RESEARCH DOSSIER
Annuity Demand Under Bequest and Emergency Liquidity Preferences
Abstract. The economic demand for lifetime annuities is jointly determined by survival-contingent income needs, the desire to transfer wealth after death, and the need to mobilize assets while alive when expenditure or income shocks occur. The three objectives draw on different states of the world and different payment dates. This monograph reconstructs their interaction within established expected-utility and life-cycle theory, evaluates the principal empirical findings concerning retirement bequests and late-life medical expenses, separates precautionary liquidity from intentional inheritance, and examines heterogeneity in observed and experimentally elicited annuitization choices. A central analytical conclusion is that a bequest preference and a liquidity requirement can reduce annuitization through distinguishable channels, while also reinforcing one another when retained savings simultaneously fund emergencies and potential estates. The sign and magnitude of each channel depend on survival, annuity pricing, preexisting guaranteed income, contract death benefits, expenditure timing and the availability of post-purchase liquidity. A transparent two-period marginal analysis demonstrates how moderate annuity loadings shift the willingness-to-annuitize threshold; a separate cash-flow stress illustration locates the shortfall arising when extraordinary spending precedes the first annuity payment. The documented scholarship provides evidence for conditional rather than universal explanations of limited annuity demand: structurally estimated bequest utility, self-insurance against health expenditure risk, annuitant selection, contract characteristics, cognitive constraints and choice framing each operate under different observational and identification conditions. The integrated research result is a disciplined interpretation of these established mechanisms and a reproducible agenda for distinguishing them in household panels and administrative contract records. The work introduces no new named theory, doctrine or causal finding from a new dataset.
Keywords: Lifetime annuities; retirement decumulation; bequest motives; intentional bequests; accidental bequests; inheritance utility; emergency liquidity; precautionary saving; medical expense risk; long-term care; longevity insurance; mortality credits; annuity pricing; actuarial present value; adverse selection; annuity equivalent wealth; heterogeneous preferences; framing; pension payout design; household panel identification; consumer welfare.
Introduction
A lifetime annuity transforms part of a household's financial wealth into a payment stream contingent upon survival. It transfers longevity risk to an insurer, pension pool or another lifetime-income arrangement while changing the amount of wealth available for discretionary withdrawal and inheritance. A pensioner choosing between a life annuity and liquid retirement assets is therefore allocating resources across time, states of survival, unexpected spending needs and beneficiaries. An economic account based only on a desire for guaranteed income omits material dimensions of that decision.
The classical starting point is Yaari's uncertain-lifetime analysis (1965). Its economic insight is that an asset paid only when its owner survives can redistribute consumption resources toward precisely those states in which the owner remains alive to consume. The welfare implications are demanding because actual retirement households also own homes, receive public or occupational pensions, face medical expenses and may wish to leave wealth. The individual-welfare analysis of Davidoff, Brown, and Diamond (2005) clarifies that positive annuitization may remain optimal under incomplete markets while complete annuitization does not follow from the same premises.
Bequest utility and precautionary liquidity are particularly difficult to separate in observed portfolios. An unspent account can be inherited if the owner dies and used to fund care if the owner survives. This is an option over future states, even if the household attaches little positive utility to the mere act of leaving an estate. Conversely, a household that values a bequest may preserve the account and thereby obtain useful emergency liquidity as a consequence. Observing high balances or low annuity purchases does not identify the psychological or economic reason for those choices.
The international relevance is immediate. The OECD Pensions Outlook 2026 emphasizes effective conversion of accumulated pension assets into income and the design of decisions at retirement. Cross-national product menus, public pensions, health financing and inheritance law differ enough that a common mechanism does not imply a common purchase rate. This dossier's scientific object is the household decision; insurer pricing and prudential matters enter when they modify the offered contract or the risk borne by buyers. The related Science fields are Annuities and Retirement Insurance, Behavioral Insurance and Risk Perception, and Insurance Economics, with one canonical dossier identity.
The investigation follows six questions. What exactly is an annuity in a state-contingent lifetime budget? What can observed estates and saving paths reveal about intentional bequests? When do emergency medical costs undermine the value of surrendering liquid principal? How do price, previous pension income and product guarantees affect demand? Which available studies distinguish genuine preferences from choice presentation? What additional data and falsifiable comparisons would improve causal inference? These questions are addressed through verified original research articles, official statistical and policy sources, transparent derivation and a critical synthesis of established economics.
