Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Annuity Discount Curve and Interest Rate Sensitivity

Abstract. The value of a life annuity is determined jointly by the contractual sequence of payments, the probability that each payment becomes due, and the discount factors applicable to those payments. This monograph investigates the third element in direct interaction with the first two: the selection and construction of annuity discount curves, the response of lifetime income liabilities to changes in the level and shape of interest rates, and the implications for insurers, beneficiaries, financial reporting, prudential valuation, and asset-liability management. An annuity has a distinctive interest-rate exposure because mortality-weighted payments extend over many possible dates while the promise remains an enforceable commitment to individual survivors. The analysis develops exact present-value, cash-flow duration, convexity, key-rate, and mortality-sensitivity identities, distinguishing fixed cash flows from benefits that react to financial markets, future policyholder behavior, or inflation. It examines the intellectual foundations of matching and immunization, then compares IFRS 17, European Solvency II, the United Kingdom's published supervisory curves, and United States principle-based valuation. The investigation uses official statutory and standard-setting documents, original actuarial and economic publications, and a fully specified constructed numerical experiment rather than insurer-confidential observations. For a synthetic annual-pay life annuity beginning at age 65, a continuously compounded 3% reference curve produces a present value of 12,227.14 per 1,000 annual payment, an effective parallel duration of 9.8814 years, and a convexity of 155.1169 years squared. A parallel increase of 100 basis points lowers the present value by approximately 9.16%, while a parallel decrease raises it by approximately 10.71%; nonparallel twists produce materially different results. The findings establish that a single annuity duration is an incomplete description of liability valuation risk, that discount-basis changes must be separated from changes in underlying obligations, and that effective hedging requires a joint view of interest-rate curve exposure, cash-flow uncertainty, capital constraints, liquidity and governance. These are independent analytic conclusions within established actuarial and financial theory, not a proposed new doctrine or an empirical claim about an actual insurance portfolio.

Keywords: Life annuity; annuity liability; discount curve; zero-coupon discount factor; survival probability; interest-rate sensitivity; effective duration; Macaulay duration; convexity; key-rate duration; yield-curve twist; asset-liability management; cash-flow matching; immunization; interest-rate risk; longevity risk; IFRS 17; Solvency II; VM-22; insurance reserves; risk-free term structure; inflation-linked annuities; regulatory valuation; model validation.

Introduction

An annuity converts a financial resource into a sequence of income payments, often conditional on continued survival. The insurer or pension provider promises to meet an obligation distributed across future years, while the eventual length and timing of those payments remain uncertain. The discount curve converts this contingent schedule into a valuation at a stated date. Changing that curve can materially change the measured obligation even when the contractual amount of every payment remains unchanged.

Interest-rate sensitivity is therefore a question of precisely what is being valued, which curve applies, which future states can affect cash flows, and which balance sheet or welfare measure is being evaluated. A fixed nominal annuity already in payment, an inflation-indexed lifetime pension, a deferred income annuity, and an annuity with investment-linked guarantees can have different responses to the same movement in observed bond yields. Two firms holding identical insurance promises may also report different amounts under different jurisdiction-specific statutory bases. Comparing their liabilities requires disclosure of the valuation regime, reporting date, currency, and cash-flow basis.

The foundations of this problem extend across actuarial science, finance and retirement economics. Redington's life-office valuation analysis (1952) formulated matching and immunization questions in the setting of long-term life insurance obligations. Fisher and Weil's interest-rate fluctuation study (1971) further situated bond-investment strategies in changing term structures. Yaari's uncertain-lifetime analysis (1965) established a foundational economic treatment of life-contingent choices. These contributions address related objects but do not make their assumptions interchangeable: investor return, household utility and insurer liability present value answer different questions.

This work brings those lines of inquiry into a single annuity-specific examination. The central estimand is the present value at an identified valuation date of a defined annuity cash-flow obligation under an explicit discount curve and survival basis. Its sensitivity is the change in this estimand under clearly specified interest-rate scenarios, with additional analysis where payment amounts or survival probabilities are themselves re-estimated. The inquiry is international because insurance accounting, supervisory discounting, contract designs and market data differ across jurisdictions; no global universal annuity curve exists.

Foundations of Annuity Valuation

The contractual object and stochastic payment obligation

For an annuity issued to a person aged x, define B_t as the payment promised at the end of year t, I_t as the indicator that the specified survival and contractual conditions for payment at t are fulfilled, and D(0,t) as the valuation-date discount factor. A simple life-contingent expected present value is

V0=∑t=1TBt E[It] D(0,t)=∑t=1TBt tpx D(0,t).V_0=\sum_{t=1}^{T}B_t\,\mathbb E[I_t]\,D(0,t) =\sum_{t=1}^{T}B_t\,{}_tp_x\,D(0,t).

Here tpx{}_tp_x denotes the probability of survival from the valuation date to time t on the stated annuitant basis. A continuous-time expression is V0=∫0Tb(t)Sx(t)D(0,t) dtV_0=\int_0^T b(t)S_x(t)D(0,t)\,dt, with appropriately defined payment density b(t), survival function S_x(t), and possibly a terminal payment. Both identities arise from linearity of expectation for a predetermined nonnegative payment and deterministic discount factors. They are valuation formulas, not claims that a real annuity portfolio has independent mortality and investment outcomes.

