Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Annuity Mathematics and Valuation

Abstract. Annuity mathematics values income streams whose timing, continuation, and amount depend on contractual rules, financial discount factors, and the survival of one or more individuals. This monograph develops the valuation of life-contingent income from a state-contingent cash-flow representation, derives discrete and continuous annuity functions, examines deferred and temporary benefits, and constructs joint-life, last-survivor, contingent-survivor, and guaranteed-payment valuations. It distinguishes financial interest sensitivity from mortality and longevity sensitivity, explains the use of period and cohort mortality information, and analyzes the consequences of dependent lives, stochastic discount rates, indexation, heterogeneous annuitant populations, and model uncertainty. The discussion connects mathematical benefit values to individual annuity pricing, insurer cash-flow projection, IFRS 17 accounting measurement, and Solvency II technical provisions without identifying these different quantities as interchangeable. Fully reproducible five-year examples and closed-form continuous results test the stated equations and comparative statics. Primary actuarial, demographic, economic, academic, and supervisory materials support an original integrated interpretation of how a benefit promise becomes a measurable contingent liability. The central finding is that an annuity's value is a property of a precisely specified payment event, mortality basis, discounting convention, and valuation purpose. The original contribution is analytical synthesis and verification of established mathematics; the work creates no new proprietary model or doctrine.

Keywords: Life annuity; actuarial present value; survival probability; mortality intensity; interest discounting; annuity-due; annuity-immediate; deferred annuity; temporary annuity; joint-life annuity; last-survivor annuity; contingent reversionary annuity; period-certain guarantee; refund benefit; mortality improvement; cohort life table; longevity risk; duration; convexity; stochastic discounting; annuity pricing; IFRS 17; Solvency II; actuarial model governance.

Introduction

An annuity is a contractual arrangement for a sequence of payments. In a life annuity, at least part of that sequence is conditional on an individual's survival. Its mathematical valuation therefore requires a joint description of future payment events and the time value of money. The payment of one monetary unit one year from now can have a lower present value than one unit paid now because of interest discounting; if it is payable only while a person survives, the probability of the benefit event also enters. For a benefit paid to a survivor of two lives, the precise relationship among their lifetimes becomes material. Every calculation ultimately answers a question about an identified stream of state-dependent obligations.

The historical association between life tables and annuities is longstanding. Edmond Halley's 1693 mortality study is cataloged by the Royal Society in its Breslau mortality paper record. In contemporary mathematical actuarial literature, Dickson, Hardy, and Waters (2012) present a systematic analysis of life-contingent annuities, including payment frequency and approximations when only integer-age mortality information is available. Their framework provides a foundation for the core identities derived in this monograph. The present study formulates those identities with explicit benefit triggers and valuation conventions and follows their consequences through financial reporting and risk analysis.

A life annuity supports long-lived beneficiaries through a transfer of longevity exposure: the beneficiary receives payments in the survival states defined by the contract, while an insurer or financing institution carries the resulting liability. Yaari's (1965) uncertain-lifetime consumer analysis explains the economic significance of lifetime income within an idealized intertemporal choice problem. Practical valuation adds operational payment dates, survivor definitions, market yields, expenses, selection in the insured population, guarantees, and the legal identity of the payment obligor. Finkelstein and Poterba (2004) documented selection across annuity contract features, reinforcing the necessity of examining the actual promise and the insured population together.

The subject of this monograph is actuarial annuity valuation. Pension funding, retirement income advice, and the legal interpretation of life insurance contracts are relevant interfaces, but their complete treatment belongs to related research. The principal published Science is Actuarial Science. A single Research Dossier identity also serves Annuities and Retirement Insurance, Life Insurance, Insurance Finance and Asset-Liability Management, Solvency and Capital, and Insurance Pricing and Reserving. This cross-disciplinary organization preserves one canonical research object.

Scientific questions and method

The research asks how survival and discounting jointly determine annuity value; what timing, mortality, dependence, and financial assumptions establish that result; how contractual guarantees transform the benefit event; and how a mathematical valuation enters distinct regulatory and accounting objectives. The method combines actuarial derivation, source-critical interpretation, independent numerical reconstruction, and a limited international comparison of supervisory measurement principles. All numerical quantities are constructed illustrations, with no claim that their mortality rates or yields describe a particular population, insurer, or current market. Primary institutional references are used for their stated scope, while papers available only through publisher abstracts support only the claims that can be verified there.

Throughout, expectation and valuation purpose must be specified. An expectation under a real-world mortality model applied to deterministic discount factors gives an actuarial expected present value. A price for bearing unhedgeable longevity risk, a contractual premium, an accounting insurance-contract liability, and a prudential technical provision can use related projected cash flows while adding or changing assumptions, risk adjustments, expenses, and recognition boundaries. Their values are accordingly distinguished throughout the text.

Life Annuity Types

Definition through payment events

Let time zero be the valuation date, let x denote a life aged x at that date, and let T_x be that person's future lifetime measured in years. The future lifetime is nonnegative; its survival function is S_x(t) = Pr(T_x > t) for t >= 0 under the model selected for this annuitant or insured population. For a payment scheduled at t that is made if the life is alive at t, the survival factor is S_x(t). Strict and non-strict inequalities are immaterial when the lifetime distribution is continuous, but a discrete mortality model requires the contract's precise treatment of deaths occurring on a payment date.

