RESEARCH DOSSIER
Insurance Liability Cash Flow Forecasting
Abstract. Insurance liability cash flow forecasting converts contractual promises, contingent insured events, operating expenditures, premium receipts, policyholder options, and recoveries into time-indexed distributions of receipts and payments. It is simultaneously an actuarial estimation problem, a conditional financial projection problem, a measurement input under several distinct accounting and prudential regimes, and an operational inquiry into the ability to settle obligations when payment falls due. This monograph develops an integrated account of those connections. It distinguishes cash-flow amounts from their timing, discounted measurement from actual funding, contractual boundaries from business plans, and conditional projections from unconditional expected values. It specifies the state variables, transition structures, data lineage, estimation techniques, scenario dependencies, and controls needed for life, annuity, health, property, and liability insurance. It examines mortality and longevity, claims development, premium persistency, surrender and partial withdrawal options, maintenance and claim-handling expenses, inflation, reinsurance recoveries, and correlated macrofinancial shocks. International comparison centers on IFRS 17, Solvency II, United States principle-based reserving and actuarial standards, and the supervisory assessment of asset-liability and liquidity risk. A series of transparent numerical illustrations demonstrates how the same aggregate expected benefit may produce different present values and funding trajectories, how scenario-responsive surrender behavior alters both present-value and immediate-liquidity exposures, and how a forecast can be internally consistent while materially deficient for a specified decision. The central analytical conclusion is that a defensible forecast is a dated, purpose-specific, evidence-traceable distribution of gross and net settlement cash flows, and that the principal governance obligation is to keep its contract scope, stochastic dependencies, assumptions, valuation conventions, and decision horizon mutually consistent.
Keywords: Insurance liability cash flows; cash flow forecasting; actuarial projection; asset-liability management; IFRS 17; Solvency II; principle-based reserving; claims payment patterns; policyholder behavior; lapse and surrender; mortality; longevity; expense inflation; stochastic scenarios; liquidity risk; model validation; reinsurance recoveries; contract boundary; technical provisions.
Introduction
The defining object of an insurance liability forecast is a dated obligation to deliver cash, or an economically equivalent contractual benefit, subject to conditions that may remain unknown until future events occur. The resulting stream is shaped by insurance risk, policy design, policyholder action, administrative process, market conditions, and the timing of settlement. A figure for the present value of the obligations is one possible function of this stream; it does not describe the stream completely. The same insurer may report a stable expected liability value while facing acute cash demand after concentrated claims, accelerated surrenders, or delayed reinsurance collections.
This work focuses on the production and interpretation of liability-side cash-flow forecasts. The design of a complete stochastic balance sheet is the subject of related research on asset-liability modeling; hedge construction and portfolio immunization also require separate treatment. Here asset cash flows enter because they are the economic means by which liability settlement is financed. The unit of scientific analysis is the liability cash-flow process and its implications for valuation, liquidity and planning.
The formal distinction matters in law and practice. Article 77 of the Solvency II Directive identifies the best estimate as the probability-weighted present value of future cash flows and requires projection of the inflows and outflows necessary to settle obligations over their lifetime, as specified in the calculation of technical provisions. IFRS 17 measures groups of contracts through fulfillment cash flows, together with the contractual service margin where applicable, according to the IFRS 17 accounting principles. U.S. life principle-based reserving establishes projection requirements in the 2026 NAIC Valuation Manual. These frameworks share financial and actuarial ingredients while specifying different valuation purposes, groupings, constraints, and outputs.
The institutional scope of this study is international. Regulatory examples identify the relevant jurisdiction and decision date, rather than treating one accounting or prudential convention as universal. The empirical evidence includes published supervisory exercises and insurance experience studies; such studies establish observations for specified samples and periods, while a forecast of another population requires independent calibration. The mathematical cases below are explicitly hypothetical. They provide reproducible demonstrations of mechanisms and support no assertion about the solvency or performance of a real insurer.
Method and analytical contribution
The research combines primary-source examination of accounting, supervisory and actuarial documents; a comparative interpretation of contractual scopes; elementary survival and multi-state mathematics; time-indexed payment development; deterministic and stochastic scenario analysis; and source-bound empirical assessment of surrender experience and liquidity stress. Original contributions are the cross-framework synthesis, a consistent cash-flow classification, the derivation and testing of illustrative cases, and the identification of conditions under which a forecast is appropriate for a decision. No new theory, supervisory doctrine or proprietary forecasting model is proposed.
A forecast is regarded as scientifically adequate only relative to its named question. The calculation of a reporting-date fulfillment value, the estimation of a six-month cash coverage requirement, the assessment of a five-year asset strategy, and the design of a regulatory stress are different inquiries. Each requires specific choices concerning the projection horizon, contract boundaries, treatment of future business, stochastic measure, gross and net recovery presentation, currencies, discounting and the inclusion of exceptional management actions. These choices should be visible at the outset.
