Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Annuity Surrender Value and Liquidity Option Pricing

Abstract. Surrender provisions convert long-duration annuity contracts into contingent claims whose financial outcomes depend jointly on contract wording, future interest rates and asset returns, survival, withdrawal decisions, liquidity needs, and insurer funding capacity. A contractual surrender value is a specified cash payment; the economic value of the surrender option is the additional value of being able to elect that payment; expected cash surrender outflows depend on how policyholders actually exercise their rights. This monograph establishes the contractual and historical foundations of surrender, partial withdrawal, and commutation; develops discrete and continuous-time option-valuation expressions; examines rational exercise and empirically observed behavior; integrates cash-flow liquidity, reinvestment, asset sales, and dynamic hedging; and evaluates international accounting and prudential requirements. It critically examines publicly accessible Society of Actuaries and LIMRA experience from 2022–2025, original financial research, Securities and Exchange Commission investor disclosures, United States actuarial standards, IFRS 17, and European Solvency II rules. A reproducible two-period illustration separates the payoff effect of surrender charges from the value of a holder's right to respond to changing interest rates; a second cash-flow identity distinguishes individual option exercise from correlated mass lapse stress. The results support contract-specific valuation and economically conditioned behavior assumptions, tested against observed exposure, while exposing unavoidable differences among market-consistent option cost, best-estimate insurer outflows, household emergency-liquidity value, and regulatory capital. The original contribution is an integrated comparative analysis, source-critical interpretation, and reproducible illustrative calculation within existing actuarial and financial theory.

Keywords: Annuity surrender value; liquidity option; partial withdrawal; commutation; surrender charge; market-value adjustment; policyholder behavior; dynamic lapses; optimal stopping; stochastic control; guaranteed lifetime withdrawal benefits; insurer asset-liability management; interest-rate risk; reinvestment risk; liquidity stress; IFRS 17; Solvency II; actuarial cash-flow modeling.

Introduction

An annuity establishes a temporal exchange. The policyholder contributes capital or premiums to an insurer in return for a future stream of contractual payments that may depend on survival, market performance, guarantee conditions, and an elected commencement date. While accumulation continues, certain annuity contracts permit the holder to withdraw funds, surrender the entire contract, exchange it, or convert the accumulated value into guaranteed income. A payout annuity already converted into an irrevocable life-contingent stream may offer no surrender right. Accordingly, the economic analysis must begin with the particular contractual promise rather than assume that every contract marketed as an annuity is freely redeemable.

The subject of this research is the monetary value of a holder's contractual exit rights and the cost of providing liquidity inside an annuity. The central question is how a surrender schedule that may appear straightforward—an account value less a stated deduction—acquires a market-dependent and behavior-dependent cost over time. A guaranteed floor on surrender proceeds may become attractive after rates rise, just as the price of fixed income assets held by the insurer falls. Conversely, when rates fall, an above-market credited return or protected income guarantee may encourage continued ownership. Emergency consumption, health shocks, taxes, advice, and contract-specific restrictions affect actual decisions independently of a narrow financial comparison.

The SEC's variable annuity guidance documents surrender periods, declining charges, contract fees, and the potential renewal of charges when an annuity is exchanged. These are examples of United States retail disclosure, not a universal description of annuity contracts. The SOA–LIMRA variable-annuity study documents a pronounced increase in surrenders in the contract year when surrender charges expire, alongside much lower observed surrender in certain guaranteed-withdrawal groups. An economically relevant model must therefore combine exercise incentives with the features of a real policyholder population.

The method here is source-critical comparative research, explicit financial and actuarial derivation, and deterministic replication of hypothetical values. The review examines financial-option scholarship, official actuarial and regulatory documents, and published experience-study summaries. Claims are limited to directly observable document provisions or to the scope of the associated original scholarly abstract where underlying full text is restricted. Figures from industry submissions are treated as exposures and descriptive experience, rather than a representative international causal experiment. The legal and institutional comparison has a cutoff of November 15, 2026. No insurer transaction-level data, market calibration, or new empirical causal estimation is represented as performed.