Let c(t) represent the contractual payment rate, D(0,t) the time-zero discount factor to t, and I(t) the indicator that the benefit is due. A general present-value random variable can be written

[ Z=\sum_{j=1}^{m} c_j D(0,t_j)I(t_j) ]

for scheduled discrete payments. In a purely deterministic interest model its expected present value is

[ V=\mathbb{E}[Z]=\sum_{j=1}^{m}c_jD(0,t_j)\Pr(I(t_j)=1). ]

This formula is the common foundation for single-life, joint-life, guarantee, and contingent annuities: contractual interpretation determines the indicator; demographic assumptions determine its probability; the valuation basis determines discounting. It holds without independence among benefit indicators, because the expectation of a sum is the sum of expectations. Dependence becomes important when calculating the probability of a particular joint-life event, or the distribution and risk of total value.

A level, annual annuity-due on x makes its first unit payment at time 0 and further payments at integer times while the life is alive. An annuity-immediate, in the traditional actuarial terminology, makes payments at the ends of periods, ordinarily at times 1, 2, and so forth. The words 'immediate' and 'deferred' in a product description may describe the date that an income contract starts, whereas 'annuity-immediate' specifies the end-of-period convention in a mathematical formula. A complete contract description therefore states dates explicitly.

The main forms are whole-life income, temporary income payable for at most n years, deferred income beginning after an interval u, annuity-certain payments independent of survival, joint-life benefits payable while two lives are alive, last-survivor benefits payable while at least one survives, and reversionary benefits triggered by the order of deaths. Escalating and inflation-linked payments alter c_j. Some variable annuity benefits also depend on investment accounts and financial options; their valuation requires a richer state-space than mortality alone.

Life tables and conditional probabilities

Write {}_tp_x = S_x(t). For integer ages, one-year survival is p_x = 1-q_x, where q_x is the probability that a life aged x dies before reaching age x+1. Conditional survival over k years is

[ {}kp_x=\prod{j=0}^{k-1}(1-q_{x+j}),\qquad {}_0p_x=1. ]

The relation is a chain of conditional probabilities and does not assert that annual death events are independent. It assumes that q_{x+j} is defined for the appropriate cohort, attained age, calendar year, and population basis. A period table captures mortality conditions of particular calendar years at different ages, whereas a cohort projection attempts to follow a birth cohort across future calendar years. The Human Mortality Database's methods and population overview documents the reconstruction of death rates and life tables from births, recorded deaths, population counts, and exposure-to-risk. Its documented coverage is selective; population data are not automatically an appropriate insured-annuitant table.

The hazard or force of mortality, when it exists, is the instantaneous conditional death intensity (\mu_{x+t}). Survival then obeys

[ S_x(t)=\exp\left(-\int_0^t\mu_{x+s},ds\right). ]

This representation clarifies that a mortality 'rate,' a one-year probability of death, and a force of mortality have different mathematical meanings. Under constant intensity (\mu) within one year, (q=1-e^{-\mu}); substituting q=\mu without explaining the approximation introduces error. A graduated mortality curve often smooths noisy observations, but graduation requires validation against relevant exposure and the objectives of valuation.

Selection and population identity

The appropriate mortality basis depends on whose lifetime triggers the payments. Annuitants may differ from the general population because individuals selecting or purchasing lifetime income have distinct characteristics. The UK annuity analysis by Finkelstein and Poterba (2002) identifies mortality differences and selection effects across voluntary and compulsory annuity markets. The Society of Actuaries' individual annuity reserving table materials document the design of the 2012 U.S. annuity reserving table, including mortality projection and margins. These instruments have different purposes: empirical annuitant experience, valuation assumptions, regulatory reserves, and population-demographic description must be labeled separately.

The Society of Actuaries' individual annuity experience catalog includes an updated study of 2020–2024 individual payout annuity mortality released in 2026. Data from industry studies can improve an insurer's assumptions when contract scope, geography, selection, exposure definition, benefit amounts, and credibility are comparable. A table adopted in a national reserving context is neither a universal international mortality forecast nor automatically the economically appropriate basis for all private valuations.

Immediate and Deferred Annuities

Discrete annuity-due and annuity-immediate

Assume a flat effective annual interest rate i > -1, v=(1+i)^{-1}, and a survival model for a life aged x. For a unit annual whole-life annuity-due with payments at k=0,1,..., its actuarial present value is

[ \ddot a_x=\sum_{k=0}^{\infty}v^k,{}_kp_x. ]

For a unit annual whole-life annuity-immediate with payments at k=1,2,...,

[ a_x=\sum_{k=1}^{\infty}v^k,{}_kp_x. ]

Whenever the sums converge, (\ddot a_x=1+a_x). This identity concerns the same life and all future integer payment times, with the annuity-due adding an unconditional initial payment. A finite sequence of n annual payments must be defined separately:

[ \ddot a_{x:\overline{n}|}=\sum_{k=0}^{n-1}v^k,{}kp_x, \qquad a{x:\overline{n}|}=\sum_{k=1}^{n}v^k,{}_kp_x. ]

For finite temporary contracts, their difference is (1-v^n{}_np_x), not necessarily 1. This distinction prevents an otherwise plausible but incorrect application of the whole-life identity to equally sized temporary schedules. The foundational Cambridge annuity chapter discusses the relationships among discrete life-contingent payment conventions.