A more general formulation is V0=EQ[∑tDω(0,t)Xt(ω)]V_0=\mathbb E^Q[\sum_t D_\omega(0,t)X_t(\omega)] when relevant market risks affect state-contingent cash flows and a suitable valuation measure Q is specified. For insurance liabilities, the use of a market-consistent risk-neutral valuation component does not automatically price all nonfinancial mortality risk; a separate mortality basis, nonfinancial risk adjustment, or supervisory margin may be required. Mixing a real-world mortality distribution with a financial pricing measure is a methodological decision requiring justification, especially where equity-linked guarantees, policyholder behavior, or financial-market-dependent death benefits produce dependence.

The contractual perimeter also matters. A payment guaranteed for a minimum period may be payable to a beneficiary after the annuitant's death. Its expectation then has a guaranteed component and a survival-contingent component. A joint-life benefit requires the probability that a contractually relevant combination of two lives survives. A deferred-income annuity adds pre-commencement survival and postpones the effective center of the liability's cash-flow distribution. Claim expenses, taxes, acquisition costs, premiums due, and reinsurance recoveries may or may not belong in a particular valuation depending on its purpose and applicable standard. A stated duration without an enumerated cash-flow boundary has limited analytical meaning.

Actuarial and historical foundations of matching

In Redington (1952), valuation of the obligations of a life office is inseparable from asset selection and the risks of changes in interest rates. His original matching and immunization discussion examines the relationship of asset and liability present values and their responses to valuation rates. At a reference interest rate, a classically immunized position seeks equality of present values, equality of first-order interest sensitivity and a suitable second-order condition. This does not make assets and obligations identical, and it does not eliminate nonparallel shifts, embedded options, credit losses or unmodeled timing uncertainty.

Fisher and Weil (1971) investigated bondholder returns under rate fluctuations, thereby demonstrating why a promised yield and realized holding-period return need to be distinguished. Their original setting is fixed-income investment rather than an insured life-contingent promise. The transfer to annuity asset-liability management is made through a clearly stated comparison: both sides of an insurer balance sheet respond to changes in the term structure, and first-order matching of a scalar yield can leave the positions exposed to other interest-rate configurations.

Modern risk management consequently evaluates economic value, accounting value, reported capital and actual liquidity needs as related but distinct quantities. The Basel supervisory treatment of interest-rate risk is directed at banking-book positions and expressly identifies nonparallel rate risk, basis risk and optionality. It provides a useful comparison of general financial mechanisms, but its regulatory prescriptions must not be transferred to insurers as though they were insurance solvency rules.

Household value, insurer value and survival assumptions

The household's reason to purchase a lifetime annuity is not reducible to the insurer's present-value calculation. Yaari (1965) analyzed consumption under uncertain lifetime, helping establish why a stream contingent on survival can have welfare value beyond its expected discounted payment. Davidoff, Brown and Diamond (2005) examined annuitization and individual welfare, including the limits to full annuitization when markets are incomplete. Illiquidity, other household resources, bequest motives and the timing of uninsured expenses affect the comparison of annuity income with liquid assets.

The survival basis has its own source and selection problem. Finkelstein and Poterba (2004), in their United Kingdom annuity-market study, found relationships between observed mortality and contract characteristics such as timing and estate-payment options. Their result establishes that annuitant contract composition can matter for mortality inference; it does not identify a universal adjustment for all countries and products. Pension and annuity modelers can use the Society of Actuaries' mortality resources as sources of specified mortality tables and improvement scales, but must examine population, observation period, occupational and selection differences before applying them to a given insured pool.

For the insurer, a longevity improvement that raises late survival probabilities tends to increase the expected undiscounted number of lifetime payments and may increase both liability value and interest-rate duration. For an individual annuitant, unusually long survival is the event against which a lifetime annuity protects. The same mortality-contingent payment thus serves household income security while constituting an uncertain insurer liability. Neither perspective overrides the other; they are different estimands based on the same contract.

Dates, currencies and cash-flow boundaries

A defensible valuation states the valuation date, contract issue and commencement dates, payment frequencies, timing of the first and last payments, currency, nominal or inflation-related benefit rule, guarantee periods, spouse or beneficiary continuation, premium status, and expense categories. The cash-flow horizon T may be a truncation convention representing an advanced maximum modeled age, rather than a genuine contractual termination. The truncation error is bounded by the omitted survival-weighted payment tail and should be estimated when material.

For cross-border comparison, valuation must also disclose calendar conventions, compounding, and whether payments are aggregated from individual policies or an assumed representative life. Annual, monthly and continuous benefit formulas cannot be interchanged without an adjustment for payment timing. Cash flows discounted in a currency different from that in which benefits are payable require a currency-consistent transformation and may introduce foreign-exchange hedging costs. These seemingly clerical definitions determine the numerical weight of future annuity payments and therefore every reported interest sensitivity.

Constructing the Annuity Discount Curve

Discount factors, yields and market observability

A discount factor D(0,t) represents the time-zero price, in a selected valuation convention, of one unit of deterministic currency payable at t. Under a continuously compounded zero-rate curve z(t), D(0,t)=e−tz(t)D(0,t)=e^{-tz(t)}; under an annual effective zero yield i(t), D(0,t)=[1+i(t)]−tD(0,t)=[1+i(t)]^{-t}. Equality of those discount factors, not equality of the quoted rate numbers, determines equivalent valuation. The instantaneous forward rate is f(t)=−∂log⁡D(0,t)/∂tf(t)=-\partial\log D(0,t)/\partial t, where a sufficiently smooth curve is assumed.

Observed government yields, interbank swaps, overnight-indexed swaps and corporate bonds have different credit, collateral, liquidity and market-microstructure characteristics. Curve choice requires a purpose-specific account of these differences. For example, the U.S. Treasury explicitly explains that its published par yield curve differs from zero-coupon rates. Inserting a quoted par yield directly as a zero-coupon rate can misprice long annuity cash flows. A reproducible market curve construction must document input instruments, day counts, coupon conventions, interpolation, bootstrap equations, collateral assumptions, bid-ask choices and dates.

