Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Discounting and Yield Curves in Actuarial Valuation

Abstract. Discounting determines the present measurement of insurance obligations whose cash flows may extend across decades, depend on survival and policyholder behavior, and respond to both financial and nonfinancial contingencies. A yield curve assigns the time value of money by maturity; its construction consequently changes reported obligations, solvency positions, asset–liability sensitivities, and the apparent economics of long-duration promises. This monograph establishes a rigorous mathematical account of discount factors, spot and forward rates, par yields, time conventions, and the valuation of contingent insurance cash flows. It evaluates market-data selection, bootstrapping, interpolation, extrapolation, nominal–real consistency, contract-specific liquidity, credit and currency effects, dynamic scenarios, and key-rate risk. Its documentary research examines IFRS 17, the Solvency II framework and the announced 2027 methodological transition, official European and United States yield-curve methodologies, the U.S. statutory valuation framework, international supervisory capital standards, and foundational econometric studies of term-structure estimation. Reproducible illustrative calculations demonstrate how seemingly modest curve perturbations can materially change distant liabilities and why a par Treasury quotation cannot be inserted into a zero-coupon present-value calculation without a convention-preserving transformation. The integrated analysis identifies five practical determinants of a defensible actuarial discount curve: the valuation purpose, the contingent cash-flow profile, the evidence and conventions used to construct the term structure, the rule for maturities lacking liquid observations, and the reporting of uncertainty under specified stresses. The contribution is an evidence-based synthesis and a set of testable interpretations of existing actuarial and financial methods, without proposing a new named theory, proprietary curve model, or universal valuation rate.

Keywords: Actuarial valuation; discount factors; time value of money; yield curve; zero-coupon rates; forward rates; nominal cash flows; real cash flows; inflation; term-structure estimation; bootstrapping; Smith–Wilson; interpolation; extrapolation; ultimate forward rate; liquidity adjustment; IFRS 17; Solvency II; insurance technical provisions; annuity valuation; duration; convexity; key-rate duration; asset–liability management; stochastic valuation; regulatory comparability.

Introduction

The actuarial problem of discounting starts with an obligation to pay an amount at a future date, frequently conditional on survival, illness, loss occurrence, surrender, economic indices, or contractually defined events. The present measurement of such an obligation combines a projection of conditional cash flows with a price for their timing and financial characteristics. Term structures therefore enter at the same conceptual level as the distribution of the insurance obligation itself. A valuation that models expected claims carefully but discounts them on an inconsistent basis can misstate both its present value and its exposure to changing markets.

The fundamental object of this dossier is the discount factor D(0,t)D(0,t), a conversion of a future unit of monetary value at maturity tt into a present amount as of valuation date 00, under a specified valuation basis. A discount curve is the family {D(0,t):t≥0}\{D(0,t):t\geq 0\} and a yield curve is a transformation of that family under a declared compounding convention. What constitutes an appropriate valuation basis depends upon the relevant financial reporting standard, prudential regime, pricing or transaction question, currency, and characteristics of the cash flows. The IFRS 17 discount-rate requirements make cash-flow characteristics, liquidity, and current observable market consistency explicit. The European Solvency II Article 77 establishes an expected-present-value best estimate using the relevant risk-free term structure and distinguishes that best estimate from the risk margin.

Financial quotations require interpretation before use. The U.S. Treasury describes its official curve as a par yield curve, while the European Central Bank publishes estimated zero-coupon, forward, and par curves. These three curve types are related but answer different questions. A zero rate prices one maturity-specific payment; a par rate solves the coupon or annuity rate that balances a coupon-bearing instrument's price; a forward rate is the term rate implied by discount factors for two dates. Their conversion demands an explicit day-count basis, compounding definition, cash-flow schedule, and choice of interpolation.

The research questions are precise. How does a term structure enter expected and path-dependent insurance cash flows? Which market observations identify the curve, and which mathematical choices fill the gaps? When is a risk-free base curve modified to reflect liability liquidity or contractual features? How do extrapolation assumptions dominate valuation beyond liquid horizons? What stress and disclosure practices make a valuation interpretable across accounting and supervisory systems?

The method combines analytical derivation, source-critical comparison, and reproducible synthetic calculations. Primary documentary sources include the IFRS Foundation and the Australian Accounting Standards Board's openly accessible IFRS-equivalent text, European Union rules and EIOPA publications, the ECB and U.S. Treasury yield-curve methodologies, the National Association of Insurance Commissioners' valuation framework, and the International Association of Insurance Supervisors' global standard-setting material. Scholarly comparison draws upon Nelson and Siegel (1987), Svensson (1994), Hagan and West (2006), Diebold and Li (2006), and the Bank for International Settlements' central-bank survey (2005). Secondary literature is used for its stated models or methodological contributions rather than as proof of an individual insurer's actuarial results. All numerical figures introduced below are hypothetical unless an official source is explicitly identified. No market observations, insurance portfolios, or stochastic simulation results are invented.

A necessary temporal distinction governs the European discussion. On the research cutoff date, November 14, 2026, EIOPA had already published a new RFR Technical Documentation, but that document explicitly states that its updated approach becomes applicable on January 30, 2027. The Solvency II review timetable confirms the same changeover. Thus current valuation and forthcoming methodology are examined separately. The global comparisons are equally conditional: an international supervisory standard can be adopted at the international level without its every computational choice becoming immediately binding in every national jurisdiction.

Interest Rate and Time Value

The valuation question

A certain cash payment of CC at a fixed future date can be valued through a suitable discount factor: PV0=CD(0,t)PV_0=C D(0,t). Insurance contracts require extending that familiar identity to uncertainty about the amount, timing, or even existence of payments. Their economic characteristics include finite or indefinite run-off, inflation indexing, deferred benefit onset, surrender options, guarantees, premium streams, and correlations between financial markets and insured events. The valuation objective determines how these properties enter the projected cash flows, discount basis, and supplementary risk adjustments.

The common expression “discount rate” can conceal distinct quantities: an internal investment return assumed for planning; a current market-implied zero rate; a regulatory risk-free term structure; a contract-specific rate reflecting liquidity; or an average equivalent yield over a particular cash-flow distribution. These quantities sometimes coincide by construction, but they are not automatically interchangeable. Discounting projected liability payments at an expected return on a risky asset portfolio, for instance, may import investment risk compensation into a liability whose future cash flows do not depend on that portfolio. Such an adjustment demands a contractual and measurement-basis justification rather than a reference to the insurer's investment strategy.