Historical, Mathematical, and Institutional Foundations
Survival risk and the annuity contract
Consider a person aged purchasing a single-premium immediate annuity of price . For annual income paid at the end of year while alive, a simplified actuarial present value is
where is the probability of survival from age to the payment date and is a discount factor consistent with the contractual currency and risk basis. An insurer's actual premium includes administration, guarantees, capital, profit and other pricing components. For a deterministic constant annual benefit and survival probabilities on the relevant basis, a stated payout rate does not alone establish value for money.
The mortality credit follows from pooling. If the insurer pays only survivors, money previously allocated to those who have died can support larger income to the survivors than an otherwise comparable non-life-contingent saving instrument could deliver, given the same discounting and assumptions. Its economic value is conditional on surviving to collect the payment. The buyer trades away a degree of liquidity and estate recoverability in exchange for that conditional income.
In the money's-worth study, Mitchell, Poterba, Warshawsky, and Brown (1999) compare discounted expected annuity benefits with premiums under specified mortality tables and interest-rate assumptions. This measure is an actuarial valuation of expected payouts; it differs from utility-based willingness to pay. A person with low subjective survival probability, large liquid reserves or a strong inheritance objective can assign low utility value to a policy even when its money's-worth ratio is high on the insurer's mortality basis. Conversely, a risk-averse person can value longevity pooling above the expected actuarial value.
Lifetime utility, liquid wealth, and estate states
A compact dynamic formulation defines age-specific liquid wealth , privately purchased annuity income , other pension income , consumption , stochastic emergency spending , and investment return . A stylized budget identity before terminal settlement is
subject to borrowing, timing, nonnegativity and contractual access constraints. The payment stream is conditioned on survival; a life-only contract provides no residual principal upon death. The estate is the after-liability residue at the stochastic time of death . An estate rebate, period-certain provision or survivor annuity changes the state-contingent contract obligations and must be modeled explicitly.
An illustrative expected-utility objective is
where is health and summarizes household beneficiaries and preference heterogeneity. This is an established preference representation, not a newly proposed utility theory. The utility of a transfer can depend on estate size, family composition and the marginal importance of wealth to heirs; the same liquid balance may also be valued because it preserves consumption options during life. A model that folds both motives into a single residual-wealth coefficient will usually lack enough structure to identify them separately.
The source of the annuity benefit matters. A preexisting social-security or defined-benefit pension already provides some longevity insurance and may lower the marginal value of additional annuitization. A liquid retirement account can serve both consumption smoothing and emergencies. A housing asset is potential bequest wealth, but its current liquidity depends on transactions, borrowing, occupancy and applicable law. These assets cannot be treated as identical simply because they appear in a household net-worth column.
Prices, pensions, and institutional choice
There are at least four separate retirement decisions: how much pension wealth is accessible, which payout products exist, whether participation or lifetime income is mandatory or defaulted, and how much discretionary liquid wealth can be exchanged for additional annuity income. Welfare analysis must state which decision is actually observed.
Brown (2001), using annuity equivalent wealth, reports heterogeneity associated with mortality, marriage, risk aversion and previously annuitized resources. A household with already sufficient baseline pension income can rationally prefer greater liquid wealth at the margin even when a household lacking reliable income would value a similar annuity more highly. Pension income floors and personal annuity purchases therefore cannot be merged into a single binary indicator.
The OECD pension-plan recommendation (2022) addresses lifetime-income protection in defined contribution plans, and its policy framework permits attention to partial and deferred lifetime income alongside programmed withdrawals. These are design choices subject to national law and plan rules, rather than uniform legal obligations across countries. For international analysis, mandatory, defaulted and voluntary annuitization must occupy distinct categories. A regime that mandates a payout pattern identifies different behavioral margins from an unrestricted voluntary retail market.
Bequest Motive Estimation
Intentional and incidental wealth transfers
The term bequest motive refers here to positive utility or a stable objective associated with the assets left to survivors or designated beneficiaries. A realized bequest is an observable transfer; it can occur without an intentional bequest motive. A person may save conservatively for longevity, die before consuming the planned balance and leave the remainder accidentally. A strong desire to bequeath need not imply a large realized estate when the person faces prolonged or costly care. The distinction runs both ways.
Hurd's study of mortality risk and bequests (1989) estimates a model in which much bequeathed wealth is incidental under its specification. Lockwood's later bequest-motive analysis (2012) asks whether preferences of plausible strength can eliminate purchases at market annuity rates. Lockwood shows in numerical examples that moderate bequest motives which may not eliminate demand for a hypothetical actuarially fair annuity can eliminate demand at available prices. These results should be interpreted together rather than treated as mutually exclusive statements: estimates of incidental estates, preference curvature, product pricing and purchase thresholds concern different observables.
