Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Joint-Life Annuity Survivor Continuation Ratios

Abstract. Joint-life annuities transform two uncertain lifetimes into enforceable contingent retirement cash flows. The survivor continuation ratio specifies the payment after a covered death relative to a contractually defined original amount. Its actuarial meaning depends on which individual dies first, which beneficiary remains entitled, whether the initial participant continues to be paid in full, payment timing, and guarantees beyond the survival contingency. This monograph develops a four-state representation of two-life benefit obligations, differentiates participant-centered qualified joint-and-survivor pensions from symmetric last-survivor contracts, and derives expected cash flows and present values for independent and dependent lifetimes. The analytical synthesis demonstrates that, for a level benefit with single-survivor payment fractions rX and rY, the bivariate survival probability enters the expected payment through the coefficient 1-rX-rY. Thus a symmetric 50-percent continuation contract can have a dependence-invariant first moment while an ordinary participant-centered 50-percent survivor pension generally remains dependence-sensitive. The investigation combines government, professional, standard-setting and scholarly sources, including empirical copula research; it examines mortality selection, regulation, accounting, risk margins, household protection, sensitivity and numerical validation. It records the precise limits of empirical transferability and identifies reproducible work for future paired-life data studies. All analytic identities arise from established actuarial probability; no new proprietary theory is proposed.

Keywords: Joint-life annuity; joint-and-survivor pension; reversionary annuity; survivor continuation ratio; last-survivor payment; actuarial present value; two-life insurance; mortality dependence; copula; longevity; life tables; survivor rights; pension choice; pension accounting; Solvency II; IFRS 17.

Introduction

The contractual object of a survivor continuation ratio is a rule changing the amount of a lifelong income stream following the death of an identified person. A pension may pay its retired member a monthly sum and subsequently pay the member's spouse 50, 60, 75 or 100 percent of that sum. A different policy may reduce its benefit following whichever of two lives dies first. Both products provide protection beyond one person's survival, but their payment state spaces, beneficiary rights and costs differ. The denominator of a ratio is equally important: the contract may reference the joint-life amount elected at retirement, an unadjusted single-life amount, a previously indexed amount or some defined benefit base.

The IRS qualified survivor-annuity guidance describes a U.S. qualified joint-and-survivor annuity, ordinarily payable to the participant for life and thereafter to a surviving spouse, within a stated survivor range of 50 to 100 percent of the participant-life amount. The Canadian federal pension guide describes joint-and-survivor protection of at least 60 percent under the pension plans within its regulatory jurisdiction. The Australian annuity consumer guidance illustrates a 60-percent reversionary benefit and separately explains a guaranteed payment period. These are descriptions of different institutional settings and do not establish a universal ratio or uniform bilateral contract.

Three research questions organize the inquiry. Which legal provisions and observable contract fields specify the covered survival states and payment triggers? How do continuation fractions, the mortality of each life, household dependence, installment frequency and interest assumptions determine present value, variance and the initial income obtainable for a fixed premium? How should pension and insurer obligations be compared across countries, and which evidence can establish that a numerical effect generalizes beyond its original study population?

The method joins actuarial probability derivation, source-specific contract interpretation, comparative institutional research and reproducible deterministic cases. Government and international institutional documents supply bounded descriptions of legal and accounting structures. Published studies document mortality dependence and selection, subject to the scope of accessible full texts and abstracts. The mathematical results are established algebraically and checked by state enumeration. Numerical assumptions are synthetic: they describe no actual insurer quotation, portfolio liability or individual household.

Scientific Foundations, History, and Research Identity

Research object and terminological boundaries

A joint-life payout concerns two measured lives and a rule determining an installment while each is alive or deceased. The term joint life may refer in actuarial usage to the joint survival of both lives, in commercial usage to a last-survivor product, or in pension language to a participant benefit followed by a survivor pension. Scientific analysis must state the legal cash flows rather than infer them from the shared term. A first-death cessation contract stops at the first death, a last-survivor contract ends only after the second death, and a participant-centered joint-and-survivor pension continues the member's own pension throughout the member's life with an elected transfer to a surviving spouse when the member dies first.

The Department of Labor retirement guide explains the U.S. pension context in which a spouse's survivorship protection generally reduces the retired participant's initial pension amount. The adjustment reflects an additional contingent benefit for a given pool of retirement resources. An ordinary price quotation must also incorporate expenses, claims administration, capital, guarantees and any policy-specific terms. An exact price cannot be derived from a continuation percentage alone.

Two jointly named annuitants may both be parties to an insurance purchase, while an occupational pension typically identifies one member and a spouse or eligible partner. A public survivor pension can derive from statute and contribution history. A period-certain rider pays for a minimum duration even if both covered lives have died. A refund feature may transfer unrecovered consideration as a lump sum or installments. These distinct payment structures require separate actuarial cash-flow terms and should not be conflated merely because payments can occur after a person's death.

The UK taxation manual for joint annuities illustrates jointly purchased annuity rights and the relevance of differing ages, contributions and health for tax treatment. The UK beneficiaries' annuity guidance separately describes related dependent or nominee annuities, potentially purchased with the member's lifetime annuity. The resulting legal rights depend on the governing instrument and arrangement. Those institutional distinctions are part of the subject's historical development toward retirement-income provisions and the empirical reason that numerical labels demand documentary verification.

Minimum contract-data requirements

For every policy or plan sample, record the name and current role of the first covered life, the second covered life, the original payment amount and its contractual denominator, the premium or actuarially converted pension, the payment frequency, the timing of the first installment, the benefit after each possible order of deaths and the precise point at which the changed amount begins. Record whether the payment while both survive differs from the payment after the spouse predeceases the member. Record any benefit pop-up, minimum payment duration, return of premium, indexation, contingent beneficiary, guarantee or beneficiary eligibility test.

Historical and legal metadata include the jurisdiction, document title, contract version, pension-plan type, commencement date, governing amendment, consent or waiver, spouse or partnership status and any qualified domestic relations order. Administrative evidence should include date-of-death verification, dates of notification, beneficiary tracing and payment adjustment. The analyst must distinguish an issued policy from a marketing example and a plan-wide default from an actual elected payment form.

A ratio obtained without those details is an incomplete exposure measure. It cannot establish which person must survive, whether a payment terminates on the first death, whether a reduced benefit follows the second death or whether a guarantee overrides mortality. In multi-country evidence synthesis, contract-state comparability is established before pooling observations.

The place of actuarial standards

The IAA international actuarial standards register identifies ISAP 1 for general actuarial practice, standards concerning employee benefits and IFRS 17, and model-governance material. Such guidance supplies a framework for assumptions, methods, disclosures and review. Its precise professional or legal applicability depends on national adoption. A documented valuation should preserve the selected mortality tables, life-pair data, valuation date, interest curve, source versions, validation diagnostics and reasons for material assumption choices.

The science of continuation ratios is therefore interdisciplinary. Actuarial mathematics gives conditional cash-flow values, insurance economics analyzes the price and household demand for those guarantees, demography tests whether pairs of lives are dependent, and pension law decides who is entitled under the specific plan. The integration rests on a common, explicitly described contract rather than an assumed globally uniform product.