Liability Cash Flow Inventory
Scientific object and measurement purposes
Let denote the valuation date and denote future payment intervals. For a portfolio , define the gross contractual and servicing net outflow in scenario as
Here denotes insured benefits and claim settlements other than surrenders; denotes surrender, withdrawal and related option payments; denotes administration and claim-handling expenses; denotes any other in-scope settlement cash outflows; denotes premiums and directly related contractual receipts; and denotes recoveries allocated to the modeled liability view. This is an analytical ledger, not a formula prescribed for every financial statement. For some reporting regimes recoveries from reinsurance held form a distinct asset and must not be mechanically offset against issued insurance liabilities. A full contractual inventory should first record gross flows, then distinguish ceded and retained views by counterparty, settlement lag and legal presentation rule.
The conditional expected time series is , with describing information available at valuation. A simple discounted expectation is , where is the discount factor appropriate to the measurement and the dependencies among rates and cash flows are retained. The expected nominal cash flow, its expected discounted value, the accounting carrying amount, the prudential technical provision, and a cash buffer for extreme outcomes are distinct statistics. Differences among them arise from the task, governing rules and distributional features.
This separation gives a practical first control: every output should state whether it is gross or net of reinsurance, positive for payment or receipt, nominal or discounted, deterministic or stochastic, and inclusive or exclusive of new business. An output without those specifications is difficult to reconcile and easy to misuse.
Contract boundaries and flow taxonomy
The forecast begins with an enforceable contract or portfolio of contracts and the applicable rules determining which future premiums and benefits belong to the measurement. An insurer's commercial business plan may include renewals, additional contracts and cross-sales; a current-contract regulatory valuation may exclude some of those flows. A liquidity projection may reasonably incorporate future new-business premiums, provided that they are distinguished and stress-tested, because those receipts can cease under adverse conditions. Contract boundary and financial forecasting perimeter therefore need separate fields.
IFRS 17 requires future cash flow estimates within the boundary of groups of insurance contracts, subject to its requirements for recognition and measurement. The IFRS 17 framework also separates insurance service results from insurance finance effects and treats certain investment components distinctly in presentation. Under European prudential rules, the Solvency II cash-flow provision lists payments for benefits, benefits in kind, expenses, premiums, intermediary flows, unit-linked-related payments, applicable recoveries and certain taxation cash flows. The list illustrates why a forecast confined to claim payments may be adequate for a claims department but incomplete for the valuation of whole contractual obligations.
A useful working taxonomy distinguishes payment purpose, insurance contract, event trigger, payment currency, legal entity, projection date, actual settlement date, accounting group, beneficiary, counterparty, tax treatment, inflation index, guarantees, surrender provisions, premium schedule, reinsurance status, and operational process. Product classes then introduce specific contingent drivers. Whole-life insurance features death or endowment benefit probabilities, premiums and possible surrender values. An immediate annuity features payments conditional on continued survival; a guaranteed-period annuity preserves payments for the stated minimum period. Long-tail casualty liabilities depend on the development of incurred claims, litigation, reporting delay, inflation and settlement practice. Short-tail property business exhibits more concentrated reporting and payment periods, with catastrophe claims capable of overwhelming normal seasonality.
The correct taxonomy accommodates different information states. A reported claim with an agreed settlement amount differs from an incurred-but-not-reported claim with an uncertain reporting date. An issued life contract with a scheduled premium differs from a flexible-premium policy whose receipts depend on account value and policyholder intention. A claim amount denominated in foreign currency carries an additional exchange-rate conversion dependency. These are economically consequential differences even if the total nominal forecasts appear similar.
Portfolio classification, data lineage and timing
Every cash-flow forecast requires a reproducible bridge from policy administration, claims, reinsurance, finance and accounting systems. Policy exposure records identify effective dates, contract state, attained age, coverage and premium mode. Claim files identify occurrence, report, assessment, adjudication, payment and closure dates. Expense systems identify acquisition, maintenance, servicing and claim-adjustment expenditures. Treaties identify cession eligibility, attachment, limits, reinstatement, collateral and reimbursement terms. Market datasets provide rates, inflation curves, equity levels, credit conditions and exchange rates. The actuarial basis stores assumptions and their approval history.
Data-source quality is itself a scientific input. The U.S. Actuarial Standards Board's Data Quality standard calls for attention to appropriateness, sufficiency, consistency, known limitations and reliance on supplied information. A large dataset can have inadequate claim settlement timestamps, outdated policy status fields or ambiguous surrender classifications. Aggregating unreliable records does not remove their bias. Conversely, a smaller credible internal experience base may be more relevant to a particular product than a larger industry sample drawn from a different market.
Time resolution should reflect the economic event. A yearly model that recognizes a catastrophe claim and related reinsurer reimbursement in the same year can conceal a month of potentially damaging cash shortfall. Daily or weekly settlement grids are appropriate for near-term funding, while annual steps may suffice for the remote tail of a simple annuity if within-year payment approximation is documented. In multiresolution models, the aggregation operator must preserve expected totals and reasonably represent intra-period timing. Claims paid at the beginning rather than end of a period can change present value and liquidity measures even when annual totals agree.