Even a precisely fitted market curve does not determine a universal insurance discount basis. Insurance cash flows can have a different degree of liquidity from traded instruments; the value of some participating benefits depends on underlying asset returns. Additional risk adjustments and margins also belong to certain accounting and regulatory calculations, but should be identified separately rather than hidden inside an unexplained yield spread.

Insurance cash-flow characteristics and IFRS 17

The IFRS Foundation's IFRS 17 Insurance Contracts standard has applied for annual reporting periods beginning on or after January 1, 2023. Its measurement framework combines estimates of future cash flows with treatment of insurance risk, an applicable contractual service margin when required, and separate presentation of insurance service and insurance finance effects. The appropriate discount rates are tied to the time value of money, characteristics of the cash flows and liquidity characteristics of the insurance contracts; they must be consistent with relevant observable market information.

The detailed IFRS 17 discounting requirements, particularly paragraph 36 and application guidance B72–B85 as adopted in Commission Regulation (EU) 2023/1803 (2026 consolidated text), prevent the mechanical substitution of an insurer's expected asset yield for an appropriate liability discount rate. Where a fixed annuity's payments do not vary with the return on its backing assets, a credit spread earned on those assets does not thereby become a discount benefit belonging to the policyholder obligation. Where cash flows do vary with the returns on underlying items, their valuation must reflect that financial variability and applicable guarantees.

IFRS 17 also makes it necessary to distinguish current rates used for fulfilment cash flows from rates fixed at initial recognition for certain contractual service margin and presentation calculations. The accounting movement in an insurance contract liability can therefore differ from a pure economic present-value change under a new curve. Classification of insurance finance income or expenses between profit or loss and other comprehensive income, where allowed and elected, influences reported volatility without changing the underlying insured promise. The International Actuarial Association's ISAP 4 on IFRS 17 provides an actuarial practice reference for this reporting environment.

A disciplined analysis reports the base curve, any illiquidity adjustment, whether the benefit is linked to underlying assets, the relevant accounting measurement component, and the route by which the yield shift enters current fulfilment cash flows. It also records how changes in nonfinancial assumptions and risk adjustment enter separately. Reporting a single figure labeled "IFRS 17 duration" without those definitions is analytically incomplete.

Solvency II, extrapolation and supervisory adjustments

European Solvency II provides a distinct prudential reference. Under Directive 2009/138/EC Article 77, technical provisions equal a best estimate plus a risk margin, and the best estimate uses a probability-weighted present value of future cash flows with the relevant risk-free interest-rate term structure. Article 43 of Delegated Regulation (EU) 2015/35 sets term-structure requirements describe currency- and maturity-specific criteria and governance of data and parameters.

The EIOPA risk-free curve publication service supplies supervised term structures, technical documentation and extrapolation information. As of November 15, 2026, its public record includes technical documentation dated November 6, 2026. Extrapolation beyond liquid maturities, ultimate forward-rate assumptions and the treatment of currency-specific instruments can influence very long annuity liabilities disproportionately. Curve sensitivities should therefore distinguish the traded-data region from extrapolation and policy parameters. An upward change in observed long yields need not translate into an identical shift at every extrapolated horizon.

Certain eligible portfolios may apply a matching adjustment subject to stringent conditions. Solvency II Article 77b addresses asset assignment, management restrictions and prior approval; Article 77c's calculation rules specify adjustments and retained credit-risk elements. This regime differs structurally from merely increasing a discount rate by an insurer's portfolio yield. Its applicability depends on contractual cash-flow characteristics, eligible assets and supervision.

A separate British implementation can be observed through the Bank of England's published supervisory risk-free curves. The page records, for example, October 2026 curves released on November 9, 2026 together with relevant technical parameters. United Kingdom supervisory rules and European Union Solvency II arrangements should be described by jurisdiction and valuation date rather than represented as one unchanged framework.

United States valuation and jurisdictional distinctions

United States statutory annuity reserves use a different architecture. The National Association of Insurance Commissioners' principle-based reserving overview describes the use of prescribed reserves and economic scenarios, and identifies VM-22 for non-variable annuities. The official Valuation Manual editions include a current 2026 edition and separate future editions; the version relevant to a particular valuation requires a jurisdiction- and effective-date check. The NAIC notes a phased transition for VM-22, with implementation beginning in 2026 and mandatory application to new issues targeted for January 2029.

For payout annuities, economic scenarios and prudent-estimate assumptions have a different institutional purpose from a single IFRS 17 current discount rate or a Solvency II regulatory best-estimate curve. The reserve can involve asset and liability projections, floor requirements and prescribed scenarios. A market yield curve is therefore part of the analytical information set, not necessarily the entire statutory valuation operator. Comparing a U.S. statutory reserve with a European best estimate without explicitly reconciling scope, margins, time horizons and scenarios can produce a false inference about the economics of the same promised benefit.

The IAIS supervisory principles give the international reference that valuation should be reliable, transparent and useful for solvency assessments, but they do not prescribe a single detailed interest-rate curve to replace jurisdiction-specific requirements. International comparison can be informative only after separating economic valuation, statutory reserves, accounting liabilities, regulatory capital and internal pricing.

Nominal, real, indexed and market-linked benefits

A fixed nominal annuity has contractual amounts B_t measured in nominal currency. A fixed escalating annuity may promise Bt=B1(1+g)t−1B_t=B_1(1+g)^{t-1} with contractually stated g. An inflation-indexed benefit can refer to a price index, with lag, caps, floors and survivorship rules. The latter has uncertain nominal cash flows and should be valued with consistent inflation expectations, nominal cash flows or real discount factors. A real rate is not interchangeable with a nominal rate without an inflation and convention transformation.