A model for the value of estates often specifies a bequest-utility term such as for , with a logarithmic limit at , where controls scale and shifts the marginal value at low wealth. A shift parameter can create luxury-good behavior, although it is the full preference and budget specification that determines the resulting demand. A fitted value of is meaningful only relative to the scaling of consumption utility, the survey elicitation and the tax/estate definition.
Lockwood's self-insurance interpretation (2018) emphasizes that bequests treated as luxury goods can support higher savings and lower annuity and long-term-care insurance purchases. When the marginal utility of leaving wealth rises in economically relevant states, retaining a liquid account becomes more attractive. The mechanism works through the opportunity cost of spending down assets and the resulting valuation of insurance. It does not imply that all households intend to maximize inheritances; a different sample or utility shape can generate different demand.
Structural identification and revealed choices
Three categories of evidence must be distinguished. Estate tax, probate, beneficiary and household follow-up records observe realized transfers, often incompletely and with strong legal selection. Longitudinal asset and expenditure observations reveal saving and decumulation conditional on survival, health and income. Elicited preference data ask respondents about hypothetical spending or bequests under controlled tradeoffs. Each category is informative, while each leaves different identification gaps.
The model estimated by De Nardi, French, and Jones (2010) uses the AHEAD longitudinal evidence and simulated moments to jointly allow uncertain longevity, medical expenses and bequest preferences. It finds that the risk of expensive medical care during long lives is an important driver of saving for many higher-income retired singles. The implication for annuity studies is methodological: residual wealth after retirement cannot be assigned wholly to an inheritance motive before the analyst measures health expenditure exposure and public insurance.
Ameriks and colleagues (2020) address preference separation using strategic survey questions alongside wealth information. Their findings assign a substantial portion of late-life wealth preservation among many studied Americans to self-insurance against long-term-care needs. Elicited tradeoffs supply identification that ordinary transaction data alone lack. Their sample and constructed hypothetical questions still require checks for selection, interpretation and consistency with subsequent realized behavior.
An empirical design should compare at least the following evidence in the same units: stated intended transfer to named beneficiaries, actual estate conditional on death, accessible balances before death, expected and realized medical costs, baseline public and occupational pensions, liquidity restrictions, and the offered annuity menu. Heterogeneous mortality is especially consequential. Dead households are selected out of later observed panels; healthier households both live longer and face different annuity valuations. Estate comparisons without survival correction can overstate or understate intentional bequest effects.
Price sensitivity and the bequest margin
The key purchase threshold is marginal. At a fixed price and in a constrained portfolio, a household may value an additional dollar of liquid wealth for inheritance and emergency use more than the incremental conditional annuity payout. Once annuity loadings, survivor guarantees and access restrictions are added, positive demand at a hypothetical fair price need not imply positive demand at actual quotes. This is the central reason that statements about the optimality of some annuitization under restrictive pricing conditions cannot automatically predict universal observed purchase.
A bequest penalty also depends on contract features. A life-only annuity extinguishes income after death. A period-certain annuity or refund guarantee may provide some post-death value to beneficiaries; a joint-life contract continues benefits for a survivor. These options can reduce inheritance concerns while reducing initial or periodic income for a given premium and pricing basis. Classifying all such contracts as identical 'annuitization' destroys an important observed margin.
Bequest language is sometimes used to rationalize every refusal to purchase an annuity. This conflates the motive with the outcome. A testable bequest explanation predicts how choices change when the contract retains a refund, beneficiaries change, tax treatment differs, or a planned transfer is already separately financed. A testable liquidity explanation predicts how choices change with emergency reserve resources, withdrawal rights, insurance for catastrophic costs and the time until the first payout. Those comparative statics guide identification.
Emergency Liquidity Demand
Precautionary balances and spending dates
Emergency liquidity demand is the value of being able to pay for a shock when payment is due. Its objects include unexpected health or care bills, urgent repairs, family support, temporary income interruption, out-of-network expenses, or contractual obligations. It is measured by the amount and timing of immediately mobilizable resources net of available borrowing and contingent insurance. A large asset holding that cannot be sold or borrowed against when a bill arrives provides limited emergency liquidity.
For a life-only annuity, initial premium is typically irreversible after a contractually defined free-look or other available option period, and periodic income cannot generally be accelerated on demand. Actual liquidity depends on the particular product: some annuities provide commutation, withdrawal features, surrender values, guaranteed benefits or borrowing rights, often with costs and risks. Thus 'annuity illiquidity' is a contract-specific variable rather than a property of every retirement income arrangement.