The inventory also identifies ownership of assumptions. Mortality tables, claim severity inflation, operational settlement delays, surrender elasticity, expenses and discount curves should have explicit source, vintage, scenario dependency and review date. An assumption imported from a prior valuation without its economic conditions can create a seemingly precise projection built on an incompatible environment. Forecast quality requires not only a calibrated number but a traceable explanation of the system that generates it.
Contractual and Economic Drivers
State-contingent liabilities
For a basic life policy, define the policy state at time as active, deceased, surrendered, matured or terminated for another reason. Transitions govern benefit and premium cash flows. For an active life with annual conditional death probability and surrender probability , a simple ordered competing-decrement approximation has survival-in-force probability if death is evaluated before surrender. The corresponding death and surrender probabilities are and . The order is a modeling convention; another event-order rule requires appropriately changed transition probabilities. With continuous transition intensities, competing events can instead be integrated over each period without arbitrary priority.
Let death benefit , surrender amount , maintenance expense and premium depend on time and possibly on scenario. One period's expected net outflow is approximately
when expenses and premiums are assessed at the period's opening. Payment timing and event order must be revised if contracts specify different cash settlement dates. This expression reveals a structural relation: a higher surrender rate increases surrender outflow immediately, reduces later death benefits and premiums, and changes future expenses. The net present-value result is therefore ambiguous without policy terms and discount assumptions.
For annuities, payments typically follow survival or guaranteed-payment conditions. If an annual immediate annuity pays at year end conditional on survival, , with the probability of survival from issue age . A joint-life policy, survivor continuation clause, cost-of-living adjustment or guaranteed term requires a modified contingent benefit indicator. Mortality improvement over time alters long-duration annuity payments and should not be confused with a flat change in today’s mortality rate. The scientific literature on systematic mortality risk shows why mortality trends shared across many lives cannot necessarily be diversified through an enlarged portfolio (Dahl and Møller, 2006).
For property/casualty insurance, the projection often starts with a stochastic or expected ultimate liability and then introduces reporting and payment development. This distinguishes incurred claim amount from its cash settlement calendar. A forecast can use individual claim simulation, loss-development and payment-pattern estimation, or a hybrid of case-reserve and aggregate development methods. The choice depends on maturity, claim count, severity distribution and available event-level history. Large, slow legal claims can dominate the tail, making a method calibrated solely on stable average triangles inadequate under major legal or inflation regime shifts.
Financial scenarios and economic dependencies
Scenario variables typically include short and long interest rates, yield-curve shape, inflation, wage growth, equity and real-estate returns, credit conditions, currency rates, unemployment and system-wide liquidity. Product-specific exposures determine which variables matter. Inflation-linked injury or medical claims transmit inflation into severity and settlement amount; wage-indexed disability benefits transmit wage growth. A unit-linked benefit responds to asset value, fee deductions and guarantees. An endowment with a fixed nominal sum may be insensitive in amount to inflation while still being exposed through expense inflation, withdrawals and the insurer's purchasing-power position.
It is essential to distinguish the physical probability distribution used for forecasting likely settlement experience from market-consistent valuation conventions under an applicable measurement framework. Observed risk premiums, investment strategies and managerial expectations may enter an economic planning distribution; a market-consistent valuation of options may require an appropriate calibrated valuation distribution. These purposes need not produce identical expected paths. Stochastic asset-liability research demonstrates the importance of modeling mortality, inflation and asset-return dependencies jointly; stochastic asset-liability modeling offers one published example of multivariate treatment of such variables (Maffra, Armstrong, and Pennanen, 2021).
For a scenario-dependent cash flow and discount factor , generally when rates and liability payments covary. For a variable annuity guarantee, an equity downturn may increase guarantee benefits at the same time that market value of backing assets falls. For a surrenderable life portfolio, an interest-rate rise may reduce fixed-income asset prices while increasing the opportunity cost of maintaining low-crediting-rate policies. A model that shocks interest rates but holds the policyholder behavior process fixed may understate the cash demand associated with rate competition.
Risk-factor calibration should maintain coherent inflation and interest-rate dynamics. A high nominal-rate scenario associated with high inflation differs economically from a high real-rate scenario associated with strong competing deposit returns. The first may raise claim costs and some nominal policy cash flows; the second may more directly encourage surrender for products with comparatively low credited yields. Surrender sensitivity should reflect contractual penalties, tax consequences, age, channel, policy duration, product guarantees and availability of alternatives. A single universally applied lapse multiplier has no empirical basis across all products and markets.
Insurance options and management actions
Insurance contracts often embed rights exercised by policyholders: lapse, surrender, partial withdrawal, premium holidays, contribution changes, annuitization, guaranteed minimum benefits and transfer of account value. The existence of a right does not imply its exercise under every favorable condition. A proper projection therefore distinguishes the financial incentive from behavior conditional on demographics, contract friction, information, advice, tax considerations and uncertainty. Exercise assumptions should permit both ordinary persistency and correlated mass-action events when supported by the product and scenario.
