RESEARCH DOSSIER
Dependency Structure and Life Insurance Demand
Abstract. The demand for life insurance depends on the economic relationships that persist after an insured person's death. Household dependency is a multidimensional relation comprising the distribution of earnings, the value of unpaid labor, legal maintenance obligations, debt exposure, access to collective protection, anticipated transfers, and preferences concerning the welfare of survivors. These components differ in duration, beneficiary, liquidity, enforceability, and sensitivity to mortality. Accordingly, marriage, the presence of children, income, debt balances, and age are incomplete indicators of the death-contingent financial loss that life insurance may address. This monograph examines how dependency structure enters the theory, measurement, estimation, and welfare evaluation of individual death-benefit demand. It critically reconstructs foundational consumer models and the dependents extension developed by Lewis; confronts the divergent empirical findings of Auerbach and Kotlikoff, Bernheim and coauthors, Lin and Grace, Hong and Ríos-Rull, and Inkmann and Michaelides; and places these findings alongside primary household-survey documentation, social-protection institutions, insurance supervisory guidance, and international market statistics. A household-state present-value calculation demonstrates why identical incomes or family counts can imply radically different protection needs. The study distinguishes policy ownership from coverage on a particular life, term death benefits from investment-linked reserves, purchasing propensities from protection adequacy, and associations from causal effects. It evaluates debt survivorship, public survivor income, dependent-care replacement, household equivalence scales, bequest preferences, affordability, underwriting, distribution channels, and the changing organization of families. Its central analytical conclusion is that dependency structure influences potential insurance value through survivor-specific post-death resources and obligations, whereas observed purchases also reflect prices, access, beliefs, institutions, contract design, and market frictions. The monograph provides an original comparative and methodological synthesis of established scholarship and evidence; it establishes no new proprietary model, principle, or theory.
Keywords: life insurance demand; economic dependents; household dependency structure; survivor welfare; life insurance adequacy; financial vulnerability; bequest motives; mortality protection; unpaid care; household debt; life-cycle consumption; insurance ownership; beneficiary designation; insurance penetration; survey measurement; selection bias; causal identification; social security survivor benefits; risk aversion; policy lapse; international insurance markets
Introduction
Life insurance addresses the economic consequences of the uncertain timing of death. The individual whose death triggers a payment may be a primary wage earner, a co-earner, a caregiver, a debtor, a parent with dependent children, an older person wishing to leave a bequest, or someone whose household commitments cannot be represented by earnings alone. The financial loss following death consequently depends on how resources and obligations are organized among the people involved. The scientific task is to reconstruct that organization and examine its influence on the propensity to acquire death-contingent coverage and on the welfare produced by coverage already held.
In his landmark account of dependents and life insurance demand, Lewis (1989) extended intertemporal insurance analysis to include the welfare of household members. That extension exposes an important ambiguity in the phrase "life insurance demand." An insured person may purchase protection motivated by the survival consumption of others, by the preservation of an estate, by contractual debt requirements, or by financial incentives attached to policies with cash value. The same observed premium or insurance amount may therefore represent very different household decisions.
The empirical record complicates simple narratives. Auerbach and Kotlikoff (1991), in their analysis of insurance adequacy, found substantial underinsurance among some married American households under model-based living-standard criteria. Bernheim, Forni, Gokhale, and Kotlikoff (2003) reported substantial mismatches between life insurance and vulnerability in older-worker data, whereas Lin and Grace (2007) identified a positive relation between some vulnerability measures and the amount of term and total life insurance in a household portfolio analysis. These results differ in population, product definition, vulnerability measurement, and model specification. Their comparison is integral to any scientifically defensible inference.
The research question is therefore specific: through which pathways do household dependents, debt obligations, and bequest motives change the private value, observed take-up, coverage amount, and social consequences of life insurance? This question connects Life Insurance, Insurance Economics, Behavioral Insurance and Risk Perception, and Insurance Distribution and Intermediation, while preserving this dossier's distinct focus on household dependency structure.
The investigation uses critical comparison of original peer-reviewed research, publicly accessible data documentation, primary normative instruments, and supervisory communications. Historical findings are attributed to their original study populations and dates. Where an original article is access-restricted, conclusions are bounded by its publisher-supplied abstract or reliably documented original research record. All numerical household examples later in the manuscript are hypothetical calculations, not estimates from the underlying surveys. References and the source-to-claim ledger distinguish verified source content, mathematical identity, authorial deduction, and unresolved questions.
Definitions, Historical Foundations, and Research Method
The scientific object: death-contingent household dependency
Dependency structure describes the composition and timing of relationships through which one person's continued life supports another person's consumption, care, housing, debt repayment, or expected inheritance. A dependent need not be a legally recognized child, and an insured parent need not have the household's highest wage. Dependency may arise from regular transfers to an elderly parent, an adult family member's disability-related support, an informal caregiver's labor, a jointly financed mortgage, or an anticipated future educational obligation.
For research purposes, the unit of observation must be specified. An individual insured life is different from a policy owner; a policy owner is different from a designated beneficiary; and the beneficiaries named on a contract may differ from the persons economically dependent on the insured. A household is also a survey construct whose membership may change between waves or after a death. Each distinction determines what can be inferred from a record of insurance ownership.
The official NAIC life insurance explanation recognizes named beneficiaries, temporary term policies, cash-value policies, and the importance of income replacement, debts, caregiving expenses, and changing household circumstances. This guidance establishes the relevance of those categories to consumer decisions; the underlying obligations and policy rights remain dependent on the particular contract and governing jurisdiction.
Demand can be observed on several margins: an indicator that some life insurance is held; the amount of death benefit on a specified insured person; premium expenditure; type and duration of coverage; employer or group participation; continuation, renewal, conversion, and lapse; and the degree of coverage relative to a survivor-resource benchmark. These outcomes are not substitutes. A family may own an investment-oriented policy with a substantial cash value and a relatively small additional death benefit; another family may hold a large temporary term benefit with no savings component. Market statistics that aggregate both forms are unsuitable as unqualified measures of dependency protection.
The intellectual origins of the inquiry
Yaari's (1965) uncertain-lifetime consumer framework places death uncertainty and insurance within intertemporal consumption choice. The framework identifies the economic importance of being able to transfer resources across uncertain survival states. Subsequent household applications incorporate the preferences of survivors and the resources at their disposal. Lewis (1989) explicitly made dependents central to the life-cycle demand formulation. In this literature the death benefit can be valued because it raises survivor consumption when the insured person's contribution has ceased.
