintage rather than combine values under the same label. Out-of-pocket ratios can also be low because care is inaccessible; therefore surveys of unmet medical need and service use belong alongside expenditure metrics.

Disaster insurance and the geography of affordability

For flood insurance, property value, flood hazard, historical pricing discounts, and household income do not move together. GAO's 2023 Risk Rating 2.0 assessment found that aligning U.S. National Flood Insurance Program premiums more closely with individual property risk improves actuarial pricing while intensifying affordability concerns for some policyholders. As of the report's December 2022 data, median annual premium was $689 and estimated median full-risk premium was $1,288, while about 9% of policies faced ultimate increases above 300%. Those are date-specific portfolio statistics, not a forecast of any household's renewal price.

A later GAO flood mitigation and affordability review reported the continuing fiscal exposure of discounts and the rationale for targeted mitigation. The distributional question extends beyond current policyholders: households outside the program may rely on disaster assistance, renters may receive little benefit from property insurance subsidies, and future purchasers face altered prices and property valuations. A geographic map of average premium burden cannot answer within-county distribution by tenure or wealth. The report should distinguish household means testing from neighborhood proxies and should not presume that higher premium burdens always belong to poorer households.

Contract performance as a source of unequal protection

Equal nominal coverage limits can yield unequal practical protection if evidence of loss, claims handling, documentation needs, and legal enforcement differ. Distributional analysis must measure conditional settlement rates, payout delays, disputed exclusions, effective deductibles, and residual financial losses. In index insurance, these outcomes depend on the correspondence between an observable trigger and actual individual damage. Cole and coauthors' field evidence on index insurance barriers and Karlan and coauthors' Ghana insurance experiments caution against evaluating policy effectiveness from nominal coverage or headline subsidy levels alone.

Causal Designs and Welfare Weights

Identifying the effect of a policy

The primary empirical estimand is the difference between a group's potential outcome under policy treatment and its potential outcome under the relevant alternative. For group g, an average treatment effect may be written E[Y_i(1)-Y_i(0)|g]. What can be identified depends on assignment, compliance, spillovers, timing, and data quality. Insurance coverage is often endogenous: households select into plans according to expected loss, risk aversion, health, income, and access. Simply comparing insured and uninsured households is therefore not an estimate of the policy's causal effect.

Random assignment of an insurance offer identifies an intention-to-treat effect on offer recipients. Actual enrollment may be imperfect. Under instrument relevance, exclusion, monotonicity, and appropriate independence, assignment can identify a local average effect for persons whose participation changes because of assignment. Imbens and Angrist's local average treatment effect framework supplies the identification logic. The resulting estimate applies to compliers under the experiment's instrument, not mechanically to all low-income households or existing enrollees.

The Oregon lottery is especially instructive because lottery selection increased, rather than guaranteed, enrollment. Its randomized offer design supports estimates for a carefully defined access margin. Analysis must preserve the distinction between the offer effect and the induced enrollment effect, verify exclusion restrictions for instrumental-variable interpretation, and avoid describing a local response as an effect of nationwide universal enrollment.

Difference-in-differences and staggered reform

Insurance policies are often adopted at different dates by different jurisdictions. Standard before-after differences can be contaminated by economic cycles, disease outbreaks, migration, changes in catastrophe exposure, or other reforms. Difference-in-differences approaches identify effects under assumptions about untreated potential trends and appropriate comparison groups. Callaway and Sant'Anna's multiple-period difference-in-differences methods provide estimators for varied adoption timing and heterogeneous effects. Sun and Abraham's event-study treatment-effect analysis demonstrates why conventional two-way fixed-effects leads and lags can mix effects from different cohorts when timing and effects vary.

The empirical design should announce treatment dates, exposure definitions, comparison regions, pretrend diagnostics, policy anticipation, cohort composition, and spillovers. In a national fiscal reform, simultaneous tax changes are part of the treatment if enacted together; isolating a premium rule may require additional variation. The mere existence of fixed effects and a statistically significant coefficient does not settle identification.

