RESEARCH DOSSIER
Yield Curve Shape and Key Rate Risk in Insurance
Abstract. Yield-curve risk is a vector of valuation and financing exposures distributed across maturities, currencies, valuation bases, and contractual states. An insurer can match the aggregate duration of assets and liabilities while maintaining economically material sensitivity to changes in curve slope, curvature, and individual tenors. This monograph reconstructs the intellectual development of nonparallel interest-rate measurement, defines key-rate and factor-based representations, derives their connection to asset-liability surplus, and evaluates fixed-cash-flow, stochastic, and behaviorally contingent applications. An explicitly synthetic five-tenor experiment reproduces a zero first-order parallel-rate exposure alongside losses of approximately $1.74 million from a nonparallel butterfly shock on a $100 million asset and $100 million liability balance sheet. The case separates an exact repricing result from its linearized approximation and shows the dependence of measured risk on shock basis, interpolation, and convexity. International supervisory evidence from the IAIS, EIOPA, the Bank of England, and the NAIC is examined with attention to the distinction between economic interest-rate risk, regulatory valuation, accounting presentation, and liquidity. The central finding is that effective insurance ALM requires the coordinated identification of term-structure exposure, contingent liability dynamics, basis risk, and feasible hedge instruments. No new theoretical model or proprietary risk doctrine is asserted.
Keywords: Yield curve; key rate duration; partial duration; DV01; PV01; term structure; slope risk; curvature risk; twist; butterfly; insurance asset-liability management; economic surplus; life insurance; annuities; nonparallel shocks; convexity; stochastic valuation; regulatory capital; liquidity risk; hedging; insurance solvency.
Introduction
The term structure of interest rates connects the timing of a payment with its present economic value. Insurance makes this connection unusually consequential. Premiums, benefits, expenses, annuity installments, surrender payments, reinsurance recoveries, and investment proceeds occur at different dates and respond differently to market conditions. A five-year interest-rate movement can influence the replacement yield of invested premiums; a twenty-year movement can dominate the value of deferred annuity benefits; a sharp change in short rates can alter policyholder withdrawals before long-term assets mature. A single statistic for the portfolio's average maturity exposure therefore gives an incomplete description of its financial position.
This dossier examines the shape dimension of interest-rate risk: the degree to which asset and liability values react to rates at distinct tenors when the curve does not shift uniformly. The subject is deliberately narrower than the construction of a discount curve, general duration theory, the actuarial valuation of an option, or the full optimization of an insurer's investment portfolio. Those subjects supply essential inputs, but this investigation focuses on the measurement, attribution, interpretation, and control of maturity-specific net exposure.
The original argument proceeds by connecting financial mathematics with institutional constraints. First, cash flows determine a map from discount factors to values. Second, a representation of admissible curve movements determines which sensitivities are observable. Third, insurance contract behavior makes some cash flows themselves functions of the scenario. Fourth, accounting and prudential conventions may alter the reported impact of the same economic movement. The resulting synthesis is an analytical interpretation of established methods and evidence, with all numerical illustration explicitly constructed for this manuscript.
Historical foundations include Fisher and Weil's interest-rate immunization study, the factor analysis of Litterman and Scheinkman (1991), and Ho's original key-rate duration method. Later parametric term-structure work by Nelson and Siegel (1987), Svensson (1994), and Diebold and Li (2006) provides a separate language for level, slope, curvature, estimation, and forecasting. These publications address related but different questions. A factor model summarizes curve variation, whereas a key-rate measurement attaches valuation sensitivity to selected yield tenors. Neither representation eliminates the insurer's obligation to specify what is shocked and which cash flows are repriced.
The evidentiary approach combines original mathematical derivations, a fully specified hypothetical balance sheet, methodological comparison, and official regulatory and supervisory sources. The quantitative experiment is a structural demonstration, not a claim about observed insurer portfolios or market probabilities. As of November 14, 2026, the revised European Solvency II delegated requirements scheduled for January 30, 2027 must be distinguished from the regime currently applicable; the 2026 amending regulation supplies the authoritative dates.
Term Structure Factors
Discount factors, zero rates, and economically comparable curves
Let denote the price today of one unit of currency payable at time with the risk characteristics specified by the selected curve. In a continuously compounded zero-rate convention, . For deterministic, positive payments at times , . An insurance liability can contain negative premium cash flows, stochastic benefits, and contract options; the simple expression is only a starting point. When payment amounts depend on the economic state, both and the discount factors change with a scenario.
A nominal government curve, overnight-indexed swap curve, accounting discount curve, and prudential risk-free curve can assign distinct values to the same nominal date. They embody different instruments, liquidity, collateral, credit, and regulatory treatments. Curve-risk figures are economically interpretable only after these bases are defined and reconciled. The Federal Reserve's nominal yield-curve estimates illustrate a government-securities research curve derived from selected off-the-run Treasury coupon issues. The Federal Reserve expressly describes the series as a staff research product rather than an official statistical release, and documents revisions to the estimation method. The associated Gurkaynak-Sack-Wright study explains the use of fitted discount curves and forward rates. Such estimates provide an empirical benchmark; they are not automatically the curve appropriate to insurance contract measurement.