A separate but overlapping research tradition studies bequests. Bernheim (1991), examining the relationship between annuitization and the demand for bequeathable resources, interpreted evidence on Social Security annuity income and private insurance holdings as supporting bequest motives. This work concerns older households and an intertemporal portfolio allocation problem, not simply replacement of the income of a young parent. It illustrates why an insurance policy's existence does not, by itself, reveal a narrowly defined protection deficit.
Hong and Ríos-Rull (2012) use life insurance and household consumption to estimate household equivalence scales and bequest parameters within a structural model. The estimation explicitly links demographic stages, household production, and insurance choice. Such results depend on the maintained preferences and market environment used in the model. They demonstrate how life insurance may contain information about household valuations without proving that every insured family's preferences or labor allocation conforms to those estimates.
A source-critical method
The research separates four evidentiary categories. First, normative documents determine what specific institutional rules or official classifications say, within their legal scope. Second, original surveys and data-provider documentation determine what information exists and how sample design affects inference. Third, economic studies estimate relationships conditional on their samples, assumptions, and methods. Fourth, independent authorial analysis compares mechanisms and draws logical implications without claiming new empirical measurements.
This hierarchy prevents common errors. A theoretical prediction is not treated as an estimated coefficient. A positive correlation is not presented as an experimentally identified behavioral response. A consumer guide does not establish a universally enforceable law. An industry protection-gap estimate, whose assumptions are central to its value, is not interpreted as an observed count of unfulfilled insurance contracts. An aggregate rise in life premiums is not assumed to indicate improved survivor protection.
The primary interpretive criterion throughout the dossier is whether a statement concerns dependency need, insurance demand, observed holdings, or resulting welfare. Four distinct objects require four distinct measurements. The reader can then compare findings without collapsing them into an uninformative single indicator.
Scope and temporal limits
The literature covers several decades of U.S. and U.K. microeconomic research and contemporary international supervisory and market evidence through November 14, 2026. The country comparisons focus on institutional differences supported by available sources, rather than a synthetic worldwide microdataset. Original longitudinal data were not downloaded or re-estimated for this manuscript. The calculations are deterministic illustrations, and the study makes no original causal claims. The purpose is to evaluate the structure of existing knowledge and identify where specific research designs can produce credible answers.
Household Dependency Dynamics and Death-Contingent Resources
Earnings and financial support
Household earnings matter to life insurance because the death of an earning member may eliminate a stream of resources. The loss need not equal gross salary: taxes, work-related costs, the decedent's personal consumption, surviving-earner adjustments, and public transfers may alter the amount available to survivors. In some families the deceased individual contributed most monetary earnings; in others both adults earned similar amounts. Earnings dependency is an ongoing net flow, not the total salary of the insured.
For a particular insured life, let the expected survivor income shortfall at future time t be the consumption resources required by survivors minus available surviving earnings, transfers, and other continuing income. The death-contingent requirement can be translated into a present value using a specified discount convention, time horizon, growth assumption, and probability scenario. The resulting amount measures a financial target conditional on death; it is not an objective optimum in the absence of information about survivor preferences, future labor supply, and mortality.
Shared expenses create economies of scale. When one household member dies, the surviving household may spend less on some goods yet continue to pay almost the same housing costs. Housing, childcare, transport, heating, and debt servicing differ in their adjustment. An equivalence-scale calculation can therefore diverge sharply from a fixed percentage-of-income shortcut. In their structural demographic estimation, Hong and Ríos-Rull (2012) show why household economies of scale and unpaid production are empirically relevant dimensions of insurance choices.
Children, education, and time-varying dependence
The presence of children changes possible loss exposure in several ways. It adds consumption, supervision, schooling, transport, and a longer planning horizon. It can affect the survivor's feasible employment and hence the ability to replace lost earnings. These components vary across infancy, school years, postsecondary education, and economic independence; children of the same age may have different financial needs.
Dependency is therefore time-indexed. A policy providing constant face value for 20 years may match a decreasing stock of educational and childcare obligations differently from a policy whose value grows with inflation. A level payment at an early date may underfund later costs in high-inflation circumstances, while the same payment near the end of the dependent period may exceed the remaining financial shortfall. Evaluating adequacy requires a trajectory, not only the count of children at the observation date.
The empirical presence of a child may also correlate with age, marital status, housing choices, credit usage, income growth, parental education, and attitudes toward risk. An estimated coefficient on child count in a simple ownership regression collects all these relationships unless the design separates them. For this reason, child count is a potentially useful proxy and an incomplete measure of dependency intensity.
Unpaid care and household production
Unpaid labor generates a crucial asymmetry in wage-only calculations. A parent who receives little salary may provide childcare or support a disabled family member; the household may need to buy replacement services or reduce the surviving adult's paid work after the caregiver dies. The economic contribution is visible through a foregone service rather than a paycheck. Its replacement cost may be substantial even where the person was not the household's nominal primary earner.
The household-portfolio evidence of Lin and Grace explicitly considers family members' nonmonetary contributions and reports some evidence that households account for them in insurance decisions. Their study therefore weakens the assumption that measured wage earnings exhaust insurable household contributions. A researcher still needs credible valuations of hours, service quality, replacement feasibility, and the possibility that household members alter their labor supply after bereavement.
A transparent account should distinguish replacement cost from opportunity cost. Commercial childcare fees may exceed the earnings a caregiver would otherwise have received; these are different economic quantities and should not be indiscriminately summed. If a surviving parent reduces paid work to provide care, the resulting earnings decline may substitute for a purchased-care cost rather than add to it. A correct model avoids counting the same lost service twice.
Housing, mortgages, and other debts
Debt generates financial consequences beyond a surviving household's lost consumption stream. A mortgage may leave substantial periodic payments against an illiquid housing asset, and the ability to remain in a home may depend on continuity of income. Other obligations include jointly assumed consumer debt, business guarantees, education loans, and family loans. Legal survivorship, joint liability, estate settlement, collateral enforcement, and creditor insurance arrangements all require separate jurisdictional treatment.
The U.S. Federal Reserve defines the household debt-service ratio as required household debt payments relative to disposable income. It is useful as a macro-financial constraint indicator but cannot directly measure an individual survivor's contractual obligations. The Bank for International Settlements likewise cautions in its debt-service methodology guidance that estimated ratio levels rely on assumptions and that cross-country absolute-level comparisons are problematic.