The critical temporal distinction is between an expenditure shock before the next annuity payment and a shock financed by contemporaneous annuity cash flow. Even when the expected discounted lifetime value of annuity benefits is attractive, a retiree may face a cash shortfall before the first payment or during a month with concentrated care expenses. A well-funded liquid reserve can prevent that timing mismatch. Conversely, when a chronic expenditure risk is recurrent and covered by ongoing income, annuitization may support rather than undermine resilience.
Davidoff, Brown, and Diamond (2005) explicitly discuss how the timing of uninsured medical expenses affects annuity demand. The lesson is to model arrival dates, payment dates and liquidity constraints rather than treating medical spending as a one-time deduction from expected lifetime wealth.
Medical and long-term-care expenditure risk
Medical-risk economics gives the liquidity channel concrete substance. De Nardi, French, and Jones (2010) find from their elderly saving model that out-of-pocket costs rise steeply with age and permanent income in their observed U.S. data; the interaction of survival and serious medical cost risk helps explain late-life saving. The program architecture of Medicaid affects both the downside of low assets and the insurance available at extreme costs. Such institutional features cannot be imported directly into other countries' welfare functions without reconstructing eligibility and access.
Peijnenburg, Nijman, and Werker (2017) analyze health cost risk at retirement and show that large expenses early in retirement can lower optimal annuitization, because the individual lacks time to rebuild a liquid buffer. They also find relatively small utility differences around some optimized annuitization allocations in their calibration. This second finding matters: a difference in observed purchase proportions may correspond to a modest welfare difference near a flat optimum, even when the contract choice appears highly salient.
Ameriks et al. (2020) examine long-term-care utility and saving, distinguishing preferences for consumption when needing care from ordinary spending and inheritance. The willingness to pay for retaining assets during a care episode is affected by the quality and availability of formal and informal care, the utility of additional care expenditures, family caregiving and public financing. A study that models 'emergency expenses' without specifying how the money changes actual consumption or care will fail to measure the relevant preference margin.
Long-term-care risk is multidimensional: it includes probability of entry, duration, severity, available substitute family care, price inflation, insurer reimbursement, coverage limits and the timing of cash obligations. Care risk can produce a sequence of correlated expenses rather than a single independent lottery. Its interaction with longevity is particularly important, since living longer can raise cumulative exposure to both ordinary consumption and high-cost dependency states. A liability estimate based on average costs alone understates the role of upper-tail liquidity needs.
Contractual sources of flexibility and friction
Retirement products can divide the financial objectives across instruments. A household may secure a floor of lifetime income while retaining investable reserves. A deferred annuity may insure very advanced age while preserving liquid funds for the earlier retirement years. An annuity with a certain period or death benefit may preserve some transfer capacity, though it pays less annual income at the same premium than a comparable life-only contract under otherwise equal assumptions. A flexible withdrawal strategy can preserve access but does not automatically pool longevity risk.
These choices create a spectrum, not a binary annuity-versus-cash distinction. The scientifically relevant design variables are guarantee duration, refund basis, survivor share, withdrawal or surrender conditions, income commencement, inflation indexing, insurer credit exposure and annuitization fraction. The retirement-income program should compare the full contingent cash-flow profiles rather than compare marketing labels.
In evaluating hybrid products, financial options carry economic cost. A return-of-premium provision diverts part of the pool's mortality credits into surviving estates, and a withdrawal guarantee changes the insurer's funding and capital exposure. Adding these benefits and keeping the same promised periodic income at the same price would violate the actuarial budget unless some compensating subsidy, assumption change or risk transfer existed. Consumer and insurer valuations therefore need a consistent contract identity.
Annuitization Choice Heterogeneity
Household composition and pension endowments
Annuitization demand varies by age, household composition, wealth, preexisting pension income, health, perceived survival, dependents, access to credit, housing, tax position and cultural inheritance patterns. These variables enter both the budget and the preference system. Marriage can change the desired survivor protection and the size of an effective household annuity pool, while children or other beneficiaries may change estate objectives. Age and health change survival probabilities; their observed correlation with annuity choices need not imply a pure preference effect.
The Brown (2001) household analysis uses a utility-based annuity valuation that depends on risk aversion, mortality, marriage and prior annuitization. This suggests a key control: comparing a household receiving substantial guaranteed pension benefits with one whose retirement income depends entirely on withdrawal from investments confounds the incremental value of an annuity. Research should record both mandatory and voluntary income streams and their survivor conditions.
The OECD longevity tables report an OECD-average remaining life expectancy at 65 in 2024 of 18.5 years for men and 21.6 for women. These are population period measures, not the mortality table of selected voluntary annuity purchasers. The distinction matters for both household expected payouts and insurer premiums. Cohort projections and annuitant selection must be modeled separately rather than assigning a general life-expectancy average to a priced contract.