A fully specified actuarial discount calculation records at least the valuation timestamp, the cash-flow currency, exact payment timing, projection convention, term structure, interpolation and extrapolation rules, compounding and day-count convention, and the applicable accounting or prudential adjustments. This information is essential for reproduction. It also separates market measurement from insurer management assumptions, which may be relevant to scenario-based asset adequacy but govern a different question.

Mathematical foundations and conventions

For a continuously compounded annualized zero rate z(t)z(t) and time in year fractions tt, define

D(0,t)=exp⁡[−tz(t)],z(t)=−ln⁡D(0,t)t,t>0.D(0,t)=\exp[-t z(t)],\qquad z(t)=-\frac{\ln D(0,t)}{t},\quad t>0.

For an annually effective rate i(t)i(t) defined as the single-maturity accumulation equivalent, D(0,t)=(1+i(t))−tD(0,t)=(1+i(t))^{-t}. For an annual nominal rate j(m)(t)j^{(m)}(t) compounded mm times per year, D(0,t)=(1+j(m)(t)/m)−mtD(0,t)=(1+j^{(m)}(t)/m)^{-mt}, provided the quotation is applicable under that convention. When DD is identical, these rate representations encode the same discounting economics. Numerically reusing a quoted percentage under another convention produces a different discount factor. For irregular payment times, the definition of tt requires a day-count convention and treatment of dates, coupons, settlement, and leap years.

A curve of continuously compounded spot rates z(t)z(t) determines the instantaneous forward rate when differentiable:

f(0,t)=−∂∂tln⁡D(0,t)=z(t)+tz′(t).f(0,t)=-\frac{\partial}{\partial t}\ln D(0,t)=z(t)+t z'(t).

Conversely, D(0,t)=exp⁡{−∫0tf(0,u) du}D(0,t)=\exp\{-\int_0^t f(0,u)\,du\}. The corresponding simple forward rate between dates t1<t2t_1<t_2 over accrual period δ\delta is F(0;t1,t2)=[D(0,t1)/D(0,t2)−1]/δF(0;t_1,t_2)=[D(0,t_1)/D(0,t_2)-1]/\delta. Each equation assumes an internally consistent measurement of maturity and year fractions. A negative forward rate is mathematically possible; it should be identified as a model or market-implied rate rather than incorrectly treated as a computational failure. Conversely, accidental oscillations or discontinuities that are inconsistent with input instruments warrant investigation.

The U.S. Treasury interest-rate FAQ explains that its constant-maturity par yields are bond-equivalent quotations based on semiannual coupon conventions. Treating a published 20-year par yield as a continuously compounded 20-year zero yield introduces a convention mismatch even before considering market or liquidity adjustments.

Insurance-contingent cash flows

Let XtX_t denote the net cash flow from a liability at time tt, positive for an outflow and negative for an inflow, under a stated projection basis. When discount factors are deterministic under that basis and the relevant valuation definition permits separation, a discrete actuarial present value takes the form

V0=∑k=1nD(0,tk)E[Xtk].V_0=\sum_{k=1}^n D(0,t_k)\mathbb{E}[X_{t_k}].

For a continuous payment stream with expected density m(t)m(t), V0=∫0TD(0,t)m(t) dtV_0=\int_0^T D(0,t)m(t)\,dt. For a life annuity paying one monetary unit annually at t=1,…,nt=1,\ldots,n conditional on survival, a simplified value is ∑t=1ntpxD(0,t)\sum_{t=1}^n {}_tp_x D(0,t), where tpx{}_tp_x is the survival probability for a life aged xx. The appropriate contract model may additionally require expense, lapse, benefit-escalation, premium, and guarantee projections. This expression does not imply that mortality and financial variables must be independent in every actuarial problem: it states the assumptions of a particular deterministic-curve calculation.

Where interest rates and contingent payments are jointly stochastic, the pricing or economic valuation object can instead be written V0=EQ[∑kM0,tkXtk]V_0=\mathbb{E}^{\mathbb{Q}}[\sum_k M_{0,t_k}X_{t_k}] for a suitable pricing measure and stochastic discounting kernel, with exactly which measure and risk adjustments are appropriate depending upon the claimed valuation basis. The valuation of a guarantee activated during low interest-rate or poor market-return states generally cannot be reduced to a single deterministic average payment discounted at an arbitrary average rate. In particular, E[DX]\mathbb{E}[D X] equals E[D]E[X]\mathbb{E}[D]\mathbb{E}[X] only if their covariance is zero under the chosen probability law. The difference equals Cov⁡(D,X)\operatorname{Cov}(D,X). A cash-flow model that ignores this term implicitly assumes away a potentially material interaction.

Financial and nonfinancial uncertainty also need separate accounting treatment. Under IFRS 17 paragraph 37, the risk adjustment for nonfinancial risk captures the entity's compensation for bearing uncertainty about the amount and timing of cash flows from nonfinancial risk. Its purpose differs from the time-value adjustment under paragraph 36. An insurance projection that embeds an extra prudence margin within expected cash flows and then applies a second equivalent margin in the discount rate or risk adjustment may double count the same risk.

Economic meaning of alternative discount assumptions

A simple comparison illuminates the conceptual problem. One payment of 1,000 in 20 years has a value of approximately 548.81 under a constant 3% continuously compounded zero rate and 449.33 under 4%. The one-percentage-point difference changes the present value by about 18.1% relative to the 3% valuation. This arithmetic is agnostic about whether either rate is an appropriate insurance liability rate; its significance is that long-duration cash flows amplify the effect of rate selection.

A reporting basis may prescribe the use of a prescribed term structure, while a pricing exercise may reflect capital costs, expenses, risk appetite, and market competition in a separate profit or risk loading. An asset-adequacy exercise considers projected investments, reinvestment paths, and adverse scenarios; the U.S. ASOP No. 22 guidance addresses the relationship between reserves, assets, and scenario testing. These distinctions make the valuation purpose an indispensable input to selecting a discount basis.

Discount Factors

Discount-factor mathematics

The sequence of discount factors is the computational foundation of a curve. Each dated liability cash flow should be multiplied by a factor applicable to its currency, maturity, and valuation basis. When projected cash flows are tabulated monthly, quarterly, or annually, payment dates must be retained rather than silently mapped to a coarse integer maturity. Claim settlement delays, monthly annuity installments, and premium receipts near reporting dates are examples of where apparently small timing approximations become systematic across large portfolios.