The distinction between stock and cash flow is fundamental. A $200,000 mortgage balance does not necessarily imply that a $200,000 policy is necessary to preserve survivor housing. The household may continue affordable payments, use a survivor benefit, downsize, refinance, or redeem part of the debt from liquid savings. Conversely, a small mortgage balance can coexist with acute immediate liquidity needs or a loss of access to housing. Protection calculations should model both serviceability and payoff options, subject to legal rules and preferences.
A mortgage-linked insurance product may satisfy a lender's condition while leaving no residual payment for dependent children. A personally owned term policy payable to selected beneficiaries may offer broader uses but demands attention to contract wording and beneficiary designation. The scientific comparison concerns net survivor resources, not merely gross face amount on a policy certificate.
Nonresident and multigenerational dependents
Household membership in a survey may omit an elderly parent supported through remittances, a child living with another guardian, or an adult dependent receiving informal care. Dependency also crosses national borders. Migrant households may support relatives in a different currency and legal system, complicating beneficiary access, documentation, transfer costs, and the valuation of survivor income.
An economy with extended-family support networks may substitute informal transfers for some market coverage, yet network assistance may be contingent, delayed, and positively correlated with other household risks. An unmeasured family support promise may reduce perceived demand without offering the liquidity or enforceability of an insurance claim. The existence of a family network is consequently not proof of adequate survivor resources.
Family transitions and changes in the insured life
Marriage, divorce, separation, remarriage, adoption, widowhood, caregiving changes, and retirement modify the dependency network. So do job loss, relocation, new debt, a dependent parent's illness, and changes in household benefit eligibility. Coverage purchased at an earlier life stage may persist even when the original dependency disappears or increases. Beneficiary instructions may also remain unchanged.
The NAIC consumer questions on life insurance specifically refer to changes in responsibility, beneficiaries, future obligations, and the timing of death benefits. Such guidance is consistent with the research finding that household structure is dynamic. Scientifically, the relevant variable is the state of the household at the date of exposure and the states possible over the policy's remaining term, with explicit recognition that future states are uncertain.
Bequest motives and protection motives
A death benefit can support dependent consumption and also transfer wealth for reasons unrelated to a financial shortfall. Altruistic bequests may reflect a desire to help descendants; strategic or exchange-related transfers may reflect family arrangements; accidental bequests arise from unused assets after an uncertain lifespan. A policy with a bequest objective can remain desirable even where the surviving household faces little measured income loss.
Bernheim's bequest-motive evidence links insurance holdings to other forms of retirement resources. Inkmann and Michaelides (2012), using U.K. microdata, find that term insurance participation and bequests are positively associated with marriage, children, and explicit indicators of stronger bequest motives. These works show that bequests and dependency can overlap empirically without being identical conceptual objects.
The distinction matters for welfare measurement. If researchers label every positive bequest-motivated policy amount as "excess coverage" relative to basic survivor consumption, they may impose an external view of household preferences. If they label every bequest intention as an essential protection need, they may overstate hardship in the absence of coverage. The researcher should identify whose preferences are being evaluated and which outcomes those preferences concern.
Insurance-Demand Estimation and Identification
From risk exposure to a decision rule
A household faces uncertain death, a budget constraint, and possible transfers across states of the world. Under a simplified expected-utility framework, an insured amount B changes survivor consumption in the insured-death state and changes current resources through the premium. A conventional starting expression is:
where q is the probability of the relevant death during the considered period; W represents resources in the living state; S represents resources of survivors in the death state before the benefit; p(B) denotes premium payments under the chosen timing convention; and U_L and U_D represent valuations of the respective states. This is an illustrative two-state expression. Actual multi-period analysis must account for premiums paid only while contracts remain in force, survival-dependent income, timing of benefits, taxes, expenses, and other contingent states.
The marginal condition in an interior differentiable version equates the premium's marginal sacrifice across relevant states to the marginal value of the incremental death-state transfer. It is altered by bequest preferences, the curvature of utility, administrative loads, liquidity constraints, and contract-specific exclusions. A binary purchase indicator therefore reflects preferences and market frictions as well as underlying dependency.
Lewis's dependent-inclusive life-cycle analysis is decisive because preferences and resources of household survivors enter the demand problem. This logic gives an analytical reason to expect a difference between the coverage placed on the lives of equally paid individuals with different numbers and circumstances of dependents, holding other conditions constant. It does not imply that the sign of every observed coefficient is fixed in a setting where income, savings, insurance prices, and household composition change together.
Decomposing the purchase process
Observed demand can be decomposed descriptively into awareness, eligibility, application, insurer acceptance, offered premium and terms, final purchase, and continuation. Group coverage may insert an employer or association into this sequence. These stages differ across countries and socioeconomic positions. Failure to observe a positive policy amount can arise from an explicit preference to self-insure, from affordability constraints, from limited information, from exclusion, or from a change in employment-related eligibility.
A useful statistical design distinguishes the extensive margin of any coverage from the conditional face amount and from the survival of coverage over time. Separate analyses are needed when policy holdings display zeros, very large upper tails, multiple policies, or purchased terms longer than observed household needs. A two-part model is often more informative than a single linear regression on total face value. However, it is still descriptive unless its identifying assumptions are defended.
For a household i at time t, an illustrative empirical specification is:
where D collects time-relevant dependency characteristics, X includes financial and demographic controls, M records access, prices, and institutional features, and F is a probability link. A second equation may model the conditional death-benefit amount. The coefficient beta identifies a partial statistical association under the chosen controls and functional form; by itself it is not the causal effect of having a child or an elderly dependent.
Measurement of the dependent
The dependent variable in a demand study can mean ownership of any life insurance, insured death benefit, term coverage only, an individual's benefit, a household aggregate, annual expenditure, or the value of a product with investment features. Each choice represents a different target. In a two-earner household, adding benefits on both lives can conceal a large shortfall for the death of one person and excess coverage for the other. The economically relevant unit is often the particular death contingency and the associated beneficiaries.
Lin and Grace (2007) expressly compare different definitions of life coverage in their life-cycle analysis. This helps explain why estimating the probability of holding any policy and estimating the amount of term coverage can yield different empirical relationships. Studies using cash values alone or premium volumes should be interpreted as investigations of portfolios or expenditure, with restricted claims about dependent protection.
Reported face amounts may also be subject to missingness, heaping, misunderstanding of employer benefits, or imperfect knowledge of coverage owned by a spouse. Policies can have beneficiaries outside the survey household or benefits pledged to creditors. Thus, the transfer actually available to a survivor can differ materially from the headline death benefit.