Selection, prices, and product attributes
Finkelstein and Poterba (2004) document UK annuity policyholder selection by detailed contract characteristics, including payment timing and possible estate payments. They find patterns consistent with asymmetric information, but not a substantive mortality difference by annuity size. This is a caution against treating premium amount as a complete measure of private mortality information. It also shows that product attributes correlated with inheritance preferences may simultaneously correlate with longevity expectations.
Premiums, taxes, guarantee options, asset quality and prevailing market yields influence offered payout rates. A high nominal payout rate during an era of elevated nominal yields does not alone imply high real guaranteed income. A life annuity's effective benefit is also affected by currency, fees, insurer default protection and the applicable taxes on premiums, investment accumulation, payments and estates. A household-facing expected present value must be built from contractual terms and the purchaser's information at the date of choice.
Annuity demand estimates are especially vulnerable to choice-set omission. A survey may report 'not purchased' when the respondent had no access to an economically attractive product or had already received compulsory lifetime income. Availability, advice distribution and account restrictions must be observed. Contracts with death benefits should be coded as distinct payout states rather than aggregated with pure life-only contracts.
Behavioral experiments and international observations
Behavioral work adds a separate explanation for the distance between modeled optimal allocations and stated or observed demand. Brown, Kling, Mullainathan, and Wrobel (2008) find in their framing experiment that respondents preferred an annuity over a savings account much more often under a consumption framing than under an investment framing; the NBER version reports 72% versus 21% in those experimental presentations. The variation demonstrates sensitivity of stated choice to representation. It does not establish that 51% of a national population would switch actual retail purchases.
Beshears and colleagues (2014) report from hypothetical annuitization surveys that allowing partial annuitization increases stated demand compared with an all-or-nothing choice; respondents also react to inflation protection, flexibility and control language. Their work includes findings in which self-reported bequest desires are less influential than spending flexibility or counterparty concerns. This can coexist with Lockwood's structurally estimated bequest results: a directly stated reason for choice is a different empirical construct from an inferred marginal utility of transfers.
Agnew and colleagues' controlled annuity experiment tests framing, defaults and participant characteristics under controlled choice conditions. Brown, Kapteyn, Luttmer, and Mitchell (2017) additionally identify large annuity valuation spreads between buying and selling, associated with measures of cognition and financial literacy. A robust empirical explanation should therefore allow preference heterogeneity and decision-process heterogeneity, including perceived complexity, trust and numeracy.
Actual market statistics provide scope but not motive identification. The UK Financial Conduct Authority's 2025/26 retirement-income statistics report 100,144 annuity purchases and 401,137 plans entering drawdown for the year ending March 2026. The source counts plans and distinct product events, not unique persons with equivalent choice sets. It does not establish a causal fraction attributable to inheritance preferences or emergency risk. Comparing these numbers with other countries requires harmonizing access rules, ages, mandatory coverage, products and reporting methods.
Scientific Validity and Sensitivity
A reproducible two-period marginal calculation
A two-period model isolates the tension between bequests and survival payments without presenting a new theoretical system. Let be wealth remaining after period-one consumption; let be the annuity premium, so the liquid investment is . One-period investment gross return is , survival probability is , and the gross annuity payoff per unit premium conditional on surviving is . An early-death estate equals . If alive, terminal consumption equals , abstracting initially from other income and shocks.
With differentiable increasing concave consumption utility and estate utility , the household maximizes
over . The one-sided derivative at zero is
This is a local criterion at the no-annuity boundary, not a claim that the global optimum is uniquely determined without curvature and feasible-set analysis. Under the chosen concavity assumptions is concave, so a nonpositive derivative at zero supports zero annuitization for this simplified two-period objective.
If the annuity is actuarially fair in this single payment model, . Then , and positive marginal demand requires . For a payout factor with proportional reduction , , the condition becomes
provided . The right side declines with . This calculation shows exactly how pricing and the estate marginal-utility ratio interact. It is an illustrative mathematical identity; it is not an estimate of any insurer's load or any population's preferences.
For independent reproduction, set , , and examine the following synthetic quote reductions. The entries are computed directly from the displayed equations.
| Proportional reduction | Conditional gross payoff | Threshold for |
|---|---|---|
| 0.00 | 1.27500 | 1.00 |
| 0.05 | 1.21125 | 0.75 |
| 0.08 | 1.17300 | 0.60 |
| 0.12 | 1.12200 | 0.40 |
For example, a household with a marginal estate-utility ratio of 0.70 in this model has positive marginal demand at the 5% quote reduction but not at 8%. This statement is a constructed comparison, not an empirical forecast. If and , evaluated at the same , the ratio is . With endogenous period-one consumption, borrowing, different estate and consumption levels, illiquid housing, health costs or a residual end-of-period bequest after surviving, the simple threshold changes.
