RESEARCH DOSSIER
Credit Spread Risk in Insurance Investment Portfolios
Abstract. Credit spread risk in an insurer's investment portfolio is the possibility that changes in the market compensation demanded for bearing issuer credit, liquidity, and related bond-market risks will diminish the value, usable liquidity, earnings capacity, or regulatory solvency position of assets backing insurance obligations. This monograph distinguishes the market revaluation of surviving securities from rating migration and realized default; derives spread duration, convexity, key-rate sensitivity, transition probabilities, and portfolio concentration effects; and embeds those measurements in insurance asset–liability management. It compares economic valuation with IFRS financial reporting and the distinct prudential treatments of the European Union, the United States, and the International Association of Insurance Supervisors. Primary regulatory texts, contemporaneous supervisory assessments, and foundational asset-pricing research provide the evidence. Transparent numerical illustrations show that the effect of an identical bond-market shock on an insurer's surplus depends on asset composition, liability measurement, policyholder liquidity, capital rules, and allowable adjustments. The central analytical finding is that spread exposure is a joint property of securities, financing and claims cash flows, valuation basis, and institutional constraints; no standalone bond spread statistic is a sufficient proxy for insurer solvency. The analysis establishes no new proprietary theory and makes no empirical inference from the illustrative scenarios.
Keywords: Credit spread risk; insurance investment portfolios; spread duration; option-adjusted spread; market liquidity; rating migration; default intensity; loss given default; asset-liability management; matching adjustment; Solvency II; Insurance Capital Standard; risk-based capital; IFRS 9; IFRS 17; insurer liquidity stress; credit concentration.
Introduction
Insurance investment portfolios connect financial-market promises to claims obligations that may extend over decades. The contract holder's right to payment is financed partly through assets whose issuers can deteriorate long before any contractual default occurs. Insurers therefore experience a distinctive sequence of credit-risk manifestations: market value changes immediately as investors revise required spreads; observable issuer quality may then migrate; and some issuers eventually fail to pay. Each stage can affect the insurer through a different balance-sheet, liquidity, financial reporting, or capital channel. A policyholder need not experience an asset default for an insurer to face an immediate solvency or liquidity problem.
Scale makes this question economically material. The National Association of Insurance Commissioners reported $5.728 trillion of bonds, or 59.8% of U.S. insurers' $9.581 trillion cash and invested assets, at year-end 2025 on the book/adjusted carrying-value basis in its 2025 asset composition report (NAIC, 2026). That industry-wide proportion is not the portfolio weight of every insurer and does not represent a fair-value measure of spread loss. The European Insurance and Occupational Pensions Authority's October 2026 risk dashboard reports median bond exposures near 49.7% of assets for its reporting population and a medium aggregate credit-risk assessment alongside increased bond-market volatility (EIOPA, 2026). Different bases, populations and denominators explain why these observations cannot be combined mechanically.
The study examines publicly traded bonds and comparable market-observable fixed-income claims held for insurer general accounts. Private loans and opaque illiquid credit instruments enter only when necessary to identify boundaries of market-spread measurements. Credit insurance underwritten by the insurer, mortgage underwriting, catastrophe losses, and asset-intensive reinsurance are adjacent subjects rather than alternative definitions of the object here. The research question is how a change in the observable price of credit risk becomes an economic loss, a capital requirement, an accounting item, or a liquidity demand for a specific insurance balance sheet.
The research method is comparative and analytical. Sources comprise the original asset-pricing literature, the published IFRS descriptions, the Solvency II rulebook and stress-test materials, IAIS capital-standard records, NAIC sector data, and BIS financial-stability research. The mathematical analysis uses explicit valuation assumptions and small synthetic examples. Unless stated otherwise, prices and exposures in examples are hypothetical, shocks are instantaneous parallel changes in option- and currency-consistent spreads, and derivatives, tax and management actions are excluded. No proprietary insurer dataset, counterparty-level holdings, or model calibration is claimed.
Credit and Spread Fundamentals
Economic meaning of a spread
For a conventional fixed-income security, the quoted yield may be represented schematically as a reference risk-free yield plus a spread. The spread is a price-derived composite. Compensation for expected default losses, risk premia for unexpected losses, trading illiquidity, taxes, security design, collateral, optionality, and market segmentation can all affect it. The distinction is fundamental because a 100-basis-point spread change cannot automatically be interpreted as a corresponding change in actuarial default probability. Huang and Huang's credit-risk decomposition study (2012) finds that structural default-risk models explain only a minority of investment-grade corporate–Treasury spreads in their calibrated comparisons, with a larger default-risk component for high-yield debt. Their estimates concern specified bonds, models and sample periods rather than a timeless numerical conversion factor.
Reference curves matter. Government yields can contain convenience and sovereign-credit components; swap curves contain their own collateral and market conventions. A spread measured to one curve need not equal a spread measured to another. Similarly, yield spread, zero-volatility spread, asset-swap spread, credit default swap premium and option-adjusted spread encode different cash flows, collateral rules, options and conventions. An option-adjusted spread relies on an interest-rate and prepayment/call model, and its error may masquerade as a credit move. A careful insurer risk inventory records quote source, reference curve, currency, seniority, embedded options, pricing date, and treatment of accrued interest for each position.
The distinction between market information and economic cause has a long foundation. Merton's corporate debt valuation analysis (1974) relates risky debt to firm assets, obligations, asset volatility and an effective default boundary. In contrast, Duffie and Singleton's defaultable term-structure models (1999) incorporate default into reduced-form discounting under chosen recovery assumptions. These frameworks answer different questions. Structural models connect credit deterioration to firm-value economics; reduced-form models are often convenient for pricing and risk sensitivities. Neither directly supplies the insurance liability valuation, realizable-sale proceeds, or supervisory capital impact.