Distributional estimands beyond the mean

For income or risk group g, report an incidence table combining (a) participation change, (b) expected and realized premium change, (c) subsidy and attributed tax change, (d) severe-loss protection, (e) utilization or productive investment, and (f) money-metric welfare where estimable. Quantile treatment effects require care because unconditional quantiles under two policies generally refer to different individuals. A change in the tenth percentile of spending is a population distribution effect, not necessarily the treatment effect of the people initially at the tenth percentile. Conditional distribution models should disclose covariate overlap and selection; attrition can be especially damaging if uninsured households with severe losses exit the sample.

Interference is consequential in insurance. Enrollment of one group can alter premiums for others through the risk pool; tax-financed expansion can change provider prices; subsidized flood coverage can change land use and expected public relief. Under spillovers, an evaluation of only treated and untreated individuals within the same price environment may miss general-equilibrium incidence. Analysts should choose the unit of randomization or comparison at the scale of the relevant market where feasible and describe unmeasured channels otherwise.

Welfare aggregation and normative transparency

One policy may deliver greater average willingness to pay while placing large costs on a disadvantaged subgroup. Another may reduce severe hardship but require a larger fiscal transfer. Standard aggregate surplus totals omit explicit distributive priorities. Saez and Stantcheva's generalized social marginal welfare weights articulate a method for aggregating money-metric gains and losses using chosen social weights. Their method concerns tax policy and welfare weights; using it for insurance is a transparent application of existing welfare-economics tools, not the announcement of a new theory.

A policy comparison can report W(k)=sum_i w_i CV_i(k) - lambda R(k), where CV_i is a consistent money-metric measure of household benefit, w_i is an openly stated social weight, R is the net resource or fiscal cost not already assigned in CV, and lambda is the relevant shadow valuation. The arithmetic must avoid double counting. If household CV already includes all financing taxes, subtracting the same budget cost again is incorrect. When social welfare weights are based on household income or vulnerability, their definition, normalization, and sensitivity should be published. A choice of equal weights, poverty-sensitive weights, or catastrophic-loss-sensitive weights can legitimately change policy rankings; the evaluation should display these rankings rather than conceal the normative assumption.

Willingness to pay is also shaped by risk beliefs, information, and liquidity. A financially constrained household may express low market willingness to pay for valuable insurance. Researchers should explain whether the objective is observed consumer surplus, normative equivalent variation under an economic model, health and poverty outcomes, or a composite policy score. Different metrics answer different questions and cannot be collapsed by an unexplained index.

Robustness and error control

A serious design registers or clearly reports the primary and secondary outcomes, data sources, exclusions, missingness, weighting, multiple-testing concerns, confidence intervals, and sources of uncertainty. Standard errors should reflect clustered policy assignment where applicable. For catastrophic losses, tail outcomes may be sparse; a zero significant result can reflect limited precision. For incidence over several years, discounting, inflation, migration, and changing benefit rules matter. All monetary variables should have stated currency, base year, and purchasing-power basis where international comparisons are attempted.

The claim-to-source audit must distinguish observed results from theoretical mechanisms. A study showing that insurance changed farm investment does not alone estimate the income distribution of lifetime benefits. A claims report cannot estimate take-up effects without the eligible population. A fiscal forecast is not a causal treatment effect. This discipline protects the validity of comparative synthesis.

Comparative Policy Evaluation

Policy options and their evaluable consequences

Universal eligibility, income-tested subsidy, fixed premium credit, risk-based rating, mandated purchase, deductible adjustment, catastrophe backstop, and targeted mitigation each change different margins. Evaluating them requires policy-specific counterfactuals, including what replaces an existing program. A redistribution measure can improve affordability yet weaken incentives to mitigate hazard; risk-based pricing can reveal hazard information yet threaten continuity of protection for cash-poor high-risk households. The policy choice cannot be inferred from the insurer's actuarial balance alone.