A second source of risk cannot be appended as a constant 'medical-cost haircut.' If a random expense occurs after survival and can be paid from the same disposable amount, the derivative uses , not . The difference follows from curvature. If the expense must be paid before the annuity benefit becomes available and borrowing is constrained, the payment may be infeasible irrespective of a favorable ex ante expected present value. The model has to specify both consumption states and timing.
Cash-flow stress and timing counterexample
Consider an illustrative household with $240,000 of liquid retirement assets and a policy that converts $160,000 into an income stream first payable three months later. It retains $80,000 initially. Suppose an urgent, uninsured expense of $95,000 falls due within the first month, with no credit or other liquid resource. Its immediate liquidity gap is $15,000. The future stream may be valuable for longevity protection, but its later dates make it unavailable to meet the current liability.
If instead the household had annuitized $140,000 and retained $100,000, the same stylized bill would leave a $5,000 cash balance. The comparison says nothing about which allocation maximizes lifetime welfare until the reduction in future annuity income, survival probabilities, spending needs, marginal utility and estate objectives are incorporated. Its narrow analytical purpose is to exhibit a financing constraint whose existence cannot be inferred from annual expected income.
This example reveals three distinct tests: could an emergency be paid at all, what guaranteed income is forgone by holding the buffer, and what happens to potential beneficiaries in short-survival states? The liquidity reserve is valuable for the first, longevity insurance for the second, and estate rights for the third. A product with withdrawal privileges would alter the first test; a refund benefit would alter the third. This state-by-state decomposition is a comparison within established actuarial and life-cycle reasoning.
Evidence hierarchy, bias, and replication
Evidence quality depends on what a source actually observes. A mathematical theorem establishes implications conditional on assumptions. A structural model identifies parameters conditional on functional forms, information sets, measurement and fitted moments. A randomized hypothetical-choice experiment isolates aspects of the decision frame while lacking a binding purchase. Administrative sale counts describe realized transactions but can omit available alternatives and buyers' unobserved objectives. A policy recommendation expresses a design objective rather than empirical identification.
The Health and Retirement Study data catalog lists recent U.S. waves and tracker files that can support longitudinal mortality and wealth studies. The Federal Reserve's 2025 Survey of Consumer Finances portal offers household financial cross-sections and related documentation. They are credible prospective research datasets; this manuscript has not downloaded restricted microdata, constructed individual variables or estimated coefficients. Dataset availability must not be represented as completed original empirical analysis.
A credible replication plan would specify the unit (person, household, pension plan or contract), observation origin, retirement age, mortality horizon and left-truncation conditions. Annuitization should be coded as a fraction of accessible discretionary resources and separately as an extensive-margin purchase indicator. Liquid balances should net out assets unavailable for timely emergency use. Estate values should account for survivors, debt, nonprobated transfers and changes in residence. Medical costs need real-currency treatment, payment timing and insurance reimbursement; pension income needs real and nominal benefit definitions.
Selection into an annuity quote sample, survey response, private pension membership and survival to interview must be documented. Estimates should include sensitivity to mortality bases (population versus annuitant), preference utility curvature, price loadings, medical expenditure tail severity, interest rates, inflation, prior guaranteed income and assumed family transfer motives. Analytical reporting should distinguish an association, a calibrated welfare comparison, an experimental stated response and a causal effect identified by a defensible exogenous shock.
International comparisons require additional care. Health-insurance systems change catastrophic expense exposure; pension conversion rules alter the choice set; estate and inheritance taxation affect transfers; inflation-linked benefits, guaranty schemes and insurer solvency regimes alter the risk of an offered promise. A cross-country regression with only a country dummy and annuitization proportion would aggregate these mechanisms without identifying any one of them.
Integrated Results and Research Agenda
Synthesis of identified mechanisms
The evidence supports an integrated, conditional explanation of annuity demand. Survival-contingent income is most valuable when continued consumption in long-life states has high marginal utility and when prices, guarantees and counterparty quality are favorable. Retained liquid wealth is valuable when current or early retirement shocks require immediate spending, and it can finance longer-term care after the loss of earning power. Estate preservation is valuable when beneficiaries matter, and this can lower the marginal opportunity cost of avoiding annuity purchases. Market availability, previous pension entitlements and presentation of the choice shape the observed outcome.