The research is associated with the published Science monograph Annuities and Retirement Insurance, and it has independent, complementary relationships to Insurance Finance and Asset-Liability Management and Life Insurance. The primary inquiry here is the option and liquidity behavior of surrenderable annuity contracts; wider longevity pricing, pure annuity mathematics, guarantees, and household retirement-income selection retain their separate research identities.

Historical, Mathematical, and Institutional Foundations

Economic object and product boundaries

A deferred fixed annuity may have a credited account value, a minimum cash surrender value, a surrender-charge schedule, a guaranteed crediting rate for a specified period, and optional partial withdrawals. A variable annuity may reference investment-linked subaccounts and separately provide death benefits, minimum withdrawal benefits, or lifetime income guarantees. An indexed-linked product introduces participation or buffered return features. Such contracts present different relationships between account values, embedded guarantees, withdrawal opportunities, and insurer asset backing.

Three values need precise definitions. The contractual surrender payment StS_t is the cash the insurer must remit at time tt if an eligible holder validly exercises a full surrender, after the specified deductions, adjustments, and any legally required minimum. The continuation value CtC_t is the economic value to the holder of the remaining contract rights if the contract stays in force, based on a specified valuation approach and decision strategy. The surrender option value O0O_0 is the increment attributable to the right to choose exercise rather than be bound to the otherwise identical continuation-only contract. In a frictionless, market-consistent model with the same contractual cash flows, O0O_0 is nonnegative because the holder may always refrain from exercising. The insurer's incremental expected costs, including fees, expenses, selection, and funding frictions, can follow a different accounting decomposition.

These objects further differ from the reported surrender liability, the actuarial reserve, a seller's acquisition cost, and the consumer's realized net proceeds after applicable taxes. A cash surrender amount of $95,000 is neither the market-consistent value of a $95,000 optional payment at an uncertain date nor proof that the insured guarantee has a $95,000 cost today. It is the contractually payable amount at the actual exercise date, conditioned on the applicable account and adjustment rules.

The SEC investor bulletin on variable annuities explains that annuity fees, surrender schedules, and optional insurance features can interact. Contract wording must establish whether a charge is applied to the amount withdrawn, original premiums, an account balance, an excess-withdrawal amount, or a cohort of deposits. Without that denominator, the apparently simple phrase “five percent surrender charge” does not define the contractual payoff.

A policy with a guaranteed lifetime withdrawal benefit (GLWB) illustrates the stakes. The account value may reach zero while a separate insured withdrawal entitlement continues. A partial withdrawal may reduce a guarantee base on a pro rata, dollar-for-dollar, or other contract-defined basis. The policyholder can therefore have an economically valuable remaining lifetime benefit when the cash account is depleted. The possibility of full surrender changes the value of the right to continue the insurance guarantee, and the effect of a withdrawal depends on whether it exceeds the permitted annual amount.

Contractual surrender and actuarial continuation value

Consider an accumulation contract issued at time zero and held by an eligible policyholder at tt. Let AtA_t be the account or prescribed accumulation value; let dt(At)d_t(A_t) be a deduction under the surrender-charge schedule; let MtM_t be a contractually specified market-value adjustment; and let GtG_t be a statutory or contractual minimum, where applicable. A simplified surrender payment is

St=max⁡{Gt, At−dt(At)+Mt},S_t=\max\{G_t,\ A_t-d_t(A_t)+M_t\},

subject to the contract's definition of eligible values, offsets, taxes withheld, and any upper bounds on market adjustments. This is a generic representation, not a claim that all contracts implement the maximum operator or allow an unrestricted positive adjustment. Some products use a fixed guaranteed surrender value unrelated to a marked-to-market account, while payout contracts may provide commutation under an independently defined formula or none at all.

A surrender charge may have several economic functions: recouping acquisition costs, discouraging premature disintermediation, protecting other policyholders against the transfer of transaction costs, and preserving asset-liability matching. A market-value adjustment can further transmit part of an interest-rate change to the departing holder. From the policyholder's perspective, a deduction lowers immediate cash available. From the insurer's perspective, charges and adjustments alter the value of the written option and the ability to finance surrender payments by liquidating assets.