A variable annuity or participating income stream may have benefits linked to investment returns. Minimum income guarantees introduce asymmetric payoffs: when market performance is poor, a contractual floor may become binding, changing both the amount and interest-rate sensitivity of expected future payments. The IFRS 17 application guidance explicitly addresses cash flows linked to underlying items and guarantees. For these contracts, holding payments fixed while shocking the curve is useful as a partial diagnostic but may significantly understate total economic sensitivity.

The resulting classification is substantive: discounting a predetermined payment schedule, discounting payments indexed to inflation, and valuing payments contingent on market returns solve distinct stochastic problems. A monograph on annuity curves must make those distinctions before estimating duration.

Measuring Interest-Rate Sensitivity

Exact revaluation under arbitrary curve shocks

Write expected payment weight Ct=Bt tpxC_t=B_t\,{}_tp_x, and assume initially that these weights are independent of the discount-rate perturbation. For a zero-rate curve z(t), the present value is

V[z]=∑t=1TCtexp⁡[−tz(t)].V[z]=\sum_{t=1}^{T} C_t\exp[-t z(t)].

For any scenario perturbation δz(t)\delta z(t), the exact difference is

ΔV=∑t=1TCte−tz(t)(e−tδz(t)−1).\Delta V=\sum_{t=1}^{T} C_t e^{-t z(t)} \left(e^{-t\delta z(t)}-1\right).

This identity directly reveals why the response to a nonparallel shift depends on the entire distribution of present-valued payments over maturity. It also shows that a change in the level of rates and a change in the shape of the curve are not interchangeable. Under positive promised payments, a positive shift at all payment maturities reduces the value, provided cash-flow weights remain fixed. With benefits varying in response to rates, that conclusion must be re-examined using the full state-dependent valuation.

Duration and convexity

For a parallel continuous-rate shift a, V(a)=∑tCte−t[z(t)+a]V(a)=\sum_t C_t e^{-t[z(t)+a]}. Differentiating gives

dVda=−∑ttCtD(0,t),d2Vda2=∑tt2CtD(0,t).\frac{dV}{da}=-\sum_t t C_tD(0,t),\qquad \frac{d^2V}{da^2}=\sum_t t^2 C_tD(0,t).

The continuously compounded effective parallel duration and convexity are

DV=−1VdVda=∑ttCtD(0,t)V,ΓV=1Vd2Vda2=∑tt2CtD(0,t)V.D_V=-\frac{1}{V}\frac{dV}{da} =\frac{\sum_t t C_tD(0,t)}{V},\qquad \Gamma_V=\frac{1}{V}\frac{d^2V}{da^2} =\frac{\sum_t t^2 C_tD(0,t)}{V}.

For a fixed positive expected cash-flow schedule these are respectively the present-value-weighted mean and second raw moment of payment time. They reflect the payment schedule and discount factors jointly, not just the contractual maturity. The variance of discounted payment times is ΓV−DV2\Gamma_V-D_V^2. For a small parallel change, ΔV/V≈−DVa+12ΓVa2\Delta V/V\approx-D_V a+\tfrac12\Gamma_V a^2. Because convexity is nonnegative for fixed positive cash flows, equal positive and negative shifts produce asymmetric present-value changes: the value gained from a rate decrease is generally larger than the value lost from an equal rate increase.

A one-basis-point dollar sensitivity is DV01=VDV(0.0001)DV01=V D_V(0.0001) as a first-order positive magnitude. A cash-flow duration defined for annual effective yields requires the corresponding derivative and cannot be substituted unchanged for continuous zero-rate duration. For an annuity with expenses, premiums, surrender payments or financial guarantees, negative net cash-flow components and rate-dependent weights may destroy the simple positive-weight interpretation.

Key-rate and shape exposure

Divide the curve into mutually exclusive maturity buckets K1,…,KmK_1,\ldots,K_m. If δz(t)=aj\delta z(t)=a_j for t∈Kjt\in K_j, the key-rate duration of a bucket is

Dj=∑t∈KjtCtD(0,t)V.D_j=\frac{\sum_{t\in K_j}t C_tD(0,t)}{V}.

The first-order value change is ΔV/V≈−∑jDjaj\Delta V/V\approx-\sum_j D_j a_j. With piecewise-constant, non-overlapping bucket shocks, the sum of key-rate durations equals the parallel duration exactly. Interpolated tenor bumps used on a market-observed knot curve generally overlap; their component sums must be interpreted according to the selected interpolation and bump operator. Stating the bump basis is as important as stating the duration number.

The utility of key-rate sensitivity follows from the balance sheet's exposure to curve shape. A portfolio can have the same total dollar duration as its liabilities and yet have a different distribution of duration by maturity. A steepening, flattening or localized long-end move then changes the economic surplus. Redington's immunization analysis supplies a historical foundation for matching, and modern supervisory interest-rate guidance identifies why nonparallel curve changes merit separate attention. Neither source removes the need for insurer-specific key-rate measurement.

A general continuous-curve sensitivity can be expressed as a functional derivative: δV/δz(u)=−uCuD(0,u)\delta V/\delta z(u)=-u C_uD(0,u) for a discrete cash-flow point u, with appropriate distributional interpretation. This is a useful mathematical reminder that the interest-rate risk of a lifetime annuity is a maturity profile, not intrinsically a single number.