The mechanisms cannot be inferred interchangeably. A high balance may reveal medical self-insurance, a desired estate, low perceived longevity, a credit-market constraint or an inability to convert a pension product. A zero purchase can reflect an intentional portfolio choice or a missing choice set. A guarantee election can reveal protection of heirs, risk perceptions, product defaults or the consumer's expected survival. The empirical task is to identify which information excludes each alternative in a particular dataset.
The intellectual contribution of this dossier is an evidence-indexed comparison of the available causal and noncausal claims, a transparent marginal purchase derivation, a dated-liquidity counterexample and an explicit set of identification requirements. It does not claim priority for the mechanisms, invent a named model or report a novel estimated coefficient. Foundational results belong to the attributed literature; the synthesis organizes the conditions under which those results can be used together.
Consumer, insurer, and public-policy implications
At the household level, a lifetime-income floor and a separate emergency buffer can serve complementary objectives. This is a portfolio-design implication of contingent cash flows, subject to affordability, contract rates, health needs and existing protection. A default lifetime income option with partial or deferred purchase may preserve options whose value is especially high around retirement. A death-benefit option may raise the attractiveness of a contract to someone concerned with beneficiaries while reducing income for the same premium. These tradeoffs should be measured and disclosed in contract-specific terms.
At the insurer level, observed demand interacts with adverse selection, mortality assumptions, guaranteed-benefit costs, investment risks and capital. Contract design changes the pool of policyholders as well as the benefit obligations. Inferences about demand from provider profitability or contract sales should distinguish household utility from the insurer's cost of promises. Regulatory consumer protection benefits from clear comparisons of life-only benefits, guarantees, withdrawal rights, costs, risk transfer and real-income protection.
At the public-policy level, the relevant target is retirement security across longevity and health states. The OECD's retirement-income assessment emphasizes payout decisions in asset-backed pensions. National authorities can evaluate whether policies permit useful partial annuitization, fair comparison of products and a meaningful liquidity reserve while maintaining protection against very long survival. No numerical optimal reserve or universal annuity fraction follows from the cited evidence, because the institutional and individual inputs differ.
Testable research designs and cross-border scope
A future matched longitudinal study can estimate bequest-intention effects by pairing stated intended transfers with later estate values and recorded medical shocks. A quote-linked design can compare how the same household values life-only, refund and survivor annuities at observed offers, holding actuarial value and transparent contract characteristics in view. A randomized information design can change presentation of liquidity rights, real benefit risks and longevity protection without changing the payout and test both immediate comprehension and verified later purchases.
Researchers should publish an auditable inventory of mortality assumptions, policy-option definitions, availability restrictions, sample recruitment, nonresponse and attrition, censoring, contract quote dates, currency and taxes. For preferences, a mixture of choice experiments and observed behavior can test whether a respondent's stated desire to bequeath predicts the incremental willingness to pay for an annuity refund guarantee after conditioning on liquid buffers. For liquidity, a difference in demand following an exogenous insurance-coverage change or emergency credit availability could be informative, provided the identification design rules out coincident changes in income, health expectations and estate values.
The next stage of comparative research should examine selected countries with materially different health financing and pension rules rather than impose one pooled global price elasticity. A defensible protocol reports country-specific estimates first, harmonizes the economic objects, tests transportability and then examines justified pooled relationships. It preserves the distinction between published scholarly evidence, authorial mathematical calculations and hypotheses offered for future tests.
Conclusion
Demand for lifetime annuities emerges from the allocation of retirement wealth among survival income, immediate consumption flexibility and future beneficiaries. Economic valuation must follow the timing of payments and needs, the household's pension endowment and the exact rights conveyed by a contract. Bequests are both a preference-sensitive outcome and a possible by-product of precautionary savings. Emergency liquidity is valued for the ability to meet obligations at their due date. Their joint effect can make liquid wealth attractive even in the presence of longevity risk and a favorable annuity's contingent return.
The principal studies impose important limits on simple interpretations. Yaari's (1965) baseline identifies the economic attraction of survival-contingent payments under disciplined assumptions. Davidoff, Brown, and Diamond (2005) extend the welfare inquiry and distinguish positive from complete annuitization. Lockwood (2012, 2018) demonstrates how bequest preferences can interact with realistic prices and self-insurance. De Nardi, French, and Jones (2010), Ameriks et al. (2020) and Peijnenburg, Nijman, and Werker (2017) make medical spending and long-term-care exposure central to retained savings and annuity choice. Finkelstein and Poterba (2004), Beshears et al. (2014), and related behavioral experiments require attention to selection and decision architecture. Taken together, these are conditional findings about distinct settings and estimands.