A yield curve may be represented by spot rates, instantaneous forwards, par swap rates, or market security prices. A parallel increase in zero rates is not identical to a parallel increase in continuously compounded forward rates under every parameterization, while a constant shift in par yields generally induces a nonconstant shift in implied spot yields. Risk reports should therefore distinguish quote-space shifts from state-variable shocks. In a multi-curve market, the discount curve, reference-rate forwarding curve, bond credit spread curve, and inflation curve may need separate shocks even when each has nominal tenor labels.
Level, slope, and curvature
The first intuitive decomposition of curve shape consists of level, slope, and curvature. A level factor moves rates across maturities in broadly the same direction; a slope factor changes shorter and longer yields differently; a curvature factor moves an intermediate sector relative to the wings. The empirical common-factors study by Litterman and Scheinkman (1991) made this decomposition central to fixed-income risk analysis. It is a statistical account of bond-return covariance under the sample and instruments studied, not an identity asserting that every future curve movement is three-dimensional.
The Nelson-Siegel family represents a spot or forward term structure using smooth loading functions and a small set of parameters. In the later Diebold-Li dynamic formulation, time variation in those parameters is interpreted as level, slope, and curvature dynamics for bond-yield forecasting. Such models are valuable for scenarios, interpolation, and the communication of factor exposures. Their limitations become significant for insurers with benefits far beyond the liquid market: factors identified predominantly from traded tenors can be insensitive to a pricing assumption or regime change at forty, fifty, or sixty years.
For a finite grid of tenors, write the shocked zero curve as , where is a tenor-by-factor loading matrix, is a factor-shock vector, and contains movements excluded by the chosen factors. If is the vector of first derivatives of surplus with respect to the tenors, the modeled first-order surplus movement is ; its omitted component is . This decomposition reveals an important empirical distinction: explaining most historical yield variance need not explain the largest prospective loss of an insurer whose exposure is concentrated at a low-variance long maturity.
Historical development of immunization and factor risk
Macaulay's duration tradition treats a fixed-income instrument's discounted payment dates as a summary of interest-rate exposure. Fisher and Weil (1971) explicitly addressed bond investment performance under interest-rate fluctuations and the limits of naive maturity approaches. The subsequent development of multifactor curve analysis and Ho's 1992 key-rate approach responds to the fact that maturity distributions and shape changes both matter. The intellectual transition is from exposure to one scalar shock toward exposure to a function across time.
The distinction has direct insurance meaning. Consider two annuity books with the same present value and the same aggregate modified duration. One pays predominantly between years fifteen and twenty; the other combines very near-term and very remote benefits. Equal aggregate duration says that the first derivatives coincide along the specific parallel-shift direction. It says nothing about their derivatives along a five-year versus thirty-year twist. For nonparallel market movements, the two books may require different assets, different derivatives, and different liquidity provisions.
Scientific boundaries of factor reduction
Factor-based modeling is a controlled simplification. A decomposition needs an observation universe, date range, currencies, curve definitions, sampling frequency, smoothing method, and explicit eigenvector or functional loadings. Principal components may change orientation or magnitude when the sample includes monetary-policy reversals, changing collateral regimes, or market disruptions. Fixed factor labels are explanatory conventions, while their realized correlations are estimated objects. An insurer should review both the share of market variation explained and the share of liability sensitivity represented, because the two criteria can rank factor sets differently.
For structural exposures, the practical questions are which curve inputs affect cash flows and which shocks change net wealth. A statistical three-factor hedge may be efficient under the historical covariance matrix yet fail under an unusual regulatory extrapolation change or one-sided policyholder exercise. That observation motivates a separate, explicitly maturity-based risk representation.
Parallel versus Nonparallel Changes
First-order price and surplus responses
For fixed cash flows discounted at continuously compounded zero rates, the first derivative with respect to the tenor's zero rate is . A small continuously interpolated perturbation gives . The familiar parallel duration case sets at every relevant maturity. The resulting derivative collapses the full weighted cash-flow distribution into one number.
Write the insurer's market-consistent economic surplus as , with asset value and positive liability value . Then . A positive duration mismatch defined as produces a first-order surplus decline under a parallel rate increase; the sign is reversed if the insurer reports the liability-minus-asset exposure instead. Reporting conventions therefore require an explicit sign statement. The current monograph defines PV01 at a key tenor as the positive valuation loss from a one-basis-point increase for a positive fixed payment, and net PV01 as asset PV01 minus liability PV01.
For a full zero-curve shock vector , let and be the vectors of dollar PV01 exposures. Expressing in basis points yields
The scalar parallel PV01 is the sum of the components only when the chosen tenor shock functions form a partition of unity over the valuation cash flows. An arbitrary set of rate-bump conventions may not have this property.