For a mortality-contingent continuation, a simple discrete present-value expression is

Ct=∑u=t+1TD(t,u) u−tpx+t E[Bu∣Ft]C_t=\sum_{u=t+1}^{T} D(t,u)\,{}_{{u-t}}p_{x+t}\,\mathbb{E}[B_u\mid\mathcal F_t]

for a fixed income benefit BuB_u payable only while the relevant life survives, with state-appropriate discounting and dependence assumptions. For a fixed accumulation product or a variable contract with several options, CtC_t must instead encompass the full array of future credits, fees, account settlements, death benefits, guarantees, further withdrawals, and contract-owner actions. A guarantee-period benefit remains payable to beneficiaries even after death if the contract so provides, so the survival factor above is not a universal formula.

An irrevocable immediate lifetime annuity without a commutation option has a continuation value but no equivalent full cash-surrender exercise. An annuity with optional commutation may allow an exchange of specified remaining income installments for a lump sum, subject to discount, medical evidence, or other conditions. The commutation payment must be evaluated using the exact benefit scope. It may extinguish only part of the income stream or all future obligations. Its liquidity value to the holder can coexist with an adverse financial exchange rate.

Development of option-based insurance valuation

Option-based analysis established that surrender rights are economically linked to guarantees and dividend or crediting policies. Grosen and Jørgensen (2000), in their participating-policy contingent-claims analysis, decompose an insurance contract into an underlying bond-like claim, a bonus mechanism, and a surrender option, and examine how the interaction affects value and solvency. The model concerns participating life insurance and supplies an intellectual foundation; its parameter outcomes cannot be transplanted directly into modern American annuity portfolios.

Bacinello (2003) treats a participating contract's embedded American-style surrender claim with a recursive binomial approach. The work makes clear that a surrender right is exercised by the holder at a chosen eligible time rather than settled automatically at a predetermined terminal date. This is the central link between a legal early-exit provision and a mathematical stopping-time problem. Later variable-annuity work developed stochastic models incorporating account performance, mortality, living guarantees, fees, and surrender, with the unifying valuation approach of Bacinello, Millossovich, Olivieri, and Pitacco (2011) illustrating the importance of modeling distinct contract specifications consistently.

In an optimal-stopping framework, the holder acts to maximize value given available information and admissible actions. Real policyholders are heterogeneous. The existence of a contractual American-style option does not assert that all holders exercise with full financial efficiency. Separating a rational exercise benchmark from observed probabilities is therefore a necessary part of research design.

Insurance contract and household perspectives

Lifetime retirement income and emergency liquidity meet at the household budget constraint. The annuitization welfare analysis of Davidoff, Brown, and Diamond (2005) shows how market completeness and annuity features affect welfare in uncertain lifetimes. Peijnenburg, Nijman, and Werker (2016), in their study of the annuity puzzle, examine mechanisms surrounding annuitization preferences and constraints. Neither study makes a specific insurer's surrender charge or commutation rate optimal by implication. Household liquidity should be studied jointly with consumption risk, outside savings, health expenses, bequests, and the age at which the insured income stream begins.

A cash option can increase perceived and actual flexibility. Yet surrendering a guaranteed income contract can eliminate benefits that would otherwise protect consumption late in life. The policyholder's optimal action is therefore conditional on the complete balance sheet and welfare objective. The insurer's option price measures contingent contractual cost under specified assumptions, while the household's willingness to pay for liquidity includes personal risks that are usually unhedgeable in traded markets.

Surrender Rights, Penalties, and Mathematical Valuation

Cash surrender schedules and market-value adjustments

A schedule may begin with a relatively high percentage deduction and decline to zero after a number of contract years. The SEC disclosure on surrender charges provides an illustrative declining percentage schedule and notes that replacement or exchange can begin a new schedule. This is an investor education example, not a prescribed universal industry tariff. The duration of surrender charges, the treatment of free withdrawals, and the basis of calculations can differ across product generations, tax-qualified accounts, and jurisdictions.