The derived marginal criterion makes the price interaction transparent: when an annuity's survival payoff is reduced relative to an actuarially fair benchmark, the marginal estate-utility ratio required to avoid purchasing becomes lower. The constructed early-expense example independently shows that a future income stream may be unable to fund a present emergency. Both calculations follow established budget and utility mathematics; their parameters are illustrative and have been reproduced arithmetically.
The resulting scientific judgment is that household demand can be explained only after survival states, estate preferences, health-cost exposure, available liquidity, existing guaranteed income, contract options and choice-setting conditions are specified jointly. Empirical research capable of disentangling these objects should combine panel outcomes, contract-specific offers and structured preference elicitation with transparent survival and expenditure assumptions. This dossier supplies a completed theoretical and source-critical synthesis for that work, while maintaining the distinction between checked evidence, reproducible illustration and future identification research.
References
Yaari, M. E. (1965). Uncertain Lifetime, Life Insurance, and the Theory of the Consumer. The Review of Economic Studies, 32(2), 137–150. Verified source record
Davidoff, T., Brown, J. R., & Diamond, P. A. (2005). Annuities and Individual Welfare. American Economic Review, 95(5), 1573–1590. Verified source record
Lockwood, L. M. (2012). Bequest Motives and the Annuity Puzzle. Review of Economic Dynamics, 15(2), 226–243. Verified source record
Lockwood, L. M. (2018). Incidental Bequests and the Choice to Self-Insure Late-Life Risks. American Economic Review, 108(9), 2513–2550. Verified source record
De Nardi, M., French, E., & Jones, J. B. (2010). Why Do the Elderly Save? The Role of Medical Expenses. Journal of Political Economy, 118(1), 39–75. Verified source record
Ameriks, J., Briggs, J., Caplin, A., Shapiro, M. D., & Tonetti, C. (2020). Long-Term-Care Utility and Late-in-Life Saving. Journal of Political Economy, 128(6), 2375–2451. Verified source record
Peijnenburg, K., Nijman, T., & Werker, B. J. M. (2017). Health Cost Risk: A Potential Solution to the Annuity Puzzle. The Economic Journal, 127(603), 1598–1625. Verified source record
Brown, J. R. (2001). Private Pensions, Mortality Risk, and the Decision to Annuitize. Journal of Public Economics, 82(1), 29–62. Verified source record
Finkelstein, A., & Poterba, J. (2004). Adverse Selection in Insurance Markets: Policyholder Evidence from the U.K. Annuity Market. Journal of Political Economy, 112(1), 183–208. Verified source record
Mitchell, O. S., Poterba, J. M., Warshawsky, M. J., & Brown, J. R. (1999). New Evidence on the Money's Worth of Individual Annuities. American Economic Review, 89(5), 1299–1318. Verified source record
Brown, J. R., Kling, J. R., Mullainathan, S., & Wrobel, M. V. (2008). Why Don't People Insure Late-Life Consumption? A Framing Explanation of the Under-Annuitization Puzzle. American Economic Review, 98(2), 304–309. Verified source record
Beshears, J., Choi, J. J., Laibson, D., Madrian, B. C., & Zeldes, S. P. (2014). What Makes Annuitization More Appealing? Journal of Public Economics, 116, 2–16. Verified source record
Agnew, J. R., Anderson, L. R., Gerlach, J. R., & Szykman, L. R. (2008). Who Chooses Annuities? An Experimental Investigation of the Role of Gender, Framing, and Defaults. American Economic Review, 98(2), 418–422. Verified source record
Brown, J. R., Kapteyn, A., Luttmer, E. F. P., & Mitchell, O. S. (2017). Cognitive Constraints on Valuing Annuities. Journal of the European Economic Association, 15(2), 429–462. Verified source record
Hurd, M. D. (1989). Mortality Risk and Bequests. Econometrica, 57(4), 779–813. Verified source record
OECD. (2026). OECD Pensions Outlook 2026. OECD Publishing. Verified source record
OECD. (2022). Recommendation of the Council for the Good Design of Defined Contribution Pension Plans (OECD/LEGAL/0467, adopted February 23, 2022). Verified source record
OECD. (2025). Pensions at a Glance 2025: OECD and G20 Indicators. OECD Publishing, chapter 'Life expectancy'. Verified source record
Financial Conduct Authority. (2026, September 24). Retirement Income Market Data 2025/26. United Kingdom. Verified source record
Health and Retirement Study. (2026). Public Survey Data [data-product catalog]. University of Michigan, Institute for Social Research. Verified source record
Board of Governors of the Federal Reserve System. (2026). Survey of Consumer Finances: 2025 Survey and Supporting Documentation. Verified source record