Nonparallel scenarios and aggregate-duration blind spots
A steepener increases long rates relative to short rates; a flattener narrows that difference. Either can occur with the level rising, falling, or remaining approximately unchanged. A twist can pivot around a specified maturity without preserving average yield. A butterfly shock raises or lowers intermediate tenors relative to shorter and longer maturities. The labels identify relative shapes, not standardized shock vectors. Tenor coordinates, magnitude, direction, and interpolation must accompany every scenario name.
If an insurer has zero aggregate parallel PV01, its first-order surplus exposure to a uniform rate move is zero. This condition mathematically restricts to the subspace whose elements sum to zero. It does not force the whole vector to vanish. Any nonparallel shock with a nonzero dot product with that vector changes the surplus at first order. A matched scalar duration is therefore a one-direction hedge, not a universal interest-rate hedge.
A related problem concerns dynamic hedging. A parallel immunization can be accurate today, while market-driven revaluation moves the portfolio's partial exposures tomorrow. Convexity changes dollar sensitivity as yields shift; maturities roll down the curve; new premiums, claims, lapses, and asset purchases alter cash flows. A duration gap is a date-stamped measurement, and the rebalance schedule is part of its risk interpretation.
Second-order terms and the source of apparent protection
For fixed zero-coupon cash flows, an exact local expansion adds approximately . For surplus, the asset and liability second-order terms are subtracted. In derivative notation, , where is the Hessian of surplus. Maturity-specific cross-gamma appears when curves are interpolated jointly, products depend on several rates, or cash flows are stochastic and respond to an entire economic scenario. Matching first derivatives does not establish second-order neutrality.
Traditional immunization arguments introduce duration matching and convexity conditions under specific assumptions about cash flows and the yield movement. Fisher and Weil's classical immunization analysis is relevant to understanding the assumptions behind such local protection. In insurance ALM, policyholder options, reinvestment, and liquidity needs complicate those assumptions. Asset and liability cash flows may not remain fixed as rates move.
Curve basis and institutional measurement
Three meanings of a "rate shock" frequently coexist. An economic desk scenario changes a liquid-market curve while holding contractual terms fixed unless behavior is explicitly modeled. A reporting scenario changes accounting discount assumptions and may alter insurance finance results or other comprehensive income. A regulatory scenario reprices specified assets and liabilities according to prudential rules and aggregations. Equal shock labels can produce different reported capital changes. The IFRS 17 overview states that insurance contract measurement incorporates current estimates of future cash flows and recognizes insurance finance income or expenses under its presentation requirements; the more detailed IFRS 17 discount-rate criteria require rates consistent with payment timing, currency, liquidity, and relevant market information. These are valuation constraints, not instructions to treat a single market-yield curve as every insurer's accounting curve.
An ALM risk report should identify the observed instrument family, curve construction, currency, valuation basis, behavior assumption, reporting objective, and position date before presenting aggregate duration or factor PV01. This metadata is analytically necessary because interest-rate risk cannot be interpreted independently of the underlying repricing convention.
Key Rate Duration
Definition and shock-basis construction
Key rate duration measures a value's sensitivity to a localized movement around a designated maturity, with the movements at other tenors determined by an explicit rule. Ho's Key Rate Durations (1992) supplies the canonical foundation. An implementation commonly chooses knot maturities , constructs piecewise-linear tent functions that equal one at knot , fall to zero at neighboring knots, and specify boundary treatment at the shortest and longest modeled maturities. A one-basis-point upward bump to key changes by .
For a valuation function , define the monetary key-rate PV01 by a centered finite difference:
This sign convention makes positive for a conventional positive fixed payment exposed to an upward yield bump. Its corresponding normalized key-rate duration is when . For liability net cash flows that may change sign, a normalized duration can be unstable or misleading, so monetary PV01 should be retained alongside any ratio. The denominator and the sign of every metric must be explicit.
An alternative derivative definition is , where denotes a dimensionless rate shift. Multiplying by recovers the local dollar PV01. For an insurer's surplus the vector is . The first-order impact of a scenario in basis points is . These units also prevent a common error: treating one percent, one percentage point, and one basis point as numerically interchangeable.
Recovery of parallel PV01
If the basis functions satisfy for every cash-flow maturity in the modeled domain, then a unit shock to each key reproduces a unit parallel shift at every maturity. Consequently approximates, and in the derivative limit equals, parallel PV01 for fixed cash flows under the same curve and valuation engine. The equality can break when the curve beyond the final knot is frozen, when endpoints use different extrapolation rules, when the same market quote affects several bootstrapped rates, or when the liability engine changes stochastic assumptions inconsistently between bumps. The failure may reveal a documented basis convention or a model implementation defect.
Curve interpolation deserves the same status as the key-rate formula. Localized bumps on par swap quotes, continuously compounded zero rates, and instantaneous forwards correspond to different effective discount-factor distortions. A key-rate hedge ratio is meaningful only when instrument and liability calculations employ the same basis, or when cross-basis sensitivities are separately represented. This is particularly important when insurer assets are corporate bonds valued on benchmark plus credit spread while liabilities are valued using a regulator-prescribed risk-free curve with separate adjustments.