Measuring vulnerability rather than assuming it
Financial vulnerability is a counterfactual outcome: how consumption, wealth, or financial security would differ for a surviving person if a particular member died under a defined set of assumptions. It may be operationalized through the present value of lost earnings, a threshold decline in sustainable consumption, a probability of poverty, a shortfall in liquid resources, a subjective perceived hardship, or a household-specific protection gap. Each measure embodies different normative and empirical choices.
The Health and Retirement Study analysis by Bernheim and coauthors (2003) examines the match between holdings and simulated vulnerabilities in older-worker households. Their results cannot automatically be applied to households with infants or extended-family dependents. Their distinct Survey of Consumer Finances analysis examines a different household frame and uses an elaborate life-cycle model; its abstract characterizes insurance as essentially uncorrelated with measured vulnerability across life-cycle stages. These are model-conditioned findings from particular historical U.S. samples.
Lin and Grace's financial-vulnerability index study finds a relation with term or total face amount and reports that older consumers use less life insurance for a given level of estimated vulnerability. One cannot infer that Bernheim's studies have been "disproved" or that Lin and Grace establish universal adequacy. The studies are evidence about differing estimands, household populations, measurement strategies, and contract definitions.
Confounding and reverse causation
Several factors can produce a positive relationship between dependents and insurance without a separable causal effect of dependents. Households with children may have different income trajectories, mortgage balances, access to workplace coverage, and risk attitudes. Purchasing insurance may also facilitate later choices about housing or borrowing. Wealth can increase coverage demand through bequests while simultaneously reducing the need for survivor income replacement.
Prices are especially difficult to separate from preferences because life premiums depend on age, underwriting classifications, product characteristics, and market conditions. People who expect higher risk may seek more coverage, face different offered prices, or be excluded. Unobserved health information and expected family transitions can affect both demand and insurer acceptance. Simple demand estimation may conflate demand with supply and selection.
Policy-level administrative data can clarify premiums and face amounts but may lack informal transfers, dependent-care labor, expectations, and external insurance. Household surveys can measure finances and family composition while suffering from self-reporting errors and incomplete product detail. Linked survey and insurer data would improve coverage but introduce privacy, linkage, and selection challenges. Methodological superiority depends on the question, the sampling process, and the provenance of each variable.
Identification designs and their limits
Panel designs can examine changes before and after marriage, birth, retirement, debt repayment, widowhood, or policy lapse, controlling for stable household traits. Yet timing of these transitions is often anticipated and correlated with income changes, employment decisions, and eligibility. Event-study graphs must evaluate pretrends, staggered timing, control-group comparability, and measurement intervals. A difference-in-differences design identifies a causal change only when its counterfactual assumptions are credible.
Natural variation in taxes, employer coverage, or benefit eligibility might offer identification but often affects household saving and the actuarial value of policies simultaneously. Policy changes also differ across places and cohorts. Instruments require a persuasive exclusion restriction and a relevant first stage; the mere availability of an administrative rule does not make it a valid instrument. Structural models may estimate preference parameters from multiple insurance and asset decisions, but their interpretation hinges on specified household production, mortality expectations, and opportunity sets.
Hong and Ríos-Rull (2012) explicitly estimate consumption scales and bequest parameters within a complete demographic model. Inkmann and Michaelides (2012) use a life-cycle portfolio framework to rationalize observed U.K. participation. These are valuable because their assumptions can generate testable joint implications. Their parameter estimates are conditional on model design, country, dataset, and observation period.
Survey provenance and statistical uncertainty
The Federal Reserve's Survey of Consumer Finances documentation describes probability sampling, public data, multiple imputations, replicate weights, and codebooks. Valid inference must respect these design features; treating every imputed implicate as an independent household exaggerates sample size and underestimates uncertainty. The 2025 survey materials were posted in October 2026, but historical studies cited here used earlier survey waves and should not be retroactively represented as findings from the latest dataset.
The HRS provides a different population frame and longitudinal structure. The RAND HRS Longitudinal File 2022, with its October 2026 version information, documents harmonized family, pension, income, asset, health, and employment variables. That product is not an exhaustive copy of every original interview field, and an older-worker result requires appropriate cohort, selection, survival, and attrition considerations. Combining published HRS and SCF findings without their sampling and age distinctions would erase their main methodological differences.
The literature review by Zietz (2003) collects conflicting life insurance determinants and provides a useful reminder that empirical signs depend on the studies reviewed. This dossier accepts the contradictions as information about specification, rather than forcing a single cross-study coefficient.
Heterogeneous Welfare Effects and Institutional Mediation
Survivor welfare and the identity of the beneficiary
The welfare effect of a death benefit depends on the person who receives it, the timing of payment, the resources previously available, and the alternatives after a death. A dollar transferred to a low-liquidity family member facing essential expenses can have a different marginal value from a dollar transferred to a wealthy heir. This is compatible with standard concave utility and does not require a new theory of insurance. It does, however, require explicit identification of the beneficiary and the consumption state.
A household aggregate can hide distributional differences. A couple may hold enough combined coverage in nominal terms while the benefit on the higher-income spouse is inadequate and the benefit on the lower-income spouse is excessive under a consumption-maintenance criterion. Conversely, a lower-wage caregiver may create a substantial replacement-cost exposure. Welfare accounting should be repeated separately for each insured life, each relevant survivor, and each contractual payment scenario.
The early adequacy study by Auerbach and Kotlikoff reports that a significant minority of American wives in its studied population would experience large losses in sustainable consumption after a husband's death. That finding is historically and demographically specific. It illustrates that nominal insurance ownership can coexist with material widowhood risk and that the appropriate denominator for protection is the survivor's expected resources, not the household's previous gross earnings alone.
Gendered economic roles
Gender enters life insurance demand through differences in labor-market participation, wages, pension entitlements, unpaid care, longevity, financial control, and the likelihood of widowhood. These institutional and behavioral pathways vary across countries and cohorts. Insurance data may reproduce historical norms about who is considered a breadwinner, whose care work is valued, and who receives policy proceeds.
The SCF vulnerability study by Bernheim and colleagues identifies a gender asymmetry in protection conditional on the study's vulnerability measure. That is evidence regarding the surveyed households and their modeled financial circumstances. It does not justify projecting the same gap, its sign, or its magnitude into every contemporary jurisdiction.
Evaluating a gender gap requires comparing like exposures and contractual rights. Higher coverage on one spouse may reflect higher lost earnings; unequal coverage at an equal risk-adjusted survivor loss may suggest differential valuation, pricing, access, or institutional treatment. The decomposition should be empirical, and the label "bias" should be tied to an identified comparison rather than asserted from raw face values.