Suppose a one-year zero-coupon claim pays 100, a five-year claim pays 200, and a twenty-year claim pays 500. The present-value contributions under illustrative continuously compounded zero rates of 2%, 3%, and 4%, respectively, are $100e^{-0.02}, \200e^{-0.15}, and \500e^{-0.8}$. The sum differs from discounting the total nominal 800 at a single arithmetic average of the three zero rates. A weighted equivalent single rate can always be solved for a fixed cash-flow set when the relevant monotonicity conditions hold, but that equivalent rate is specific to the cash-flow distribution; it cannot replace the underlying term structure for liability changes or key-rate risk management.

The arbitrage-free pricing identities commonly used in fixed-income calculations presuppose appropriate traded instruments, collateral, funding, credit, liquidity and other conditions. Their financial-instrument setting must not be casually transplanted into insurance reporting as if any traded security were an exact liability replicator. A government-bond curve can be an input to market observation without automatically representing a universally risk-free insurance discount curve. The contractual features and measurement rules determine which market factors should be retained.

Bootstrapping from market instruments

Bootstrapping infers maturity-specific discount factors from observable prices and known contractual payments. For a hypothetical risk-free annual-pay bond priced at par, with annual coupon rate cc and maturity nn, the pricing identity is

1=c∑j=1n−1D(0,j)+(1+c)D(0,n).1=c\sum_{j=1}^{n-1}D(0,j)+(1+c)D(0,n).

If shorter discount factors are already identified, then D(0,n)=[1−c∑j=1n−1D(0,j)]/(1+c)D(0,n)=[1-c\sum_{j=1}^{n-1}D(0,j)]/(1+c). This familiar recursive formula applies only to the stated coupon schedule, price normalization, and clean/dirty price convention. In live markets, coupon dates may be irregular and the unknown factors may require interpolation within the bootstrap. Observations for swaps, government securities, overnight indexed swaps, and inflation-linked bonds also have distinct cash-flow and credit conventions, which must be translated before pooling them.

The official Treasury yield-curve methodology reports indicative, bid-side quotations for specified on-the-run government securities, a bootstrap of instantaneous forward rates at input maturities, and monotone-convex interpolation between those points. The Treasury explicitly does not publish the full computer formulation of its curve derivation program. A practitioner can reproduce the general technique, but exact numerical equality with the official series must not be asserted from the published high-level description alone.

The European Central Bank similarly documents an estimation pipeline: selecting bonds, applying outlier controls, and fitting euro-area curves, including sovereign composite spot, forward, and par estimates. Its yield-curve methodology describes a Svensson specification; the yield curves cover selected fixed-coupon and zero-coupon government instruments in a defined maturity interval. The resulting curves are estimated official data products, not a raw, complete cross-section of individual transactions. Curve identification also depends on the retained population of instruments.

Data quality and curve calibration

Data input defects propagate nonlinearly into long-horizon values. Stale prices, inconsistent settlement calendars, bid–ask conventions, thin trading, benchmark switches, unexpected coupon features, and interpolation across gaps can produce discontinuous or economically implausible forwards. A robust process records each quote's observation date, source, instrument type, maturity, price/yield side, accrued-interest convention, exclusion rule, and fallback hierarchy. When multiple curves serve different currencies or collateral conventions, cross-currency conversions and basis relationships must be separately documented.

A curve-fitting objective might minimize weighted squared price errors ∑iwi(Pimodel−Piobserved)2\sum_i w_i(P_i^{model}-P_i^{observed})^2 or errors in quoted yields. These objectives are not equivalent. Price errors overweight or underweight some maturities differently from yield errors depending on instrument duration and coupon structure. A term structure fitted to low-duration bills can appear excellent under an unweighted price metric while misrepresenting a very long-dated insurance obligation. Suitable weighting, realistic market bid–ask tolerances, and out-of-sample checks should be chosen in relation to the measurement question.

The BIS survey of central-bank curves documents diversity in estimation techniques, data coverage, and reported spot and forward representations. It supplies an important methodological lesson: a label such as “government yield curve” or “zero-coupon curve” does not by itself identify a uniform data-generating procedure across countries or years. For cross-jurisdiction empirical studies, a change in official fitting methodology can be mistaken for a market shift unless the series' construction history is controlled.

Probability, dependence, and optionality

A deterministic discount curve can be appropriate for expected fixed benefits where the selected reporting basis permits that simplification. It becomes inadequate as a complete market-consistent description of an embedded option whose exercise depends on the discount curve itself. Consider a guaranteed annuity conversion rate exercised preferentially when market annuity rates have fallen. In lower-rate states, the guarantee becomes more valuable, and future option-related outflows can rise while discount factors also rise. The product of state-dependent payments and discount factors should be evaluated under a model consistent with the instrument and reporting context.

The conceptual test is whether projected cash flows are independent of financial outcomes in a manner sufficient for the chosen valuation approximation, or whether a recalculation is needed in each market scenario. Surrender, premium persistency, guaranteed minimum benefits, and dynamic policyholder decisions may couple insurance behavior to financial trajectories. The calibration target may involve option prices, yield-curve instruments, or other market observations. Mathematical consistency and representativeness matter more than nominal model complexity. Where a model is simplified, scenario-bound approximation error can be described rather than silently transferred into an unexplained discount-rate margin.

Term Structures and Forward Rates

Spot, forward, and par curves

A term structure translates maturity into the discounting terms applicable to a future payment. The spot curve z(t)z(t) gives the rate for a single payment at tt under its quoted convention. The forward curve describes the rate embedded between future dates. A par curve assigns the fixed coupon which, with a specified schedule and floating or fixed principal, prices a benchmark security at par. Distinct curve representations encode common underlying discount factors only after their cash-flow definitions and market bases have been reconciled.

For a hypothetical annual-pay par bond, the coupon cnc_n at maturity nn satisfies

cn=1−D(0,n)∑j=1nD(0,j).c_n=\frac{1-D(0,n)}{\sum_{j=1}^n D(0,j)}.

The denominator is the present value of the unit annual coupon stream. This identity makes the dependence of a par yield on the entire preceding discount curve explicit. Hence one cannot use an observed par rate c20c_{20} as the 20-year spot rate z(20)z(20) merely because both are quoted in percent per annum. The difference can be significant when the curve is steep, humped, or inverted, and it becomes increasingly material as the maturity and number of coupon dates increase.

The forward curve is likewise a transformed discount function, not an unqualified forecast of the future short rate. Its interpretation as an expectation of future short rates depends on assumptions about term premia, risk aversion, market segmentation, and the relevant pricing measure. The U.S. Treasury expressly advises that its par yields are not forecasts of future market rates. Svensson's forward-rate analysis of Sweden discusses the use of the forward curve as a monetary-policy indicator and illustrates why separating time horizons can be informative, while recognizing that such an interpretation is model-dependent.