Numerical differentiation and conditioning
Finite differences require a bump large enough to overcome numerical noise and small enough to represent a local derivative. A central difference usually reduces leading truncation error relative to a one-sided bump for smooth valuation functions. Yet embedded options, discontinuous management rules, reinsurance triggers, and abrupt cash-flow changes can make the function locally nonsmooth. In those cases, reporting separate up and down sensitivities adds information that a symmetric average conceals.
Model governance should test bump size, key-grid density, interpolation method, curve bootstrap, cash-flow replication, and valuation reproducibility. The baseline value must be evaluated with the same calibration and assumptions as the stressed values. Recalibration choices are part of the experimental design: reoptimizing a model's stochastic parameters after each bump answers a different question from holding calibrated parameters fixed. The relevant quantity depends on whether the purpose is market hedging, accounting sensitivity, regulatory solvency, or full management stress.
A sensitivity matrix across curves and instruments can be ill-conditioned. Very long-duration instruments may have nearly proportional exposures over the available liquid tenors. A nominally exact hedge then requires offsetting large positions, high transaction costs, or instruments with limited market depth. Regularization and liquidity constraints make the achievable hedge different from the unconstrained linear-algebra solution. The desk should report residual tenor exposures and the assumptions supporting any netting.
Statistical factors and key rates as complementary coordinates
Let map factor moves into tenor moves and represent net PV01. The equivalent factor sensitivity is , subject to the sign convention adopted. A factor hedge constrains exposure along the columns of ; a full key-rate hedge constrains more directions but can amplify measurement noise. Key rates are therefore appropriate for maturity attribution and stress testing, while parsimonious factors are often appropriate for covariance estimation and communication. Neither is universally more accurate. Their quality follows from the shock space relevant to the insurer.
If is a covariance matrix of basis-point movements, the first-order variance approximation to economic surplus changes is . This expression assumes a stable covariance estimate, local linearity, and the specified rate basis. It omits credit losses, lapse feedback, and nonlinear liquidity effects unless those variables are represented. A low variance under one historical sample cannot be equated with a bound on losses under rare but economically plausible curve shapes.
Audit-ready reporting
An independently reviewable key-rate report should expose tenors, currency, zero/par/forward basis, interpolation, bump size, sign, valuation timestamp, liability scenarios, assets, liabilities, net amounts, and parallel reconciliation. Residual PV01 beyond the longest liquid tenor should be a visible category. The report should also preserve gross asset and liability exposures when netting masks large positions; gross sensitivity can determine collateral flows and trading requirements even when economic surplus risk is small.
Twist and Butterfly Scenarios
A fully specified synthetic insurance balance sheet
Consider a stylized insurer with assets and positive liability obligations each valued at $100 million. The horizon knots are 2, 5, 10, 20, and 30 years. Assume a baseline continuously compounded flat discount curve of 3.5 percent, fixed deterministic payments exactly at the five knots, no credit spread, no tax, no policyholder options, and no trading during the instantaneous shock. The table gives present-value weights, so the underlying promised cash flow at each maturity is C(t)=$100\,\mathrm{million}\times w(t)e^{0.035t}. This definition makes the initial valuation transparent and reproduces the stated totals exactly.
| Maturity | Asset PV weight | Liability PV weight | Asset minus liability weight | Net PV01 per +1 bp |
|---|---|---|---|---|
| 2 years | 8% | 8% | 0% | $0 |
| 5 years | 22% | 12% | +10% | +$5,000 |
| 10 years | 10% | 25% | -15% | -$15,000 |
| 20 years | 40% | 35% | +5% | +$10,000 |
| 30 years | 20% | 20% | 0% | $0 |
| Total | 100% | 100% | 0% | $0 |
The asset and liability weighted present-value maturities both equal years. Under this zero-coupon, continuously compounded construction, this is also the common modified duration. Aggregate parallel PV01 is $162,600 per side for a one-basis-point move. However, the maturity-specific net vector is dollars per basis point. Its sum is zero but its entries are economically material. The example isolates exactly the dimension hidden by the scalar duration match.
Exact repricing method
Let the shock at maturity , in basis points, be . Because the initial portfolio weights are present values, repricing yields the exact surplus change:
The first-order approximation is
There is no curve interpolation ambiguity inside this deliberately simplified example because cash flows and stressed zero rates occur at the same five maturities. In a real portfolio with intermediate payments, the interpolation between shock knots becomes a model input.
| Synthetic scenario | 2y | 5y | 10y | 20y | 30y | Linear surplus change | Exact surplus change |
|---|---|---|---|---|---|---|---|
| Parallel upward shift | +100 | +100 | +100 | +100 | +100 | $0 | +$33,387 |
| Front-end up, long-end down | +100 | +80 | +20 | -30 | -40 | +$200,000 | +$214,097 |
| Belly upward hump | 0 | +20 | +80 | +20 | 0 | +$900,000 | +$857,700 |
| Reverse butterfly | 0 | +80 | -40 | +80 | 0 | -$1,800,000 | -$1,743,548 |
All shock entries are basis points. Dollar results are rounded to the nearest dollar. Exact changes use exponential repricing, not an invented convexity multiplier. The sign of a gain or loss is specific to the constructed portfolios; another insurer with a reversed key-rate profile would have a different outcome.