Joint and Last-Survivor Payout Schedules

A complete four-state model

Let X be the participant's remaining lifetime and Y the remaining lifetime of the second covered individual, both measured from the valuation date. At a future installment date t, exactly one of four states holds: both individuals survive; X alone survives; Y alone survives; or neither survives. Let B be the initially promised amount per payment period. Define r_X as the fraction of B payable when X alone survives, and r_Y as the fraction payable when Y alone survives. The subscripts identify the individual still alive, not the deceased person whose death triggered a transfer. Payment is B while both survive and zero after both have died, unless a separate guarantee or refund applies.

Survival state Event at t Basic payment
Both alive X > t and Y > t B
X alone alive X > t and Y <= t r_X B
Y alone alive Y > t and X <= t r_Y B
Neither alive X <= t and Y <= t 0

Write S_X(t)=Pr(X>t), S_Y(t)=Pr(Y>t), and J(t)=Pr(X>t,Y>t). The first is the marginal survival of X, the second is the marginal survival of Y, and the third is their joint survival probability. The four mutually exclusive state probabilities are J, S_X-J, S_Y-J and 1-S_X-S_Y+J. Their sum is precisely one, and each must be nonnegative. The immediately testable bounds are therefore

max⁡{0,SX(t)+SY(t)−1}≤J(t)≤min⁡{SX(t),SY(t)}.\max\{0,S_X(t)+S_Y(t)-1\}\le J(t)\le\min\{S_X(t),S_Y(t)\}.

These Fréchet probability bounds apply regardless of which statistical dependence model has been selected. A reported dependent-mortality scenario that violates them cannot correspond to a valid joint distribution. They do not, however, guarantee that independently chosen values of J(t) at different times form a dynamically coherent bivariate lifetime distribution. The time dimension requires its own nonincreasing survival and consistency checks.

The benefit indicator is

C(t)=B{1X>t,Y>t+rX1X>t,Y≤t+rY1X≤t,Y>t}.C(t)=B\{\mathbf{1}_{X>t,Y>t}+r_X\mathbf{1}_{X>t,Y\le t}+r_Y\mathbf{1}_{X\le t,Y>t}\}.

Taking expectations yields

E[C(t)]=B{J(t)+rX[SX(t)−J(t)]+rY[SY(t)−J(t)]},E[C(t)]=B\{J(t)+r_X[S_X(t)-J(t)]+r_Y[S_Y(t)-J(t)]\},

and collecting terms gives

E[C(t)]=B{rXSX(t)+rYSY(t)+(1−rX−rY)J(t)}.E[C(t)]=B\{r_X S_X(t)+r_Y S_Y(t)+(1-r_X-r_Y)J(t)\}.

The identity follows from the partition of the sample space and requires no parametric mortality distribution. It is the principal analytical bridge between beneficiary-state wording and joint-life actuarial valuation. It remains valid if marginal survivals come from different legitimate valuation bases, provided J is a coherent joint survival distribution for those marginals and all amounts refer to the same payment instant.

Three commonly confused product forms

A joint-survival, or first-death cessation, annuity pays B only while both X and Y are alive. It corresponds to r_X=r_Y=0 and has expected installment B J(t). The benefit is most strongly connected to the period of shared survival; once either dies, the payment ceases.

A full last-survivor annuity pays B while at least one covered life remains alive. It corresponds to r_X=r_Y=1 and has expected installment B[S_X(t)+S_Y(t)-J(t)]. Its state event is the union of two survival events. This algebra also verifies the inclusion-exclusion identity behind traditional last-survivor pricing.

A symmetric partial-continuation contract pays B while both survive and rB when either alone survives. It has r_X=r_Y=r and expected benefit

E[Csymmetric(t)]=B{r[SX(t)+SY(t)]+(1−2r)J(t)}.E[C_{\rm symmetric}(t)]=B\{r[S_X(t)+S_Y(t)]+(1-2r)J(t)\}.

At r=0.5, the coefficient on J vanishes. The resulting expected installment is half the sum of the two single-life marginal survival probabilities. This is a mathematical cancellation for one particular two-direction state-payoff structure, not a generic feature of a 50-percent survivor pension.

Participant-centered survivor pensions

The ordinary participant-centered structure differs sharply. The original plan member X is paid B for as long as X survives, including when Y dies first. Only if X dies while Y survives is a fraction rB paid to Y. The correct state fractions are therefore r_X=1 and r_Y=r, and the expected installment is

E[Cparticipant(t)]=B{SX(t)+r[SY(t)−J(t)]}.E[C_{\rm participant}(t)]=B\{S_X(t)+r[S_Y(t)-J(t)]\}.

The final term represents the probability that the participant has died and the named survivor is alive. Increasing the ratio increases this particular contingent benefit. It does not cause the pension to be reduced when Y predeceases X unless the issued contract explicitly specifies a different design or administrative adjustment. A pop-up clause following the death of the intended beneficiary is another separate feature requiring a clause-specific cash-flow representation.

The IRS qualified survivor-annuity explanation gives the U.S. legal setting for one-direction pension protection. The UK lifetime-annuity tax guidance distinguishes a member's lifetime annuity from associated dependants' benefits purchased in the same joint-life arrangement. Such institutional definitions explain why r is insufficient to identify a contract and why the state schedule is the relevant object of comparison.

Timing and nonstandard contractual provisions

An annuity-immediate normally depends on survival at end-of-period payment dates, whereas an annuity-due depends on entitlement at earlier payment instants. Real policies may test life status at the beginning of a payment month, make a full installment covering a month containing a death, or change the amount at the first payment date after notification. If the date-of-death rule differs from the payment-date rule, the valuation must model the triggering event and any contractual lag rather than simply substitute the end-of-month survival probability.

A contract can contain an initial guaranteed period, cash refund, installment refund, cost-of-living index, joint-life escalation, minimum survivor amount, deferred commencement or designated death-benefit beneficiary. Guaranteed-period installments may continue when both X and Y have died and are therefore absent from the four-state table. A deferred annuity may require survival to the elected commencement date before its subsequent state payments begin. State-dependent indexation may require distinct B(t) functions. Each addition should be valued as an explicitly defined contract clause instead of being absorbed into an unexplained survivor fraction.

In the event of simultaneous deaths or uncertain ordering, a directional reversionary annuity may have no surviving beneficiary, while a guaranteed period or refund clause may still require a payment. The administrative and legal treatment of missing death times matters if the two individuals died in a common catastrophe. A mortality dependence model must not silently decide a legal question about simultaneous deaths.

Continuous lifetime value

Assume for the moment that a payment accrues continuously, at rate B per year in the both-alive state, with constant fractions in the two survivor states. Let v(t)=exp[-integral from 0 to t of delta(s) ds] be the deterministic discount factor and T the specified end of the projection horizon. The net present value of promised benefits is

V0=B∫0Tv(t){rXSX(t)+rYSY(t)+(1−rX−rY)J(t)} dt.V_0=B\int_0^T v(t)\{r_X S_X(t)+r_Y S_Y(t)+(1-r_X-r_Y)J(t)\}\,dt.

For full lifetime coverage, T is the limiting maximum duration allowed by the modeled lifetimes, and convergence should be established from the chosen discount and mortality assumptions. A zero floor to the conditional payment after both deaths is part of the derivation. Expenses, reserve risk adjustments, taxes, capital requirements and compensation for uncertainty are additional costs and reporting components.