Mortality, cash-flow and discount interactions

Now let pt(θ)p_t(\theta) depend on mortality parameters θ\theta, and permit Bt(θ,z)B_t(\theta,z) to vary with contractual or behavioral assumptions. A first-order differential becomes

dV=∑tDt[pt dBt+Bt dpt]+∑tBtpt dDt.dV=\sum_t D_t\left[p_t\,dB_t+B_t\,dp_t\right] +\sum_t B_t p_t\,dD_t.

The last sum is the pure discount-factor effect holding expected payment amounts constant. The first two sums represent changes in payment amount and survival. For continuously compounded zero yields, dDt=−tDt dz(t)dD_t=-tD_t\,dz(t). Reporting these effects separately avoids attributing an increase in liabilities caused by improved annuitant longevity to falling interest rates.

The survival distribution may affect interest-rate sensitivity even without being changed in a scenario. When mortality improves primarily at higher ages, additional late payments generally raise duration. Different populations can therefore have different interest-rate sensitivity for the same initial benefit amount and discount curve. Finkelstein and Poterba's evidence on annuity selection reinforces the empirical importance of distinguishing annuity purchasers from broad population mortality. The SOA mortality table catalogue identifies multiple possible datasets but does not license universal transfer from one population to another.

In general a discount-rate stress might also change future inflation, investment-linked payouts, asset returns, surrender incentives and insurer behavior. Sensitivity calculated with frozen C_t is then a partial derivative, while a full model revaluation is a total derivative or scenario difference. Both are legitimate if clearly labeled. The analyst should not report a partial DV01 as a complete measure of stochastic guarantee risk.

Guarantees, embedded choices and nonlinear exposure

Life-contingent benefit guarantees convert otherwise smooth payment schedules into contingent claims. A guaranteed minimum income benefit may become more valuable when rates fall, with an associated change in optimal or actual exercise behavior. A policy that permits commutation, deferral, or an alternative lump sum may exhibit rate-sensitive elections. Management actions and investment returns can change the expected payment schedule through economic feedback.

The necessary computational distinction is between a deterministic cash-flow valuation and a stochastic contract-value calculation. The latter may require interest-rate models calibrated to option prices, dynamic lapse or annuitization assumptions, Monte Carlo path generation, and explicit treatment of minimum benefits and mortality correlation. The insurer's curve can be one input to that model, but a scalar duration computed at one date cannot describe the entire nonlinear exposure. The NAIC principle-based reserving framework illustrates why multiple economic scenarios are important in some annuity valuation contexts.

For any model, convexity is a local measure: a second derivative at a specific reference state. It is not a universal upper bound on losses from large shocks, especially when cash flows change by scenario or when an extrapolation regime changes. Full revaluation across sufficiently broad interest scenarios is essential for material nonlinear exposures.

Reproducible Valuation Experiments

Constructed portfolio and calculation protocol

The following experiment is deliberately synthetic and fully deterministic. It is an illustration of mathematical sensitivity, not a finding about any insurer, mortality cohort, actual market yield curve or regulatory reserve. The hypothetical contract pays 1,000 nominal currency units at each calendar year-end from t=1 through t=40 if an initial 65-year-old annuitant remains alive. Annual survival probabilities are specified by the synthetic survival function

pt=exp⁡(−0.01t−0.002t2),t=1,…,40.p_t=\exp(-0.01t-0.002t^2),\qquad t=1,\ldots,40.

The conditional hazard implied in the continuous-time analogue is 0.01+0.004t0.01+0.004t. This is an invented smoothly increasing late-life hazard for computational illustration; it is not estimated from any mortality table. The reference zero-rate curve is flat at a continuously compounded 3%, with Dt=exp⁡(−0.03t)D_t=\exp(-0.03t). The baseline is therefore

V0=1000∑t=140exp⁡[−0.01t−0.002t2−0.03t].V_0=1000\sum_{t=1}^{40} \exp[-0.01t-0.002t^2-0.03t].

The calculation uses no tax, insurer expense, risk adjustment, mortality improvement, surrender, guarantee period, reinsurance or profit margin. Every scenario holds p_t and the nominal payment amount fixed. Rates in the experiment are hypothetical and are intentionally not represented as observed November 2026 quotations. The 40-year limit is a modeling truncation; extending it would require documenting the omitted positive tail.

A minimal independent reproduction is possible with the following pseudocode:

T = 40
p(t) = exp(-0.01*t - 0.002*t*t)
V(shift, escalation=0) =
  sum(1000 * p(t) * (1+escalation)^(t-1)
      * exp(-(0.03 + shift(t))*t) for t in 1..T)
duration =
  sum(t * 1000 * p(t) * exp(-0.03*t) for t in 1..T) / V(0)
convexity =
  sum(t*t * 1000 * p(t) * exp(-0.03*t) for t in 1..T) / V(0)

The values below were calculated independently by direct summation of the displayed formulas and checked for duration-bucket additivity and the direction of rate responses.

Parallel shocks and approximation error

At the baseline assumptions, the expected present value is 12,227.140706 currency units, the continuous-rate duration is 9.8813556 years, the convexity is 155.1168819 years squared, and the first-order DV01 is 12.082073 currency units per basis point. A parallel 100-basis-point increase gives an exact revaluation of 11,107.721020, a decline of 9.1552%. A parallel 100-basis-point decrease gives 13,536.988065, an increase of 10.7126%.

The second-order approximation for a 100-basis-point increase is ΔV≈−VDV(0.01)+12VΓV(0.01)2=−1,113.375456\Delta V\approx -V D_V(0.01)+\frac12 V\Gamma_V(0.01)^2=-1,113.375456, compared with the exact difference of -1,119.419687. The difference arises from third and higher derivatives and is small relative to the full modeled liability in this deliberately smooth fixed-cash-flow case. It would be unsafe to extrapolate this approximation accuracy to option-rich annuities or larger shocks.