Evidence on common movement and liquidity
Credit spread shocks combine issuer-specific news and market-wide repricing. Collin-Dufresne, Goldstein, and Martin's study of spread changes (2001) reports substantial common unexplained variation in the industrial bond data they examined, beyond standard issuer variables. Its historical sample is evidence for joint market drivers, not a demonstration that the residual factor has a single invariant cause. In an insurer portfolio, apparently diversified issuers can therefore experience simultaneous negative mark-to-market returns when discount premia, liquidity preferences or dealer intermediation conditions shift.
Liquidity and default risk should be measured separately even though their market prices interact. An investment-grade bond held for long-term cash flow collection may remain capable of honoring every payment while suffering a price reduction as market liquidity deteriorates. The price reduction is economically relevant to an insurer that needs collateral or must sell assets to fund surrender requests or catastrophe claims. Conversely, if the insurer can match liabilities without forced sales, short-run sale-price fluctuations can have a different cash-flow significance; they remain relevant to market-value solvency and some financial reporting regimes. Hold-to-maturity intent is not a substitute for testing actual cash-flow requirements.
The BIS life-insurance analysis (2024) describes the increased significance of less-liquid assets and interconnected reinsurance arrangements for financial stability. This monograph does not transfer empirical conclusions about private funds to publicly traded bonds. It uses the BIS findings to establish why funding needs, collateral structures and asset encumbrance can transmit credit-market stress into the insurer even when the nominal credit-quality distribution is unchanged.
Bonds, insurers and the time horizon
Insurers invest across sovereign, municipal, corporate, covered, structured and securitized bonds. Identical quoted spreads across these categories need not imply equivalent underlying loss distributions or capital treatment. A government obligation may carry political, inflation and currency risk; a senior corporate bond depends on borrower leverage and covenants; structured securities depend on collateral pools, payment waterfalls and prepayment behavior. Ratings are summary opinions with rating-scale differences and migration timing. Portfolio governance needs security-level cash flows and legal terms, with rating class used as one dimension rather than the full measurement system.
Life insurers typically confront long-dated contractual cash flows and options such as surrender, guarantees and policyholder participation. Property and casualty insurers may need faster liquidity after claims shocks. An insurer's effective investment horizon is a joint output of liability runoff, premium inflows, collateral calls, reinsurance balances, capital constraints and management discretion. The contractual maturity of an asset is only one input. This observation becomes decisive when reconciling fair-value spread loss with realized credit loss and when designing a credible liquidation stress.
Spread Duration
From price to sensitivity
Let P(s) be a bond's dirty market price when the chosen spread curve is shifted in parallel by s, holding the reference yield curve, expected cash flows, option model and other state variables fixed. Spread duration is defined by D_s = -(1/P)(∂P/∂s). For a small shock Δs expressed as a decimal, the first-order proportional price change is ΔP/P ≈ -D_s Δs. Thus 100 basis points equals 0.01 and a bond with spread duration five loses approximately 5% for an increase of 100 basis points before convexity and any change in the expected cash flows. This is a sensitivity identity, not an estimate of the likelihood or magnitude of the shock.
Define spread convexity C_s = (1/P)(∂²P/∂s²). Then ΔP/P ≈ -D_s Δs + 0.5 C_s (Δs)². In a simple fixed-cash-flow discount model, both quantities follow directly from discounted cash-flow derivatives. For a fixed-rate option-free security with continuously compounded discount factors, P(s)=Σ_t CF_t exp[-(r_t+s)t]; D_s is the present-value-weighted average payment time and C_s is the present-value-weighted average squared payment time. For callable or mortgage-related instruments, expected cash flows depend on the state and the spread convention; analytic fixed-flow sensitivities can then fail, requiring full repricing under specified option assumptions.
A portfolio's spread dollar duration is Σ_i P_i D_{s,i}. The relative duration is this dollar duration divided by current portfolio market value, provided prices and exposures use a consistent base. For a nonparallel shock, define spread key-rate sensitivities to the tenor points of a chosen credit curve or bucketed issuer-sector curves. The approximate loss is a sum over key-rate shocks, with cross-gamma and migration effects added where material. One aggregate duration conceals whether exposure is concentrated in short-maturity subordinated banks, long-dated utilities, or callable securities whose sensitivity changes after a shock.
Transparent duration calculations
Consider a hypothetical $100 million market-value bond portfolio with spread duration 5.0. A 150-basis-point parallel spread widening gives a first-order loss of $100m × 5.0 × 0.015 = $7.50 million. A 50-basis-point widening gives $2.50 million. A 200-basis-point widening gives $10.00 million. These are first-order valuations, not realized default losses or calibrated market forecasts. Convexity and curve shape change the precise answer; they can matter materially under severe shocks.
If the portfolio instead consists of $60 million with duration four and $40 million with duration seven, aggregate spread dollar duration is $60m × 4 + $40m × 7 = $520 million-years, and portfolio spread duration is 5.2. A uniform 100-basis-point shock therefore produces an approximate $5.20 million loss. However, if short-sector spreads increase by 50 basis points and long-sector spreads by 200, the two-position approximate loss is $1.20m + $5.60m = $6.80 million. The difference from the $5.20m parallel result is a direct demonstration of curve and sector composition risk.
These quantities are most informative when linked to data quality. A pricing yield for a stale infrequently traded bond, an incorrect currency curve or a missing embedded option can produce a seemingly precise duration with little forecasting meaning. Internal valuation governance should store calculation date, vendor data, model version, curve identity, spread convention, instrument eligibility and sensitivity bump sizes. Independent full-repricing checks against finite-difference duration reduce coding and sign errors, although they do not validate the economic assumptions of the model.