For actuarial purposes, forward rates perform three roles. They make mathematical coherence of a fitted discount curve visible; they support construction of forward-starting discount relations where cash flows are projected between dates; and they help represent the interest-rate exposure of liability tails. They need not be used as deterministic predictions of reinvestment returns. A projected scenario for insurer asset earnings, by contrast, may require explicit forward dynamics, credit assumptions, and the evolution of spreads. The valuation basis determines whether risk-neutral, real-world, regulatory, or other distributions are appropriate for a given component.

Market-curve estimation

The classic Nelson–Siegel specification expresses the zero-coupon curve as a parsimonious combination of a long-run level, a decaying slope component, and a hump-shaped curvature component. One common representation, for τ>0\tau>0, is

z(t)=β0+β11−e−t/τt/τ+β2(1−e−t/τt/τ−e−t/τ).z(t)=\beta_0+\beta_1\frac{1-e^{-t/\tau}}{t/\tau} +\beta_2\left(\frac{1-e^{-t/\tau}}{t/\tau}-e^{-t/\tau}\right).

Its appeal is interpretability and its ability to fit a range of common yield-curve shapes with few parameters. Nelson and Siegel (1987) tested the method on U.S. Treasury bill yields in 1981–1983; their historical fit results do not constitute a guarantee of adequate extrapolation for modern insurance liabilities. The assumed functional form influences distant maturities even when short-end fit is satisfactory.

Svensson (1994) extended the family with an additional curvature term, enabling a more flexible shape for forward and spot curves. The ECB's published curve method uses the Svensson approach, making it a relevant example of an official central-bank estimation method. Diebold and Li (2006) subsequently analyzed a dynamic factor interpretation of yield-curve level, slope, and curvature in their term-structure forecasting study. This forecasting model answers a different question from discounting a set of market-consistent contingent insurance liabilities as of one valuation date. A forecasting improvement does not itself define the permitted valuation rate.

Parametric models trade flexibility for stability, interpretability, and parsimony. Local splines and direct discount-factor interpolation may fit market quotes more tightly but can create irregular forward rates between sparse pillars. Models fitted to rates directly can also behave differently from models fitted to prices or discount factors. The estimation domain, constraint set, loss function, and boundary conditions require explicit specification, because two curves can match the same observed prices within bid–ask spreads yet produce different present values of distant claims.

Interpolation and no-arbitrage checks

Interpolation defines the curve between observed or calibrated maturities; extrapolation extends it beyond them. Their mathematics are related but their evidential status differs. Between deeply traded pillars, interpolation uses nearby market information. Beyond the liquid horizon, the result increasingly depends on a structural rule or regulatory parameter. Confusing these regimes overstates the precision of long-duration liability measurements.

Linear interpolation in continuously compounded zero rates z(t)z(t), linear interpolation in log discount factors ln⁡D(0,t)\ln D(0,t), and linear interpolation in instantaneous forwards produce different intermediate curves. Linear log-discount interpolation between two pillars produces a constant forward rate within that interval; linear zero-rate interpolation generally does not. Polynomial splines can improve smoothness but require careful constraints to control overshooting and derivative behavior. A differentiable fitted curve can still contain economically implausible forward shapes.

Hagan and West's interpolation-method comparison studies practical yield-curve algorithms, including monotone-convex interpolation and their implications for hedging. It demonstrates why interpolation should be evaluated together with instrument bootstrapping, rather than chosen as a purely cosmetic step after the curve has already been constructed. For insurance liabilities, this coupling affects key-rate hedges because a localized shock to one market pillar is transmitted to the liability through the interpolation basis and can change adjacent forward-rate segments.

A useful validation suite checks whether the curve reprices calibration instruments within documented tolerance, whether discount factors and forward rates are mathematically coherent, whether negative rates (if present) follow from an accepted market representation rather than numerical artifacts, whether currency and collateral conventions agree, and whether small quote perturbations produce proportionate valuation changes. Monotonic discount factors are often intuitive under strictly nonnegative instantaneous forward rates; they are not a universal requirement in environments admitting negative market rates. Thus QA should test mathematical and economic consistency against the permitted rate domain instead of imposing an historically contingent positivity assumption.

International valuation regimes

The same insurance cash-flow projection can lead to distinct recognized measurements under financial reporting, prudential supervision, statutory reserve calculation, pricing, and economic capital evaluation. This is not an error if the purposes, discount bases, and risk margins are separately disclosed. It becomes a comparability problem when two values are presented without identifying their measurement regimes.

Under IFRS 17, paragraph 36 and application guidance direct entities to use discount rates that reflect time value, relevant financial risk, characteristics of cash flows, and contract liquidity, while remaining consistent with observable market prices for comparable financial instruments. The standard accommodates a bottom-up construction from an adjusted liquid risk-free curve and a top-down construction from a reference-asset portfolio after removing irrelevant factors; its published IFRS text describes these possibilities in paragraphs B80–B82. Neither approach means that an arbitrary portfolio yield can be imported without adjustments. The entity must also avoid capturing the same financial risk in both cash flows and discount rates.

For Solvency II, Article 77 best-estimate valuation uses the relevant risk-free interest-rate term structure for probability-weighted projected cash flows, and the framework separately recognizes a risk margin. Article 77a extrapolation distinguishes maturities with deep, liquid, transparent markets from the extrapolated region and provides for convergence toward an ultimate forward rate. Where permitted and approved, the matching adjustment mechanism modifies the relevant liability discount basis for eligible matched portfolios under specified conditions. A market-consistent supervisory best estimate, a financial-reporting fulfillment cash-flow measure, and an economic valuation for an insurer's own decisions remain distinct constructs.

In the United States, reserve and cash-flow scenario requirements depend on insurance product, applicable state law, and the adopted NAIC Valuation Manual. The 2026 edition contains product-specific principle-based requirements and scenario machinery, including treatment of interest-rate and investment return paths; the 2026 Valuation Manual is a document to consult in the context of statutory scope, not a single prescribed continuously compounded risk-free discount curve for every insurance contract. The Actuarial Standards Board's ASOP No. 22 asset-adequacy guidance similarly makes the asset and liability cash-flow interaction relevant to specified appointed-actuary opinions. It does not erase differences between cash-flow testing and market-consistent financial reporting.

The International Association of Insurance Supervisors adopted the Insurance Capital Standard in December 2024 as the quantitative component of its Common Framework for internationally active insurance groups. The standard advances supervisory comparability, while actual application and detailed jurisdictional implementation must be assessed within their respective supervisory systems. It should not be cited as evidence that all insurers worldwide use one identical yield curve or one universal discounting methodology.