Affordability, liquidity, and policy persistence
A household can face a large death-contingent exposure and still rationally prioritize current food, housing, medical costs, and debt service when liquidity is scarce. In that setting the marginal benefit of coverage competes with immediate expenditures. A low purchase rate among financially vulnerable households therefore does not demonstrate that the insurance mechanism has little economic value; it may reveal a binding budget constraint or an expensive policy offer.
Acquisition is only the start of protection. A household under financial strain may later stop paying premiums, lose employment-linked benefits, borrow against cash values, or reduce face amount. A cross-sectional snapshot of ownership can miss periods of exposure after lapse or before new coverage is effective. The distinction between initial purchase and sustained coverage is essential to understanding whether dependency needs are protected through time.
The NAIC term and permanent policy guidance explains different premium and cash-value characteristics and warns that abandoning certain policies early can be costly. A research model that compares only headline face amounts may overlook the premium burden, renewal rights, conversion options, and the effects of cash-value accumulation on the household portfolio.
Public survivor protection
Private insurance operates alongside statutory survivor pensions, employer arrangements, occupational schemes, and social assistance. The ILO Social Security Convention No. 102 sets out survivors' benefits in Part X, including loss of support caused by the death of a breadwinner, conditional on the Convention's applicability and ratification context. It is a primary normative benchmark, not evidence that identical entitlements exist in every country.
The OECD's Social Expenditure Database includes survivors among its social policy areas and documents the difficulty of cross-country expenditure comparisons. Public survivors' benefits should enter household resource projections only when eligibility, payment period, indexation, and fiscal treatment are correctly specified. A general national expenditure ratio cannot be converted into an individual's guaranteed benefit.
These institutions can substitute for some privately insured income replacement, but the relationship is not mechanically one-for-one. Benefit ceilings, waiting periods, means tests, marriage conditions, orphan entitlements, and interactions with pensions may matter. Private death benefits may address immediate liquidity where public income is periodic, or provide additional support when public benefits preserve only a floor. The relevant question is the combined package of enforceable resources available to actual survivors.
Distribution channels and advice
Insurance is purchased through individual advisers, digital interfaces, workplace schemes, and other intermediaries. The effective choice set is shaped by information, commission structures, eligibility, defaults, disclosure, policy illustrations, and the ability to compare face amount, duration, and cost. A household can misstate its needs without realizing that group coverage may end with employment or that a beneficiary designation no longer reflects family circumstances.
EIOPA's 2025 insurance consumer survey covers 25,846 EU adults through online interviews in May 2025 and reports substantial use of both online and adviser-mediated distribution. That survey concerns a broader insurance and pensions market, not causal effects of life-insurance advice on dependent welfare. It supports a narrower claim: channel heterogeneity is a real feature of contemporary consumer markets and must be captured when studying access and behavior.
Consumer research should test whether products are explained as death protection, saving, investment, debt protection, or retirement provision. Because these purposes can be bundled, sales statistics alone are ambiguous. Surveys asking only "Do you own life insurance?" fail to reveal which financial obligation the purchaser believed was being protected.
Temporal heterogeneity, mortality, and portfolio choices
Expected death probabilities rise with age in conventional mortality patterns, but the duration of dependency may decline as children become independent, mortgages amortize, and retirement resources accumulate. Consequently, the desired face amount can increase, peak, and decline across the life cycle. This pattern also depends on the insured's health, availability of affordable coverage, time of purchase, and the household's willingness to retain protection for bequests.
The distinction between risk and loss matters. A higher probability of death can raise the actuarial cost of coverage while shrinking the present value of future earnings at risk if remaining work years are fewer. Health changes can affect the supply price as well as demand. A claim that every increase in mortality risk should increase the face amount confuses these different economic channels.
Insurance is also one asset within a portfolio. Liquid savings, pensions, real estate, debt, and death benefits are imperfect substitutes because they differ in timing, taxation, investment risk, access, and legal ownership. The portfolio findings of Lin and Grace reinforce the importance of modeling joint decisions rather than isolated insurance purchases.
Comparative Evidence, Robustness, and International Context
Empirical comparison across established studies
The most informative comparison begins with the estimand, not the reported coefficient. Lewis (1989) offers a theoretical dependent-inclusive demand mechanism and encouraging empirical estimates. Auerbach and Kotlikoff (1991) assess financial adequacy through sustainable consumption in older survey cohorts. Bernheim and colleagues (2003) compare insurance holdings to modeled financial vulnerabilities in an HRS population; related SCF work also examines a different survey population and vulnerability calculation. Lin and Grace (2007) examine life-cycle protection and product definitions through a vulnerability index. Hong and Ríos-Rull (2012) infer preferences and household equivalence scales within a structural model. Inkmann and Michaelides (2012) use U.K. participation evidence and a portfolio-choice model to study bequests.
These contributions ask overlapping but nonidentical questions. The discrepancy between "insurance is weakly related to vulnerability" and "term or total insurance is related to vulnerability" can arise without logical contradiction because risk measures, sample ages, outcome construction, control sets, and assumptions differ. A methodologically sound synthesis therefore reports the relevant conditional finding and evaluates which question it answers.
A transparent comparative table captures the source-defined emphases:
| Study | Population or method | Principal outcome | Direct inferential limit |
|---|---|---|---|
| Lewis (1989) | Theoretical life-cycle model with dependents; empirical illustration | Demand and dependent preferences | Model-dependent demand mechanism |
| Auerbach and Kotlikoff (1991) | Historical U.S. household financial surveys | Sustainable-consumption adequacy | Historical cohorts and annuity assumptions |
| Bernheim et al. (2003), HRS | Older-worker life-cycle vulnerability analysis | Match between insurance and modeled loss | Age selection, counterfactual model |
| Bernheim et al., SCF working paper | Survey of Consumer Finances life-cycle analysis | Vulnerability and protection mismatch | Survey vintage and model assumptions |
| Lin and Grace (2007) | U.S. survey vulnerability index and portfolio | Term and total life insurance amounts | Product and index construction |
| Hong and Ríos-Rull (2012) | Structural demographic household model | Equivalence scales and bequest parameters | Maintained preferences and model calibration |
| Inkmann and Michaelides (2012) | U.K. microdata and life-cycle portfolio model | Term participation and bequest indicators | Country, sample, structural assumptions |
Each row identifies the study's question rather than a universally transferable estimate. The table does not pool coefficients because the source outcomes and definitions are not commensurate.