Insurers also control some future actions: credited rates, discretionary bonuses, asset sales, expense programs, hedging, premium repricing where legally available, and management of new business. Such actions alter projected cash flows subject to contract, law, governance and execution capacity. EIOPA's revised valuation guidelines explicitly address future management actions, expense modeling, options and guarantees, economic scenario generators and policyholder behavior. Projected actions should be feasible under the exact stress being modeled. Assuming rapid asset sales at unstressed prices while testing a liquidity shock generates internally favorable results by construction.
The distinction between contractual and discretionary flows is equally important in variable benefit products. A promised guaranteed minimum creates an obligation that remains when an underlying account declines; a non-guaranteed dividend or bonus is governed by its contract and applicable management discretion. Benefit classification affects both scenario responses and the financial statements to which projected amounts are reconciled. Interactions are often nonlinear at guarantee thresholds, surrender-charge expiry dates or regulatory intervention points. Linear sensitivities may serve local diagnosis but cannot substitute for simulation across those discontinuities.
Surrender and Benefit Contingencies
Behavioral response and empirical evidence
Surrenders convert future policy cash flows into immediate cash demands according to contractual value, charges and administrative processing. The option's economic attractiveness varies with alternative yields, credited rates, embedded guarantees, taxes and policyholder needs. Termination also changes future premium receipts, death exposure, maintenance costs, and, in some products, profit participation. Accordingly, surrender is a transition mechanism affecting the whole future cash-flow vector.
Evidence supports heterogeneity. The SOA Research Institute and LIMRA term and whole life experience study drew from 28 companies and approximately 135.9 million exposed policies covering 2015–2022 experience. The same organizations' fixed-rate deferred annuity study drew from 23 companies and roughly 13.4 million contracts exposed to surrender. The universal life persistency study reports different participating company sets and policy-year exposures for premium persistency and lapse behavior. These source populations differ in contract features, calendar periods, selection and available results. Their aggregate study sizes establish scope and credibility of observations; they do not authorize a uniform surrender parameter for every insurer.
Macrofinancial evidence reveals a further distinction. The Bundesbank's 2024 financial stability assessment reports incentives for German life-policy lapses associated with competing returns, while describing the observed risk of a broad lapse wave as limited in its stated setting. The Bank for International Settlements' 2024 annual economic report discusses surrender incentives and funding profiles after rate increases. These findings are compatible: financial incentives can rise without producing the same realized behavior across jurisdictions or books. Forecasting should capture the conditional relationship and its friction, rather than declaring an automatic universal lapse shock.
Guarantees, annuities and competing events
A guaranteed surrender value can create cash demand precisely when assets backing the policy have declined in market value. Whether this mismatch is borne by the insurer depends on surrender charges, market-value adjustments, policyholder account provisions, hedging and product design. Expiry of surrender charges can cause a discrete shift in behavior. Some contracts preserve a valuable embedded guarantee if the policy continues, counteracting the incentive to leave; others permit partial withdrawals that maintain part of the protection. Modeling surrender as a scalar increase in all product payments fails to represent these terms.
A competing-decrement model also changes mortality exposure. If a group of policyholders exercises surrender after a rate shock, fewer policies remain to pay death benefits and premiums. For whole life, the value impact depends on the relative sizes of cash value and mortality-contingent benefit, and on future margins. For fixed deferred annuities, the surrender benefit can be substantial relative to near-term account assets, while the avoided future annuity obligation depends on conversion rights and guaranteed rates. For variable products, fund value and guarantee moneyness can make behavior strongly state dependent. An appropriate forecast reproduces the contract's state transitions before it aggregates company-wide liabilities.
The 2024 IAIS Global Insurance Market Report describes surrender exposure and liquidity considerations across monitoring populations and emphasizes differences in economic penalties and redemption time. Such evidence is especially relevant to forecast horizons measured in days and months. However, supervisory monitoring aggregates cannot be substituted for the insurer's own product-state data and exposure distribution. The materiality of an identical lapse percentage depends on how much cash each terminating contract requires, the timing of payments and the liquid resources available at the legal entity that owes them.
Worked scenarios and sensitivity analysis
The following cases are deliberately stylized. They demonstrate mathematical and timing effects with invented parameters; their outputs are neither industry estimates nor recommended regulatory assumptions.
Illustration: identical total claims, different payment calendars
Suppose a portfolio is expected to pay a total nominal in a two-year period. Under pattern A the insurer pays at the end of year one and at the end of year two. Under pattern B it pays at the end of year one and at the end of year two. With a constant annual effective discount rate of ,
Thus the same nominal expected claim total produces a present-value difference of approximately , solely from assumed settlement timing. If liquid cash available for year-one claims were , pattern A would create a year-one funding gap of before other receipts; pattern B would not. A correct aggregate reserve amount cannot resolve that first-year funding question without payment dates.
Illustration: scenario-conditioned surrender and future flows
Consider otherwise identical active policies. Each policy is expected to produce one annual premium of at the beginning of year one, maintenance cost at the same date, a surrender payment of at the end of year one if surrendered, and a hypothetical end-of-year-two payment of only if the policy remains in force and the modeled benefit condition occurs with probability . For a simple scenario comparison, let end-of-year-one surrender probability be in the baseline and in a high-rate stress; ignore deaths in year one and all other events.