This comparison produces a substantive methodological conclusion. A defensible actuarial rate is not determined solely by the financial instrument used for curve calibration. Its defensibility rests on the correspondence between the instrument's observed characteristics, the obligation's relevant financial characteristics, and the valuation rule that governs their connection. That three-way correspondence is the natural explanatory unit of international actuarial discounting.

Nominal and Real Cash Flows

Purchasing power and currency consistency

Nominal amounts are expressed in monetary units at their payment dates; real amounts are stated in units of purchasing power at a specified price-base date. A nominal cash-flow projection should be discounted with a nominal term structure consistent with its currency and financial-risk content. A real projection should be discounted with a correspondingly defined real basis. Mixing a cash-flow inflation escalation with a real discount rate without converting one side will generally double count the purchasing-power effect.

Under a deterministic constant annual inflation assumption π\pi and nominal effective annual yield ini_n, the exact Fisher-consistent effective real rate is

1+ir=1+in1+π.1+i_r=\frac{1+i_n}{1+\pi}.

The commonly used approximation ir≈in−πi_r\approx i_n-\pi omits the compound interaction term. As a numerical illustration, consider a future real benefit of 100 units of today's purchasing power payable after 20 years, nominal inflation of 2% per year, and a nominal discount rate of 4% per year, both annually effective. Its projected nominal payment is $100(1.02)^{20};discountingat4; discounting at 4% yields \100(1.02/1.04)^{20}\approx67.82.Alternatively,theexactrealrateis. Alternatively, the exact real rate is(1.04/1.02)-1\approx1.9608%,anddiscountingtheconstantrealbenefitatthatratealsoyields, and discounting the constant real benefit at that rate also yields67.82$. This equality is a consistency check on definitions, not a claim that actual inflation is deterministic or a real yield curve can be deduced from a single assumed inflation number in all market conditions.

In capital markets, a nominal–real comparison may embed inflation risk premia, market liquidity differences, indexation lags, and other instrument-specific effects. An inflation-linked government bond is not automatically a pure estimate of expected future inflation when compared with a conventional bond. The U.S. Treasury publishes distinct nominal and real yield series, and differences between them require interpretation rather than immediate substitution as an actuarial inflation forecast. A projection process should also specify the inflation index, lags, caps, floors, and escalation dates actually relevant to policy benefits or expenses.

Inflation-sensitive insurance obligations

Long-term disability payments, indexed pensions, healthcare costs, legal settlement benefits, workers' compensation, and some insurance expenses illustrate how inflation enters actuarial valuation. Their indexation can follow a general consumer index, medical cost trends, wages, fixed contractual escalation, or discretionary changes. A policy with a guaranteed 2% increase is not equivalent to one paying actual inflation, and neither is automatically equivalent to an unindexed benefit in a currency exposed to inflation.

The actuarial projection should first determine what changes the contractual cash amount and when. Medical cost inflation can exceed general price inflation; wage-related benefit replacement can respond to wage growth; nominal coverage limits can cap escalation. Some inflation effects are correlated with interest rates or with claim behavior, complicating the deterministic decomposition of amount and discounting. Where possible, financial options on inflation or contractually indexed payments should be valued using a coherent model for index levels and discount factors, supplemented by clearly distinguished nonfinancial assumptions.

The ASOP No. 27 assumption guidance provides a useful adjacent discipline example: in pension measurement, discount-rate, inflation, and other economic assumptions must be selected with attention to measurement purpose and consistency. Its scope is defined benefit pension obligations rather than general insurance liabilities; the comparison illustrates methodological consistency without transferring a pension standard into another regulatory domain as binding authority.

Liquidity and credit adjustment

A cash-flow stream's liquidity characteristics can affect the discount rate relevant to its valuation under a particular accounting or supervisory framework. A policyholder's inability to demand immediate payment for an illiquid, predictable annuity differs from a claim likely to be settled promptly or a surrenderable policy with accessible cash value. The value of illiquidity depends on the measurement rule and on how closely the contractual obligation matches reference assets.

The IFRS 17 measurement guidance requires rates to reflect the liquidity characteristics of insurance contracts; its bottom-up method begins with a liquid risk-free curve and adjusts for differences in liquidity. Its top-down method starts with a reference-asset return curve and eliminates factors irrelevant to the liabilities. In either method, credit risk inherent in a reference-asset portfolio that does not affect insurance-contract cash flows should not silently increase the liability discount rate. The result cannot be justified by merely citing an insurer's expected asset yield or a portfolio spread above government bonds.

The European prudential regime has its own separately specified mechanisms. The Solvency II matching-adjustment conditions concern an eligible portfolio of insurance obligations and assigned assets, subject to supervisory approval and regulatory conditions. This framework is not identical to contract-liquidity adjustment under IFRS 17. Even where the two calculations produce rates of similar magnitude, they differ in eligibility, measurement purpose, governance, and the decomposition of relevant risk.

A twenty-year fixed payment illustrates sensitivity to a seemingly modest additional discount spread. If a continuously compounded zero rate is increased from 3.00% to 3.50%, the discounted value changes by the multiplicative factor e−0.005×20≈0.90484e^{-0.005\times20}\approx0.90484, a reduction of about 9.52%. This is a mathematical scenario, not a proposed or permitted liquidity adjustment. Its lesson is that choosing a long-run spread without identifying its economic content can dominate the reserve calculation. Uncertain spread estimates warrant separate stress testing and disclosure, and an already projected credit-related cash flow should not be duplicated in a rate adjustment.

Comparative actuarial examples

Consider three distinct hypothetical liabilities: a short-tailed claim paid in six months, a fixed annuity payment due in 20 years, and a 20-year inflation-linked benefit. A parallel change in the relevant curve has much greater present-value effect on the long-dated fixed payment than on the short-tailed claim, all else equal. The inflation-linked payment adds a second source of variation: the level and joint dynamics of inflation. A fixed long-dated benefit can have substantial interest-rate duration without any explicit index sensitivity, whereas an indexed benefit can have both. Thus the exposure categories follow the contract and projection, not the class label “insurance liability.”

A cross-currency example makes the discipline even clearer. If a claim is contractually payable in euros, its euro-denominated expected cash flows should ordinarily be discounted under a euro-consistent basis and converted to reporting currency under the applicable reporting convention, or equivalently measured using a coherent set of forward exchange and discount relationships. Discounting nominal euro cash flows directly using a U.S.-dollar Treasury curve can produce an inconsistent value unless a complete currency conversion or basis transformation is specified. Foreign-exchange risk and the distinction between insured-currency obligations and reporting-currency presentation are analytically separate from term structure selection.