Constructed interest-rate scenario Present value Change from baseline
Flat 3% reference 12,227.141 0.000%
Parallel +100 bp 11,107.721 -9.155%
Parallel -100 bp 13,536.988 +10.713%
+100 bp only after year 15 11,728.836 -4.075%
-50 bp at years 1–10, +100 bp at years 11–40 11,664.636 -4.600%
+100 bp at years 1–10, -50 bp at years 11–40 12,298.862 +0.587%

These changes are computed under the same fixed cash-flow basis. A positive net effect under the final twist does not mean that a general rate rise increases the liability; it means the chosen combination of short-end increase and longer-end decrease has a net positive present-value effect on this payment pattern.

Key-rate decomposition and curve twists

For three mutually exclusive maturity buckets, the present-value-weighted contributions to parallel duration are 1.030406 years for payments in years 1–5, 4.287309 years for years 6–15, and 4.563641 years for years 16–40. Their sum equals 9.881356 years, within rounding. The corresponding positive DV01 magnitudes are approximately 1.259892, 5.242153 and 5.580028 currency units per basis point. This decomposition shows that almost half the reference parallel duration contribution comes from payments beyond year 15, despite the survival probabilities of those payments being lower than earlier ones.

The actual revaluation under a long-end-only 100-basis-point shock is a decline of 4.075%, not the 9.155% resulting from a whole-curve 100-basis-point shock. A short-end decline combined with a longer-end increase reduces the value by 4.600%; reversing the direction in the stated twist increases the value by 0.587%. These precise results follow from the declared bucket definitions. A market implementation using interpolated tenor knots would require a different sensitivity matrix and a stated interpolation method.

The analytical lesson is that the direction and magnitude of annuity revaluation depend on where the curve moves. A single modified duration, by design, compresses the maturity distribution to its first moment and cannot distinguish these cases.

Payment escalation and alternative cash-flow profiles

Holding all other synthetic assumptions fixed, an escalating benefit of 1000(1.02)t−11000(1.02)^{t-1} has a reference value of 14,752.815161, compared with 12,227.140706 for level benefits. Escalation increases the relative weight of later years and therefore changes interest-rate exposure; its exact duration must be recomputed under that payment schedule rather than copied from the level annuity. The fixed 2% escalation is a contractual assumption, not a scenario for actual consumer-price inflation.

For comparison, a deterministic 20-year certain annuity paying 1,000 annually has present value 14,815.145902 at the same 3% discount rate. Its higher value is not evidence that certain annuities are categorically more expensive than lifetime annuities: the two example contracts make different payment promises and have different modeled terms. The example demonstrates that a valuation comparison without matching the promise and survival conditions can be misleading.

If a hypothetical policy instead pays an inflation-adjusted amount, the analyst must specify an inflation process, possible index lags, caps and a nominal or real curve framework. The preceding fixed-escalation experiment does not measure inflation hedge effectiveness or real household purchasing power.

What the numerical results establish

The reproducible example establishes four facts of the stated mathematical experiment. First, an annual life-contingent liability is highly sensitive to discounting because a significant portion of expected payment value is located years into the future. Second, positive convexity makes parallel increase and decrease shocks asymmetric. Third, the rate sensitivity is distributed across maturity buckets and can produce opposite signs under different nonparallel shocks. Fourth, changing benefit escalation changes the expected cash-flow schedule even with identical survival and discount factors.

Its external validity is deliberately narrow. The mortality basis is synthetic; the 3% curve and prescribed shocks are illustrative; the annual 40-year convention is a truncation; and the analysis excludes insurer expenses, policyholder behavior, capital margins, observed investment portfolios and stochastic asset returns. No causal claim about actual pensioners or market prices follows from these constructed results. Their contribution is independent verification of the analytical mechanics and a transparent test case for comparing implementations.

Annuity Assets, Solvency and Financial Reporting

Asset-liability matching and economic surplus

Let A[z] denote the present value of assets under a market curve and L[z,\theta] the economic value of insurer obligations, where theta collects mortality, expenses and contractual assumptions. Define economic surplus S=A−LS=A-L. For a small parallel shock with the obligations held at fixed cash-flow weights,

ΔS≈(−ADA+LDL)Δr+12(AΓA−LΓL)(Δr)2.\Delta S\approx (-A D_A+L D_L)\Delta r +\frac12(A\Gamma_A-L\Gamma_L)(\Delta r)^2.

First-order parallel duration matching sets ADA=LDLA D_A=L D_L, usually together with a separate target for baseline surplus. If the portfolio value and dollar duration of assets are matched to those of liabilities, the immediate first-order effect of an infinitesimal parallel shift cancels. That statement does not ensure equal cash flows at each future date, eliminate changes in insurer credit quality or guarantee reserve requirements, or neutralize mortality and inflation developments.

More granular matching seeks small differences in bucket dollar durations ADA,j−LDL,jA D_{A,j}-L D_{L,j} and ideally cash-flow timing consistency. This follows the logic of Redington's original valuation study while accommodating the practical limits of a finite set of available investments. Because annuity obligations can extend beyond liquid bond maturities, perfect market replication may be impossible. For illiquid, long-tail liabilities, extrapolation assumptions and the availability of actual matching assets may dominate precision in a scalar duration statistic.