For an annual amount B instead of one unit, multiply each unit-value function by B. If the amount differs by year, replace the unit by the scheduled benefit (B_k). The resulting value is linear in promised benefit amounts under fixed probabilities and discount factors. This linearity remains useful in decomposing complex contractual income schedules into simpler benefits; nonlinear features, such as account-dependent guarantees or behavioral exercise, may also change the event probabilities and underlying projections.

Deferred commencement

For an annuity-due deferred u integer years, with payment at u if the person survives to that date and later annual payments while alive,

[ {}{u|}\ddot a_x=\sum{k=u}^{\infty}v^k,{}_kp_x =v^u{}up_x\ddot a{x+u}, ]

where the final equality uses a coherent mortality model and (\ddot a_{x+u}) conditional on survival to u. The factor (v^u{}_up_x) combines discounting and survival until commencement. It provides the bridge between the purchased deferred promise and its value at the commencement date. For n payments at dates u through u+n-1,

[ {}{u|}\ddot a{x:\overline{n}|}=\sum_{k=u}^{u+n-1}v^k{}_kp_x =v^u{}up_x\ddot a{x+u:\overline{n}|}. ]

A deferred income contract may require premiums before commencement, pay a death benefit during deferral, carry an account or surrender value, and impose special vesting conditions. The preceding formula prices only its identified survival-contingent payout component. Refund and accumulation provisions enter as additional cash flows under their own triggers.

Decomposition and payment timing

Discounting a benefit from its date of actual payment is essential. A monthly payment of annualized amount B paid at the end of each month is B/12 at times j/12, not B every month and not a single B payment at year-end. Its temporary value over n years is

[ V_{m}=\frac{B}{m}\sum_{j=1}^{mn}D(0,j/m),{}_{{j}/{m}}p_x. ]

For an m-times-per-year annuity-due, payment times are j/m for j=0,...,mn-1. Fractional-year survival requires an interpolation or direct mortality intensity estimate. The choice among a uniform distribution of deaths within a year, a constant force, or a fitted age/calendar-time hazard can matter when interest rates, mortality, and payment frequency are high. The established actuarial text by Dickson, Hardy, and Waters (2012) treats UDD and Woolhouse approximations; an approximation should be labeled and checked against its input resolution.

For a nominal annual rate convertible m times, the period discount factor uses the rate i^(m)/m; for an effective annual rate i, the fractional-year discount factor under constant force of interest is (1+i)^(-t). Nominal and effective quotation conventions should never be combined without conversion. Contractual day-count, anniversary rules, business-day adjustments, and end-of-month treatment can introduce smaller timing differences that remain consequential for large portfolios.

Net single premiums and product prices

The actuarial net single premium of a standalone benefit under the selected basis is its expected present value. The commercial consideration charged by an insurer must address a fuller set of promises and obligations, including administration, distribution, adverse deviations, financing costs, investment returns, and capital. When the benefit is purchased through periodic premiums, each premium's timing and any persistency contingency must be incorporated. The equivalence of present values can yield an illustrative break-even premium but does not establish a unique market price because investment strategy, contract options, profit targets, regulation, and demand affect the transaction.

In an idealized benefit-only illustration, a single premium P supporting a certain amount B per year satisfies (P=B\ddot a_x). Thus (B=P/\ddot a_x) when there are no other benefits, expenses, loads, or margins. Increasing assumed future survival typically increases the annuity factor and reduces B for a fixed P under otherwise equal conditions. A quoted payout rate should therefore be interpreted relative to the mortality basis, the guarantee structure, any inflation increase, and the insurer's financial obligations.

Survivor and Joint-Life Benefits

Both-lives and at-least-one-alive events

Let x and y be two insured lives with future lifetimes T_x and T_y. A joint-life annuity pays only while both are alive, with event ({T_x>t,T_y>t}). Its due value is

[ \ddot a_{xy}=\sum_{k\ge0}D(0,k)\Pr(T_x>k,T_y>k). ]

If the future lifetimes are independent under the selected mortality model, their joint survival probability is ({}_kp_x{}_kp_y). The independence assumption is substantive and may be inappropriate when household members share environmental, socioeconomic, infectious-disease, or catastrophic exposures. Under dependence, multiplication of marginal survival factors is unjustified.