Separating interest rate and spread risk
Rate duration and spread duration are distinct partial derivatives: D_r = -(1/P)∂P/∂r and D_s = -(1/P)∂P/∂s under the chosen curve construction. A government rate shock can affect insurance liabilities through their discount curve, whereas a corporate credit spread shock may primarily affect assets unless a specific accounting or prudential liability adjustment is legally available. For an option-free bond with a flat reference rate and an additive constant spread, rate and spread duration can numerically coincide. It is still necessary to shock the factors separately because the insurer's liability response is different and correlation between rate and credit spreads is state-dependent.
Market valuations may use spread-to-swap or spread-to-government curves, and a change in the reference instrument itself changes the measured spread. These benchmark-basis effects complicate a clean two-factor attribution. In practice, a full risk decomposition should reconcile the sum of factor effects with observed security price changes and identify residual contributions from curve interpolation, cash-flow assumptions, accrued interest, model error and transaction pricing. A decomposition that cannot explain material unexplained price moves has not yet become a dependable management measurement.
Hedging and residual exposure
Credit default swaps, bond index hedges and total-return swaps can transfer specified spread or default exposure, but their payoff conventions introduce counterparty, basis, collateral and rollover risks. A derivative may hedge an issuer's hazard premium while leaving the cash bond's liquidity discount, funding spread and bond–CDS basis. An index hedge may respond to benchmark index constituents differently from a bespoke insurer portfolio. Duration neutrality on the initial curve does not ensure neutrality after rating migrations or nonlinear option exercise. Hedge effectiveness requires scenario-level valuation of both cash assets and hedges using consistent settlement assumptions.
The insurer's liabilities impose further constraints. A hedge requiring variation margin during adverse spread moves can create a cash demand before an asset appreciation is realizable or recognized on the same basis. Collateral transformation, repo haircuts and eligible collateral restrictions can therefore convert a hedged market-value position into a liquidity-sensitive one. A successful hedge policy is assessed by post-hedge economic surplus, liquidity buffers, counterparty credit risk and regulatory treatment, not by one reduction in reported spread duration alone.
Rating Migration and Default
Migration as a stochastic state transition
Market spreads reprice anticipated future credit conditions. Rating migration records changes in an assigned credit-quality state; default is one possible terminal event. Jarrow, Lando, and Turnbull's Markov credit-rating model (1997) formalizes rating-state transitions and their role in term structures of credit spreads. With a discrete transition matrix T whose rows sum to one, an initial rating-state distribution p_0 evolves to p_n = p_0 T^n when the matrix is time-homogeneous and the period is fixed. A continuous-time transition intensity matrix Q produces T(t)=exp(Qt). Default may be represented as an absorbing state. The matrix must be conditional on the relevant population, geography, sector, issuer type and historical period, because none of those characteristics is automatically homogeneous.
Transition matrices estimated from physical rating histories are not risk-neutral probabilities. They are useful for expected loss and financial planning under a specified real-world scenario. Market-implied default intensities incorporate risk premia and recovery assumptions. In a simplified intensity model with constant hazard λ and fractional recovery R, a heuristic credit component of spread is approximately λ(1-R) for short horizons and restrictive assumptions. The relationship is model-dependent, particularly for nonzero recovery timing, taxes, risk aversion, term structure, liquidity premia, discounting and dependence between default and rates. Using market spreads to infer λ and then adding the full spread again as an independent expected loss would count part of the same risk twice.
Merton (1974) explains credit deterioration through the relation of firm asset value to obligations, whereas reduced-form default pricing (Duffie and Singleton, 1999) conditions prices on default arrival and recovery conventions. Both are compatible with a disciplined insurance risk inventory provided the portfolio engine identifies whether it is reporting market-consistent price, physical expected loss, capital tail loss or realized cash flow. Rating categories simplify communication but do not fully identify recovery, liquidity, security seniority or joint economic sensitivity.
Worked migration calculation
Take a synthetic $50 million bond position initially assigned to BBB. Suppose that over a one-year horizon it has probabilities 88% of remaining BBB, 8% of migrating to BB, 3% of upgrading to A and 1% of defaulting. These invented probabilities sum to 100% and are used only to make the arithmetic auditable. Suppose conditional BB migration entails an additional 180-basis-point spread widening relative to the BBB state, with modified spread duration 4.5. The first-order conditional price loss is 4.5 × 0.018 = 8.1%, and the scenario-probability-weighted contribution is $50m × 0.08 × 0.081 = $324,000. If default recovery is 40% of principal, the gross expected default shortfall is $50m × 0.01 × 0.60 = $300,000.
The $324,000 and $300,000 figures describe mutually exclusive synthetic end states, not two incremental charges that can be added uncritically to an already market-implied spread valuation. A full expected one-year return would additionally require the cash coupon, pull-to-par, initial market value, upward migration gain, changes in reference rates and credit risk premia, and proceeds at the exact recovery timing. The two terms illustrate why migration loss and default shortfall must be specified precisely before aggregation. They are not an actual historical migration table or a recommendation for regulatory calibration.
Frequency, severity and dependence
For name i, expected discounted credit loss can be written schematically as exposure-at-default EAD_i × probability-of-default PD_i × loss-given-default LGD_i, modified for discounting, recoveries and timing. The expression is a planning approximation; it is not equivalent to the observed spread or to an IFRS 9 expected-credit-loss amount without applying the relevant accounting model. Seniority, collateral and covenants affect LGD. Exposure may rise with accrued interest, derivative settlement, reinvestment or drawn commitments. Sovereign restructuring introduces distinct legal and recovery mechanisms.