Baseline expected surrender cash at year one equals ; the stress produces , an additional of immediate cash demand. Year-two expected covered benefits, absent further decrements, equal in the baseline and in the stress. The year-one premium and expense flows are identical by construction because they occur before surrender. A model that increases year-one withdrawals but leaves year-two in-force benefits unchanged violates the transition mechanics. If the premium were due at the beginning of year two instead, premiums would fall with persistency, reinforcing the need to encode dates accurately.
The surrender event is conditioned on the economic state rather than assigned a universal elasticity. The example shows that reducing long-run expected benefit payments can coexist with sharply increased near-term cash outflow. It does not determine net present-value change without a full contractual horizon and an appropriate discount curve.
Illustration: reinsurance collection lag
Suppose an insurer must pay of catastrophe claims immediately, and is contractually recoverable from reinsurers three months later. The ultimate retained claim cost is before other costs and counterparty contingencies. The peak gross financing need before reimbursement remains if there are no other cash inflows and the firm begins with no designated liquid resources. Recording the claims on a net basis of in a quarterly financing test would omit the temporary requirement. The amount of required liquidity reduces if collateral or advances can be used, and increases if recoveries are delayed, contested or impaired. Thus collection timing is a first-order forecast input, even where the reinsurance agreement substantially reduces the final economic burden.
Illustration: interest rate and timing sensitivity
For one fixed nominal payment due after years, under a constant annual effective discount rate . Its sensitivity is . A payment of in year ten has a present value of approximately at and at . That sensitivity is a valuation effect. The actual cash due at maturity remains in the example. For floating or option-dependent benefits, may itself change with the rate scenario, and the total valuation effect must include this dependence.
Scenario Validation
Internal consistency and independent checks
Forecast validation begins with contract-level accounting identities and ends with evidence that the resulting distribution is suitable for the intended decision. The insurer should reconcile opening policy states, entries, deaths, maturities, lapses, surrenders and closing policy counts for each segment. Each projected benefit must correspond to an event and benefit basis; a premium receipt must correspond to a policy state and billing rule. Gross claims, ceded recoverables, paid claims and outstanding balances must be reconcilable without double counting. Each cash flow must be assigned a period and currency. Currency conversion should preserve the distinction between original-currency obligation and reporting-currency value.
In non-life business, incremental paid triangles derived from cumulative data should reconcile back to historical payment records, and projected ultimate claims should agree with independently approved estimates within explained methodological differences. A backtest should compare projections made at prior valuation dates with payments subsequently observed, separated into emergence, timing and severity components. A forecast that predicts the correct total payout but systematically delays payments can pass an ultimate loss-ratio test while failing a liquidity test. Backtests should preserve the information set available when the original forecast was produced; using later-adjusted data would overstate predictive performance.
Actuarial ASOP No. 56 on Modeling identifies model structure, appropriateness, data, options, uncertainty and validation as relevant matters. Its practical implication is that output plausibility alone provides inadequate assurance of a multi-module cash-flow projection. The validation process should test modules independently: data mapping, policy state transitions, claim settlement distributions, premium schedules, inflation, expense allocation, reinsurance treaties and scenario feeds. It should then test interactions and aggregated totals. Versioning of model code, assumptions and observed data supports repeatability, especially when results are used across accounting, ALM and management forecasts.
Independent model review also requires a clear distinction between deterministic sensitivity and stochastic sampling uncertainty. A model may be implemented correctly and remain exposed to parameter uncertainty, structural error, extreme-event misspecification or behavioral regime change. Governance should specify materiality thresholds, validation frequency, escalation criteria, approved overrides and the exact use limitations for outputs. A release gate that checks JSON structure and hyperlinks provides editorial and technical assurance but cannot by itself demonstrate actuarial predictive validity.
Stress calibration and uncertainty
A liability forecast should contain at least a baseline conditional projection and plausible adverse scenarios suited to the portfolio. Relevant scenarios may include higher and lower rates, yield-curve shifts, inflation shocks, catastrophe clusters, unexpected mortality or longevity changes, lapse concentration, claims payment acceleration, operational disruption and delayed reinsurance recovery. Combining shocks requires examination of realistic dependence: severe inflation can coincide with elevated claims costs and high interest rates; a market crisis can depress asset realizations while increasing collateral calls and withdrawals. Stress severity must be stated, including whether it is a supervisory prescribed shock, an empirically estimated distribution tail, or a management-designed reverse stress.
In the EIOPA 2024 Insurance Stress Test, the adverse scenario produced substantial modeled liquidity strain among participating insurers, with the mass lapse assumption a major driver of outflows in life insurance. The report distinguishes fixed-balance-sheet and constrained-balance-sheet views and examines liquid assets as well as immediate cash. Its sample and prescribed scenario are a supervisory analytical exercise, not a measured probability distribution for the entire European industry. The analysis nonetheless illustrates why liability cash forecasts require coordination with an asset liquidity inventory: an insurer can have substantial investments and still need to mobilize them rapidly to meet immediate payments.