Interpretation of the four experiments
The first row demonstrates second-order surplus sensitivity under a perfectly matched first-order parallel duration. The exact gain of about $33,387 arises from different weighted convexities, despite zero linear exposure. The amount is small relative to the large gross present values but is materially different from the linear estimate.
The front-end-up and long-end-down shock gives a positive surplus effect here because the liability's relative concentration at ten years offsets the asset's net five- and twenty-year exposures. It should not be generalized into a claim that flattening or steepening helps all insurers. Scenario language alone never determines the sign of exposure.
The belly-hump scenario increases the ten-year rate much more than adjacent tenors, reducing the more heavily weighted ten-year liability value and increasing surplus. The reverse butterfly moves the five- and twenty-year rates up while the ten-year rate falls, producing a sizeable surplus loss even though the parallel PV01 is exactly zero. The latter case provides a particularly clear falsification of the assumption that equal scalar durations imply first-order insulation against plausible yield-curve shapes.
A reproducibility check requires only the weights and the equations. For each row, calculate and compare it with the linear expression. The differences quantify nonlinear repricing within the deterministic construction; they do not represent stochastic option value, mortality, or reinvestment.
Connection with observed supervisory stresses
Actual stress exercises combine more variables than this clean interest-rate experiment. EIOPA's 2024 insurance stress test included a narrative of larger expected-rate increases at short than at long maturities, further yield-curve inversion, widening credit spreads, and insurance-specific effects. Its results cannot be used to calibrate the probabilities or dollar exposures of the hypothetical balance sheet; the empirical point is that real supervisors consider nonparallel term-structure disturbances and joint balance-sheet consequences.
The Bank of England's 2025 life insurance stress results address the opposite direction of an important market channel: declining risk-free rates accompanied by credit deterioration and asset-price changes under Solvency UK. The difference between the two exercises illustrates why rate-direction assumptions, accounting or prudential regime, and correlated shocks should be documented separately. The UK exercise is evidence about participating insurers under that prescribed severe scenario, not a general statistical frequency of yield-curve events.
For practical scenario design, useful families include parallel up/down shifts, steepeners, flatteners, front-end-only shocks, long-end-only shocks, intermediate-tenor butterfly movements, zero/forward basis changes, changes to extrapolation assumptions, and joint rate-credit-inflation scenarios. Each should be supplemented by a clear valuation convention and liquidity treatment rather than identified solely by a stylized chart.
Long-Dated Benefit Exposures
Why insurers are distinctive curve-risk institutions
A life insurer's contractual obligation may remain outstanding well beyond thirty years. Survivorship, longevity, premium continuation, guaranteed-crediting terms, inflation indexation, and renewal rights determine which cash flows remain and when they become payable. A non-life insurer with short-settlement claims has a different duration profile, but even that insurer may hold long bonds, fund delayed liability payments, or face inflation-linked claim costs. Term-structure risk is thus a question of product design and portfolio financing, not a property confined to conventional fixed annuities.
The NAIC's interest-rates and insurance analysis explains the dual effects of rising rates on insurer investment values, replacement yields, surrender incentives, and liquidity. Those mechanisms show why an interest-rate sensitivity test should not stop at market-valued bonds. Policyholder behavior changes projected liability cash flows, while a surge in near-term withdrawals may require selling assets before maturity.
For participating contracts, future bonuses or crediting decisions respond to asset returns and market competition. Minimum guaranteed rates create asymmetric exposure: falling rates can make old guarantees more valuable, while rising rates may accelerate disintermediation or trigger repricing. Universal life, fixed indexed annuities, and variable products with financial guarantees require product-specific cash-flow and option assumptions. The American Academy of Actuaries' 2026 fixed indexed annuity paper discusses cash-flow matching, surrender risk, derivatives, and collateral liquidity as connected ALM considerations.
Endogenous liability cash flows
A more realistic liability valuation is under a model-consistent pricing measure when market-consistent valuation is intended; other actuarial purposes may use a different projection and discount basis. The derivative with respect to the yield curve contains the direct discounting effect and the effect of changing projected payments, probabilities, contract options, and management actions. For example,
when differentiation and expectation can be interchanged under suitable regularity conditions. This expression is schematic: stochastic discounting, path dependence, measure dependence, and recalibration may introduce additional components. Its central lesson is that a fixed-cash-flow liability key-rate report captures only part of the full economic response.
The maturity of a predicted surrender payment is itself an outcome of rates and policyholder behavior. A ten-year fixed bond can become inadequate collateral for benefits that suddenly accelerate into year two. Economic surplus, statutory surplus, and liquidity survival therefore require distinct but coordinated scenarios.
Long-end extrapolation and curve incompleteness
Many insurance obligations extend beyond the longest maturity with a deep, liquid, and transparent market. Extrapolation rules determine how observable market information is extended to those commitments. The European Solvency II Article 77a establishes the conceptual role of observable liquid maturities, extrapolation, and ultimate forward rates. EIOPA's risk-free term-structure releases provide official technical datasets and methodological documentation, including material updated in November 2026.