Concentration is a second dimension beyond name-level expected loss. If D_i is a Bernoulli default indicator, Cov(D_i,D_j) contributes to aggregate default-loss variance, and common macroeconomic shocks can also change recoveries. For two equal Bernoulli names with each annual default probability 1%, independence yields joint default probability 0.0001 (0.01%). If their default-indicator correlation is hypothetically 0.20, the binary covariance identity gives 0.0001 + 0.20 × 0.01 × 0.99 = 0.00208 (0.208%), more than twenty times the independent joint frequency. This arithmetic is a dependence illustration, not an estimated insurer correlation. Loss severity dependence could increase stress further.
Sector, country, currency, financial-parent and counterparty group should be inspected simultaneously. Security-level diversification may conceal exposure to one issuer through different debt tranches or to a single collateral pool through multiple structured instruments. Reinsurer receivables and investment bonds can also create a shared counterparty concentration. For life insurers, the joint distribution of defaults, surrenders and funding conditions may matter more for survival than the marginal rating mix. Correlated default and asset-liability liquidity need a coherent systemic stress.
Recovery and issuer distress
Recovery is economically a distribution, not merely a constant percentage. Legal seniority, collateral values, bankruptcy regime, workouts, restructuring costs, exchange-rate controls and time to payment matter. Near-default credit spreads can reflect recovery uncertainty as well as rising default intensity, and different securities of the same issuer may have distinct terms. A practical credit loss simulation should recognize recoveries when funds can realistically be collected and should avoid assuming immediate par-value cash settlement.
A downgrade can induce an insurer to sell before ultimate default because it breaches a board investment limit, affects eligible capital, triggers collateral terms or increases risk-based capital. Such policy-mediated selling can crystallize an otherwise unrealized loss. The correct stress horizon therefore follows both issuer cash-flow dynamics and the insurer's governance constraints. A modeled position that remains on the books despite a mandatory eligibility violation is not a complete institutional scenario.
Asset-Liability Spread Mismatch
The insurer's economic surplus
Let A be the economic value of invested assets and L the economic value of insurance liabilities under a consistent reporting basis. Economic surplus is E=A-L. For changes in market spreads, ΔE = ΔA - ΔL. This identity is simple; determining either component under a valid set of assumptions is the substantive problem. Assets may incorporate a quoted spread while liability discount rates are governed by separately prescribed rules. Thus an insurer can be duration matched against risk-free interest rates yet remain materially exposed to credit spread widening. The relevant mismatch is not only between asset and liability maturities; it concerns which market factors enter their respective valuations and how their cash flows respond to policyholder behavior.
Assume a hypothetical insurer begins with $100 million of bond assets and $92 million of liabilities valued on an economic basis, leaving $8 million surplus. If assets have spread duration five and spreads rise by 150 basis points, the first-order asset loss is $7.50 million. If the liability discount curve and projected liability cash flows are unaffected by that credit shock, the resulting surplus is approximately $0.50 million. In this illustration, a spread change equal to 1.5% in yield absorbs 93.75% of starting surplus. The ratio demonstrates leverage of market-value capital to credit premia; it has no estimated probability attached.
Now consider a simultaneous upward parallel risk-free rate shift of 100 basis points. If asset rate duration is five, the asset rate loss is approximately $5 million. If liability rate duration is seven, the liability value decreases approximately $6.44 million, giving a first-order $1.44 million benefit to surplus from the rate shift alone. With the independent $7.50 million credit-spread asset loss also applied, combined surplus change is approximately -$6.06 million, leaving $1.94 million. In the absence of nonlinearities and covariance terms, this is simply $1.44m - $7.50m. Actual liability options, negative convexity, spreads correlated with rates, taxes and market-consistent discount adjustments may change the result materially. The arithmetic explains why it is analytically incorrect to treat a single observed bond yield movement as either an interest-rate or a credit-spread shock without decomposition.
Cash-flow matching and financial strength
For an insurer managing long-duration general-account liabilities, bond cash flows are useful because they can provide contractual payments aligned to predictable policy benefits. Alignment reduces exposure to liquidation at an unfavorable time, although it does not eliminate the risk of missing the contractual cash flows through default or restructuring. A matched portfolio needs sufficient granularity across payment dates and scenarios, not simply a common duration or weighted average life. Premium lapses, mortality changes, policyholder exercise, inflation-linked claims and reinsurance payment delays can all change the needed asset cash-flow schedule after market stress.
Insurance asset–liability analysis should distinguish three horizons. Immediate market-value solvency asks how market prices and technical provisions revalue at the shock date. The liquidity horizon asks whether cash and eligible collateral can be delivered during the next days, weeks and months. The ultimate cash-flow horizon asks whether the bonds will pay enough on time to meet contractual obligations over years. A portfolio can look sound on the third horizon while suffering a severe first-horizon capital decline and a second-horizon liquidity squeeze. Conversely, a temporary spread recovery cannot erase a genuine contractual default loss. This distinction governs both the design of stress tests and the interpretation of accounting reserves.
Discount-rate adjustments under Solvency II
The EU Solvency II framework contains regulated mechanisms for recognizing specified components of asset spreads in liability discounting. The matching-adjustment calculation under Article 77c explicitly deducts a fundamental spread designed to represent retained risks including default and downgrade effects; the permitted adjustment depends on qualifying portfolios and other statutory conditions (EIOPA, current rulebook). The volatility-adjustment provisions under Article 77d provide another defined mechanism based on risk-corrected representative spreads and are not interchangeable with the matching adjustment. These are regulatory valuation rules, not arbitrary company-chosen recognition percentages.