A household protection gap is a constructed counterfactual
Industry protection-gap studies often calculate required survivor resources minus available support. The Swiss Re New Zealand investigation reports a national mortality protection gap estimate and describes dependency, debt, income, savings, and social protection inputs. The estimate is meaningful within its stated assumptions and sampling design; it is neither an inventory of unpaid insurance claims nor a causal estimate of welfare from buying more coverage.
There are multiple possible protection-gap concepts. An income-replacement gap uses an expected consumption flow. A liquidity gap measures cash required before other resources are accessible. A social floor gap identifies resources needed to avoid deprivation. A preference-adjusted gap concerns the specific utility of survivors and potential bequests. These concepts will yield different figures for the same family, and policy conclusions depend on the chosen criterion.
In practical scientific terms, the protection gap is an accounting and behavioral object. It should incorporate asset liquidity, beneficiary rights, remaining debt, insurance already in force, public entitlements, expected family adjustments, and time horizons. The computation becomes misleading when a single fixed multiple of annual gross income is presented as a universal measurement of economic dependency.
International market statistics and interpretation
The OECD's Global Insurance Market Trends 2026 reports variation in 2025 life-premium growth across jurisdictions, noting that investment-linked products, pension reforms, and demographic developments all played roles. The report states average nominal life-sector premium growth of 8.7% and real growth of 4.0% across its 52 reporting jurisdictions. These measures describe the insurance sector; they do not indicate a corresponding percentage increase in death protection for dependent families.
Likewise, the OECD's life insurance penetration discussion distinguishes the shares of life and non-life premiums in GDP and the effects of investment-linked insurance. Life insurance penetration can rise because of savings or annuity inflows rather than greater coverage on parents or wage earners. A market with a large savings-based life sector may have a substantial dependency protection gap, while a market with relatively modest premiums can offer meaningful public survivor benefits.
For cross-country comparison, researchers must harmonize what counts as life insurance, the level at which premiums are measured, population age, household structure, currencies, price levels, access to group insurance, public benefits, and the reliability of administrative data. Without such adjustments, ranking countries by aggregate premium penetration is a category error for the household-demand question.
Legal and social-protection variation
ILO Convention No. 102 and OECD SOCX illustrate why institutional resources must be explicitly modeled. Some systems provide comparatively substantial survivor pensions; others emphasize private or occupational protection. Eligibility can depend on past contributions, survivor age, family composition, income, and legal marital status. Social protection may be periodic and indexed, whereas private life benefits may be lump sums with contractual limits and claims-administration requirements.
Cross-country household dependency is also affected by different family norms, elder-care arrangements, housing finance, retirement ages, and labor-market participation. These dimensions are legitimate hypotheses for empirical investigation. The present evidence does not support a single estimated international elasticity of life insurance demand with respect to a dependent child or a mortgage.
Checks on counterevidence and sensitivity
The principal counterevidence to a simple dependency-demand story is observed underinsurance in some vulnerable households, the continued purchase of insurance by less vulnerable households, and sensitivity of empirical results to definitions. The first can arise from liquidity, information, supply, and institutional limitations. The second can reflect bequests, portfolio demand, or residual risks absent from a narrow vulnerability model. The third indicates the importance of data construction and research design.
An adequate sensitivity analysis should vary at least the expected support horizon, real discount rate, survivor earnings, survivor benefits, childcare replacement, debt-service assumptions, asset liquidity, inflation, and the treatment of bequest motives. It should also distinguish gross from net benefits and hypothetical from contractually enforceable resources. Age-heterogeneous and gender-heterogeneous results are best interpreted using within-population contrasts rather than unadjusted cross-country extrapolation.
Researchers should make explicit whether they evaluate insurance at an annual point in time, across possible future death dates, or across an entire stochastic life course. A household can be well insured against immediate death while poorly protected ten years later when coverage expires or a new dependent emerges. Static coverage adequacy and long-run reliability therefore represent distinct outcome dimensions.
Reproducibility and research standards
For any future original empirical study, the reproducibility package should contain a variable dictionary linking each insured life, household member, dependent, debt instrument, benefit, and policy to the corresponding exposure date; instrument wording when legally accessible; documented sample exclusions; missingness and imputation procedures; survey weights; code to estimate all models; and sensitivity tables for alternative vulnerability calculations. Public code should avoid exposing identifiable household or health records.
The present monograph retains a different but explicit standard: every source-based substantive claim is linked contextually to its source; full bibliographic identity is retained; historical and contemporary periods are separated; and the illustrative calculations below disclose all arithmetic assumptions. A source-to-claim ledger accompanies the archived manuscript. No unobserved microdata regression, simulated experimental estimate, or independent field validation is represented as completed.
Integrated Scientific Findings
Dependency is relational and contingent
The primary finding from the analytical synthesis is that the economic significance of a life depends on the resources and services that would be lost to specified survivors following death. A gross salary, a marital label, or a count of children can be informative but captures only a subset of the relation. Care responsibilities, nonresident dependents, survivor earning capacity, inherited obligations, and public benefits can change the value of death-contingent protection without changing a survey's conventional household size.
This conclusion is supported by the logic of Lewis's dependent-inclusive theory, by empirical attention to nonmonetary household production in Lin and Grace, and by model-based valuations in Hong and Ríos-Rull. The original contribution here is the explicit reconciliation of these established mechanisms in a source-critical accounting structure. It is a synthesis of existing concepts rather than a new theory or named model.
Demand, holdings, and adequate protection are different quantities
The second finding is the need to preserve the distinction among a household's desire for coverage, its ability to buy, the contract it actually owns, and the welfare the payment may create. Premium prices, underwriting, distribution, household budgets, and policy renewals affect observed outcomes. A single regression of ownership on child count cannot establish the full path from dependency to welfare.
Historical research demonstrates this distinction. Auerbach and Kotlikoff document model-defined coverage shortfalls; Bernheim and coauthors find weak correspondence between some vulnerability measures and holdings; and Lin and Grace find a relation for certain insurance and vulnerability definitions. The disagreement becomes analytically productive once each paper's population and estimand are specified.
The insured-life perspective is indispensable
The third finding is that household-level totals conceal the identity of the person whose death is insured and the person who will receive the payment. Two households with the same combined face amount may have very different protection because benefits are distributed across insured lives or payable to different beneficiaries. This insight follows from ordinary contingent-state accounting and contract design. It should shape both empirical data structures and the interpretation of consumer disclosures.
The replacement cost of a caregiver's work and a breadwinner's foregone earnings require separate calculation. They may coexist, but their interactions need to be addressed to prevent double-counting. Household exposure is not exhausted by formally employed workers or by dependents resident in the same dwelling.