A reported forty-year key-rate duration can be a composite of true market exposure and sensitivity to the prescribed extrapolation algorithm. If the last liquid point, convergence rule, or ultimate forward rate changes, the repricing direction is different from a tradable forty-year swap quote. This distinction matters for economic hedging: a synthetic regulatory valuation movement cannot always be hedged one-for-one with market instruments. The insurer should identify tenor exposures arising from observed instruments and the separate sensitivity to structural extrapolation parameters.
In a market without liquid ultra-long bonds or swaps, a portfolio may have a measured long-end mismatch despite the absence of a feasible offsetting hedge. That exposure should be reflected in capital, liquidity strategy, product design, or risk appetite instead of being hidden by interpolation. The globally oriented IAIS ICP 16 ALM requirement explicitly contemplates the valuation of assets and liabilities under different scenarios and the effective management of mismatches. Its standard does not require one universal duration or grid.
Interaction with credit, illiquidity, and measurement
Insurer bonds usually contain credit spreads, while prudential discount curves may include separately regulated adjustments. In the European framework, Article 77c on matching adjustment prescribes the relation between assigned assets, liability cash flows, and fundamental spreads for retained risk. Consequently, a risk-free curve twist can have a different own-funds effect from a corporate-spread twist of the same number of basis points. The value of matching-adjustment eligibility and the consequences of asset downgrades introduce further contingent sensitivities.
The adopted EU Delegated Regulation 2026/269 entered into force in March 2026 but applies from January 30, 2027. At the November 2026 editorial date, it is incorrect to present its future revised interest-rate submodule as already applicable. The temporal distinction affects supervisory interpretation, while the analytical need to test nonparallel exposures is continuous across the change.
Comparative treatment also matters. The United States combines statutory reserve and risk-based capital methodologies with scenario-based cash-flow testing for relevant life business; its financial reporting, risk-based capital, and economic-hedge numbers need not coincide. The European Union applies Solvency II technical-provision and capital rules; the United Kingdom operates the separately amended Solvency UK framework; insurers reporting under IFRS apply the relevant IFRS 17 measurement and presentation rules. These are separate regimes, and their differences should be preserved in any cross-country empirical comparison.
Interest-rate risk and cash availability
An accounting or market-value gain may be unavailable as cash when benefits are due. An economic hedge may offset a liability change but create immediate variation margin on derivatives. An illiquid asset may have a nominally adequate long-duration profile but cannot fund concentrated surrenders without discounts or financing. The BIS discussion of life insurance and financial stability analyzes the consequences of less-liquid insurer investments and complex reinsurance structures. Its argument reinforces the need to connect balance-sheet sensitivity with funding resilience.
The analytic implication is that long-term key-rate exposure must be reported alongside at least three other time axes: expected contractual outflows, stressed policyholder outflows, and collateral or financing settlement dates. A long-end hedge can improve market-value matching while worsening near-term collateral demands. Neither outcome can be inferred from duration alone.
Hedging and Validation
Hedge construction from key-rate sensitivities
Suppose a portfolio has net key-rate vector , and a set of available hedge instruments has a sensitivity matrix , with one column per instrument and one row per tenor. Hedge notionals produce residual exposure . A perfect first-order match requires , but a solution may not exist, may be unstable, or may involve impractical trading. If there are fewer independent hedge directions than risk tenors, some exposures necessarily remain. If instruments are nearly redundant, exact matching can be numerically fragile.
A constrained hedge program may minimize a specified expression such as , subject to legal, liquidity, collateral, concentration, asset-eligibility, and trading limits. Here is an assumed yield-shock covariance, represents a cost or penalty structure, and determines the trade-off. This is a standard illustration of quadratic hedging, not a proposed proprietary insurance optimization theory. Its solution is meaningful only after validation of risk estimates, exposures, market depth, and the loss scenarios outside the covariance estimate.
Government bonds, interest-rate swaps, futures, caps, floors, and swaptions offer different maturity patterns and liquidity characteristics. Vanilla swaps may efficiently redistribute long-end DV01 but introduce collateral, counterparty, and basis exposures. Swaptions and other options can address nonlinear guarantees but add volatility risk and require model-consistent calibration. Cash bonds deliver contractual inflows and potential liquidity or eligibility benefits that a derivative cannot reproduce. The choice of instruments should follow the liability structure and governance constraints.
Continuous maintenance rather than point-in-time matching
Roll-down, new business, premium receipts, claim settlement, surrender activity, changing market conditions, and hedge settlement continuously alter key-rate exposure. An insurer can maintain a tolerance band and rebalance when limits or forecasts require action. Such a process must specify who owns the liability assumptions, who validates the curve and derivative pricing, who monitors collateral, and who authorizes trades. A position limit stated only in parallel-duration units is insufficient when the portfolio's risk is concentrated in curve-shape directions.