For payments B_k at deterministic dates t_k with discount factors v_k, the discrete analogue is

V0=∑k=1nvkBk{rXSX(tk)+rYSY(tk)+(1−rX−rY)J(tk)}.V_0=\sum_{k=1}^{n}v_kB_k\{r_X S_X(t_k)+r_Y S_Y(t_k)+(1-r_X-r_Y)J(t_k)\}.

The index k identifies contract dates rather than consecutive years; monthly pension benefits are commonly represented with a monthly schedule. An inflation-indexed B_k can be used if the actual indexation applies to all relevant states. Where the survivor benefit escalates differently from the participant benefit, state-specific B_{X,k} and B_{Y,k} must replace the shared B_k.

With stochastic discount factors D_k and market variables correlated with mortality or benefit adjustments, the correct expression is E[sum_k D_k C_k]. The factorization into E[D_k]E[C_k] needs independent cash-flow and discount-factor risks, or another justified pricing or conditioning argument. The simple deterministic identity is a valuation building block rather than a complete stochastic asset-liability framework.

Monotonicity and accounting for the initial payment

For fixed B, mortality and discounting, the derivative of the payment value with respect to r_Y is B times the discounted expected occupancy of the Y-only state. It is nonnegative. Therefore greater contingent protection for surviving Y cannot decrease promised-benefit present value if all other terms stay fixed. In a fixed-premium comparison, however, the initial amount B generally falls as the unit benefit factor rises. An apparent price contradiction vanishes when the analyst states which quantity is held constant.

The Department of Labor's retirement explanation notes the ordinary reduction in the initial monthly pension associated with a survivor benefit. That policy-level observation is consistent with the mathematical inequality; it is not a claim that a fixed statutory percentage always produces the same reduction in every plan.

The identity also makes clear that a survival-state present value is not the same as the policy's customer price. Product loadings, longevity capital, guarantees, distribution cost, asset returns and other contractual options require independent measurement.

Survivor Continuation Fractions and Valuation

The meaning of a percentage

Suppose a contract's original installment is B, the participant alone receives B, and the surviving spouse receives rB after the participant dies first. The stated continuation percentage r is a conditional benefit fraction. It says nothing about the single-life benefit that could have been purchased with the same premium, and it must be distinguished from the percentage by which an initial pension is reduced when survivor protection is elected. If a single-life option pays 1,200 and a joint-and-survivor option pays 1,050 with a 60-percent survivor benefit, the continuation amount is 630, measured as 60 percent of 1,050; the initial election reduction is 150/1,200, or 12.5 percent. The figures are constructed examples rather than observed premiums.

These quantities answer different questions. The first measures the conditional transfer of income after a covered death. The second measures the cost of purchasing protection as reflected in the initial income amount, with the exact comparison influenced by the insurer's or plan's conversion factors. A third measure, the actuarial money's-worth ratio, relates expected discounted payments to a purchaser's consideration and needs separate assumptions on expenses, selection, default risk and discounting.

The U.S. IRS survivor-annuity explanation specifies a ratio in a legally defined qualified pension structure. The Canadian pension-members' guidance independently describes an at-least-60-percent lifetime continuation for an eligible spouse or common-law partner under federally regulated pension arrangements. These percentages apply to the relevant participant pension and may coexist with a reduced initial benefit. They are not universal premium loading rates.

Fixed benefit versus fixed premium

Let a(r_X,r_Y) denote the net unit present value of the state-payment schedule at a valuation date. Under a deliberately simplified actuarially fair single-premium contract with premium P, the initial payment per stated period is B=P/a(r_X,r_Y) when all time units and purchase dates have been consistently specified. Holding B constant, higher r_Y generally increases expected promises. Holding P constant, an increase in the present-value factor generally reduces B. Insurer expense charges, risk provisions and commercial margins break the exact equality but ordinarily preserve the direction of the economic trade-off.

For the directional participant-centered case, the unit continuous payment factor decomposes as

aQJSA(r)=aX+r(aY−aXY),a_{\rm QJSA}(r)=a_X+r(a_Y-a_{XY}),

where a_X and a_Y are the discounted single-life survivor functions of X and Y and a_{XY} is the discounted joint-survival factor. The difference a_Y-a_{XY} is the present value of payments in the state where Y survives X. It is nonnegative for coherent joint-life probabilities. Its size responds to the spouse's age, relative age, mortality dependence, interest rates and benefit commencement conditions. Consequently, two households electing an identical r may face materially different initial-income conversion factors.

A symmetric continuation structure has factor

asymmetric(r)=r(aX+aY)+(1−2r)aXY.a_{\rm symmetric}(r)=r(a_X+a_Y)+(1-2r)a_{XY}.

These formulas make it possible to diagnose a misclassified contract from the dependency of its quoted value on joint-survival assumptions. If a purported ordinary QJSA with r>0 is valued as 0.5(a_X+a_Y) merely because its marketing sheet says 50 percent, the model has priced a different state-based obligation. Contract classification is a material valuation control.

A one-year numerical experiment with exact inputs

To demonstrate every state and discounting step without claiming a population mortality estimate, set the one-year survivor probabilities S_X=0.94 and S_Y=0.90. Choose the independent-survival benchmark J=S_X S_Y=0.846. The four state probabilities are 0.846 for both alive, 0.094 for X only, 0.054 for Y only and 0.006 for neither. Their sum is one. The feasible joint probability interval at those fixed marginals extends from 0.84 to 0.90.

For a symmetric-direction contract in which X alone receives 60 percent and Y alone receives 50 percent of an initial unit installment, the expected amount at the end of one year is

0.846+0.60(0.094)+0.50(0.054)=0.9294.0.846+0.60(0.094)+0.50(0.054)=0.9294.

Using an illustrative deterministic annual discount rate of 4 percent, its present value is 0.9294/1.04 = 0.8936538462 per initial unit. For a participant-centered survivor pension paying X in full while X survives and Y 60 percent following X's earlier death, the expected payment is 0.94+0.60(0.054)=0.9724. Its present value at 4 percent is exactly 0.935. The second schedule costs more in this illustrative fixed-B comparison because the original participant's payment does not fall when the other life dies first. These two values are illustrative state-contingent benefit factors, not premiums quoted by a carrier.

Now consider a symmetric 50-percent continuation contract. At time one, its expected amount is 0.5(0.94+0.90)=0.92. This value is independent of the admissible choice of J when both marginal survivor probabilities and contractual fractions are fixed. The result is algebraically exact. In contrast, a participant-centered 50-percent benefit has expected amount 0.94+0.50(0.90-J) and therefore changes as J changes. The same nominal continuation ratio is compatible with materially different actuarial sensitivities.

A controlled dependence sensitivity

Holding S_X and S_Y fixed, increase J from 0.846 to 0.860, still inside the feasible interval. The corresponding four state probabilities become 0.860, 0.080, 0.040 and 0.020. For r_X=0.60, r_Y=0.50, the expected installment changes from 0.9294 to 0.9280, a decline of 0.0014. The one-year discounted difference is -0.0013461538 per initial unit. For a participant-centered 60-percent pension it changes from 0.9724 to 0.9640, a decline of 0.0084; its discounted difference is -0.0080769231.

The relevant derivative at any time, with fixed marginals and payment amounts, is

∂E[C(t)]∂J(t)=B(1−rX−rY).\frac{\partial E[C(t)]}{\partial J(t)}=B(1-r_X-r_Y).