Public and private arrangements interact through survivor resources
The fourth finding is that the relevant comparison uses the combined streams of private benefits, occupational protection, liquid assets, and public survivor transfers. Official social-protection instruments establish that survivor claims form part of many systems, while national details determine their actual value. A study that assumes all households have the same statutory replacement rate will misclassify protection needs.
Similarly, the gross balance of household debt cannot automatically be treated as a death-contingent loss in full. Estate settlement, joint liability, serviceability, and asset disposition determine the realized burden. Debt changes the timing and liquidity of protection needs, which can matter even where the long-run present value of survivor resources appears adequate.
Heterogeneity is an empirical requirement
The fifth finding is that welfare and demand effects vary by survivor resources, household composition, age, employment, gendered labor allocation, insurance access, and social policy. The same death benefit has different distributional implications depending on need and timing. Group averages conceal large tails of financial vulnerability, as the older U.S. studies demonstrate. International premium statistics do not resolve this heterogeneity because they mix protection and investment functions.
No universal direction can be imposed on every coefficient. Additional dependents can raise a household's potential need while simultaneously tightening its current budget. Higher wealth can increase bequest demand while reducing the survivor-income shortfall. More generous survivor pensions can reduce some private needs while generating offsetting saving and portfolio effects. These implications are consequences of recognized economic mechanisms, not claims of newly discovered laws.
A disciplined empirical agenda
The sixth finding concerns research design. Strong future studies should separately observe death-contingent needs, offered insurance contracts, effective ownership, beneficiaries, public entitlements, and persistence. They should identify variation that is credibly exogenous to the household's preferences and financial trajectories, use proper survey design, and report product-specific estimates with uncertainty. Policy evaluation should measure changes in survivor welfare or exposure, not only uptake or premium revenue.
The comparison of contemporary HRS and SCF documentation also establishes a practical constraint: microdata availability does not by itself validate an estimate. Variable comparability, sampling weights, implicate structure, linkage, sample survival, and contract measurement must be scrutinized before regression results can be treated as evidence about dependency structure.
The integrated interpretation
Dependency structure gives economic content to the insured death event. Insurance becomes an effective household protection mechanism when contractually available proceeds align in timing, amount, and beneficiary with the resources and obligations that death would change. Observed demand reflects a separate allocation process under prices, imperfect knowledge, institutional options, and limited budgets. A complete scientific evaluation requires both sides of this relationship.
This conclusion synthesizes verified scholarship and institutional documents. It neither attributes an original theorem to existing authors nor announces a new authorial doctrine. Its significance is methodological precision: specifying the dependent, the insured life, the survivor state, the relevant resources, the product, the counterfactual, and the historical jurisdiction before interpreting a coefficient or a market statistic.
Conclusion
Dependency Structure and Life Insurance Demand examines the way household relationships shape the financial consequences of death and the acquisition of life-contingent protection. The research finds that household dependency includes monetary earnings, care labor, debts, beneficiary rights, social transfers, and bequest preferences, each with distinct time horizons and institutional constraints. These elements make death-contingent exposure specific to the insured life and to the actual survivors.
The literature offers compelling theoretical reasons for dependency to matter and substantial empirical evidence that some households remain underprotected. It also demonstrates that observed insurance holdings can deviate from model-based need and that the strength of the empirical association depends on product definition, vulnerability index, population, and research design. The differences among Bernheim and coauthors, Lin and Grace, and other foundational studies are retained as scientifically meaningful evidence of measurement and identification challenges.
The international evidence confirms the importance of institutional mediation. Public survivor pensions and occupational benefits alter available resources; underwriting and distribution affect access; household liquidity affects sustained purchases; investment-linked life premiums complicate aggregate market indicators. A rise in life-insurance premiums cannot be equated with an improvement in household mortality protection.
The defensible synthesis is that life-insurance demand originates in a household's preferences and constraints over contingent future resources, while effective survivor protection is assessed by the benefits actually payable against specified post-death needs. Strong research records the mapping between them. The monograph ends with that established analytical conclusion, a transparent worked example, and a fully identified bibliography. No external peer review is claimed, and no new theoretical system is asserted.
Methodological Appendix: Transparent Survivor-Resource Calculations
Scope and conventions
This appendix provides a reproducible arithmetic example to show how dependency, debt, and resources interact. Every input is hypothetical, denominated in abstract currency units, and expressed in real terms at time zero. The example is not financial advice, an estimated actuarial premium, or a calibrated risk model. It assumes the death of one specified adult at time zero, immediately available survivor resources, constant annual real needs, and a constant annual real discount rate of three percent. Payments occur at the end of each year; taxes, mortality after the event, and uncertainty of future incomes are omitted for clarity.
For n annual payments of amount C discounted at real rate r, the present value is:
Suppose a survivor's net annual resource shortfall after continuing earnings, public benefits, and necessary expenses is 34,000 for 12 years. The discounted shortfall is:
Assume further that a separately assessed mortgage or debt payoff of 160,000 and immediate transition costs of 20,000 are to be covered, while 85,000 in available liquid assets and 60,000 in death-contingent benefits already payable to the appropriate survivor can be applied. The arithmetic gap is:
The figure is conditional on the assumption that the 160,000 payoff is additional to, rather than already embedded in, the annual 34,000 consumption shortfall. If the annual shortfall already includes full mortgage payments, adding the payoff without deducting future payments double-counts debt. The numerical conclusion must then be recalculated. This single modeling choice can substantially change the purported protection gap.
Sensitivity and identification
At a lower real rate, the same annual shortfalls have a higher present value; at a higher real rate, lower value. Extending the duration of dependency increases the present value. Reducing survivor earnings increases the shortfall, while verified survivor benefits reduce it. An illiquid family home should not be treated as immediately spendable cash simply because it increases net wealth.
Consider a second hypothetical household with the same current earned income but without a dependent child's long remaining care horizon, with a paid-off dwelling and sufficient liquid savings. Its loss exposure may be much smaller. The comparison does not require a new behavioral theory; it shows that income alone fails to determine required protection. Conversely, a person with little earned income may create sizable dependency exposure through essential care work that must be replaced.
The computed G is a scenario-specific accounting result, not a recommendation to purchase insurance of that exact amount. Coverage valuation also depends on the probability of the event during the policy term, death timing, premium affordability, beneficiary rights, and possible future adjustments. A full actuarial simulation would vary the year of death and discount contingent resources from that year, with explicit mortality assumptions and insured-life selection.