A last-survivor annuity pays while at least one member is alive, with event ({T_x>t}\cup{T_y>t}). Inclusion–exclusion gives

[ \Pr(T_x>t\text{ or }T_y>t) ={}_tp_x+{}_tp_y-\Pr(T_x>t,T_y>t). ]

Consequently, for otherwise identical unit payment times and discount factors,

[ \ddot a_{\overline{xy}}=\ddot a_x+\ddot a_y-\ddot a_{xy}. ]

This identity holds with dependent lifetimes; its joint-life component must, however, reflect the actual dependence. At every time point the joint-life event is a subset of each single-life survival event, and the last-survivor event is their union. Therefore the joint-life factor cannot exceed either single-life factor, and the last-survivor factor cannot be smaller than either. Such ordering provides a strong implementation check.

Reversionary and contingent-survivor benefits

Consider a reversionary payment to y that begins only after x has died, provided y is then alive. At a scheduled date t the relevant event is ({T_x\le t,T_y>t}), assuming the contract treats death and benefit commencement at that date consistently. Its probability is

[ \Pr(T_y>t)-\Pr(T_x>t,T_y>t). ]

The corresponding due-value function is (\ddot a_y-\ddot a_{xy}) with aligned dates, starting times, and benefit amounts. If instead a benefit is payable after the first death to whichever life survives, the event and resulting payment schedule differ. A survivorship option may pay 100%, 75%, or 50% of the original pension after the first death. Its value can often be decomposed into a component while both survive plus a reduced component while either survivor remains, but the actual contract must determine both the base amount and the date of reduction.

For a primary life x receiving B during joint survival and a survivor y receiving rB after x dies, a unit-date representation is

[ V=\sum_k D(0,k)B{\Pr(T_x>k,T_y>k) +r\Pr(T_x\le k,T_y>k)}. ]

The contract may alternatively pay x for life whether y survives or not and then rB to y; that is a different event representation. These examples show why 'joint and survivor' is an insufficient mathematical definition without specifying original-payment continuation, survivor percentage, priority, and first-payment timing.

Dependence sensitivity without a fitted joint model

A useful model-agnostic bound follows from the Fréchet inequalities. For marginal survival probabilities a and b at one payment date, the joint probability j satisfies

[ \max(0,a+b-1)\le j\le\min(a,b). ]

These inequalities bound joint-life and last-survivor values when marginal survival is specified but their dependence is uncertain. Using the bounds separately at every date produces conservative termwise envelopes; the sequence of extreme joint probabilities across dates need not correspond to a single globally feasible lifetime dependence model. Such bounds are diagnostic rather than a replacement for a coherent joint distribution. Mortality correlation may be especially material for small portfolios or household benefits, while population-level systematic longevity effects require a separate model.

Continuous and Discrete Valuation

Continuous payment rate and lifetime representation

Suppose a life annuity pays at a continuous rate b(t) per year while the life is alive. Its discounted present-value random variable is

[ Z=\int_0^{T_x}b(t)D(0,t),dt. ]

For nonnegative benefits, Tonelli's theorem permits exchange of expectation and integration. Under deterministic discounting,

[ \bar a_x[b]=\mathbb{E}[Z] =\int_0^\infty b(t)D(0,t)S_x(t),dt. ]

For b(t)=1, constant mortality force (\mu>0), and constant force of interest (\delta) satisfying (\delta+\mu>0), the value becomes

[ \bar a_x=\int_0^\infty e^{-(\delta+\mu)t},dt =\frac{1}{\delta+\mu}. ]

This result is an exact closed-form value within the constant-hazard and constant-interest illustration. With rate (\delta=\ln(1.03)) and hazard (\mu=0.02), it equals about 20.178050 years of unit annualized payments. It does not describe the expected number of payments of a finite, annual annuity or an empirical annuity factor at any particular age. Replacing the constant hazard by a realistic attained-age hazard generally requires numerical integration.

For a continuously paid annuity deferred u years, the value is (\int_u^\infty b(t)D(0,t)S_x(t)dt). A term-certain n-year continuous payment guarantee instead separates the integral over the guaranteed interval from later survival-dependent payments. The structural operation is the same as for the annual payment case: the contract event is specified before any integration occurs.

Integral and summation are different cash-flow instruments

A continuous annualized income rate and annual payments of the same nominal annual amount do not make identical promises. In the five-year example below, an annual annuity-due pays immediately and at successive annual anniversaries, whereas a continuous annuity spreads payments over time. Comparing their present values requires harmonized cash-flow units. Under smooth survival and discount functions, the m-thly annuity with total annual amount B approaches the continuous-rate value as m grows; the limiting result is a Riemann-sum convergence statement under the assumed regularity. A numerical interpolation scheme should be selected for accuracy and convergence, not because a named approximation is conventional.

For a continuously paid life benefit, the lifetime random variable Z may carry substantial variance even when its expected value is computed precisely. In the constant positive discount-force case with unit payment and lifetime T, (Z=(1-e^{-\delta T})/\delta). Thus calculation of the mean alone does not measure uncertainty. A second moment can be obtained by integrating the distribution or directly through simulation. Portfolio aggregation reduces diversifiable individual mortality fluctuation under suitable independence and moment conditions, but common mortality-improvement shocks, interest risk, and model error remain.