The forecast should distinguish four uncertainty layers. Process uncertainty concerns which insured events occur. Parameter uncertainty concerns the mortality, lapse, inflation or severity parameters inferred from evidence. Model uncertainty concerns the choice of representation, including event order, claim tail structure or policyholder response. Scenario and regime uncertainty concerns whether the future resembles the processes on which parameters were calibrated. These layers need different tests. Bootstrap analysis can examine parameter stability under a specified model; alternative plausible model structures probe model risk; stress scenarios examine conditions beyond observed calibration periods. Reporting one high percentile without explaining which uncertainty sources it includes overstates its completeness.
Stress results should be evaluated by timing and contractual effect. In a high-rate scenario, the valuation of fixed nominal liabilities may fall; mass lapse can nevertheless increase short-term funding demands, and assets sold at market value may realize losses. In a high-inflation scenario, long-tail claim payments may rise; if policy premiums are contractually fixed, the near-term premium cash stream does not automatically compensate. In a catastrophe scenario, claims may accelerate alongside operational bottlenecks, and reinsurance reimbursement may follow after a delay. These effects should be represented consistently in the same scenario path.
Jurisdiction-specific valuation and governance
International comparisons require identifying the rule that controls each output. Under IFRS 17, fulfillment cash flows are used in the measurement of insurance contract groups, subject to discounting, nonfinancial risk adjustment, and contractual service margin requirements. The IFRS Foundation's standards summary identifies the standard's effective date as annual periods beginning January 1, 2023. A forecasting engine may produce several IFRS 17 reporting outputs, but the numerical amount shown as the liability cannot in general be equated with an undiscounted cash forecast or a Solvency II technical provision.
Under EU Solvency II, Article 77 describes a best estimate plus a separate risk margin. Article 28 of Delegated Regulation (EU) 2015/35 specifies cash-flow categories for the best estimate. The current technical framework must be read alongside the regulatory transition. The European Commission's Delegated Regulation (EU) 2026/269, published February 18, 2026, applies from January 30, 2027. Its changes are therefore prospective for this November 2026 research date, rather than requirements already applied to 2026 reporting. Treating prospective measures as present duties would alter a forecast's legal assumptions without authority.
For U.S. life insurance, the NAIC 2026 Valuation Manual includes detailed VM-20 cash-flow modeling provisions, including projections over a horizon long enough to encompass obligations and treatment of benefits, premiums, expenses and relevant product features. Application depends on product, statutory scope and applicable state adoption. The U.S. ASOP No. 7 cash-flow standard was revised in December 2025 and is effective for actuarial work performed on or after June 1, 2026; ASOP No. 20 applies to relevant property/casualty work products issued on or after that date. These standards define professional conduct for relevant actuarial engagements and should be interpreted with their precise effective-date wording.
International supervisory principles also matter. In its Insurance Core Principles, the IAIS establishes a broadly accepted supervisory framework, with ICP 16.5 addressing the insurer's explicit asset-liability management policy and the interaction of asset and liability strategies. The December 2024 ICP and ComFrame text explains that liability cash flows should be considered alongside funding inflows and valuation under different scenarios. The principle does not require every insurer to use the same highly complex projection model; proportionality and materiality remain relevant. The internationally shared objective is credible analysis of obligations and their financing, while the binding details depend on national implementation.
Asset-Liability Decision Use
From valuation to cash settlement
A liability-side forecast enters insurer decision-making through at least three mechanisms: valuing obligations, planning cash settlement and managing economic exposures. Valuation uses time value and risk adjustments under specified conventions. Settlement planning tracks actual dates, gross payments, recoveries and legal-entity funding. Economic risk management connects liabilities to assets, financial options and reinvestment. Confusing these mechanisms leads to misleading management conclusions, especially where a present-value reduction accompanies an immediate increase in cash needs.
One useful management presentation starts with an undiscounted monthly gross liability run-off, decomposed into claims, surrenders, premiums, expenses and recoveries; extends the horizon into annual tails; and supplies a parallel stochastic distribution and selected stress paths. It then adds separately identified asset maturities, coupons, other receipts, readily available collateral and realizable liquid assets after stress-appropriate discounts. The resulting cash coverage calculation should reveal periods of positive and negative net liquidity. Market values of long-duration securities cannot be treated automatically as same-day cash without sale capacity, operational settlement timing and haircuts.
Liquidity and economic valuation may also respond differently to management actions. Selling assets can generate cash while crystallizing adverse market valuations. Restricting discretionary credited rates may conserve future liability costs but increase surrender incentives. Purchasing an interest-rate hedge may reduce a valuation mismatch while adding derivative margin requirements. Reinsurance may improve long-run net claim cost while introducing recovery timing, collateral and counterparty risk. A coherent forecast shows these conditional consequences within the same state and period, allowing decisions to be assessed across several horizons.