A market-value adjustment (MVA) may be linked to changes in a stated reference interest rate between issue and surrender. In broad economic terms, an increase in market rates reduces the value of a fixed-rate asset portfolio. Without an adjustment, a contract whose guaranteed surrender payment remains largely unchanged can transfer the interest-rate loss to the continuing insurer or other holders. A well-specified MVA can partially align exit proceeds with the market value of the supporting fixed-income position. Its accuracy depends on the contract's prescribed duration, reference index, caps or floors, and correspondence between asset and liability characteristics.

A lower cash payment should not automatically be interpreted as an economically fair adjustment. The insurer may hold assets with different durations, credit spreads, liquidity, and embedded optionality from the contractual reference. The instrument must therefore be assessed both as a consumer payment formula and as an imperfect hedge of the underlying portfolio. A purportedly neutral MVA can introduce basis risk when the credited yield or adjustment index diverges from actual funding costs.

Partial withdrawal rules change the exercise set. A holder may be allowed a percentage of the account balance each contract year without surrender charge; additional amounts may carry charges, reduce an income guarantee, or re-set benefit bases. Contract anniversaries, waiting periods, notice requirements, and cutoffs matter for exact option valuation. Surrender values should be reconstructed as functions of the complete contract state, rather than read as a single static number.

Stopping-time representation and valuation measures

Let XtX_t represent relevant state variables: prevailing interest rates, separate-account value, credit spreads, benefit base, surrender-charge duration, age, health information, liquidity needs, and market-volatility state. Let τ\tau be a stopping time chosen from legally permissible exercise dates T\mathcal T. Let DQ(0,τ)D^{Q}(0,\tau) be a stochastic discount factor under an appropriate pricing measure QQ for hedgeable financial risk. In a stylized contract without intermediate withdrawals,

V0exercise=sup⁡τ∈TEQ ⁣[DQ(0,τ)Sτ 1{τ<ζ}+DQ(0,ζ)Hζ 1{ζ≤τ}],V_0^{\mathrm{exercise}}= \sup_{\tau\in\mathcal T} \mathbb E^{Q}\!\left[D^{Q}(0,\tau)S_\tau\,1_{\{\tau<\zeta\}} +D^{Q}(0,\zeta)H_\zeta\,1_{\{\zeta\leq\tau\}}\right],

where ζ\zeta is death time, HζH_\zeta a contract-defined death settlement, and the survival, mortality pricing, and hedgeability assumptions are stated separately. This expression is schematic. It excludes periodic income benefits and any product-specific automatic continuation value unless those are inserted into the cash-flow functional. In a full contract, optimal stopping compares the immediate exercise payoff with a conditional expectation of all continuation cash flows.

On a discrete event grid, let Jt(x)J_t(x) be the holder's option-inclusive value. With an immediate cash payoff St(x)S_t(x) and continuation operator Ct\mathcal C_t, a fundamental Bellman representation is

Jt(x)=max⁡{St(x),Ct[Jt+1](x)},J_t(x)=\max\{S_t(x),\mathcal C_t[J_{t+1}](x)\},

at times when surrender is allowed. If partial withdrawals are permitted, replace the binary maximum with a supremum over admissible withdrawal amounts ww, including contract-defined account and guarantee-base transitions:

Jt(x)=sup⁡w∈Wt(x){Pt(w,x)+Ct[Jt+1](Tt(x,w))}.J_t(x)=\sup_{w\in\mathcal W_t(x)} \left\{P_t(w,x)+\mathcal C_t[J_{t+1}](T_t(x,w))\right\}.

Here PtP_t is the net immediate withdrawal payoff and TtT_t is the state transition after the withdrawal. Equations describe existing dynamic programming and stochastic control ideas, not a newly proposed theory.

The valuation measure matters. A risk-neutral expectation prices hedgeable market components only under specified complete-market or calibrated approximation conditions. Mortality, idiosyncratic liquidity needs, and behavioral choices may be incompletely traded. A prudential best estimate uses projected behavior in the real-world measure together with the applicable financial valuation rules and risk allowances. A household expected-utility evaluation employs individual preferences and constraints. They can be internally coherent yet yield different numerical values.

The GMWB stochastic-control treatment of Dai, Kwok, and Zong (2008) demonstrates why the ability to withdraw at varying rates, pay a penalty on excess withdrawals, or surrender creates a continuous control problem. This is especially important where withdrawal choices interact with future guaranteed benefits; replacing the state-dependent strategy with one fixed average annual withdrawal rate loses information.