Discrete numerical reconstruction: five annual payments

To make all values reproducible without a proprietary mortality table, assume a life with annual conditional death probabilities q at successive attained ages of 0.010, 0.012, 0.014, 0.016, and 0.018. Take a flat effective annual interest rate of 3%, unit nominal benefits, and at most five payments. The five annual survival probabilities are computed by multiplication of successive one-year conditional survival probabilities. These rates are constructed purely to test the formula.

Time k (years) Conditional q during year k to k+1 Survival {}_kp_x Discount v^k
0 0.010 1.000000000 1.000000000
1 0.012 0.990000000 0.970873786
2 0.014 0.978120000 0.942595909
3 0.016 0.964426320 0.915141659
4 0.018 0.948995499 0.888487048
5 — 0.931913580 0.862608784

The survival row at k=5 is included to evaluate an end-of-year temporary annuity; there is no sixth due payment in the five-payment contract. Direct evaluation yields:

Defined benefit Formula and unit payment dates Actuarial present value
Five-payment temporary annuity-due Sum k=0,...,4 of v^k S(k) 4.608894
Five-payment temporary annuity-immediate Sum k=1,...,5 of v^k S(k) 4.412771
Three payments deferred two years Sum k=2,3,4 of v^k S(k) 2.647729
Three-year guarantee, five total due payment dates Sum k=0,1,2 of v^k plus k=3,4 of v^k S(k) 4.639227
Five-payment joint-life due (two independent identical marginal lives) Sum k=0,...,4 of v^k S(k)^2 4.504707
Five-payment last-survivor due (same independent lives) Sum k=0,...,4 of v^k [2S(k)-S(k)^2] 4.713081

For the first two rows, the exact difference is (1-v^5S(5)), about 0.196123. The three-year guaranteed row exceeds the ordinary temporary annuity-due because payments at k=1 and k=2 are made even after the annuitant's earlier death; the k=0 payment is already certain. The last-survivor annuity exceeds the single-life factor, and the joint-life factor is smaller. These inequalities follow from the event definitions and are simultaneously numerical validation checks. All six outputs were independently recomputed using direct sums rather than inferred from a published price table.

Continuous–discrete diagnostic identity

There is a complementary identity between certain annual life-annuity and life-insurance values. Define a unit whole-life insurance benefit payable at the end of the policy year of death, with (A_x=\sum_{k=0}^\infty v^{k+1}{}kp_xq{x+k}), and let the annual discount rate be i with (d=i/(1+i)=1-v). Provided death eventually occurs under the model and terminal terms vanish, summation by parts yields

[ A_x+d\ddot a_x=1. ]

For a temporary n-year annuity-due the corresponding identity is

[ A_{x:\overline{n}|}^{\mathrm{term}}+v^n{}np_x +d\ddot a{x:\overline{n}|}=1, ]

where the first term is death benefit payable at the end of the year of death if death occurs within n years. The second term is the pure-endowment payment of one at n if alive. This equation also follows by partitioning all trajectories between death before n and survival to n and accounting for the annual financing difference d. It is useful for reconciling independently developed insurance and annuity valuation routines, provided the death-benefit timing is identical. Changing any timing convention changes the identity.

Income Guarantees

Period-certain payments

A life annuity with a guarantee of g payments makes those payments even if the covered life dies before the gth scheduled date. After the guaranteed period, payments are survival-contingent. For an annuity-due with at most n unit annual payments and 0 <= g <= n,

[ V_{g,n}=\sum_{k=0}^{g-1}v^k +\sum_{k=g}^{n-1}v^k{}_kp_x. ]

The increment relative to the corresponding life-only annuity is

[ V_{g,n}-\ddot a_{x:\overline{n}|} =\sum_{k=0}^{g-1}v^k(1-{}_kp_x)\ge0. ]

This algebra identifies precisely which states produce extra value: the annuitant dies before an otherwise scheduled payment inside the guarantee. The result depends on the same discounting and the same ordinary survival basis as the unguaranteed comparison. If payments following death transfer to an estate or beneficiary, the contract must establish payment ownership and continuation rights, even though the arithmetic guarantee component is independent of later survival.

A period-certain term can coexist with a whole-life benefit, in which case the second sum extends to the final lifetime horizon. A guarantee lasting through a fixed term n turns the whole n-payment schedule into an annuity-certain. Such a contract ceases to depend on the original annuitant's mortality for those specified payments, even if delivery is operationally administered as an annuity.

Refunds, death benefits, and deferred guarantees

A cash-refund annuity can promise an estate payment equal to part of the premium not recovered through prior annuity payments. The benefit is dependent on the cumulative paid amount and on the time of death. A general benefit component is (R(T_x,\text{history})D(0,T_x)), whose expectation must be computed with the correct death-time distribution and contractual restrictions. It is not, in general, equal to a period-certain guarantee of similar headline length. An installment-refund annuity continues payments until a specified amount has been recovered; its cash-flow schedule may be represented by an additional contingent survivor-independent tail.

For deferred annuities, a guarantee might apply to premiums paid during the deferral period, to a stated account, or to a minimum number of post-commencement payments, and those promises have very different valuation triggers. One should specify the state-space before valuing the options: alive/dead, commencement date reached/not reached, benefit elected/not elected, and, if relevant, account balance or market-index performance. There is no single universal discount factor or death probability that prices every guarantee.