Credit, liquidity and reinvestment distinctions

Assets backing annuities may include government securities, investment-grade credit, mortgages, private debt and other investments. Their contractual yields compensate for credit, liquidity and other exposures in varying proportions. A high observed asset yield is a source of potential investment earnings and associated investment risk; it does not, by itself, justify valuing a fixed promised annuity at that yield under every accounting or solvency regime.

A fall in rates can raise the present value of both fixed-income assets and fixed nominal liabilities. Whether the insurer's surplus rises or falls depends on relative exposure. If asset duration is shorter than liability duration with similar baseline value, a rate decrease generally raises liability value more than asset value and weakens economic surplus. Conversely, rising rates may improve this simplified surplus while simultaneously creating liquidity challenges, unrealized losses on assets sold early, or other accounting effects. These outcomes are compatible because economic value, immediate cash availability and realized accounting earnings are different dimensions.

Reinvestment risk matters when assets mature or cash is received before the expected annuity outgo. Credit spread widening can reduce bond prices independently of the reference risk-free curve; where an adjustment to liability discount rates is permitted, it must be computed under the applicable rules rather than presumed to offset all spread losses. The Solvency II matching-adjustment provisions illustrate the institutional conditions imposed on such liability valuation treatment.

Prudential, accounting and management views

IFRS 17 reports insurance contract groups under an accounting framework intended for financial statement users. Solvency II evaluates insurance obligations and prudential capital within a supervisory framework. U.S. statutory valuation follows state-adopted manuals and other applicable rules. Internal economic management may add hedging sensitivities and stressed surplus metrics that are not identical to any reported statutory amount.

The distinction is substantive. Under IFRS 17's current measurement approach, changing market rates can affect fulfilment cash-flow measurement and insurance finance results; the precise statement presentation depends on the standard and the entity's elections. Under Solvency II's best-estimate rule, the relevant currency-specific supervisory term structure directly informs discounted projected payments. Under the NAIC's valuation-manual framework, prescribed methodologies and scenario reserves can determine statutory obligations. A calculation called "liability duration" without specifying which of these quantities it differentiates may conflate values with different margins or risk-neutral versus real-world assumptions.

For meaningful comparison the analyst should provide a reconciliation table identifying the contract set, gross or net reinsurance position, mortality and longevity assumptions, discount curve origin and adjustments, cash-flow horizon, nonfinancial risk provisions, capital treatment, and treatment of current versus locked-in rates. Differences remaining after reconciliation can then be assigned to known components rather than attributed generically to "regulation."

Product design, policyholders and cross-border comparability

An annuity buyer's immediate-income quote responds to market conditions and insurer pricing practices. Higher available investment yields can support higher newly priced contractual payouts in some circumstances, conditional on expenses, competition, solvency requirements and product design. Existing fixed nominal annuitants do not generally receive an automatic increase in contractually guaranteed payments when rates rise. New-business pricing and valuation of in-force promises are separate analyses.

The welfare implications also depend on the individual household. Davidoff, Brown and Diamond's study of incomplete-market annuitization provides grounds to consider liquidity needs and consumption risk alongside lifetime income security. Contract options and mortality selection affect both pricing and who purchases annuities. Finkelstein and Poterba's UK contract-selection evidence supports this specification concern. It does not establish quantitative interest elasticity of all annuity buyers in other jurisdictions.

International comparisons must distinguish current capital market conditions, regulatory valuation rules, observed annuitant mortality, inflation indexing and product options. Comparing currency amounts without currency and purchasing-power normalization can mislead; comparing percentage sensitivities without reconciling discount-basis conventions and benefit terms can likewise mislead. A valid cross-country analysis preserves the legal and contractual source dates.

Stress management and governance

A robust annuity interest-rate risk program considers parallel upward and downward changes, nonparallel twists, local long-end shocks, extrapolation parameter changes, credit spread shocks, collateral and liquidity stress, and mortality improvement scenarios. It evaluates both stand-alone and combined stresses, including scenarios in which falling yields accompany worsening longevity assumptions or changes in policyholder behavior.

Governance should specify data owner, curve construction owner, mortality assumption approval, valuation code version, shock definitions and change control. Each reported duration must be traceable to a dated cash-flow dataset and curve input. Independent calculation, including full-revaluation checks against local derivatives, is especially important when extrapolation or embedded guarantees are material. The IAA's actuarial standards register and IAIS supervisory framework supply international professional and supervisory context, while detailed operational requirements remain dependent on each jurisdiction and institution.

Scientific Validity and Integrated Findings

Evidence hierarchy and reproducibility

This research differentiates four evidence classes: original legal and regulatory text; official technical and standard-setting publications; original peer-reviewed research; and constructed mathematical demonstrations. Each establishes different kinds of knowledge. A regulation establishes the applicable written requirement in its defined jurisdiction and version, not an actual insurer's implementation. An original economic article establishes the work's documented research setting and reported findings, not universal empirical validity. A synthetic calculation establishes the result of explicitly supplied formulas, not actual mortality or trading behavior.

The sources have been consulted at the level available through official HTML pages and publisher descriptions. Access to the full text of some historical journal articles is restricted; historical claims are accordingly confined to bibliographically confirmed topics and accessible summaries. No insurer contract microdata, audited statutory reserve files or individual annuitant records were accessed. This prevents empirical estimation of population-specific rate elasticities or actual hedge effectiveness and is explicitly part of the evidence boundary.

The numerical case is reproducible from its stated formulas and constants. The interpretation follows from independent derivatives and direct summations. Source-specific links are embedded in the relevant scholarly claims; all cited external documents are listed with full bibliographic identity in the References and recorded with precise claim checks in the research evidence ledger.