IAIS ICP 16.5 links the ALM policy to product development, pricing, investments, liquidity, and scenario valuation. This supervisory perspective gives an institutional rationale for joint accountability. A hedge that improves one investment desk's DV01 while increasing the group's liquidity exposure may fail the intended risk-management objective.
Scholarly verification and model risk controls
Scientific evaluation requires distinct verification levels. A mathematical identity such as the linear key-rate decomposition can be checked by derivation. A numerical example can be recomputed from declared inputs. A statistical claim about curve factors needs the original sample, instruments, estimation method, and competing specifications. A supervisory claim requires the actual dated document and jurisdiction. A management recommendation requires clarity about its assumptions and limits. Combining these levels without attribution invites false precision.
A strong technical validation suite includes: independent baseline repricing; finite-difference convergence at several bump sizes; the sum-of-key-rates to parallel-PV01 reconciliation; comparison between zero-rate and quote-rate shocks; sensitivity to interpolation and extrapolation; separate up and down shocks; full repricing of twist and butterfly scenarios; reconciliation of first- and second-order estimates; stochastic revaluation of contract options; model-versus-observed cash flows; and collateral liquidity under the same market shocks. Any discrepancy should be quantified and traced to a documented assumption or defect.
The actual EIOPA 2024 stress exercise evaluates insurer capital and liquidity with fixed and constrained balance-sheet assumptions. Its separation of management-action regimes supplies a useful validation lesson: a response observed after hypothetical management actions is not the passive, unhedged financial sensitivity of the initial balance sheet. Similarly, the UK LIST 2025 analysis documents material interest-rate and credit channels alongside management responses. Scenario assumptions should remain reconstructible across institutional reports.
Evidence quality, counterexamples, and remaining limitations
The illustrative table proves that the aggregate-duration match does not imply shape immunization under the stipulated fixed-cash-flow zero-rate model. It does not estimate the distribution of future curve changes; does not show how actual insurer obligations respond to lapses, longevity, or guarantees; and does not establish that a particular hedge is optimal. Those are separate empirical or institution-specific questions. The magnitude of a given insurer's net key-rate exposure depends on its contracts, assets, valuation conventions, management actions, and reporting perimeter.
Counterexamples clarify the proper scope of the findings. A small short-tail insurer with near-term claims and highly liquid short assets may be well served by coarse tenor granularity. A long-duration annuity writer with large thirty- to sixty-year promises may require detailed long-end stress despite limited trading information. An insurer with perfectly matched contractual cash flows may still be exposed to credit defaults, surrender acceleration, inflation, or derivatives collateral calls. The maturity grid should be fitted to risk materiality, not selected for visual uniformity.
For international research, empirical comparisons must account for currency, product mix, market liquidity, accounting classifications, regulatory discounting, and observation dates. A factor estimated from U.S. Treasury yields is not automatically a valid factor for euro swap rates or an emerging-market government curve. Time-varying policy regimes can also change the economic meaning of a short-rate surprise. The scientifically defensible approach is to retain these boundaries rather than generalize a single calibration.
Integrated interpretation
The evidence supports four connected findings. Curve-shape measurement extends the one-dimensional protection of parallel immunization by resolving maturity-specific exposure. The key-rate and factor approaches provide complementary representations whose adequacy depends on the shocks and cash flows of interest. Insurance-specific optionality, long-end extrapolation, credit bases, and liquidity can dominate a clean fixed-cash-flow calculation. Finally, governance quality depends on whether the reported risk can be traced from contracts and market inputs to sensitivity, hedge, capital treatment, and cash funding.
These findings are an original synthesis of established financial and actuarial methods, source-based supervisory evidence, and the reproducible case constructed here. They do not establish a new named theory, a new actuarial principle, or a universal numerical risk limit.
Conclusion
Yield-curve shape is a central dimension of insurance interest-rate risk because the maturity distribution of asset cash flows rarely replicates the timing, contingencies, and valuation bases of the insurer's obligations. Aggregate duration describes sensitivity in a specified parallel direction, while key-rate PV01 resolves the distribution of first-order exposure across tenors. Factor representations of level, slope, and curvature provide a complementary basis for interpreting empirical movements and designing stress scenarios. The choice of curve, tenor grid, bump rule, and interpolation determines what the reported sensitivities actually mean.
The fully specified hypothetical balance sheet established the key mathematical result. Asset and liability values of $100 million each and equal 16.26-year aggregate duration produced zero first-order parallel PV01, but a reverse-butterfly shock resulted in an exact economic surplus loss of about $1.74 million. This demonstration shows how maturity concentration can dominate a scalar duration hedge. The divergence between exact and first-order results also documents convexity effects without attributing real-market probabilities to an artificial scenario.
International evidence reinforces the structural relevance of the result. IAIS supervision requires deliberate ALM and scenario-based valuation; EIOPA and UK stress testing combine interest-rate changes with capital, credit, and liquidity consequences; U.S. regulatory discussions describe the effect of rate levels on investment value, surrender behavior, and cash-flow mismatch. Accounting and prudential rules determine how portions of these economic changes are recognized. In the European Union, the January 30, 2027 application date for the 2026 Solvency II amendments remains a specific legal boundary at the manuscript's November 2026 date.