A useful comparative specification is a matrix of effects rather than a single unqualified ranking:

Policy instrument Primary immediate channel Principal distributional test Material identification concern
Income-tested premium subsidy Lower net household premium Take-up and subsidy per eligible income group Take-up barriers, phaseouts, supply pass-through
Uniform percentage subsidy Price reduction proportional to premium Dollar benefit by income and risk Composition and insurer pricing
Community rating with risk adjustment Redistribution within price pool Premium and enrollment by risk Risk adjustment adequacy, selection
Mandated insurance benefit Coverage expansion or richer contract Total compensation, consumer protection Wage and price shifting
Public coverage expansion Eligibility and financing change Financial hardship, use, health, tax burden Enrollment compliance and counterfactual trends
Risk-based catastrophe pricing Alignment of premium with expected loss Risk price, residual affordability, coverage gaps Property capitalization and heterogeneous hazard
Index insurance subsidy Improved access to contingent payout Joint actual-loss/payout distribution Basis risk, trust, liquidity, spillovers
Hazard mitigation grant Expected-loss and capital changes Protection per dollar by vulnerability Targeting, project timing, avoided-loss modeling

The matrix specifies empirical questions; it does not imply any instrument is universally preferred. Insurance line matters. A public health program has provider capacity, diagnostic outcomes, and tax financing; an agricultural index contract has correlated weather loss and trigger performance; a flood insurance scheme has immobile property, housing wealth, and land-use consequences.

A source-grounded cross-sector interpretation

The Oregon randomized evidence demonstrates that making health insurance available can materially reduce medical financial strain among a constrained eligible population. The two-year Oregon follow-up cautions that financial protection and changes in measured physical-health biomarkers must be evaluated separately. The Ghana agricultural experiments show another channel: insurance can support investment in risky productive activity before a claim occurs. The GAO NFIP ratemaking assessment instead emphasizes how a shift toward actuarial pricing redistributes premium burdens over time and creates an explicit need to examine affordability and fiscal subsidy design. These findings are complementary because they address different contracts and counterfactuals.

A cross-country statement such as “insurance subsidies are progressive” would require harmonized evidence on recipient income, risk, enrollment, contract value, provider access, and tax financing. The WHO financial protection indicator helps establish observable household hardship; it cannot on its own measure an insurer's causal contribution. The CBO federal insurance subsidy baseline establishes fiscal magnitudes in the United States, not global distributional incidence. External validity therefore rests on matching mechanisms and institutions, not on treating results from unlike programs as interchangeable effect sizes.

Evaluation sequence for governments and researchers

An operational study first defines the intervention, legal eligibility, financing, dates, and counterfactual. It then maps all affected populations, including nonparticipants, employers, providers, owners of insured assets, insurers, and taxpayers. Next it measures direct household transfers and coverage, estimates causal effects on risk, utilization, claims, or behavior using a credible design, and calculates final incidence after market and tax adjustments. Finally it tests policy rankings under declared welfare weights and distributional measures, assesses sensitivity to tail risk and fiscal assumptions, and reports data and identification limits. Each stage adds information that a simple claims-to-premiums ratio cannot provide.

A credible international comparison also records institutional differences: social insurance eligibility may depend on employment; private contracts may allow risk selection; government catastrophe pools may have explicit or implicit guarantees; claim settlement may be influenced by enforcement capacity; and purchasing-power differences complicate monetary comparisons. Comparisons should use common definitions where possible and preserve national policy context where a common indicator would erase material differences.

Research boundaries and the status of the synthesis

The conceptual decompositions and illustrative tables in this dossier are original analytical arrangements of established economic ideas. They are not empirical estimates, a proposed named theory, or an assertion of priority over earlier scholars. The source base includes journal publications, original institutional reports, and publisher or institutional summaries. Paywalled original texts are interpreted at the bibliographic and abstract level where full access was unavailable; the monograph does not present inaccessible results as independently replicated. Some methodologies are cited as tools whose applicability to insurance policy must be established case by case.