Identification, calibration and uncertainty

When an observed insurer liability changes between reporting dates, changes in the curve are only one possible cause. Contract runoff, new business, deaths, benefit changes, revised mortality, expenses, reinsurance and valuation basis changes may all affect the number. An attribution exercise first aligns the contract population and uses a controlled roll-forward. It then reprices old cash flows with the new curve, changes survival and expense assumptions one by one or through a documented multivariate attribution, and separately measures interactions. Without that sequence a reported liability movement does not identify an interest-rate effect.

A market-based curve is sensitive to which bonds or swaps were observed, whether raw input quotes were stale, how credit and liquidity features were handled, the interpolation convention, and how the extrapolated tail was formed. The U.S. Treasury's par-curve methodology explanation exemplifies why identifying the published yield type is essential. EIOPA's term-structure technical records demonstrate the relevance of versioned parameters. Annuity liability models that extend substantially beyond observable assets should disclose sensitivity to the ultimate-forward-rate and extrapolation methodology in addition to conventional yield shocks.

Mortality basis uncertainty adds a second, structurally distinct set of parameters. It affects the expected number of late payments and the weight of the curve tail. The interaction can be analyzed with cross partial derivatives or paired scenarios. A rate-only duration should always be accompanied by sufficient description of the survival basis to make its financial meaning reproducible.

Model validation and sensitivity governance

A complete internal validation program checks contractual eligibility, payment dates, mortality exposure and aggregation; reconciles curve input dates, currencies, compounding and instrument types; independently reproduces discount factors and present values; tests parallel and nonparallel shocks; compares bump-based DV01 with analytical derivatives; reconciles bucket DV01 with aggregate DV01 under the specified basis; and checks extrapolation, benefit indexing and embedded option revaluation.

Several elementary tests catch consequential mistakes. A fixed positive-cash-flow liability should decrease when every zero rate increases; the baseline duration should be a present-value-weighted payment-time average; duration contributions from disjoint time buckets should sum to total duration; and a second-order approximation should approach exact revaluation as the bump shrinks. If an implementation violates one of these, the source of failure may be an input or convention error, a changed cash-flow assumption, a code defect or an inappropriate application of a fixed-cash-flow identity.

The control process also tests scenario independence. Mortality and behavior responses should be held fixed where a partial interest-rate derivative is sought, but recalculated where a full economic stress is intended. Financial guarantees should be simulated under consistent risk-neutral or real-world assumptions for the specific measure. Results should identify whether they are gross or net of reinsurance and whether a capital or risk margin is included. A validated calculation's scope must be clearly marked so that a useful partial indicator is not mistaken for total enterprise solvency.

Integrated synthesis and research agenda

The literature and regulatory comparison support an integrated analytical conclusion: annuity discount-curve risk has three linked dimensions, namely the definition of the payment obligation, the construction of the valuation curve, and the response of the combined asset-liability position under relevant states. This is a descriptive synthesis of established actuarial and financial constructs, not a newly named theory. None of the dimensions is reducible to a single model output without an explicit assumption. Mortality-weighted payment dates determine the liability's underlying maturity profile; the selected curve weights those dates; accounting and solvency rules determine which reported quantity is affected; and assets and capital determine the insurer-level consequence.

Two apparently conflicting findings can therefore both be correct. A declining interest-rate curve can increase the economic value of a fixed in-force annuity liability, yet the insurer may also experience asset gains that partly or wholly offset that liability change. Likewise, a higher current market rate can support a higher payout quote for new annuity sales while reducing the present value of an existing fixed promise. These are different time, contract and balance-sheet comparisons.

Further empirical research should use matched contract-level cash flows, mortality histories, dated regulatory curves and underlying assets. A comparative panel could analyze how bucket interest sensitivities vary across deferred, immediate, guaranteed-period and index-linked annuities; another design could estimate how observed insurer practices differ from a common valuation model. Reliable empirical inference requires access to actual portfolio data, legal effective dates and identifiable changes in model assumptions. Such projects would extend the present synthesis through evidence rather than through speculative doctrinal innovation.

Conclusion

Annuity obligations are intrinsically long-horizon and contingent. Their discount-rate sensitivity is determined by the timing and amount of enforceable future payments, the survival probabilities attached to those payments, and the exact discount factors applied to each date. A rate shock can change a reported liability without changing the promise itself, while changing longevity assumptions can change both the value and its interest sensitivity.

The monograph's independently derived equations establish that the duration of a positive fixed-cash-flow annuity is its present-value-weighted expected payment time under the selected basis, convexity captures second-order curvature, and key-rate sensitivities express the maturity structure that an aggregate duration conceals. The constructed example confirms these properties numerically: a synthetic annual-pay life annuity has a duration near 9.88 years at the stated 3% curve; a 100-basis-point increase lowers its measured liability by about 9.16%, and a 100-basis-point decrease raises it by about 10.71%. Nonparallel shocks produce different outcomes and may reverse the sign of the value change.

The international source review establishes the importance of valuation purpose. IFRS 17, Solvency II and U.S. statutory reserving incorporate different rules for insurance obligations and their discount or scenario bases. Within each system, cash-flow consistency, observable data, risk treatment, maturity extrapolation, source documentation and independent model validation determine the credibility of reported sensitivity.

The principal practical consequence is methodological: annuity interest-rate risk must be evaluated with an explicitly defined cash-flow perimeter, term structure, survival basis, accounting or prudential measure, and asset-liability context. This allows rate-driven changes, demographic changes and model-basis changes to be analyzed with traceable attribution. The findings are an original critical synthesis and reproducible mathematical demonstration of established methods; no separate authorial theoretical system or empirically unverified discovery is asserted.

References