Article 181 of the delegated regulation also addresses the impact of a spread stress on matched portfolios: the matching-portfolio spread scenario requires both shocked asset values and a recalculation of technical provisions accounting for the changed matching adjustment. This makes a generic asset-loss-only approximation inadequate when the insurer actually qualifies for and uses the mechanism. It also means that applying hypothetical liability relief to every insurer would be incorrect. Matching-adjustment eligibility, assigned-asset cash flows, permissible obligations, fundamental-spread assumptions, supervisory approval and the legally applicable date must be recorded before reporting any compensated surplus loss.
In the earlier $100m / $92m illustration, imagine a strictly hypothetical modeling sensitivity in which an additional 25 basis points enters the liability discount curve and liability duration is seven. The first-order liability decrease would be $92m × 7 × 0.0025 = $1.61 million; the net credit-spread surplus loss would be $7.50m - $1.61m = $5.89m. The 25-basis-point liability effect is an invented response parameter for sensitivity teaching. It is not a prescribed Solvency II volatility adjustment, matching adjustment, regulatory factor, or statement of entitlement. The only scientifically valid use of this illustration is to show the direction and magnitude of valuation-basis sensitivity.
Liquidity, surrenders and market amplification
The aggregate effect of spread widening on insurer resilience depends on liquidity access. If surrender behavior, collateral calls or claim settlements require asset sales after bond prices decline, the insurer realizes losses, potentially concentrates the remaining book in illiquid names and forfeits future contractual coupons. Haircuts and spread-sensitive repo collateral values further complicate funding. A portfolio with diversified issuers can be concentrated in one dealer market or one liquidity channel. The insurer's management policy should model asset sales net of transaction costs and determine what remains after meeting the liquidity requirement.
The BIS analysis of life insurers (2024) emphasizes the interaction among insurer business models, less-liquid investment exposures and interconnected reinsurance relationships. Its institutional evidence supports testing second-round liquidity and counterparty effects without presuming that all insurers hold the same investment mix. For property and casualty insurers, catastrophe claims can generate accelerated liquidity demands; for annuity insurers, rate-sensitive lapses and guarantees may change expected outflows. Stress transmission should respect contract-specific policyholder incentives rather than assuming one universal lapse response.
Accounting measurements of bond risk
IFRS accounting uses classification and measurement rules that may generate profit-and-loss, other comprehensive income, and impairment outcomes different from an insurer's economic surplus. The IFRS Foundation's IFRS 9 summary states the contractual-cash-flow and business-model criteria relevant to amortized cost, fair value through other comprehensive income and fair value through profit or loss, together with the expected credit loss impairment approach. A spread shock in an otherwise performing bond can create a visible fair-value movement under a fair-value classification, while expected-credit-loss recognition follows its own conditions and measurement requirements. A financial asset's accounting category must not be inferred solely from its rating or maturity.
The IFRS 17 overview explains the measurement and presentation of insurance contracts, including insurance finance income and expenses and the permitted accounting policy choice concerning presentation in profit or loss or other comprehensive income. Matching the presentation of insurance assets and liabilities may reduce certain accounting volatility, but presentation rules do not remove an underlying deficit of economic cash flows or cure an issuer default. The same asset spread shock can therefore produce distinct amounts for economic surplus, statutory solvency, IFRS profit or loss, IFRS other comprehensive income, and management liquidity. A comprehensive insurer risk report reconciles those views rather than declaring one of them the universal loss.
Financial Stress Tests
Stress design as a joint scenario
A stress test is informative only when it states which risks are shocked, which correlations are imposed, what the time horizon is, and which management actions are permitted. For traded-credit portfolios, the fundamental scenarios include a parallel credit spread widening, a sector- or rating-specific shock, curve twist, flight to quality, rating downgrades, issuer defaults, liquidity discount expansion, and simultaneous interest-rate and policyholder-cash-flow changes. A repeat of an observed historical spread event may improve interpretability but cannot stand alone as a bound on future outcomes. Forward-looking stresses should consider market structure, asset concentration, legal changes and collateral practices that differ from past episodes.
European supervisory practice explicitly recognizes both solvency and liquidity consequences. EIOPA's 2024 insurance stress test used a severe macro-financial adverse scenario and prompted a recommendation to improve modeling and reporting of look-through assets, liquidity flows, unit-linked shocks and marginal insurance-specific effects (EIOPA, 2024–2025). That public finding is evidence of supervisory concern with model implementation; it is not a validation of the synthetic numbers in this monograph. The same institutional lesson applies to credit spread shocks: identify what the holdings actually contain, trace indirect fund exposures and calculate cash-flow consequences alongside balance-sheet valuations.
A transparent scenario set can be defined with no claim of market calibration:
| Synthetic scenario | Risk-free rate shock | Parallel credit spread shock | Additional mechanism | Approximate first-order result for a $100m bond portfolio with D_s = 5 |
|---|---|---|---|---|
| Mild repricing | 0 bp | +50 bp | No downgrade or liquidity shock | -$2.50m |
| Broad spread widening | 0 bp | +150 bp | No additional credit event | -$7.50m |
| Severe spread widening | 0 bp | +200 bp | No additional credit event | -$10.00m |
| Compound deterioration | +100 bp | +150 bp | Rating migration, issuer-specific defaults, and redemptions must be separately modeled | -$7.50m spread component, before rate and other effects |
The table isolates the spread-duration component. It deliberately does not supply unestimated conditional default losses, insurance liability reactions, confidence levels, or regulatory capital figures. A meaningful scenario run replaces the placeholders with each insurer's current market valuations, curve sensitivities, transition model, recovery assumptions and cash-flow constraints. The last row cannot be reduced to one additive percentage because default and spread revaluation interact and because liquidations can change the assets remaining for the next time step.