Partial withdrawals, commutation, and benefit-base interactions

Suppose a variable annuity has account value AtA_t, remaining guaranteed withdrawal base BtB_t, and an annual permitted withdrawal amount gtBtg_tB_t. A permitted withdrawal wtw_t reduces the account balance, while a larger or early withdrawal may cause a contractual reduction of BtB_t beyond the cash removed. The continuation value of an excess withdrawal thus depends on the lost future insured income, not solely on today's surrender fee. The exercise rule requires modeling the after-withdrawal state and the cash now received.

A commutation option is structurally related but contractually distinct. Under full commutation, a policyholder exchanges future installments for an immediate lump sum according to a formula that may involve a discount rate, life expectancy, statutory constraints, or insurer discretion. If commutation is available only after a triggering condition, it is a contingent right with a restricted exercise set. An annuitization option, conversely, permits conversion from accumulation value into future income under guaranteed rates. The existence of one option need not imply the other. Net economic value results from their interaction where both appear in a policy.

When guaranteed benefits continue after account exhaustion, policyholder behavior can reverse the intuition from ordinary account-based surrender. If the separate account is low but the insured lifetime withdrawal promise remains generous, surrender may destroy valuable protection. If the rider is not activated and the guaranteed base is less valuable than immediately accessible funds, the same holder may have stronger reasons to exit. Model inputs must include rider status, income commencement, benefit moneyness, prior withdrawals, age, surrender-charge expiration, and tax account type.

Reproducible two-period surrender illustration

The following is an explicitly synthetic, two-period discrete calculation in nominal currency units. An eligible policyholder survives from initial valuation t=0t=0 to decision time t=1t=1 with probability 0.990.99, assumed independent of financial scenarios solely for this example. At t=1t=1, continuation produces 105105 at t=2t=2 only if the covered life survives a further year with conditional probability 0.980.98. We take a nominal year-zero discount rate of 3%. The next-year interest rate observed at t=1t=1 takes two hypothetical values: 3% with probability 60% and 12% with probability 40%. These are physical scenario weights chosen for arithmetic illustration; they are not calibrated arbitrage-free risk-neutral probabilities.

At t=1t=1, expected continuation value conditional on the low-rate scenario is

C1L=0.98(105)/1.03=99.90291262.C_1^{L}=0.98(105)/1.03=99.90291262.

Under the high-rate scenario it is

C1H=0.98(105)/1.12=91.87500000.C_1^{H}=0.98(105)/1.12=91.87500000.

Assume a reference surrender account of 100100 and a five-percent surrender charge, giving an immediate surrender payment of 9595 in both scenarios. A financially optimizing holder who uses this simplified comparison continues in the low-rate scenario and surrenders in the high-rate scenario. Relative to a commitment never to surrender, the conditional option increment is zero in the low-rate state and $95-91.875=3.125$ in the high-rate state. Its scenario-discounted value at inception is

O0illustrative=0.991.03[0.60max⁡(95−99.90291262,0)+0.40max⁡(95−91.875,0)]=1.20145631.O_0^{\mathrm{illustrative}}= \frac{0.99}{1.03} \left[0.60\max(95-99.90291262,0)+0.40\max(95-91.875,0)\right] =1.20145631.

With a zero surrender charge, the payment becomes 100100. The option can then be exercised in both scenarios: the low-rate surplus is 0.097087380.09708738 and the high-rate surplus 8.1258.125. The same calculation gives O0illustrative=3.17977660O_0^{\mathrm{illustrative}}=3.17977660. With a 10% charge, the payment is 9090 and exercise is suboptimal in both scenarios, yielding zero in this restricted two-state example. These comparisons expose the nonlinear effect of a penalty on a holder's right to switch cash flows.