No new microdata were acquired, no international incidence dataset was assembled, no meta-analysis was performed, and no quantitative welfare optimum was estimated. Numerical results in the text remain attached to their original date, sample and design. Evidence is current to November 14, 2026, for the sources actually examined. These boundaries preserve the distinction between a completed theoretical and literature-based research dossier and an empirical program requiring new data.

Conclusion

Distributional evaluation of insurance policies is the study of who gains contingent protection, who pays for it, how markets and public budgets transmit those costs, and which households remain exposed after a reform. An insurance policy has a distribution across economic resources, risk states, access, and time. Its redistributive character is established by evidence on these dimensions and by an explicitly stated social criterion.

Five conclusions follow from the examined theory and primary evidence. First, statutory subsidies, mandated benefits, and insurer claims are intermediate financial flows: ultimate incidence requires prices, wages, take-up, and financing. Second, ex ante protection from uncertain loss remains valuable even where observed short-term health or productivity outcomes do not change measurably. Third, income targeting and risk targeting address different distributions; evaluating either requires examining their joint consequences and the excluded population. Fourth, causal identification and external validity determine how strongly evidence from a program can support a broader policy statement. Fifth, an explicit welfare judgment requires social weights, a consistent resource boundary, and sensitivity to outcomes in the lower tail of household resources.

The analytical synthesis establishes a defensible evaluation practice: report household and fiscal accounts together; distinguish coverage, realized claims, ex ante risk reduction, and final welfare; preserve income-by-risk heterogeneity; estimate counterfactual effects by credible designs; and compare policy alternatives under visible normative assumptions. The research supports this practice through distinct health, agricultural, and catastrophe insurance settings. It leaves empirical policy rankings contingent on their actual data, financing, institutions, and social objectives. The resulting monograph is a source-grounded account of distributional insurance evaluation, prepared as a standalone scholarly dossier for future editorially authorized publication.

References

  • Arrow, Kenneth J. 1963. “Uncertainty and the Welfare Economics of Medical Care.” American Economic Review 53 (5): 941–973. https://www.jstor.org/stable/1812044.

  • Baicker, Katherine, Sarah L. Taubman, Heidi L. Allen, Mira Bernstein, Jonathan H. Gruber, Joseph P. Newhouse, Eric C. Schneider, Bill J. Wright, Alan M. Zaslavsky, and Amy N. Finkelstein. 2013. “The Oregon Experiment—Effects of Medicaid on Clinical Outcomes.” New England Journal of Medicine 368: 1713–1722. https://doi.org/10.1056/NEJMsa1212321.

  • Callaway, Brantly, and Pedro H. C. Sant’Anna. 2021. “Difference-in-Differences with Multiple Time Periods.” Journal of Econometrics 225 (2): 200–230. https://doi.org/10.1016/j.jeconom.2020.12.001.

  • Chetty, Raj, Adam Looney, and Kory Kroft. 2009. “Salience and Taxation: Theory and Evidence.” American Economic Review 99 (4): 1145–1177. https://doi.org/10.1257/aer.99.4.1145.

  • Cole, Shawn, Xavier Giné, Jeremy Tobacman, Petia Topalova, Robert Townsend, and James Vickery. 2013. “Barriers to Household Risk Management: Evidence from India.” American Economic Journal: Applied Economics 5 (1): 104–135. https://doi.org/10.1257/app.5.1.104.

  • Congressional Budget Office. 2026. Federal Subsidies for Health Insurance, 2026 to 2036. Washington, DC: CBO, July 23. https://www.cbo.gov/publication/62539.

  • Einav, Liran, Amy Finkelstein, and Mark R. Cullen. 2010. “Estimating Welfare in Insurance Markets Using Variation in Prices.” Quarterly Journal of Economics 125 (3): 877–921. https://doi.org/10.1162/qjec.2010.125.3.877.

  • Einav, Liran, Amy Finkelstein, and Jonathan Levin. 2010. “Beyond Testing: Empirical Models of Insurance Markets.” Annual Review of Economics 2: 311–336. https://doi.org/10.1146/annurev.economics.050708.143254.