A coherent stress engine
A robust stress engine first fixes the opening asset inventory and all relevant contractual cash-flow models. It reprices reference rates and credit spread curves under the scenario, updates instrument options and bond prices, projects issuer states and recovery cash flows, and computes resulting portfolio trades under predeclared management constraints. It then reprices or projects policy liabilities on the corresponding basis, tests eligible capital, and runs cash and collateral waterfalls. The entire calculation is repeated for relevant currency, credit sector, rating and stress horizon combinations. The opening balance sheet and starting capital denominator must remain traceable through all reports.
Credit spread moves and default events must be generated consistently. If a spread curve scenario already embeds future defaults and the resulting mark-to-market prices, adding a separate full present value of those expected defaults to the same shocked market value duplicates losses. Separate realized-default cash shortfalls can be reported in a later cash-flow horizon or treated as alternative state-contingent paths, with the valuation bridge made explicit. The same concern applies when a downgrade is used both to move the security price to a worse rating curve and to impose an additional independent accounting impairment without a justified accounting reconciliation.
Sensitivity tests should vary spread duration, recovery, rating migration, bid–ask cost, correlation, and liquidity availability over plausible ranges rather than placing all uncertainty into one haircut. For long-dated callable bonds, rerun the option model after the move. For concentrated sectors, simulate several simultaneous issuer downgrades. For funding, require a minimum cash buffer and model realistic sales against market depth. Stress results should include worst cash shortfall, dates of binding capital or liquidity constraints, percentage of assets sold, residual portfolio quality and policyholder obligations fulfilled. Such outputs are actionable because they specify the mechanism of potential failure.
Governance of quantitative models
Model governance has to address structural, parameter, implementation and data uncertainty. Structural uncertainty concerns whether issuer hazard, spread decomposition or policyholder options are correctly represented. Parameter uncertainty concerns transition probabilities, recovery distributions and correlations estimated from sparse or changing data. Implementation uncertainty concerns curves, duration calculations and coding. Data uncertainty concerns stale prices, ratings, missing look-through holdings and ambiguous issuer identifiers. Each category needs different validation evidence.
Effective controls include independent security-price sampling, reconciled holdings, dated rating and recovery data, versioned cash-flow models, direct comparison of analytical sensitivities with bumped full repricing, stress-to-capital reconciliations, and documented model-owner decisions. Backtesting observed spreads against predicted changes is useful, but a model trained on a calm market cannot be certified by performance in calm markets alone. A calculation can be numerically correct while its assumed dependencies or liquidation capacity are economically wrong. Internal validation should document both mathematics and the scope of observed evidence.
Empirical interpretation and responsible inference
The October 2026 EIOPA dashboard describes medium aggregate credit risk, modest spread increases and high market-risk conditions in the European insurance sector, with market indicators largely measured through September 2026. Its median credit quality step of about two and low-rated exposure median near 1.4% indicate high average reported bond quality in that population, not immunity from common-spread losses (EIOPA, 2026). These medians must not be multiplied as if they represented one insurer's exact holdings. Likewise, the NAIC's U.S. insurance investment figures (2026) are reported at book/adjusted carrying value and concern a different market and accounting basis. Both are authoritative for their stated measurement questions and insufficient for estimating a representative insurer's spread value at risk without granular holdings.
An original analysis should distinguish a reported statistic, a model assumption, a derived mathematical identity and a normative control recommendation. The duration loss formula is a mathematical first-order consequence of its definition. The synthetic scenario amplitudes are assumed. The sector allocation figures are empirical official reports. The instruction to reconcile stress effects across economic capital, accounting and cash is an analytical conclusion supported by the different obligations of each measurement system. Labeling these categories prevents scenario arithmetic from being mistaken for validated historical findings or new proprietary theory.
Regulatory Capital and Evidence
European Union: spread risk and liability-sensitive capital
The Solvency II standard formula includes a spread-risk sub-module. Under the Article 176 bond rule, stress factors for relevant bonds and loans depend on modified duration and prescribed credit-quality steps, subject to the detailed instrument classification and exemptions in the applicable law (EIOPA, rulebook). This is a regulatory shock translated into basic own-funds loss, not a simple assertion that the current market-implied spread is expected to rise by that number. The result must be calculated under applicable aggregation and eligible adjustments, avoiding a naïve sum of all standalone risk categories.
For qualifying matching-adjustment portfolios, the Article 181 scenario provision requires specified technical-provision recalculations when spread stress is applied. The fundamental spread rules preserve risk allowances for default and downgrade; the rules cannot be replaced with an insurer's internal market-spread estimate without statutory justification. In other portfolios, the volatility-adjustment rule operates under its own conditions. These provisions together show why capital impact is institution-specific even when two insurers own the same bonds.
The European framework is undergoing an important dated transition. EIOPA's Solvency II legislative overview and Directive (EU) 2025/2 state that the revised framework applies from January 30, 2027. At this monograph's November 15, 2026 research cutoff, those future effective provisions must be described as forthcoming rather than already effective. Legal change therefore requires dual dating of the stress-assessment rule: what applies at the balance-sheet observation date and what will apply for future projections. No numerical transition factor from the future regime is silently assumed here.
United States: asset designations and reporting bases
U.S. insurers operate under state-centered supervisory and statutory accounting systems with risk-based capital requirements. Their bond risk treatment depends on legal entity, asset designation, reporting basis and other characteristics; a corporate bond's market price reduction is not synonymous with its statutory carrying-value change. The NAIC's 2025 asset mix analysis gives a broad empirical inventory of the sector but does not specify the capital factor for a given bond. Proper analysis must connect the insurer's relevant statutory accounting guidance, designation assignment and risk-based capital instructions to its actual book of investments.