For the directional participant-centered structure this derivative equals -rB, reflecting the probability transferred from the second-life-survivor-only state into the both-alive state as J increases at fixed marginals. For a first-death cessation annuity the coefficient is positive, and for a full last-survivor benefit it is negative. For symmetric half-continuation it is zero.

These signs cannot be interpreted as a universal claim that positive spouse correlation lowers every annuity premium. A positive statistical association is a property of a complete joint distribution, the entire trajectory of J(t), its relationship to the marginals and the contractual state schedule. A scenario changing a joint probability at one date is a local algebraic sensitivity rather than a full lifetime copula recalibration. The 1996 dependent annuity valuation study by Frees, Carriere and Valdez found a sizeable valuation consequence in its particular insurer data; its approximately five-percent figure is a sample-specific finding, not a global adjustment or a validation of the synthetic figures above.

Second moments, cash-flow covariance and long-horizon risk

An equality of expected installments does not imply an equality of insurer risk. For the same four-state schedule the second moment of the one-period payment is

E[C(t)2]=B2{J+rX2(SX−J)+rY2(SY−J)}.E[C(t)^2]=B^2\{J+r_X^2(S_X-J)+r_Y^2(S_Y-J)\}.

Subtract the square of the expected payment to obtain the variance. At r_X=r_Y=0.5 the coefficient on J in the second moment is 1-0.25-0.25=0.5, even though J cancels from the first moment. Thus dependence can materially affect payment dispersion while leaving the expected installment invariant.

A lifetime present value includes temporal covariance between payments at successive dates, as well as survival covariance between individuals. Payments across months are not independent because survival from one month to another is persistent. A stochastic present-value variance requires a double sum or double integral of covariances, not the sum of separately calculated installment variances. Likewise, an insurer's aggregation across couples depends on portfolio-wide common shocks and systematic longevity improvements in addition to dependence within each couple.

If longevity improvement affects both lives' marginal distributions, the first-moment cancellation for the symmetric 50-percent product no longer implies absence of mortality-model risk; changes in S_X or S_Y change the value directly. If stochastic financial discount factors covary with mortality, E[D_k C_k] need not factor into E[D_k]E[C_k]. For a mortality-linked rider with a death-triggered lump sum, state occupancy cannot be used alone: the distribution of death times and the claim trigger must be modeled.

Household protection and benefit adequacy

Survivor protection reallocates consumption resources between the state of joint life and the state in which only one partner remains. Household expenditures typically change after death, but their composition matters. Housing costs, property taxes and outstanding debt may persist while some personal consumption disappears; later-life care and health expenditures can also be concentrated in survivor years. The pension's continuation percentage is therefore an income-insurance parameter rather than a direct measure of how much a survivor needs.

At a fixed premium, selecting a higher fraction may lower income while both beneficiaries live. At a fixed initial income, higher continuation increases the value of the contingent survivor claim. Assessing whether either option improves household welfare requires the partners' preferences, other pension rights, public survivor entitlements, assets, liquidity, expected spending, relative mortality and legal constraints. The OECD recommendation on pension design provides an institutional context for adequate and coherent decumulation, but does not prescribe one globally optimal continuation ratio.

In two-pension households, the disappearance of one pension on the first death can interact with the survivor election on the other. If each person has an asymmetric survivor provision, combined income must be calculated state by state. The resulting aggregate fraction after a death need not equal either policy's printed percentage. Age and sex differences, policy-specific inflation escalation and public-benefit offsets can further alter the effective ratio in real spending units.

Selection, pricing fairness and communication

Households may self-select into survivor options based on privately known health, family history, age disparity, bequest motives or financial dependence. A nominal 100-percent continuation election could be more attractive when the intended survivor is relatively young or expected to be long-lived. Pricing rules that only differentiate by the participant's age may therefore embed selection and cross-subsidies. Empirical confirmation requires paired-life election and survival data, not inferences from the contract terms alone.

The UK annuity selection evidence by Finkelstein and Poterba (2004) reports mortality selection associated with contract characteristics in the examined historical market. The accessible publication abstract supports that limited empirical description, while it does not establish how continuation ratios causally change selection in all current markets. A regulatory or fairness analysis should distinguish lawful contractual differentiation, group pooling, accessibility for diverse household structures and demonstrated claims experience.

Consumer communication should present three monetary examples: the amount while both parties survive, the amount when X alone survives, and the amount when Y alone survives. Showing a percentage alone obscures the relevant entitlement. A further line should say what happens when neither is alive, including guaranteed installments, and identify the exact event and payment date at which each change takes effect. These straightforward disclosures also improve actuarial data coding and later claims auditability.

Correlated Household Longevity and International Institutions

Dependence between two remaining lifetimes

If X and Y are statistically independent, J(t)=S_X(t)S_Y(t). Independence simplifies valuation but does not follow from having separate actuarial tables for each person. Household members can share living conditions, exposure to disease, socioeconomic position, accidents, inherited risks and responses to bereavement. A shared environment can produce a long-run association while bereavement may produce a short-lived change in the remaining partner's hazard. Association can also differ by duration since the first death, the age difference between partners, pension eligibility and how couples entered the observed annuity sample.

A joint-life mortality model must therefore define the unit of observation: a paired couple or other legally covered two-life arrangement. An individual annuitant life table is a marginal source and contains insufficient information to estimate the association between the lives. Likewise, a general population correlation need not describe privately purchased two-life annuities because purchasing and selecting survivor protection can be nonrandom.

The foundational dependent-mortality annuity analysis by Frees, Carriere and Valdez (1996) used a copula representation in empirical annuity valuation, showing that dependence can affect prices of particular joint-life products. Their sample-specific approximately five-percent valuation difference is evidence of materiality under the original empirical setting, not a general pricing constant. The 2023 Egyptian family lifetime study estimates dependence in a substantially different population and distinguishes spousal from other family relationships. These studies demonstrate the value of paired-life evidence and caution against exporting one country's dependence coefficient to another.

Bivariate lifetime and copula representation

Let F_X(t)=1-S_X(t), F_Y(u)=1-S_Y(u). For continuous marginal lifetime distributions, a copula C expresses the joint probability Pr(X<=t,Y<=u)=C(F_X(t),F_Y(u)). It follows that

J(t)=1−FX(t)−FY(t)+C(FX(t),FY(t)).J(t)=1-F_X(t)-F_Y(t)+C(F_X(t),F_Y(t)).

The joint survivor function can equivalently be specified by a survival copula. A model must respect the marginals, the probability bounds, boundary conditions and monotonic behavior of a valid joint distribution. Families such as Frank, Gumbel, Clayton and Joe differ in association and tail structure. An empirical fit can be evaluated through likelihood-based diagnostics, censoring-aware estimation, out-of-sample prediction and comparisons of implied cash flows.

The Egyptian study's open-access methodology compares multiple copulas with Bayesian Markov chain Monte Carlo and inference functions for margins. Its research design uses Egyptian social pension data, where beneficiary types include spouses and relatives, rather than a homogeneous voluntary commercial-annuity block. The authors note that individual underlying data are available upon request; the full set of person-level observations was not independently reproduced in this dossier.

A correlation coefficient does not uniquely determine a copula. Two models may produce similar rank association yet different upper or lower tail dependence and different probabilities of very long last-survivor payment periods. A model based only on a single pairwise correlation can also omit the direction of the survival dependence after the first death. The selection of a dependence family should follow diagnostic evidence and substantive properties of the sample, with sensitivity tests across plausible alternatives.