The comparison supports a practical sequence of calculation: specify the obligation and valuation basis, model cash-flow amounts and timing, select a curve consistent with their financial characteristics, and report model and assumption sensitivities. The sequence follows established valuation identities and supervisory requirements.

Curve Extrapolation

Long-horizon identification

The longest-dated insurance payments often mature far beyond the interval over which reliable instruments trade. Life-contingent annuities, longevity guarantees, long-term care exposures, and certain liability settlements can have payment tails many decades into the future. In such cases, the observed market establishes only part of the term structure. A decision is needed about how to complete it. This decision is intrinsic to the actuarial valuation and can materially change reserves even when all liquid-maturity quotes are held fixed.

The relevant distinction is between observed data, an estimated curve inside the liquid region, and extrapolated values outside that region. An extrapolated 60-year discount factor is not a directly observed market price simply because a model produces it to several decimal places. The number of decimal places expresses computation precision rather than economic identification. That distinction is particularly important when comparing regulatory curves across currencies whose markets vary in depth, instrument availability, and maturity distribution.

Let TLT_L be the last liquid maturity identified for a chosen market and valuation regime. Values D(0,t)D(0,t) with t>TLt>T_L may depend on an assumption about long-term forwards, a convergence speed, continuity conditions, or a stochastic term-structure model. Calibration at or before TLT_L is insufficient by itself to identify these long-term parameters. One can have two smooth curves that perfectly reprice every observed instrument up to TLT_L, yet deliver materially different 40-year or 60-year insurance values. The inferential uncertainty is structural, not merely measurement noise.

This problem is recognized in the European supervisory framework. Article 77a of Solvency II requires that the construction use deep, liquid and transparent market information and specifies convergence of extrapolated forward rates toward an ultimate forward rate. The corresponding EIOPA term-structure data program supplies currency-specific technical information for supervisory use. The official method should be understood as a documented convention for forming a supervisory valuation basis, not as an empirical demonstration that a particular unobserved long-term forward rate will occur.

Extrapolation families and supervisory practice

Extrapolation can be approached with constant long-end spot rates, constant long-end forward rates, parametric decay, kernel-based curve completions, and regulatory convergence mechanisms. Their mathematical boundary behavior differs. Holding the terminal observed zero rate fixed makes z(t)z(t) constant beyond the boundary; holding the terminal forward rate fixed instead makes the log discount factor linear beyond the boundary. Those two assumptions coincide only when the zero and instantaneous forward rates agree at the boundary. A smooth convergence to a separately specified ultimate forward rate adds both a target and a speed of convergence.

The Smith–Wilson family of methods is widely associated with the European risk-free curve framework. In its common form, it uses an asymptotic term-structure assumption and a kernel representation to fit observed market instruments, with convergence controlled by methodological parameters. The implementation details, applicable first smoothing point, market classification, and ultimate forward rate must be obtained from the authority's version-specific technical documentation. The EIOPA RFR documentation is particularly important because the methodology is evolving. An older description of Smith–Wilson and long-term convergence should not be treated as sufficient evidence of the exact parameters and interpolation choices for the revised 2027 regime.

Each extrapolation choice has a different effect on the present value of remote payments. Suppose a single guaranteed payment of 5,000 at year 40 is valued under two hypothetical constant continuously compounded long-end zero rates, 3% and 4%. The factors are e−1.2e^{-1.2} and e−1.6e^{-1.6}; the resulting values are approximately 1,505.97 and 1,009.48. The difference of 496.49 is entirely attributable to the assumed terminal discount basis. These figures do not replicate any EIOPA currency curve. They show the identification problem: long-tail reserves can move substantially without any change in underlying claim frequency or severity.

Convergence constraints may improve stability, but their valuation consequences depend on where extrapolation starts and how rapidly convergence is imposed. The key evidence is the distribution of liability cash flows after the liquid boundary. A portfolio whose projected cash flows largely mature before TLT_L will be less sensitive to long-end extrapolation, other things equal, than one whose expected benefit payments continue for many decades. Analysis should therefore weight extrapolation uncertainty by the liability cash-flow profile rather than treating the same curve deviation as equally material to all insurance products.

The 2027 European transition

The 2026 research cutoff demands strict version control. EIOPA's November 2026 RFR technical documentation, identified as EIOPA-BoS-26-479 and dated October 2026, describes modifications related to the Solvency II review, including interpolation and extrapolation, the volatility adjustment, deep/liquid/transparency assessments, representative portfolios, and technical parameters. Its change note states unequivocally that the updated document applies from January 30, 2027. The EIOPA July 2026 implementation announcement likewise identifies January 30, 2027, as the date of application of the revised framework.

The EIOPA RFR archive and monthly releases contain the technical term structures associated with actual valuation months, including the October 2026 release published November 5, 2026, and a separate preparatory-phase series for 2027. These resources provide a crucial distinction between a monthly curve that belongs to the currently applicable framework and a preparatory calculation illustrating forthcoming rules. A comparison across the changeover must identify the method and effective date of each curve. Otherwise a methodological discontinuity can be reported falsely as a market move, and a prospective method can be applied prematurely to a prior valuation date.

For the present monograph, the consequence is methodological rather than a numerical replication of EU regulatory rates: the analyst should archive the exact curve vintage, documentation version, currency, matching-adjustment or volatility-adjustment status if applicable, valuation date, and the date from which the rules govern the calculation. EIOPA itself separates components such as credit risk adjustment, extrapolation, volatility adjustment, and matching adjustment. The user of a curve should preserve those component distinctions, especially when analyzing whether movements in a reported best estimate derive from market prices, methodology changes, portfolio eligibility, or assumption revisions.

Uncertainty and validation

The presence of a prescribed supervisory extrapolation methodology supports legal and reporting consistency; it does not eliminate long-term economic uncertainty. A complete sensitivity assessment can vary the ultimate forward rate, first extrapolation or smoothing point, convergence speed, market-data exclusions, interpolation scheme, and, where relevant, volatility or matching adjustment. The actual menu of permitted production choices is regime-specific. A sensitivity test may be informative even for parameters that the reporting entity is not authorized to change in its official submission.

Out-of-sample validation beyond the liquid boundary is intrinsically limited. Analysts can nevertheless evaluate stability across market dates, continuity at the boundary, shape of inferred forwards, coherence with other long-term benchmarks, response to liquidity shocks, and concentration of discounted liability value by maturity. Empirical backtests should be labeled as tests of a historical implementation under observed conditions; they do not certify a future ultimate rate for horizons that have never become liquid.