Comparison with Solvency II should therefore hold the economic asset scenario fixed and compute each jurisdiction's reporting consequences separately. One should compare price sensitivity, contractual expected loss, impairment recognition, capital charges and realized cash loss in a common reporting table. A claim that one regime invariably requires a lower capital charge than another would depend on asset type, insurer type, available offsets, risk aggregation and the rule version. The evidence examined here does not establish any such universal ranking.
International consistency and the Insurance Capital Standard
The International Association of Insurance Supervisors adopted the Insurance Capital Standard in December 2024 as a globally comparable risk-based capital measure for internationally active insurance groups under ComFrame (IAIS, 2024). Its formal adoption announcement describes the framework's supervisory purpose. The adoption provides an international reference but does not make every national insurer subject to identical capital calibration or convert an ICS ratio directly into a Solvency II SCR or U.S. RBC value. Comparisons must state group perimeter, valuation method, qualifying capital and the implemented local supervisory arrangements.
Financial reporting adds another dimension. IFRS 9 classification and impairment and IFRS 17 liability presentation can affect the timing and location of credit-related amounts in financial statements. Those accounting presentations and prudential capital assessments are related but not identical. A faithful scientific comparison maintains an explicit reconciliation among economic value, accounting carrying values and statutory or prudential solvency positions.
Supervisory evidence and limitations
EIOPA's insurance stress-test findings and October 2026 dashboard demonstrate both the relevance of market shocks and the need for granular insurer data. The NAIC's investment composition study demonstrates the centrality of bonds in U.S. balance sheets. The BIS insurance review addresses vulnerabilities where investment risk, liquidity and institutional links interact. Together these sources justify an integrated research question while stopping short of yielding one cross-country calibrated distribution of insurer spread losses.
The scholarly findings also constrain inference. Huang and Huang (2012) show that default-based models do not explain all yield spreads for the assets they study. Collin-Dufresne, Goldstein, and Martin (2001) find pronounced common spread movements in their dataset. Merton (1974) and Duffie and Singleton (1999) supply different theoretical foundations for issuer valuation and defaultable pricing. The combined reading supports multiple separately labeled spread components, dependence-aware stress design and accounting reconciliation. It does not establish a new general pricing law, empirically invariant liquidity factor or claim of discovery priority.
Integrated Research Findings
A measurable sequence of losses
The first integrated finding is that spread risk must be tracked along a sequence of potential losses: immediate market revaluation, credit-state migration, default and recovery, insurer cash-flow response, accounting recognition and prudential capital impact. The sequence allows several items to occur simultaneously, but it does not permit the same price deterioration to be charged repeatedly without an explicit bridge. A security can lose market value without default; it can default after previously losing value; and it can incur an additional realized cash loss relative to its already-impaired price. The reporting question must specify which difference is being measured.
The second finding concerns asymmetry between securities and liabilities. A government-rate duration match does not generally hedge issuer spread shocks. A market-value loss on a traded bond may have no equivalent effect on the contractual payment obligations or the liability discount basis, although a legally permissible liability adjustment can change reported solvency. The same opening portfolio therefore has several meaningful exposure measures, and each is conditioned on a specified purpose and rule set. In this research, the appropriate unifying unit is a scenario-reconciled insurer balance sheet, not a freestanding average bond spread.
The third finding concerns diversification. Issuer counts and rating distributions are incomplete proxies for joint stress losses because corporate spread movements share systemic components and default dependence can amplify tail events. Financial-sector concentrations, funding channels, correlated recovery values and collateral demands can connect apparently distinct issuers. Rating migration also has institutional consequences independent of eventual cash shortfalls, including forced portfolio changes and eligibility constraints. Stress testing must recognize both distributional dependence and governance-mediated portfolio turnover.
The fourth finding concerns time. A portfolio can be strong in contractual runoff terms and fragile at the immediate liquidity horizon, or report stable amortized-cost carrying values while its economic resale value has fallen. Holding a bond to maturity may reduce forced realization when cash flows match, but it does not remove default uncertainty, investment constraints or solvency obligations. A credible insurer report includes immediate economic capital, monthly liquidity projections and longer-term default cash flows, with assumptions reconciled across them.
The fifth finding concerns regulatory comparison. Legal, accounting and supervisory regimes address different measurement objectives. Solvency II spread stresses and adjustments, NAIC statutory reporting and risk-based capital, ICS group capital, and IFRS classification and liability presentation need not yield the same number for identical instruments. Dates matter: the January 2027 revised EU framework cannot be assigned retrospective force to a November 2026 observation. The valid comparison is a documented bridge from one economic scenario to each regime's actual requirements at the stated date.
Limits of the present synthesis
This work uses authoritative public sources and explicit calculations. It does not independently estimate historic transition matrices, bid–ask liquidity costs, stochastic risk-premium dynamics, or recovery correlations from raw market observations. No microdata were collected from insurers, and the numerical cases are deliberately synthetic. Several foundational journal contributions were verified at the original publisher's abstract, publication metadata and available article presentation rather than independently extracting every equation from full-text subscription material. These access limitations restrict claims about exact derivations or quantitative calibration from those articles. Supervisory reporting populations and valuation denominators remain different across NAIC and EIOPA evidence.
The quantitative examples were checked algebraically and arithmetically using the stated assumptions. They are deterministic sensitivities or simple probability identities, not validated forecasts. The conclusions concern measurement architecture, loss transmission and research interpretation. Their merit follows from the explicit relationships between asset pricing, insurance liabilities and institutional constraints and does not require announcing a proprietary risk theory, doctrine or universal spread decomposition.