Portfolio actions, liquidity and capital
A forecast can inform strategic portfolio duration, liquidity reserves, reinsurance structure, product pricing, capital allocation, claims staffing and business continuity. The inference must follow the modeled information. A projected increase in year-one benefits is directly relevant to the maturity ladder and expected cash receipts of the asset portfolio. A longer payout tail affects duration and reinvestment needs. Wider forecast dispersion may affect risk appetite and capital evaluation under the applicable rules. A temporary cash deficit prompts assessment of funding capacity, not an immediate conclusion that the insurer is balance-sheet insolvent.
The international empirical setting reinforces this distinction. EIOPA's 2024 stress-test findings describe how available liquid assets can materially alter the interpretation of cash pressure. The IAIS 2024 market monitoring report examines surrender and liquidity characteristics across reporting groups. These reports show why forecasting needs both legal-entity data and cross-market awareness. They do not justify importing a single stress multiplier or observed ratio from a different insurer without adaptation.
For policyholder protection, immediate claims-paying capacity is a distinct object from the present value of remote liabilities. Some contracts provide rapid withdrawals; others include contractual notice, penalties or market-value adjustment. A payment forecast must therefore identify policyholder rights and operational exercise times. Asset-liability decisions should also recognize liabilities arising from incurred claims whose exact payment dates remain subject to claimant actions, courts and medical or repair processes. An insurer's contractual obligations continue to be governed by applicable law and contract terms even when its model assigns a small estimated probability to a payment event.
Decision architecture and comparative findings
An integrated forecasting process can be expressed as a sequence of explicit scientific operations. First, define the decision, entity, time grid, measurement date and lawful contract population. Second, reconstruct contract states and current exposures from reconciled data. Third, derive benefit, expense, premium and recovery rules with documented contractual boundaries. Fourth, calibrate transition rates, amounts, payment delays, inflation and behavior to credible evidence, specifying uncertainty and scenario dependence. Fifth, generate conditional cash-flow paths with correct event ordering and financial interactions. Sixth, aggregate in gross and net views while preserving currency, settlement date and correlated shocks. Seventh, apply the valuation or liquidity transformation required for the decision. Eighth, reconcile results to accounting and operational records, test sensitivities and challenge adverse scenarios. Finally, preserve the versioned model, evidence, approvals, limitations and output identity.
This sequence is a synthesis of established practices, not a new theoretical system. Its value is analytical transparency: assumptions that change the legal perimeter, the underlying cash stream, or the measurement convention remain individually observable. A single engine can produce multiple outputs only if the transformations are named and validated. A net value created for financial reporting must not silently replace the gross stress outflow required for settlement planning. A best-estimate distribution calibrated to an ordinary population cannot be claimed to describe an extreme correlated event without additional evidence.
Several comparative findings follow. Life insurers and annuity writers frequently face long horizons and policyholder options whose exercise depends on rates and guarantees. Non-life insurers may face shorter expected run-off for common claims and long, inflation-sensitive settlement tails for casualty risks, with catastrophe events concentrating payments. Health and disability obligations may contain intermediate states and payment frequencies that cannot be reduced to a simple death/survival calculation. Across these classes, the timing of payment and collection is a shared control variable, while the state transition mechanisms are product-specific. The most general forecast is therefore a distribution of dated, legally grounded transactions, with product-specific state processes and purpose-specific output conventions.
Conclusion
Insurance liability cash flow forecasting is the quantitative reconstruction of future contractual settlement transactions under uncertainty. Its core objects are the event-contingent amount, the payment date, the surviving contractual exposure, the legal and economic boundary of the obligation, and the financial conditions under which it will be met. Discounted estimates are derived results of that reconstruction. Liquidity requirements are likewise derived from the distribution and timing of obligations, together with the accessibility and timing of offsetting resources.
The research demonstrates four durable findings. First, identical nominal total claims can imply materially different present values and cash coverage requirements when the settlement calendar changes. Second, scenario-dependent surrender behavior can increase immediate payments while reducing future insured benefits, premiums and expenses; the net economic effect depends on product design and discount assumptions. Third, gross claim payments and reinsurance collections must be forecast separately when reimbursement lags, collateral conditions or counterparty risk affect financing needs. Fourth, consistent financial and behavioral scenarios are necessary because interest rates, inflation, guarantees, market values and policyholder actions interact rather than enter the projection as independent arbitrary shocks.
The comparative institutional analysis locates these mechanisms within IFRS 17, the Solvency II best-estimate framework, U.S. principle-based reserving, professional actuarial standards and IAIS supervision. Their shared demand for defensible assumptions coexists with different reporting and prudential purposes. European changes applying from January 2027 are kept distinct from those applicable at the November 2026 research date. The original analytical synthesis and numerical demonstrations advance the clarity of interpretation without claiming a novel forecasting theory or legal doctrine.
The decisive standard of a mature insurance liability forecast is traceability from contractual right or obligation to dated cash transaction, from transaction to conditional assumptions and scenarios, and from projected distribution to the particular valuation, solvency, liquidity or management conclusion for which it is used. The associated governance process must preserve the integrity of each link, including explicit uncertainty, model limitations, effective dates and evidence provenance.