Censoring, age disparity and sampling

Paired-life insurance data are frequently subject to right censoring: both individuals may be alive when observation ends, or the insurer may stop tracking a spouse once its own payment liability ceases. If a plan captures the participant's death but does not follow the beneficiary to their eventual death, the recorded data may disproportionately include deaths that triggered an immediate benefit. When the person entering the data set must die during the sampling period, the likelihood must reflect that conditional selection. These recording processes can bias naive mortality dependence estimates.

For a defensible analysis, retain the ages of both lives at entry, age difference, calendar year, sex or other authorized risk class, product form, election, commencement date, exit reason, observed death date and information about nondeath termination. Record whether the policy is in force, whether another annuity took over, and whether the beneficiary remained eligible. Use a likelihood or estimating procedure designed for exact and censored pair records; report the sensitivity to missing linkage and incomplete dates.

The Egyptian family evidence documents different dependence strengths across relationship types and warns against assuming that evidence from high-income spousal annuities identifies the relation in other household structures. Its sample is tied to specific institutional and historical conditions. The distinction between observed correlation and a causal bereavement effect also matters: shared exposure, age, socioeconomic status and survivor selection may contribute to the association.

Selection into annuity purchase and election

The empirical dependence of two insured annuitants can differ from dependence among all married couples. Couples elect joint-life coverage partly because of observed or unobserved information about relative survival and household needs. Pension members may have a default survivor right imposed by law, while private annuity buyers may choose among products. Mandatory benefits can reduce certain forms of election selection but do not erase socioeconomic and employment selection into the covered population.

Finkelstein and Poterba's UK annuity-market selection research identifies contract characteristics associated with mortality differences in a specific historical market. The evidence provides a conceptual and empirical reason to test selection in present-day continuation elections. It cannot by itself answer whether one household's elected percentage caused different mortality, nor determine a universal correction for longevity dependence.

An applied study should estimate three layers separately when data permit: marginal longevity of each covered population; within-pair dependence conditional on measured attributes; and the process by which couples enter the portfolio or choose a continuation ratio. Pooling these layers into a single crude age adjustment makes assumptions difficult to audit and may confound a high continuation fraction with an unusually long-lived covered second life.

United States: qualified pension survivorship

In the U.S. qualified plan context, the IRS QJSA definition specifies a life annuity for the participant followed by a life annuity for the qualifying surviving spouse, with the survivor amount between 50 and 100 percent of the participant amount. Applicable plans may require particular consent to waive this form, and the rules distinguish the qualified optional survivor annuity and qualifying alternative beneficiaries under a domestic relations order. Consequently the default participant-spouse contract must first be characterized under its precise statutory and plan provisions.

In the standard directional payment representation the participant X receives B while alive and the eligible spouse Y receives rB after X's death if Y then survives. A hypothetical symmetric 50-percent clause would pay X only half of B if Y died before X. Those two arrangements are not legally or economically equivalent. A practitioner examining an elected QJSA must inspect the plan's actual death-order, pop-up, beneficiary and installment timing terms to identify any deviation from the simplified directional schedule.

The U.S. labor-department retirement guide is useful for understanding why the elected form can reduce initial monthly benefits. It is participant education rather than the signed plan election, so it cannot establish all rights in a particular case.

Canada: federally regulated pensions

The OSFI Pension Members' Guide explains that the relevant federally regulated pension generally must be paid as a joint-and-survivor pension when a spouse or common-law partner exists at retirement unless that person waives the right. It provides lifetime survivor income of at least 60 percent of the pension after the member dies and acknowledges that the member's initial pension may be reduced to support the entitlement. The guide separately notes the application to relevant transfers used to purchase an annuity.

The federal scope is decisive. Canada's pension landscape includes federal, provincial and public schemes with differing statutory rules. The documented at-least-60-percent figure should be represented as an entitlement within the identified plan class, not as the universal Canadian annuity percentage. An accurate international comparison also distinguishes spouse eligibility and waiver from the model's statistical survival process.

Australia: reversionary annuities and guarantees

The Australian regulator's consumer site MoneySmart on annuities describes a reversionary beneficiary whose payment continues after the original annuitant dies; it includes an example of a 60-percent payment to the beneficiary. The same guidance discusses guarantees under which a nominated person or estate may receive payments for a guaranteed period. The latter is a payment conditional on contractual term rather than only survival of Y. Thus an Australian policy must be classified by both its survivor ratio and any independent guarantee.

Australian tax and pension interactions may influence consumer choices, but this work makes no jurisdiction-wide estimate of demand or compliance. The governing policy document, issue date, eligibility terms and indexed amounts determine the cash flows that can be compared with U.S. or Canadian examples.

United Kingdom: annuity type, beneficiaries and ownership

The HMRC joint-annuity ownership example shows why jointly purchased annuity rights, contributions and life circumstances can matter to the legal and tax analysis. The beneficiary-annuity tax manual and lifetime-annuity rules classify lifetime and related survivor entitlements within UK pension taxation. The 2025 HMRC response document records a policy discussion about pension and survivor arrangements. These sources are relevant to contract characterization, but their tax or consultation functions do not amount to an actuarial determination of a universal UK continuation ratio.

A UK joint-life policy might be a participant-centered purchased annuity, a last-survivor benefit, or an arrangement with additional guaranteed years. Each must be mapped to the actual payee and benefit after each death. Beneficiary nomination and ownership can also determine whose interest persists after the initial annuitant dies, a legal question which cannot be answered solely from survival notation.

Public survivors' pensions as an institutional boundary

The Government of Canada's CPP survivor-pension explanation describes a public benefit whose payment depends on eligibility, age and integration with other CPP entitlements. Its statutory calculations differ from a commercial joint-life annuity priced from a single purchase consideration. A fixed percentage printed on a private annuity contract cannot simply be transferred to a contributory public pension with caps and combined-benefit rules.

The OECD Pensions Outlook 2026 provides a contemporary cross-country setting for survivor pension policy and variation in eligibility and retirement arrangements. Cross-national evidence should classify the institutional source of the benefit before aggregating ratios. A comparison of insured annuities with public or occupational survivor pensions is informative if the benefit formula, tax treatment, legal beneficiary and funding mechanism are all clearly defined.

Legal entitlement, waiver and death ordering

A survivorship clause requires a legal answer to which relationships qualify, when rights vest, how waiver or consent works, what divorce or remarriage changes, and whether another individual can be named as beneficiary. Documentation may make the spouse at commencement decisive or may establish a separately designated individual. In certain plans a former spouse may be treated as a surviving spouse under a qualified domestic relations order. Such facts control benefit eligibility even if a statistical model predicts that the person will survive.

The order of deaths matters especially for reversionary rights. If the intended survivor dies first, the original participant's benefit may continue unchanged, revert to a higher amount through a pop-up provision, or be adjusted only if the issued form says so. If deaths occur in a common accident, applicable rules about proof of survivorship and guarantees may determine the outcome. Actuarial modeling should implement those rules explicitly and identify unresolved legal facts as data-quality issues rather than concealing them in a statistical assumption.

International evidence therefore supports a general conclusion of contractual heterogeneity. The common scientific object is a set of survival-state contingent cash flows; the actual percentages, qualified relationships, waivers and funding responsibilities arise from distinct institutions.