Guaranteed income and financial options

A contract promising a minimum investment-linked income can expose the insurer to both mortality and market risk. In this setting a simple life-annuity present-value formula is a component, not the whole valuation. A variable annuity guarantee may have policyholder election dates, roll-up bases, withdrawals, surrender rights, guaranteed annuitization factors, and path-dependent fees. Those features can make cash flows depend on investment returns, behavior, and longevity simultaneously. A market-consistent option valuation must specify a tradable-risk pricing approach and treatment of unhedgeable mortality, lapse, and expense uncertainty; risk-neutral discounting alone does not determine every nonfinancial component.

The distinction is also relevant to regulatory measurement. Under IFRS 17 Insurance Contracts, measurement of groups of insurance contracts includes fulfillment cash flows and, where relevant, a contractual service margin. Under the European Solvency II Article 77 provisions, technical provisions ordinarily comprise a best estimate and a risk margin, with a separate treatment for reliably replicable cash flows. A contract's stochastic actuarial present value is an input to these frameworks, while the resulting reported measure follows additional rules. IFRS 17's application also requires a contract within the standard's scope; products marketed as annuities are not automatically identical accounting instruments.

Economic interpretation of guarantees

Income guarantees change the allocation of value across survival states. A life-only benefit concentrates payments in later-life states, while a period-certain promise also directs payments to estate or beneficiary states when the original annuitant dies earlier. For the same single premium, adding a valuable guarantee generally reduces the maximum sustainable life-only periodic benefit if all other factors are fixed. Actual product comparisons require examination of the guaranteed beneficiaries, tax treatment, benefit inflation, contract liquidity, and financial strength of the obligor. Mortality selection and contract design may interact: Finkelstein and Poterba's (2004) UK annuity policyholder evidence associates mortality patterns with features such as payment timing and potential payments to estates. It cannot be generalized into a universal causal pricing coefficient for every jurisdiction or product.

Interest and Longevity Sensitivity

Interest duration and convexity

With flat force of interest (\delta) and predetermined expected survival-weighted benefit density (w(t)=b(t)S_x(t)), the continuous annuity factor is

[ V(\delta)=\int_0^\infty e^{-\delta t}w(t),dt. ]

Where differentiation under the integral sign is justified,

[ \frac{\partial V}{\partial\delta} =-\int_0^\infty t e^{-\delta t}w(t),dt, \qquad \frac{\partial^2 V}{\partial\delta^2} =\int_0^\infty t^2 e^{-\delta t}w(t),dt. ]

For nonnegative benefits the first derivative is nonpositive and the second derivative nonnegative. The first moment of discounted expected payment times defines a modified-duration-like measure with respect to the force of interest: (D_\delta=-V'(\delta)/V(\delta)). The second normalized moment gives the associated convexity. These sensitivities concern the specified benefit-only cash flows held fixed; they exclude changes in mortality assumptions, policyholder behavior, credit spreads, asset reinvestment, or regulatory discount-curve rules.

In annual models with effective rate i, derivative units differ: (\partial V/\partial i=-\sum_k k(1+i)^{-k-1}{}_kp_x) for due payments. A one-percentage-point increase in an effective annual rate is (\Delta i=0.01); a one-basis-point parallel change is 0.0001. A curve-based valuation requires a vector of zero-coupon discount factors or rate buckets, and a single flat-rate duration may mask offsetting sensitivities across maturities.

For the constructed five-payment annuity-due, changing the flat effective rate from 3% to 2% gives 4.696252, and increasing it to 4% gives 4.524827. Holding mortality fixed, lower yields raise the present value and higher yields reduce it. The shifts are asymmetric because present value is curved in the interest rate. An insurer with very long payout obligations therefore faces substantial reinvestment and asset–liability matching risk, even when the expected future benefit amounts are estimated without sampling error.

Longevity sensitivity and mortality improvement

In a deterministic survival model the annuity value generally increases as survival probabilities at future payment dates increase. Suppose (S_x(t;\theta)=\exp[-\int_0^t\mu_{x+s}(\theta)ds]). Then, under suitable differentiability and integration conditions,

[ \frac{\partial V}{\partial\theta} =-\int_0^\infty b(t)D(0,t)S_x(t;\theta) \left[\int_0^t\frac{\partial\mu_{x+s}(\theta)}{\partial\theta}ds\right]dt. ]

The sign is interpretable. If a parameter increases hazard at all maturities, annuity value falls; if a longevity-improvement scenario reduces hazard, the value rises. This expression also shows why an improvement occurring at younger future ages influences many later benefit dates through cumulative survival. There is no requirement to introduce an independent 'longevity surcharge formula' when a sensitivity can be derived directly from the established survival function.

A simple finite stress scales each annual q in the five-year example by a common multiplier while keeping q below one. At 80% of the baseline q values, the five-payment annuity-due is 4.630314; at the baseline it is 4.608894; at 120% it is 4.587583. This stress varies mortality, not the benefit schedule, interest assumptions, or original contract. The numbers are unit tests for the direction and magnitude of local model sensitivity, not a calibrated pandemic, demographic forecast, or regulator-prescribed solvency stress.