Conclusion
Credit spread risk is a central risk of insurer investment portfolios because bonds transform current premium resources into future policyholder payments while remaining exposed to changing market compensation for credit and liquidity risk. A rigorous evaluation starts with the actual securities and their cash flows, identifies reference curves and spread conventions, measures price sensitivity, and independently tracks rating transition and default recovery. It then calculates insurer-specific asset–liability effects and tests whether the available cash, collateral and capital remain adequate under coherent joint scenarios.
The evidence establishes several durable conclusions. Market spread widening is not identical to realized credit default. Spread duration measures a local pricing sensitivity and requires convexity, nonparallel curve, option and liquidity checks for severe scenarios. Rating migration and issuer default demand explicit physical or market-implied assumptions, with disciplined prevention of double counting. Insurance liabilities and liquidity needs determine how bond-market price changes are transmitted into solvency and policyholder security. Accounting and prudential consequences depend on jurisdiction, instrument classification, legal eligibility and effective date. Public aggregate indicators of strong insurer credit quality are compatible with substantial institution-specific spread losses and therefore cannot replace portfolio-level analysis.
The most defensible research practice is to keep economic valuation, contractual credit cash flows, liquidity, financial reporting and capital outcomes separately measurable and reconciled. This original analytical synthesis organizes established financial, actuarial and supervisory evidence for the insurance balance sheet. It makes no claim to a new risk theory or to empirically calibrated loss probabilities. The resulting monograph supplies a research-grounded foundation for assessing how publicly traded credit spreads interact with insurer financial strength and for specifying further data-intensive validation.
References
Bank for International Settlements. (2024, September). Shifting landscapes: Life insurance and financial stability. BIS Quarterly Review special feature. https://www.bis.org/publications/qr-202409/shifting-landscapes-life-insurance-and-financial-stability
Collin-Dufresne, P., Goldstein, R. S., & Martin, J. S. (2001). The determinants of credit spread changes. The Journal of Finance, 56(6), 2177–2207. https://doi.org/10.1111/0022-1082.00402
Duffie, D., & Singleton, K. J. (1999). Modeling term structures of defaultable bonds. The Review of Financial Studies, 12(4), 687–720. https://doi.org/10.1093/rfs/12.4.687
European Insurance and Occupational Pensions Authority. (2024–2025). Insurance stress test 2024: Results and recommendations. EIOPA. https://www.eiopa.europa.eu/insurance-stress-test-2024_en
European Insurance and Occupational Pensions Authority. (2026, October). Insurance Risk Dashboard, EIOPA-BoS-26-495. EIOPA. https://www.eiopa.europa.eu/assets/insurance-risk-dashboard/EIOPA-BoS-26-495%202026%20October%20Insurance%20Risk%20Dashboard.html
European Insurance and Occupational Pensions Authority. (n.d.). Solvency II: Regulatory framework and revision effective dates. EIOPA. https://www.eiopa.europa.eu/browse/regulation-and-policy/solvency-ii_en
European Insurance and Occupational Pensions Authority. (n.d.). Spread risk on bonds and loans: Delegated Regulation (EU) 2015/35, Article 176. Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5796_en
European Insurance and Occupational Pensions Authority. (n.d.). Application of the spread risk scenarios to matching adjustment portfolios: Delegated Regulation (EU) 2015/35, Article 181. Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-5801_en
European Insurance and Occupational Pensions Authority. (n.d.). Calculation of the matching adjustment: Directive 2009/138/EC, Article 77c. Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2163_en
European Insurance and Occupational Pensions Authority. (n.d.). Volatility adjustment to the relevant risk-free interest rate term structure: Directive 2009/138/EC, Article 77d. Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2164_en
European Parliament and Council of the European Union. (2025). Directive (EU) 2025/2 amending Directive 2009/138/EC as regards proportionality, quality of supervision, reporting, long-term guarantee measures, macro-prudential tools, sustainability risks, and group and cross-border supervision. Official Journal of the European Union. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=OJ%3AL_202500002
Huang, J.-Z., & Huang, M. (2012). How much of the corporate–Treasury yield spread is due to credit risk? The Review of Asset Pricing Studies, 2(2), 153–202. https://doi.org/10.1093/rapstu/ras011
IFRS Foundation. (n.d.). IFRS 9 Financial Instruments: Standard overview. IFRS Foundation. https://www.ifrs.org/issued-standards/list-of-standards/ifrs-9-financial-instruments/
IFRS Foundation. (n.d.). IFRS 17 Insurance Contracts: Standard overview. IFRS Foundation. https://www.ifrs.org/issued-standards/list-of-standards/ifrs-17-insurance-contracts/
International Association of Insurance Supervisors. (2024, December 5). IAIS adopts Insurance Capital Standard and other enhancements to its global standards to promote a resilient insurance sector. IAIS. https://www.iais.org/2024/12/iais-adopts-insurance-capital-standard-and-other-enhancements-to-its-global-standards-to-promote-a-resilient-insurance-sector/
International Association of Insurance Supervisors. (n.d.). Insurance Capital Standard: Framework overview. IAIS. https://www.iais.org/activities-topics/standard-setting/insurance-capital-standard/
Jarrow, R. A., Lando, D., & Turnbull, S. M. (1997). A Markov model for the term structure of credit risk spreads. The Review of Financial Studies, 10(2), 481–523. https://doi.org/10.1093/rfs/10.2.481
Merton, R. C. (1974). On the pricing of corporate debt: The risk structure of interest rates. The Journal of Finance, 29(2), 449–470. https://doi.org/10.1111/j.1540-6261.1974.tb03058.x
National Association of Insurance Commissioners, Capital Markets Bureau. (2026, May). U.S. Insurance Industry's Cash and Invested Assets Grow to $9.6 Trillion as of Year-End 2025. NAIC Capital Markets Special Report. https://content.naic.org/sites/default/files/capital-markets-special-reports-asset-mix-ye2025.pdf