Evidence Validation, Sensitivity, and Model Governance

A hierarchy of source verification

A credible study of continuation ratios should separate the legal definition of a benefit from the estimated probability of payment and from the accounting treatment of the resulting obligation. The best evidence of a particular benefit obligation is the governing policy or plan instrument and a valid election, read with legislation or regulation applicable at the relevant time. Regulator explanations can verify general legal rules; model documents and marketing brochures identify possible designs; statistical studies test actual mortality and selection for defined samples. A citation to an authoritative regulator does not substitute for the issued contract when deciding whether a particular death changes a specific installment.

This investigation compares publicly accessible primary institutional materials and scientific publications. The IRS survivor-annuity guidance is the source for the defined U.S. QJSA structure; the OSFI pension guide is the source for the federally scoped Canadian example; and Australian annuity guidance grounds the reversionary and guarantee-period distinction. Each has a different institutional purpose and jurisdiction. None provides the actual election schedule of a sampled annuity or an empirical bivariate mortality table.

The Frees, Carriere, and Valdez study and Finkelstein and Poterba's UK selection research are used within the limits of accessible bibliographic and abstract evidence. The 2023 Egyptian dependence paper is open-access full text, allowing closer inspection of sampling, copula estimation and limitations, although the source individual records are not openly reproducible. Theoretical identities and synthetic calculations in the present manuscript are separately marked as derivations and can be verified without either proprietary data set.

Mortality level, improvement and population mismatch

Two covered lives require marginal survival probabilities consistent with the contract population, duration and age at issue. Population tables, individual annuitant mortality experience, occupational pension populations and medically underwritten groups can differ through selection and observable heterogeneity. The SOA individual-annuity reserving report describes an established U.S. annuity valuation mortality basis. It does not provide a universal bivariate mortality distribution or an estimate of the survival dependence of every insured couple.

A lifetime annuity may be payable decades after the issue date. Cohort mortality improvement therefore changes the marginal probabilities S_X and S_Y over future calendar years. The SOA RPEC 2026 mortality update reports that its committee did not release a replacement improvement scale for 2026 and continued to regard MP-2021 as a reasonable best estimate for typical retirement valuations, particularly at ages 65 and older, pending additional experience. That statement is a professional research finding regarding population trends and pension assumptions; it does not override plan-specific demographic evidence, supply joint-life copula coefficients or guarantee that a particular insurer's selected assumption is suitable.

A practical calibration record should specify the mortality table edition, observed exposure dates, age definitions, whether sex or occupation classifications are legally permissible, baseline mortality adjustments, cohort projection convention, mortality improvement assumptions and upper age limit. The method for reconciling population and annuitant mortality should be documented. Advanced-age data sparsity and changes in death registration quality deserve explicit sensitivity analysis because long-lived survivors create the duration tail of the obligation.

Independent reproduction of the illustrative arithmetic

The one-period example in this monograph has input S_X=0.94, S_Y=0.90, J=0.846 and discount factor 1/1.04. The state-probability vector is

(pboth,pXonly,pYonly,pneither)=(0.846,0.094,0.054,0.006).(p_{\rm both},p_{\rm Xonly},p_{\rm Yonly},p_{\rm neither})=(0.846,0.094,0.054,0.006).

The probabilities are nonnegative and sum to one. They satisfy 0.84<=J<=0.90, the fixed-marginal feasible interval. The r_X=0.6, r_Y=0.5 contract produces expected amount 0.846+0.6(0.094)+0.5(0.054)=0.9294 and discounted value 0.9294/1.04=0.893653846153846. The r_X=1, r_Y=0.6 pension produces 0.846+0.094+0.6(0.054)=0.9724 and discounted value 0.935.

Changing J alone to 0.86 produces the alternative state vector (0.86,0.08,0.04,0.02). The r_X=0.6, r_Y=0.5 expected amount is 0.86+0.6(0.08)+0.5(0.04)=0.928, a change of -0.0014. The directional pension yields 0.86+0.08+0.6(0.04)=0.964, a change of -0.0084. Dividing by 1.04 gives respective present-value changes of -0.001346153846154 and -0.008076923076923. The symmetric half-continuation formula gives 0.92 at either admissible J. These values can be recovered by direct state enumeration or by the compact expression r_X S_X+r_Y S_Y+(1-r_X-r_Y)J.

The exercise is a unit test for code implementing benefit states. It does not test a full bivariate lifetime model across multiple years, the predictive accuracy of any particular copula or a real insurer's pricing. For multiple years, the entire sequence of state distributions must be generated by one coherent joint distribution, and cash-flow dependencies across years must be respected.

Verification of benefit-state behavior

The strongest contract-form tests are structural rather than arbitrary numerical targets. Setting r_X=r_Y=0 must reproduce the first-death cessation value. Setting both to 1 must reproduce the full last-survivor value. Setting both to 0.5 must reproduce half the sum of two single-life expected benefits under deterministic fixed benefits. Setting r_X=1 and r_Y=0 must reproduce the participant-only life annuity. Increasing r_Y with fixed B must not lower expected benefit value because the added promise is paid in a state having nonnegative probability.

When two covered lives are exchangeable under the assumptions, exchanging their labels and simultaneously exchanging r_X and r_Y must not alter value. Conversely, swapping only the label while retaining unequal beneficiary fractions can legitimately change the contract. This serves as a check against a common coding error in which the survivor fraction is stored by death-event identity rather than by the person entitled to the payment.

For a portfolio, the validator should also test policies where a named survivor predeceases the participant, where the insured participant dies immediately after commencement, where the beneficiary dies shortly after the participant, and where a guarantee period outlasts both. The model must reconcile generated payment rows with administrative benefit schedules. A valuation that passes the algebra but fails these entitlement cases does not establish that the actual liabilities have been measured.

Model validation and estimation uncertainty

A copula-based mortality analysis should compare candidate forms using pairs observed under compatible entry and censoring mechanisms. Researchers can evaluate marginal survival fits with observed-to-expected deaths and exposure reconciliation, then conditional survival and dependence with likelihood-based diagnostics and out-of-sample predicted payment states. Posterior uncertainty or confidence intervals should cover both marginal and dependence parameters. Validation can test age differences, sexes or other authorized population classes, pension commencement cohorts and calendar-time changes.

The Egyptian family-lifetime model shows why research often needs a censoring-aware estimation procedure. A naive sample composed only of observed pairs of deaths can distort the relationship between lifetimes, especially when beneficiaries remain alive beyond the observation window. Comparing estimation techniques, copula families and observed categories can reveal model instability, but the analyst must distinguish lack of power to detect an effect from evidence of independence.

Dependence sensitivity should include a model consistent with observed historical association, a plausible weaker or independent benchmark and scenarios with stronger common shocks or changes in bereavement duration. A positive joint-life association at one horizon can coexist with different relationships at other horizons. Each scenario must preserve the marginal and joint probability constraints. If an exposure source only provides independent individual deaths, a joint survival assumption is an unverified model component and should be identified as such.

Interest rates, expenses, inflation and asset-liability effects

The continuation fraction affects benefit duration and thus the timing of insurer or pension outflows. Discount rates determine the present value of each projected payment; a lower discount curve generally raises the value of a positive fixed nominal benefit stream. The timing distribution can change differently across first-death, last-survivor and directional benefits. Consequently, sensitivity should report discount-rate effects separately from mortality and dependence changes and should use consistent valuation curves for comparisons.