The broader research problem of mortality improvement requires calendar-time projection. Lee and Carter's (1992) mortality forecasting analysis models log age-specific death rates using an age pattern and a period index. Cairns, Blake, and Dowd's (2006) two-factor stochastic mortality study analyzes a different structure for mortality dynamics at older ages and parameter uncertainty. Neither paper supplies a universally correct annuity forecast; their evidence establishes that assumptions about future mortality dynamics and parameter uncertainty are material and model-dependent. Forecast backtesting, cohort consistency, and plausible treatment of extreme old ages are essential for application.

Discount rates, risk adjustments, and measurement objectives

A valuation for managing a portfolio of promised benefits begins by projecting cash flows by currency, benefit type, and contract boundary. Market interest rates then enter through a documented curve. However, an expected present value discounted at a convenient yield is not, by itself, an accounting or solvency reserve. The official IFRS 17 standard overview identifies risk-adjusted present value of fulfillment cash flows and the contractual service margin for the appropriate group of insurance contracts, along with recognition, presentation, and disclosure rules. Its discount rates reflect characteristics of cash flows and financial risks relevant to those cash flows rather than a universal fixed premium pricing rate.

The EIOPA Solvency II Article 77 text defines the best estimate as a probability-weighted future-cash-flow measure using the relevant risk-free term structure; technical provisions generally add a risk margin. The best estimate includes required future cash inflows and outflows and is calculated gross of reinsurance recoverables, which are determined separately. This differs from a unit-benefit annuity factor because actual policy obligations may include premiums, expenses, options, policyholder actions, and separate recoverables. The measurement is tied to an applicable prudential jurisdiction and its operative rules, including any authorized adjustments and effective dates.

The International Actuarial Association's ISAP standards directory identifies ISAP 1 for general actuarial practice and ISAP 4 for IFRS 17. These are model standards produced by a professional international body; they must not be described as universally binding statutes. Their relevance is methodological: clear assumptions, appropriate data, professional communication, and consistency between valuation purpose and model design. A robust annuity valuation records its own methodology and its relationship to local professional and statutory requirements.

Stochastic interest, dependence, and the measure problem

When the discount factor D(0,t,\omega) depends on the future state (\omega), expected value is generally

[ V=\mathbb E!\left[\sum_k B_k(\omega)D(0,t_k,\omega)\right], ]

not simply a sum of products of marginal expected payments and marginal expected discounts. The equality (\mathbb E[BD]=\mathbb E[B]\mathbb E[D]) needs an independence or covariance condition. Interest rates and mortality improvements can be associated through macroeconomic, inflationary, medical, or social mechanisms; even if their dependence is weak in one sample, assuming it without examination may bias a complex valuation. A financial-market-consistent approach must also state whether probabilities refer to physical demographic outcomes, a pricing measure for tradable risk, or an actuarially adjusted nonfinancial-risk treatment. Combining discount curves, hazard projections, and market prices from incompatible measures produces an internally incoherent result.

For inflation-indexed income, nominal cash flows may be contingent on a price index and the relevant discount factors must be consistent with nominal or real units. For a deterministic inflation escalation g and nominal flat interest i, an annual amount B_k=B_0(1+g)^k produces discounted factors (((1+g)/(1+i))^k); treating those projected nominal benefits as if they were constant real benefits overstates or understates value depending on the mismatch. Stochastic indexation and embedded floors require additional state modeling.

Integrated assumptions, validation, and governance

An actuarial annuity valuation can be reproduced only when its basic evidence is retained: valuation date and currency; contract and benefit triggers; insured ages and cohort; commencement and payment dates; mortality data provenance and experience adjustment; improvement assumptions; discount factors and compounding convention; dependence for multiple lives; expenses and policyholder options where relevant; calibration version; sensitivity definitions; and selected accounting, capital, or pricing purpose. A well-documented number that uses the wrong contractual event remains wrong, so review must begin with the underlying promise.

Five classes of checks follow directly from the mathematics. First, structural event tests verify that probability bounds are satisfied and that joint-life and last-survivor orderings hold. Second, time-consistency checks reconcile annual, fractional-year, and continuous schedules and the resulting units. Third, numerical tests recompute cash-flow sums and integrals independently and verify simple closed-form benchmarks. Fourth, scenario tests compare mortality, curve, guarantee, and dependence sensitivities under controlled perturbations. Fifth, source and governance tests verify mortality-table applicability, effective legal rules, and the separation of modeled cash-flow values from reported reserve or commercial prices. These checks are consistent with the IAA's general actuarial standards framework and with the risk transparency required by the official supervisory and accounting sources discussed above.

The mathematical interpretation can now be stated precisely. An annuity is valued by the probability of each legally and operationally defined benefit event multiplied by its scheduled discounted amount, with dependence and stochastic market factors integrated where required. The time structure of an annuity distributes obligation across many future dates; mortality improvement and lower discount rates increase the value of positive long-duration life payments under corresponding controlled assumptions. Guarantees change which states receive payments, while multi-life forms change the survival event. Accounting and prudential rules then select which projected cash flows and risk components must be reported. This is a synthesis of established actuarial mathematics, demographic evidence, and regulatory measurement rather than a proposed new theoretical system.