A reporting table may separate current-market effects from structural and assumption uncertainty. For example, its columns can state (1) official prescribed-base present value, (2) value under a hypothetical lower convergence target, (3) value under a higher target, and (4) differences due to the alternative starting maturity. The main text should then explain which changes are legally admissible as alternative reporting bases and which are only management sensitivity analyses. This avoids giving scenario values an authority or precision they do not possess.

Discount Sensitivities

Duration, convexity, and key-rate exposures

A curve is a collection of maturity-dependent discount factors, so a single scalar duration only partially describes the liability's interest-rate risk. For fixed positive cash flows CkC_k discounted at continuous zero rates z(tk)z(t_k), define their present-value weights wk=CkD(0,tk)/V0w_k=C_kD(0,t_k)/V_0, with V0=∑kCkD(0,tk)V_0=\sum_k C_kD(0,t_k). Under an additive parallel shift Δ\Delta applied to every continuously compounded zero rate, the shifted value is V(Δ)=∑kCkD(0,tk)e−tkΔV(\Delta)=\sum_k C_kD(0,t_k)e^{-t_k\Delta}. Differentiation gives

−1V0dVdΔ∣Δ=0=∑kwktk=D,-\frac{1}{V_0}\frac{dV}{d\Delta}\bigg|_{\Delta=0}=\sum_k w_k t_k=\mathcal D,

and

1V0d2VdΔ2∣Δ=0=∑kwktk2=C.\frac{1}{V_0}\frac{d^2V}{d\Delta^2}\bigg|_{\Delta=0}=\sum_k w_k t_k^2=\mathcal C.

Here D\mathcal D is the portfolio's duration with respect to the specified parallel shift in continuous zero rates; C\mathcal C is the corresponding convexity measure. Their derivation assumes the cash flows themselves remain unchanged under the perturbation. For interest-sensitive policyholder behavior, callable guarantees, or financially indexed benefits, the total sensitivity also includes derivatives of the cash flows. These are sometimes described as behavioral, option, or effective-duration components, but their actual definition depends on the scenario and model.

The second-order approximation for a small parallel shift is

V(Δ)−V0V0≈−DΔ+12CΔ2.\frac{V(\Delta)-V_0}{V_0}\approx-\mathcal D\Delta+\tfrac12\mathcal C\Delta^2.

The second-order term explains why a linear duration approximation can overstate or understate the response to a large rate move, depending on the cash-flow shape and embedded options. Duration calculated from a regulatory yield curve need not equal a market DV01 on an actual hedging instrument. Key-rate measures depend on how individual curve pillars are shocked and how the interpolator translates those shocks to all payment dates.

For a designated pillar jj with shock parameter ϵj\epsilon_j, define a liability's key-rate dollar sensitivity Kj=−∂V/∂ϵjK_j=-\partial V/\partial\epsilon_j, evaluated at the current curve. It can be approximated by symmetric finite differences when the cash-flow engine is rerun under appropriately localized curve shocks. Since a bump scheme itself defines a basis, the selected pillars, bump widths, units, interpolation method, and any recalibration should be documented. Summing key-rate sensitivities will approximate a parallel duration exposure when the chosen basis spans the parallel shift; it need not do so for arbitrary nonlocal or overlapping shocks.

The methodological findings of Hagan and West matter here because an interpolator can change not only fitted values but also hedging implications. The distinction is pronounced for long-dated insurance liabilities, where the sensitivity to unobserved curve segments must be allocated to observable pillars or to separate extrapolation parameters. A hedge that neutralizes current market key-rate exposure can retain significant basis, liquidity, model, and extrapolation risk.

Scenario design and asset–liability interactions

A parallel 100-basis-point shock is informative but incomplete. Insurance interest-rate exposure may depend on twists, butterflies, long-end divergence, steepening or inversion, inflation shocks, credit-spread widening, surrender behavior, and changes to asset reinvestment opportunities. The valuation engine should treat a scenario as a coherent set of financial states and specify whether dynamic contract cash flows are reprojected. Scenario pathways and one-time instantaneous shocks answer different questions and should not be reported as interchangeable.

Asset–liability management adds the asset side. Long-dated obligations can be backed by bonds, loans, derivatives, and other instruments whose sensitivity to interest rates, spreads, and liquidity differs from that of the liabilities. A crude match of weighted average maturity is not the same as a match of present-value sensitivities by curve pillar, nor does it neutralize embedded insurance guarantees. The larger research domain is treated in the related Science Insurance Finance and Asset-Liability Management; the present work restricts itself to discount-curve construction and liability valuation sensitivities.

The U.S. 2026 NAIC valuation framework includes product-specific stochastic scenario requirements for principle-based reserves, illustrating that cash-flow projection and rate evolution can be linked in a statutory regime. The separate ASOP No. 22 standard addresses professional work concerning asset adequacy, subject to applicable law. Neither source supports replacing each mandated statutory calculation with a single universal market-consistent scenario model. The reporting objective governs scenario choice, permissible assumptions, governance, and resulting reserves.

Replicable numerical case study

Consider a deliberately simple liability with three fixed outflows. At year 1 it pays 1,000; at year 5 it pays 2,000; and at year 20 it pays 5,000. The annualized zero curve uses continuously compounded rates 2%, 3%, and 4% at those respective maturities. The calculation treats these numbers as exact assumptions, omits fees, mortality, credit and policyholder option effects, and uses times measured in whole years. This makes the computation transparent and reproducible.

Payment date (years) Hypothetical cash flow Continuous zero rate Discount factor Present value
1 1,000 2.00% 0.9801986733 980.20
5 2,000 3.00% 0.8607079764 1,721.42
20 5,000 4.00% 0.4493289641 2,246.64
Total 8,000 — — 4,948.26

The portfolio's present-value-weighted continuous-shift duration is approximately 11.0181 years. Its second derivative measure is approximately 190.5061 years squared. A 1-basis-point increase in the entire zero curve therefore produces a first-order liability decrease of approximately 5.45 currency units. A 100-basis-point increase in each of the three specified zero rates yields a new exact value of approximately 4,447.30, a decline of 500.96. The linear-duration approximation is more negative; incorporating the second-order convexity term gives approximately 4,450.19 and substantially narrows the approximation error. The remaining difference reflects the higher-order terms of the exponential discount function.