A deterministic discounting sensitivity can be calculated by replacing v(t) or v_k while keeping the contract and mortality state probabilities unchanged. Interest-rate exposure is then the discounted duration or derivative of the expected payment schedule. For investment-linked or indexed benefits, benefit escalation can be stochastic and correlated with discount rates. A real-income survivor guarantee should therefore be modeled with its index definition, lag and cap, and not merely with a fixed nominal ratio.

Administrative costs include death verification, processing of beneficiary claims, payments over potentially long survivor periods, corrections and litigation. Expense inflation can be particularly relevant in late-duration benefits. Continuation provisions may also impose reinsurance and counterparty exposures depending on the insurer's risk-transfer arrangements. The net present value of gross benefit cash flows is one input to pricing and reserving, not a complete insurance liability or solvency measure.

IFRS 17 and IAS 19 classification

Under IFRS 17 insurance-contract accounting, obligations within the standard's scope are measured through a framework of fulfillment cash flows, risk adjustment for nonfinancial risk and contractual service margin, subject to the applicable recognition and measurement requirements. The expected survivor payments supplied by the four-state model form part of the cash-flow basis when the contract qualifies. They do not by themselves establish accounting profitability, the risk adjustment or the margin.

An employer's defined-benefit pension undertaking may instead be reported under IAS 19 employee-benefits accounting. The actuarial valuation of an employer's promise and the insurance measurement of an annuity issued by an insurer have distinct accounting responsibilities even when their underlying projected retiree payments look similar. A pension buy-in or buyout transaction may further change the legal entity ultimately responsible, requiring the arrangement and contracts to be examined.

The prudent analytical sequence is to identify the party legally liable, identify the relevant accounting and supervisory regime, generate the correct contingent cash flows, and then add regime-specific measurement components. Assigning IFRS 17 or IAS 19 solely from the existence of a survivor ratio is an unsupported classification.

Solvency II and capital distinction

For European Union insurers, Solvency II Article 77 provides a framework for technical provisions based on a probability-weighted present-value best estimate and a risk margin. The bivariate benefit projection is relevant to the best-estimate input. It does not produce a complete technical provision until expenses, discounting, other contingencies and risk margin are addressed.

Solvency II Article 105 identifies mortality and longevity risk categories, among other underwriting risks. Article 101 on capital establishes the solvency capital calibration framework, including a one-year 99.5-percent confidence measure. A change in a joint-life dependence parameter can affect the expected value of benefits in one direction while changing risk dispersion or tail behavior in another. Regulatory capital cannot be inferred simply from the sign of the first-moment sensitivity.

The presence of survivor payments can increase the duration of liabilities, sensitivity to mortality improvement and exposure to common longevity trends. A stochastic capital stress must examine the portfolio's loss distribution and the applicable supervisory rules. Historical sample dependence is one input among several, rather than an all-purpose capital coefficient.

Reproducible scientific and editorial controls

For documentary review, every source-linked claim should resolve to the specific official rule, published study or standard actually cited. Descriptive legal claims should state jurisdiction and scope; empirical numerical claims should state population and sampling period; methodological claims should identify limitations of abstract-only evidence. The source link should occupy a short, semantically meaningful part of the sentence and the complete bibliographic details should appear in References. External research URLs must have corresponding entries in the preserved evidence ledger.

For mathematical review, the manuscript's formulas should be reproduced independently from the state table and numeric inputs, with clear notation and explicit assumptions on payment timing and interest. The illustration should be tested as a calculation rather than treated as an observed statistic. For technical release review, the manuscript fingerprint, evidence fingerprint and Git revision should be recorded without circular self-reference; the required local repository source-link QA should be run on the exact archive before an approval is granted.

The International Actuarial Association standards register supplies a relevant reference for documented practice, including communication, data quality, assumptions and appropriate use of models within the standards applicable to the engagement. Full reproducibility of a policy calculation also depends on access to original contract terms and the source data used to identify covered lives.

Integrated Research Findings and Open Scientific Questions

First result: the benefit schedule precedes the percentage

The investigation establishes that a survivor continuation percentage is an incomplete description of the product unless accompanied by the full survival-state schedule. An ordinary participant-centered pension may pay the member 100 percent as long as the member survives and pay the survivor only when the member dies first. A symmetric last-survivor annuity may reduce the installment after the death of either life. Both can describe their survivor protection as 50 percent; only the state schedule reveals the relevant valuation and beneficiary consequences. Because a percentage can also be measured relative to different initial or unreduced pension amounts, a reproducible comparison must state its denominator.

This interpretation is consistent with the IRS qualified pension structure, the Canadian joint-survivor default and the Australian reversionary annuity example. Their legal frameworks and specific percentages differ. The comparative conclusion rests on the documented scope of those sources, rather than an assumption that their contracts are identical.

Second result: joint mortality dependence has a contract-specific first-moment effect

The derived identity

E[C(t)]B=rXSX(t)+rYSY(t)+(1−rX−rY)J(t)\frac{E[C(t)]}{B}=r_XS_X(t)+r_YS_Y(t)+(1-r_X-r_Y)J(t)

directly expresses the role of the two marginal survivals and their bivariate survivor probability. At fixed marginal probabilities and fixed B, the joint-survival sensitivity is 1-r_X-r_Y. It is positive for first-death cessation, negative for full last-survivor payments and zero for symmetric half continuation. For a participant-centered pension the state fractions are (1,r), yielding sensitivity -r rather than zero for a positive survivor percentage.

The importance of this result lies in its capacity to test contract modeling. A numerical system that assigns the same dependence sensitivity to first-death and last-survivor benefits has lost the distinction between their state payoffs. A system that asserts dependence invariance for an ordinary directional 50-percent survivor pension has mistaken the relevant cash-flow structure. These are consequences of standard probability and actuarial definitions, not an independently introduced theory.

The empirical basis for modeling dependence is supported by the 1996 annuity study and the 2023 Egyptian family study. Both use specific populations and methodologies, so their numerical findings cannot be combined into a universal correlation or premium adjustment. A transferable scientific claim must identify which features of the contracts, mortality marginals, selection and dependence models are common across studies.

Third result: an expected-value cancellation does not eliminate financial risk

At symmetric r_X=r_Y=0.5, the expected installment equals 0.5B(S_X+S_Y), independent of J under the stated deterministic benefit and fixed marginal assumptions. The second moment, however, depends on J through the coefficient 0.5B squared in the relevant expansion. A portfolio's multi-period cash-flow variance also depends on survival covariance over time and on systematic mortality shocks shared across covered lives. Stochastic discount factors can introduce further dependence even if the standalone benefit expectation cancels its bivariate term.

The distinction matters to both insurers and pension sponsors. An expected-benefit value can be sufficient for a narrow valuation identity while reserve margins, capital requirements, hedging and liquidity analyses need uncertainty information. The IFRS 17 framework and EU solvency technical-provisions rules illustrate why expected benefit payments are important but incomplete liability inputs. Their precise application remains regime and entity specific.

Fourth result: individual valuation depends on selection and population-specific longevity

The source record supports distinguishing purchased annuitants, occupational pension participants and public pension beneficiaries. The SOA annuity mortality materials provide a marginal reserving reference, whereas the RPEC 2026 mortality update addresses improvement assumptions for retirement populations. Neither alone supplies the joint survival of two selected annuitants.