Conclusion

Annuity mathematics begins with an explicit promise: an amount payable at identified times if stated survival, death, or guarantee conditions are satisfied. From that promise follow the benefit indicators, conditional survival probabilities, and discount factors used to calculate actuarial present value. The same construction applies to life-only, temporary, deferred, joint-life, last-survivor, reversionary, certain, escalating, and refund-related benefits once each contract is translated into a consistent sequence of contingent cash flows.

The analytical derivations establish several consequential results. Annual due and immediate conventions require distinct schedules; their whole-life identity must be adapted for temporary terms. Deferred commencement combines time discounting with the probability of reaching the commencement date. Joint-life and last-survivor benefits obey event inclusion–exclusion, while dependence between lifetimes changes their numerical values. Guarantees substitute unconditional payment states for survival-only states over defined intervals. The interest derivatives of positive annuity benefits exhibit decreasing value with higher discount rates and positive convexity, and lower mortality generally increases life-annuity obligations. The constructed numerical examples reproduce these results and provide transparent implementation benchmarks.

Scientific valuation also requires the mortality basis to match the insured population and forecast horizon. General population life tables, experience studies, and reserving tables answer different questions. Period and cohort mortality information must be distinguished; dependent lifetimes, selection, future mortality improvements, and stochastic market conditions can materially alter results. Classical and contemporary scholarship establishes these as recognized sources of estimation and model uncertainty. The resulting valuation is credible only to the extent that assumptions, cash-flow units, horizon, and methods are transparent and verified against their use.

Finally, the expected present value of a promised life-income stream is one important measurement quantity within a larger financial institution. Commercial prices may add expenses and capital charges; reporting under IFRS 17 applies insurance-contract measurement rules; Solvency II technical provisions follow prudential requirements. This distinction is central to comparability across jurisdictions and valuation purposes. The integrated conclusion of this monograph is that annuity value is generated by a precise relationship among contractual contingency, demographic survival, time-dependent finance, and an explicitly stated measurement objective. Its contribution is a checked, source-grounded reconstruction and synthesis of that relationship, with no claim of a new proprietary theory or independently estimated mortality dataset.

References

  • Cairns, Andrew J. G., David Blake, and Kevin Dowd. 2006. “A Two-Factor Model for Stochastic Mortality with Parameter Uncertainty: Theory and Calibration.” Journal of Risk and Insurance 73 (4): 687–718. Publisher DOI.

  • Dickson, David C. M., Mary R. Hardy, and Howard R. Waters. 2012. “Annuities.” In Actuarial Mathematics for Life Contingent Risks, 107–141. Cambridge University Press. Chapter DOI.

  • European Insurance and Occupational Pensions Authority (EIOPA). 2026, accessed November 14. “Solvency II Single Rulebook: Article 77 — Calculation of Technical Provisions.” Rulebook presentation of Directive 2009/138/EC. Official rulebook.

  • Finkelstein, Amy, and James Poterba. 2002. “Selection Effects in the United Kingdom Individual Annuities Market.” Economic Journal 112 (476): 28–50. Publisher DOI.

  • Finkelstein, Amy, and James Poterba. 2004. “Adverse Selection in Insurance Markets: Policyholder Evidence from the U.K. Annuity Market.” Journal of Political Economy 112 (1): 183–208. Publisher DOI.

  • Human Mortality Database (HMD). 2026, accessed November 14. “Project Overview: Data, Scope, and Methods for Mortality Rates and Life Tables.” University of California, Berkeley, and Max Planck Institute for Demographic Research. HMD project overview.

  • IFRS Foundation. 2026, accessed November 14. “IFRS 17 Insurance Contracts: Standard Overview and Measurement Principles.” IFRS Foundation. IFRS 17 page.

  • International Actuarial Association (IAA). 2026, accessed November 14. “International Standards of Actuarial Practice: ISAP 1 and ISAP 4.” IAA. Standards directory.

  • Lee, Ronald D., and Lawrence R. Carter. 1992. “Modeling and Forecasting U.S. Mortality.” Journal of the American Statistical Association 87 (419): 659–671. Publisher DOI.

  • Royal Society. 2026, accessed November 14. “Paper of the Breslau Bills of Mortality by Edmond Halley.” Archival record of the 1693 historical paper. Royal Society archive.

  • Society of Actuaries (SOA). 2011–2012. “2012 Individual Annuity Reserving Report & Table.” Society of Actuaries, experience study and related implementation resources. SOA table materials.

  • Society of Actuaries (SOA). 2026. “Individual Annuity Experience Studies.” Includes the 2020–2024 Individual Payout Annuity Mortality Experience Study (2026 listing) and earlier U.S. annuity research resources. SOA experience catalog.

  • Yaari, Menahem E. 1965. “Uncertain Lifetime, Life Insurance, and the Theory of the Consumer.” Review of Economic Studies 32 (2): 137–150. Publisher DOI.