A defensible insurance term-structure risk assessment consequently preserves gross and net key-rate positions, validates the parallel reconciliation, tests nonparallel and nonlinear scenarios, documents long-end extrapolation and basis risk, and measures policyholder and collateral cash-flow responses. It treats hedging feasibility and risk-governance responsibilities as part of the underlying financial analysis. These conclusions follow from mathematical derivation and inspected sources; they remain contingent on the instruments, contracts, market data, and valuation basis under investigation.
References
American Academy of Actuaries. (2026). Fixed Indexed Annuities: Product Mechanics and Risk Management (policy paper, February), section "Asset/Liability Management, Investment Strategy, and Hedging Considerations." https://actuary.org/wp-content/uploads/2026/02/life-FIA-policypaper.pdf
Australian Accounting Standards Board. (2022). AASB 17 Insurance Contracts, compiled December 2022, paragraphs 36 and B72-B85. https://standards.aasb.gov.au/aasb-17-dec-2022
Bank for International Settlements. (2024). "Shifting landscapes: Life insurance and financial stability." BIS Quarterly Review, September 2024. https://www.bis.org/publications/qr-202409/shifting-landscapes-life-insurance-and-financial-stability
Bank of England, Prudential Regulation Authority. (2025). Life Insurance Stress Test 2025: Results, November 17. https://www.bankofengland.co.uk/prudential-regulation/publication/2025/november/list-2025-results-report
Diebold, F. X., & Li, C. (2006). Forecasting the term structure of government bond yields. Journal of Econometrics, 130(2), 337-364. https://doi.org/10.1016/j.jeconom.2005.03.005
European Commission. (2015). Commission Delegated Regulation (EU) 2015/35 supplementing Directive 2009/138/EC (Solvency II), interest-rate risk provisions. https://eur-lex.europa.eu/eli/reg_del/2015/35/oj
European Commission. (2026). Commission Delegated Regulation (EU) 2026/269 amending Delegated Regulation (EU) 2015/35, Official Journal of the European Union, February 18; application January 30, 2027. https://eur-lex.europa.eu/legal-content/EN/ALL/?uri=OJ%3AL_202600269
European Insurance and Occupational Pensions Authority. (2024). Insurance Stress Test 2024: Scenario, methodology, results, and recommendations. https://www.eiopa.europa.eu/insurance-stress-test-2024_en
European Insurance and Occupational Pensions Authority. (2026). Risk-free interest rate term structures: Monthly technical information and technical documentation, including the November 2026 update. https://www.eiopa.europa.eu/tools-and-data/risk-free-interest-rate-term-structures_en
European Insurance and Occupational Pensions Authority. (n.d.). Solvency II Single Rulebook, Article 77a: Extrapolation of the relevant risk-free interest rate term structure. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2161_en
European Insurance and Occupational Pensions Authority. (n.d.). Solvency II Single Rulebook, Article 77c: Calculation of the matching adjustment. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2163_en
Federal Reserve Board. (2026 access). Nominal Yield Curve: Data, definitions, methodology, and qualification as a staff research product. https://www.federalreserve.gov/data/nominal-yield-curve.htm
Fisher, L., & Weil, R. L. (1971). Coping with the risk of interest-rate fluctuations: Returns to bondholders from naive and optimal strategies. The Journal of Business, 44(4), 408-431. https://doi.org/10.1086/295402
Gurkaynak, R. S., Sack, B., & Wright, J. H. (2007). The U.S. Treasury yield curve: 1961 to the present. Journal of Monetary Economics, 54(8), 2291-2304. Federal Reserve working-paper and data record: https://www.federalreserve.gov/econres/feds/the-us-treasury-yield-curve-1961-to-the-present.htm
Ho, T. S. Y. (1992). Key rate durations: Measures of interest rate risks. The Journal of Fixed Income, 2(2), 29-44. https://doi.org/10.3905/jfi.1992.408049
International Association of Insurance Supervisors. (2024). Insurance Core Principles and ComFrame, ICP 16.5, asset-liability management, revised December 2024. https://www.iais.org/icp-online-tool/
International Financial Reporting Standards Foundation. (2023). IFRS 17 Insurance Contracts, standard overview and history; effective from January 1, 2023. https://www.ifrs.org/issued-standards/list-of-standards/ifrs-17-insurance-contracts/
Litterman, R., & Scheinkman, J. (1991). Common factors affecting bond returns. The Journal of Fixed Income, 1(1), 54-61. https://doi.org/10.3905/jfi.1991.692347
National Association of Insurance Commissioners. (2025). Interest Rates and Insurance, updated September 29, 2025. https://content.naic.org/insurance-topics/interest-rates-and-insurance
Nelson, C. R., & Siegel, A. F. (1987). Parsimonious modeling of yield curves. The Journal of Business, 60(4), 473-489. https://doi.org/10.1086/296409
Svensson, L. E. O. (1994). Estimating and Interpreting Forward Interest Rates: Sweden 1992-1994. NBER Working Paper 4871. https://www.nber.org/papers/w4871