References
Actuarial Standards Board. (2016). Actuarial Standard of Practice No. 23: Data Quality (revised December 2016; effective April 30, 2017). American Academy of Actuaries. https://www.actuarialstandardsboard.org/asops/data-quality/
Actuarial Standards Board. (2019). Actuarial Standard of Practice No. 56: Modeling (effective October 1, 2020). American Academy of Actuaries. https://www.actuarialstandardsboard.org/asops/modeling-3/
Actuarial Standards Board. (2025a). Actuarial Standard of Practice No. 7: Life or Health Cash Flow Analysis (revised December 2025; effective June 1, 2026). American Academy of Actuaries. https://www.actuarialstandardsboard.org/asops/life-or-health-cash-flow-analysis/
Actuarial Standards Board. (2025b). Actuarial Standard of Practice No. 20: Analysis of Property/Casualty Cash Flows, Including Discounting (revised December 2025; effective June 1, 2026). American Academy of Actuaries. https://www.actuarialstandardsboard.org/asops/asop-no-20-analysis-of-property-casualty-cash-flows-including-discounting/
Bank for International Settlements. (2024). Annual Economic Report 2024 (including discussion of life insurers' surrender and funding exposures). https://www.bis.org/publ/arpdf/ar2024e.pdf?form=MG0AV3
Deutsche Bundesbank. (2024). Financial Stability Review 2024: Stability situation in the German financial system. https://publikationen.bundesbank.de/publikationen-en/reports-studies/financial-stability-reviews/financial-stability-review-2024-943034?article=stability-situation-in-the-german-financial-system--944380
Dahl, M., & Møller, T. (2006). Valuation and hedging of life insurance liabilities with systematic mortality risk. Insurance: Mathematics and Economics, 39(2), 193–217. https://www.sciencedirect.com/science/article/pii/S0167668706000473
European Commission. (2015). Commission Delegated Regulation (EU) 2015/35, Article 28: Cash flows. European Insurance and Occupational Pensions Authority, Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2441_en
European Commission. (2026). Commission Delegated Regulation (EU) 2026/269 of 29 October 2025 amending Delegated Regulation (EU) 2015/35 (OJ L, February 18, 2026; applicable January 30, 2027). EUR-Lex. https://eur-lex.europa.eu/eli/reg_del/2026/269/oj/eng
European Insurance and Occupational Pensions Authority. (2022). Revised Guidelines on Valuation of Technical Provisions. https://www.eiopa.europa.eu/publications/revised-guidelines-valuation-technical-provisions_en
European Insurance and Occupational Pensions Authority. (2024). 2024 Insurance Stress Test (EIOPA(2024)0103699). https://www.eiopa.europa.eu/document/download/f8a234b0-a84a-49ff-975e-c47f8849bfc0_en?filename=Report+-+Insurance+Stress+Test+2024.pdf
European Parliament and Council. (2009, as amended). Directive 2009/138/EC: Solvency II, Article 77, Calculation of technical provisions. European Insurance and Occupational Pensions Authority, Solvency II Single Rulebook. https://www.eiopa.europa.eu/rulebook/solvency-ii-single-rulebook/article-2160_en
IFRS Foundation. (2017, as amended). IFRS 17 Insurance Contracts (effective annual reporting periods beginning on or after January 1, 2023). https://www.ifrs.org/issued-standards/list-of-standards/ifrs-17-insurance-contracts/
International Association of Insurance Supervisors. (2024a). Insurance Core Principles and ComFrame (December 2024 adopted text, including ICP 16.5). https://www.iais.org/uploads/2024/12/IAIS-ICPs-and-ComFrame-adopted-in-December-2024.pdf
International Association of Insurance Supervisors. (2024b). Insurance Core Principles and ComFrame: Standard-setting overview. https://www.iais.org/activities-topics/standard-setting/icps-and-comframe/
International Association of Insurance Supervisors. (2024c). Global Insurance Market Report 2024. Hosted by National Association of Insurance Commissioners. https://content.naic.org/sites/default/files/inline-files/Global-Insurance-Market-Report-2024.pdf
National Association of Insurance Commissioners. (2026). 2026 Valuation Manual (including VM-20, Requirements for Principle-Based Reserves for Life Products). https://content.naic.org/sites/default/files/pbr_data_valuation_manual_current_edition.pdf
Society of Actuaries Research Institute & LIMRA. (2024a). 2015–2022 Term and Whole Life Insurance Policy Surrender/Lapse Experience Study. https://www.soa.org/resources/experience-studies/2024/15-22-twlls/
Society of Actuaries Research Institute & LIMRA. (2024b). 2015–2022 Fixed Rate Deferred Surrender Experience Study. https://www.soa.org/resources/experience-studies/2024/15-22-frds/
Society of Actuaries Research Institute & LIMRA. (2024c). 2015–2021 Universal Life Premium Persistency and Lapse/Surrender Experience Study. https://www.soa.org/resources/experience-studies/2024/15-21-ulpp-ulls/
Maffra, S. A., Armstrong, J., & Pennanen, T. (2021). Stochastic modeling of assets and liabilities with mortality risk. Scandinavian Actuarial Journal, 2021(8), 695–725. https://www.tandfonline.com/doi/full/10.1080/03461238.2021.1873176