Now apply a 100-basis-point bump to one of the three zero-rate pillars at a time and leave the other two unchanged. Since each cash flow falls precisely on its own pillar, the differences are unambiguous for this stylized example. A year-1 bump reduces portfolio value by approximately 9.75; a year-5 bump by 83.95; and a year-20 bump by 407.25. The sum of these three changes equals the exact combined-bump decrease to rounding, because in this specific example the bumps affect disjoint cash-flow contributions. In a real interpolation-based curve, a year-5 quote shock can change discount factors at many intermediate maturities; the key-rate basis must be stated to interpret the same calculation.

Illustrative stress Resulting liability PV Change from base
Base yield curve 4,948.26 —
+100 bps at year 1 only 4,938.51 −9.75
+100 bps at year 5 only 4,864.31 −83.95
+100 bps at year 20 only 4,541.01 −407.25
+100 bps in all three zero rates 4,447.30 −500.96

The case study demonstrates three separate propositions. Cash-flow timing determines rate sensitivity even when contractual payment amounts are held fixed. Curve shape matters because a maturity-specific cash flow is multiplied by its own discount factor. Model interpretation matters because a regulator's calibrated curve, a market shock to traded instruments, and an insurer's internal scenario may map the same label “100-basis-point rate change” to different computational transformations.

The complete calculation can be reproduced in any language implementing an exponential function:

from math import exp
cash_flows = [(1, 1000, 0.02), (5, 2000, 0.03), (20, 5000, 0.04)]
base = sum(c * exp(-t * z) for t, c, z in cash_flows)
parallel_up = sum(c * exp(-t * (z + 0.01)) for t, c, z in cash_flows)
duration = sum(t * c * exp(-t * z) for t, c, z in cash_flows) / base
convexity = sum(t*t * c * exp(-t * z) for t, c, z in cash_flows) / base
print(base, parallel_up, duration, convexity)

This is an elementary validation experiment, not an empirical estimate of an insurance portfolio. The values have been calculated with standard double-precision exponential arithmetic; monetary outputs are rounded to two decimal places only when displayed. The exact input assumptions and formulas are supplied so that a reader can recompute the results independently.

Integrated research synthesis and implementation controls

The documentary and analytical evidence supports a structured reading of actuarial discounting. The starting question is the measurement objective: insurance financial reporting, supervisory best estimate, statutory reserve, transaction pricing, solvency capital, or internal risk management. Next comes the insurance obligation: which cash flows occur in which currency, under what survival or claim contingencies, at which dates, with which contractual options and indexation rules? Only then can the relevant curve be selected or constructed from observed instruments and regime-specific adjustments.

The analytical synthesis yields five interconnected findings. First, curve selection is part of the definition of the measured obligation: a rate represents particular financial characteristics and valuation conventions. Second, the market evidence is transformed by pricing equations, observation filters, and interpolation rules; it never arrives as a full set of self-explanatory discount factors. Third, discounting and cash-flow projection must be mutually coherent, particularly in the presence of inflation, guarantees, options, and financial–insurance dependencies. Fourth, extrapolation is an independent source of model risk for long-duration insurance liabilities, even when all near-term market quotes are undisputed. Fifth, scalar sensitivity statistics describe only specified perturbations; reliable analysis additionally requires key-rate, curve-shape, liquidity, inflation, and scenario analysis aligned with the actual liabilities.

These findings are inferential conclusions from the mathematics and comparative sources. They do not constitute a new theoretical system or a claim of priority over established actuarial science. They explain why sound practice combines source verification, convention control, mathematical recomputation, documented professional judgments, and transparent presentation of material sensitivities.

An auditable implementation should preserve a versioned source-data inventory; exact quote observation and valuation dates; a curve-calibration report; separate spot, forward and par representations; the cash-flow projection basis; inflation, credit and liquidity treatments; the documented extrapolation parameter set; and a reproducible sensitivity report. Independent review should include sample cash-flow reconciliation, unit and day-count checks, market instrument repricing, targeted perturbations, and change-of-method analysis. For valuations governed by a supervisory or accounting regime, the auditor must additionally compare the output with that regime's effective requirements on the actual valuation date.

The conditional character of every comparison is central. IFRS 17 provides a measurement and financial-reporting architecture; Solvency II specifies a supervisory insurance-valuation framework subject to dated amendments; NAIC materials govern particular U.S. statutory contexts through adopted local requirements; and international capital standards facilitate supervisory comparability. They have overlapping scientific concerns, but their discount curves need not coincide. The strongest international conclusion is therefore an obligation to identify which question each curve answers, and to document the conversion between instruments, contingent cash flows, and measured liability values.

Conclusion

Discounting is a constitutive component of actuarial valuation because the timing and financial characteristics of insurance cash flows affect their present measurement. Discount factors, zero rates, forward rates, and par yields express related aspects of the term structure, while the quoted market instruments, rate conventions, and curve construction procedure determine how those representations are computed. A consistent valuation preserves the relationship between contingent cash flows and the economic or regulatory basis on which those cash flows are discounted.

The international evidence establishes substantive differences among major applications. IFRS 17 requires discount rates consistent with insurance cash-flow and liquidity characteristics and with appropriate observable market information. Solvency II defines a relevant risk-free discounting framework, including explicit treatment of extrapolation and specified adjustments; the revised EIOPA technical methodology dated in 2026 begins applying January 30, 2027, and must be distinguished from the contemporaneous monthly 2026 curves. U.S. principle-based reserves and professional asset-adequacy analyses have their own product-specific and jurisdiction-dependent measurement requirements. Global supervisory standards establish an international reference for capital comparability without erasing national implementation choices.

Mathematical analysis demonstrates that long-duration liability values can be dominated by relatively small changes in discount assumptions, and that interpolation, long-end extrapolation, inflation consistency, and financial option dependence are material sources of actuarial model risk. The reproducible fixed-cash-flow experiment illustrates the importance of maturity-specific rates, duration, convexity, and key-rate decomposition. Its simplifying assumptions are explicit and do not substitute for the modeling of mortality, losses, policyholder behavior, or stochastic market conditions in actual insurance portfolios.

The resulting scientific conclusion is operational and testable: a defensible actuarial discounting result can be reconstructed from its contingent cash-flow definition, valuation purpose, source instruments, quote and convention controls, curve-fitting and extrapolation procedures, permitted adjustments, and documented sensitivity tests. Such reconstruction makes financial reporting comparisons more meaningful, supports independent scholarly examination, and permits corrections when underlying standards or market methodologies change. The analytical contribution is an original synthesis of established mathematics and verified regulatory practice. No new actuarial theory, proprietary term-structure model, or universal discount rate is asserted.

References