Arithmetic verification
The 12-payment real annuity factor at three percent is approximately 9.954004. Multiplying by 34,000 gives 338,436 to the nearest unit. The immediate financing addition is 180,000, existing resources total 145,000, and the resulting gap equals 373,436. These steps can be replicated with a standard calculator. They introduce no empirical inputs beyond those stipulated.
Evidence and Interpretation Appendix
The research cutoff is November 14, 2026. Peer-reviewed source records are used at the level made accessible by the original publisher, abstract, repository version, or official author record. Original articles whose complete publisher texts were inaccessible are cited for conclusions explicitly supplied by their publisher abstracts or archival abstracts; detailed unverified parameter estimates have not been reconstructed.
Original U.S. studies using the HRS, SCF, and other historical household surveys are not direct estimates of households in 2026. Contemporary SCF and HRS source pages document available data and methods but were not reanalyzed. OECD 2026 insurance statistics refer primarily to the insurance sector in 2025, rather than to a measured individual life-insurance protection gap. NAIC guidance is a U.S. consumer-information source, and ILO instruments are interpreted according to their legal applicability. Swiss Re's reported protection gap is an industry-produced model estimate with assumption sensitivity.
The authorial contribution consists of independent critical synthesis, the reconciliation of divergent measures and results, an explicit separation of household need from observed demand and adequacy, and a fully transparent illustrative financial account. This is analytical scholarship grounded in existing literature and documentary evidence. No new theory, doctrine, or original causal estimation is claimed.
References
Auerbach, Alan J., and Laurence J. Kotlikoff. 1991. "The Adequacy of Life Insurance Purchases." Journal of Financial Intermediation 1 (3): 215–241. https://www.sciencedirect.com/science/article/pii/104295739190008N
Bank for International Settlements. n.d. "BIS Data Portal: Frequently Asked Questions" (debt service ratios and cross-country comparability). https://data.bis.org/faq
Bernheim, B. Douglas. 1991. "How Strong Are Bequest Motives? Evidence Based on Estimates of the Demand for Life Insurance and Annuities." Journal of Political Economy 99 (5): 899–927. https://www.journals.uchicago.edu/doi/pdf/10.1086/261783
Bernheim, B. Douglas, Katherine Grace Carman, Jagadeesh Gokhale, and Laurence J. Kotlikoff. 2001. "The Mismatch Between Life Insurance Holdings and Financial Vulnerabilities: Evidence from the Survey of Consumer Finances." NBER Working Paper 8544. https://www.nber.org/papers/w8544
Bernheim, B. Douglas, Lorenzo Forni, Jagadeesh Gokhale, and Laurence J. Kotlikoff. 2003. "The Mismatch Between Life Insurance Holdings and Financial Vulnerabilities: Evidence from the Health and Retirement Study." American Economic Review 93 (1): 354–365. https://www.aeaweb.org/articles?id=10.1257%2F000282803321455340
Board of Governors of the Federal Reserve System. 2026. "Survey of Consumer Finances" (2025 survey release and technical documentation; prior surveys catalog). https://www.federalreserve.gov/econres/scfindex.htm
Board of Governors of the Federal Reserve System. n.d. "Household Debt Service Ratios: About." https://www.federalreserve.gov/releases/DSR/about.htm
European Insurance and Occupational Pensions Authority. 2025. "Eurobarometer 2025: Consumer Trends in Insurance and Pension Services." https://www.eiopa.europa.eu/tools-and-data/eurobarometer-2025-consumer-trends-insurance-and-pension-services_en
Health and Retirement Study. 2026. "RAND HRS Longitudinal File 2022" (October 2026 version 2). University of Michigan Institute for Social Research. https://hrsdata.isr.umich.edu/data-products/rand-hrs-longitudinal-file-2022
Hong, Jay H., and José-Víctor Ríos-Rull. 2012. "Life Insurance and Household Consumption." American Economic Review 102 (7): 3701–3730. https://www.aeaweb.org/articles?id=10.1257%2Faer.102.7.3701
Inkmann, Joachim, and Alexander Michaelides. 2012. "Can the Life Insurance Market Provide Evidence for a Bequest Motive?" Journal of Risk and Insurance 79 (3): 671–695. https://onlinelibrary.wiley.com/doi/full/10.1111/j.1539-6975.2011.01455.x
International Labour Organization. 1952. "Social Security (Minimum Standards) Convention, 1952 (No. 102)," Part X, Survivors' Benefit. https://www.ilo.org/dyn/normlex/en/f?p=NORMLEXPUB%3A12100%3A0%3A%3ANO%3A12100%3AP12100_ILO_CODE%3AC102
Lewis, Frank D. 1989. "Dependents and the Demand for Life Insurance." American Economic Review 79 (3): 452–467. https://ideas.repec.org/a/aea/aecrev/v79y1989i3p452-67.html
Lin, Yijia, and Martin F. Grace. 2007. "Household Life Cycle Protection: Life Insurance Holdings, Financial Vulnerability, and Portfolio Implications." Journal of Risk and Insurance 74 (1): 141–173. https://onlinelibrary.wiley.com/doi/10.1111/j.1539-6975.2007.00205.x
National Association of Insurance Commissioners. n.d. "Life Insurance: Consumer Information and Policy Guidance." https://content.naic.org/consumer/life-insurance.htm
Organisation for Economic Co-operation and Development. 2026. Global Insurance Market Trends 2026. Paris: OECD Publishing, November 2. https://www.oecd.org/en/publications/global-insurance-market-trends-2026_c8be6585-en/full-report.html
Organisation for Economic Co-operation and Development. 2026. "Insurance Penetration," in Global Insurance Market Trends 2026. https://www.oecd.org/en/publications/global-insurance-market-trends-2026_c8be6585-en/full-report/component-3.html
Organisation for Economic Co-operation and Development. n.d. "Social Expenditure Database (SOCX)." https://www.oecd.org/en/data/datasets/social-expenditure-database-socx.html
Swiss Re Institute. 2021. "Closing the Mortality Protection Gap in New Zealand." September 15. https://www.swissre.com/institute/research/topics-and-risk-dialogues/economy-and-insurance-outlook/expertise-publication-new-zealand-mortality-protection-gap.html
Yaari, Menahem E. 1965. "Uncertain Lifetime, Life Insurance, and the Theory of the Consumer." Review of Economic Studies 32 (2): 137–150. https://academic.oup.com/restud/article-abstract/32/2/137/1555438
Zietz, Emily Norman. 2003. "An Examination of the Demand for Life Insurance." Risk Management and Insurance Review 6 (2): 159–191. https://onlinelibrary.wiley.com/doi/full/10.1046/j.1098-1616.2003.030.x