A purchaser choosing continuation protection can differ from one purchasing single-life income. The Finkelstein and Poterba analysis provides historical evidence that annuity contract characteristics can be associated with mortality selection. A modern study of continuation-ratio choice should measure selection into annuity purchase, choice among benefit forms and observed paired-life outcomes in a shared data framework. Drawing a causal inference about the effect of an election on survival from a cross-sectional association would exceed the evidence.

Fifth result: institutional comparisons require separate entitlement and valuation layers

The same economic objective of survivor income is implemented through multiple institutions. An insured annuity transfers a specified benefit liability to a carrier; a pension employer may retain a defined benefit promise; a public survivor pension derives from statute and contribution conditions. The CPP survivor-pension rules give a case where public-benefit calculations include eligibility and combined-pension rules rather than a simple commercial-annuity transition from B to rB.

It follows that international research cannot compile bare percentages and present their average as a scientific measure of protection. A credible comparative table must report the plan or policy type, the survivor identity, the statutory or contractual entitlement, the denominator of the percentage, the date or version of law, the full benefit-state schedule, the payment horizon and the funding entity. The OECD Pensions Outlook 2026 offers relevant international context, while the OECD pension-design recommendation underscores retirement-income adequacy as a broader policy objective. The appropriate continuation ratio remains a problem to be assessed under local law and household circumstances.

Original analytical synthesis and its limits

The original contribution of this monograph is an integrated comparison of contractual survivor rights, state-contingent value and dependent-mortality sensitivity, together with a reproducible diagnostic distinction between symmetric and directional continuation contracts. The analytical findings are derived from established probability and classical actuarial cash-flow valuation, contextualized by source-specific international legal and empirical literature. No proprietary statistical model, legal doctrine, actuarial theory or newly estimated mortality effect is claimed.

The synthesis identifies four useful analytical levels that should remain visible in every study: the legal obligation defining payment states; the statistical probabilities of occupying those states; the financial valuation of the resulting cash flows; and the distributional consequences for the surviving household. Each layer has its own evidence base and validation standard. Combining them improves the precision of research, but a strong result in one layer cannot stand in for missing evidence in another.

Open questions suitable for empirical investigation

A large linked insurer-policyholder data set could establish how the mortality association of purchased-annuity couples differs from that of other married or partnered households after adjusting for selection and measured health. Such a study should distinguish households choosing 50-, 60-, 75- and 100-percent benefits while recording age differences and socioeconomic variables. It would test whether election groups have different paired-survival experiences rather than assuming a common copula.

A second program would compare static copulas with models allowing a short-term change in mortality after bereavement. The model should predict both the timing of first deaths and the subsequent survival duration, with validation under right censoring. Testing different copula families for the tail of survivor payments would clarify how dependence uncertainty affects exceptionally long benefit durations.

A third program would link actuarial projections to claims administration and benefit documentation. It could compare expected and actual payment commencement after a death, the effect of delayed notification on overpayments, changes in beneficiary eligibility, and differences between promised and operationally realized payments. The appropriate performance measure would be error in lawful payment entitlement and discounted outflow, rather than merely error in a mortality estimate.

A fourth program would study the interaction of individual annuity income, public pension survivor rights, indexation and survivor spending needs across jurisdictions. Public and employer-sponsored income can change following death by formulas that are not proportional to an original private annuity. Household-level benefit sufficiency therefore requires joint measurement of several sources and the timing of outlays.

These are scientific research questions that lie beyond the current documentary and mathematical record. Their resolution requires access to original life-pair data or contract texts and appropriately documented estimation, and their existence does not weaken the reproducible state-based results established here.

Scope limitations and reproducibility statement

The empirical claims in this research are attributed to their named studies and jurisdictions. There is no original household survey, linked mortality panel, fitted new copula, confidential contract review, insurer quote, stochastic economic scenario generator or calculated regulatory capital requirement. Numerical examples use explicitly hypothetical annual survival inputs and a 4-percent discount factor. A second analyst can reproduce every value using the displayed state probabilities, benefit fractions and formulas.

The cited international standards establish the character of relevant accounting or prudential frameworks, rather than a universal legal conclusion about any unexamined policy. The legal examples are source- and jurisdiction-specific; subsequent statutory amendments or contractual variations require a new effective-date check before using a conclusion for a real payment claim. Publisher abstracts for restricted studies are used for limited conclusions and are not represented as reproduced microdata analysis.

These boundaries preserve a distinction between completed scholarly analysis, source verification to the extent available and empirical generalizations still requiring original data. The resulting monograph is a standalone theoretical, documentary and comparative research artifact with explicit pathways for later independent replication.

Conclusion

Joint-life annuity survivor continuation ratios are legally defined transformations of contingent retirement-income payments. Their scientific evaluation begins with the actual state-dependent promise: the initial amount, the two covered persons, the payment after each possible order of death, the point when a changed benefit becomes effective, and any separate payment guarantee. A ratio without those conditions is insufficient to determine either beneficiary entitlement or actuarial value.

The four-state formulation developed in this monograph distinguishes first-death cessation, full last-survivor payment, symmetric partial continuation and participant-centered pension reversion within a common mathematical language. For a level benefit B and fixed fractions r_X and r_Y paid in the two single-survivor states, expected cash flow equals B[r_X S_X+r_Y S_Y+(1-r_X-r_Y)J]. This direct probability identity clarifies how joint-survival dependence interacts with contract design. A symmetric 50-percent continuation provision has an expected installment determined by the two marginal survival curves under the stated fixed-benefit assumptions, whereas a participant-centered 50-percent survivor pension retains sensitivity to the joint survival probability. The benefit amounts and legal payment states, rather than the label assigned to the percentage, govern the conclusion.

A reproducible one-period case verifies the derivation and shows that a two-direction 60/50-percent schedule and a participant-centered 60-percent survivor schedule can have different discounted values under the same survival probabilities and interest rate. The ability to calculate and explain that difference is an important validation control for pricing and reserving systems. The general formula also shows why dependence-invariance of a first moment cannot be treated as independence of volatility, systematic longevity risk or liability capital requirements.

The international documentary evidence locates continuation ratios in distinct legal institutions. The U.S. qualified joint-and-survivor framework, federally regulated Canadian pension survivor protection, Australian reversionary annuity illustrations and UK annuity and beneficiary classifications all reflect different rules, beneficiaries and funding arrangements. Public pension survivor benefits establish a related but institutionally separate category. Comparative science requires explicit jurisdiction, benefit denominator, effective document version, plan coverage and legal entitlement before interpreting the percentage.

Dependence between lives is a legitimate empirical research problem with documented actuarial applications. Published copula studies show that its strength and pricing consequences vary across samples, family types and institutional conditions. The evidence supports careful paired-life estimation, censoring-aware model choice and population-specific validation, rather than the transfer of a single numerical adjustment to every joint annuity. Mortality improvement, contract selection, discounting, household spending needs and administration each remain separately identifiable components of the assessment.

The central practical conclusion is that survivor continuation must be measured as a complete two-life contingent cash-flow schedule and evaluated through appropriately matched actuarial, legal and financial assumptions. The resulting framework makes comparisons transparent, identifies model misclassifications and provides a sound foundation for later empirical studies using original contract and household-lifetime data. Its analytical statements are applications of established science, with no new authorial theory or unsupported novelty claim.

References