  • Finkelstein, Amy, and Robin McKnight. 2008. “What Did Medicare Do? The Initial Impact of Medicare on Mortality and Out of Pocket Medical Spending.” Journal of Public Economics 92 (7): 1644–1668. https://doi.org/10.1016/j.jpubeco.2007.10.005.

  • Finkelstein, Amy, Sarah Taubman, Bill Wright, Mira Bernstein, Jonathan Gruber, Joseph P. Newhouse, Heidi Allen, Katherine Baicker, and the Oregon Health Study Group. 2012. “The Oregon Health Insurance Experiment: Evidence from the First Year.” Quarterly Journal of Economics 127 (3): 1057–1106. https://doi.org/10.1093/qje/qjs020.

  • Government Accountability Office. 2023. Flood Insurance: FEMA’s New Rate-Setting Methodology Improves Actuarial Soundness but Highlights Need for Broader Program Reform. GAO-23-105977. Washington, DC: U.S. GAO, July 31. https://www.gao.gov/products/gao-23-105977.

  • Government Accountability Office. 2026. Flood Risk Mitigation: Reducing Fiscal Exposure and Improving Affordability. GAO-26-109045. Washington, DC: U.S. GAO. https://files.gao.gov/reports/GAO-26-109045/index.html.

  • Gruber, Jonathan. 1994. “The Incidence of Mandated Maternity Benefits.” American Economic Review 84 (3): 622–641. https://www.nber.org/papers/w4157. (Earlier working-paper treatment: “The Efficiency of a Group-Specific Mandated Benefit: Evidence from Health Insurance Benefits for Maternity,” NBER Working Paper 4157, 1992.)

  • Imbens, Guido W., and Joshua D. Angrist. 1994. “Identification and Estimation of Local Average Treatment Effects.” Econometrica 62 (2): 467–475. https://www.gsb.stanford.edu/faculty-research/publications/identification-estimation-local-average-treatment-effects.

  • Karlan, Dean, Robert Osei, Isaac Osei-Akoto, and Christopher Udry. 2014. “Agricultural Decisions after Relaxing Credit and Risk Constraints.” Quarterly Journal of Economics 129 (2): 597–652. https://doi.org/10.1093/qje/qju002.

  • Mossin, Jan. 1968. “Aspects of Rational Insurance Purchasing.” Journal of Political Economy 76 (4, part 1): 553–568. https://doi.org/10.1086/259427.

  • Rothschild, Michael, and Joseph Stiglitz. 1976. “Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information.” Quarterly Journal of Economics 90 (4): 629–649. https://doi.org/10.2307/1885326.

  • Saez, Emmanuel, and Stefanie Stantcheva. 2016. “Generalized Social Marginal Welfare Weights for Optimal Tax Theory.” American Economic Review 106 (1): 24–45. https://doi.org/10.1257/aer.20141362.

  • Sommers, Benjamin D., Katherine Baicker, and Arnold M. Epstein. 2012. “Mortality and Access to Care among Adults after State Medicaid Expansions.” New England Journal of Medicine 367: 1025–1034. https://doi.org/10.1056/NEJMsa1202099.

  • Sun, Liyang, and Sarah Abraham. 2021. “Estimating Dynamic Treatment Effects in Event Studies with Heterogeneous Treatment Effects.” Journal of Econometrics 225 (2): 175–199. https://doi.org/10.1016/j.jeconom.2020.09.006.

  • World Health Organization and World Bank. 2025. Tracking Universal Health Coverage: 2025 Global Monitoring Report. Geneva: World Health Organization, December 5. https://www.who.int/teams/health-financing-and-economics/global-monitoring-report/2025.

  • World Health Organization. 2025. “Financial Hardship Due to Out-of-Pocket Health Spending and Its Components (SDG 3.8.2, 2025 Definition).” Global Health Observatory. https://www.who.int/data/gho/data/themes/topics/financial-protection.