Illustrative surrender charge Cash at exercise Low-rate optimal action High-rate optimal action Scenario PV of option increment
0% 100.00 Surrender Surrender 3.17977660
5% 95.00 Continue Surrender 1.20145631
10% 90.00 Continue Continue 0.00000000

This calculation is a controlled comparative illustration, not a tradable option quote or an insurer reserve. A consistent financial pricing model would specify a risk-neutral interest-rate transition and stochastic discounting; a full insurance model would also allow for premium and fee cash flows, guaranteed death benefits, correlated mortality if material, transaction frictions, and further exercise dates. The example establishes the logic of switching and the sensitivity to a surrender charge; it deliberately makes no empirical prediction about actual surrender frequency.

Policyholder Behavior and Empirical Evidence

Financial incentives and liquidity motives

Observed surrender decisions arise from multiple mechanisms. The interest-rate hypothesis predicts increased incentive to replace a contract paying a relatively low credited rate when alternative investments become more attractive, after allowing for surrender charges and taxes. The emergency-fund hypothesis predicts withdrawals when cash is required for consumption or income disruption. A guarantee-value hypothesis predicts retention when living or death benefits are economically favorable. Financial advice, inertia, attention, product complexity, and replacement commissions can affect realized decisions. These hypotheses can operate simultaneously.

Kuo, Tsai, and Chen (2003), in their cointegration study of lapse rates, examine interest rates and unemployment in life insurance, providing historical evidence relevant to competing financial incentives. Its population and contract category differ from contemporary surrenderable deferred annuities. Eling and Kochanski (2013), in a review of life-insurance lapse research, demonstrate the breadth of explanations and research designs. The inference for annuity science is methodological: an interest-rate coefficient estimated for a historical life insurance portfolio cannot be inserted into variable-annuity behavior projections without validation.

A behavioral rate should distinguish exposure to a surrender opportunity from actual exercise. Conditional full-surrender probability is not the same as any withdrawal probability. A partial withdrawal may fund consumption while preserving insured guarantees; annuitization may end the accumulation phase; death may terminate a policy before voluntary surrender; a 1035 exchange can represent replacement rather than permanent exit from insured saving. Such competing transitions require unambiguous event codes and observation windows.

United States annuity experience studies

The most directly relevant recent publicly accessible issuer-compiled evidence is the 2022–2024 variable-annuity behavior report, issued by the SOA Research Institute and LIMRA in March 2026. Seventeen contributing companies represented about 45% of industry new premium over the period as defined in the report. The study reports approximately 11.5 million contract-count surrender exposures, $1.5 trillion in contract-value surrender exposure, more than 620,000 surrenders, and $56 billion of contract withdrawals. These are period exposure measures and aggregate study totals, not 11.5 million distinct persons and not an annualized population surrender probability. The public study is restricted to qualifying United States variable annuities and excludes fixed indexed, fixed-rate, and employer-sponsored annuities.

Its strongest qualitative findings are economically informative. The report shows a pronounced surrender-rate rise in the year surrender charges expire, both with and without guaranteed lifetime withdrawal riders, while the rider-bearing group has a lower surrender rate. The report further identifies materially lower surrender rates when a GLWB rider has been activated than when it has not. Approximately half of contracts with GLWB riders took withdrawals over the study period, compared with approximately one-quarter without those riders; distribution by qualification status and systematic withdrawal behavior also differs. The study's observations therefore reject the adequacy of a single uniform lapse assumption across contracts with materially different guarantee and exercise states.

The 2023–2024 fixed-rate deferred study provides a separate contract cohort: 24 contributing insurers, approximately 65% of new sales by premium for the study definition, about 4.8 million contract-count surrender exposures, $612 billion of value exposure, and more than 567,000 surrenders. These raw totals cannot be added to the variable-annuity study to derive an industry-wide rate, since products, observation definitions, sponsors, and coverage differ. The public landing page describes proprietary detailed materials, so this monograph does not claim direct re-estimation of detailed age, duration, or distribution-channel effects.

The SOA and LIMRA released a 2024–2025 registered index-linked annuity study on October 7, 2026. Fifteen contributors represent about 75% of new premium for the studied segment, with approximately two million contract-count exposures, $420 billion in contract-value exposure, more than 100,000 surrenders, and approximately $18 billion in withdrawn value. As of the research cutoff, the announced high-level public report is described as forthcoming. Hence the published release summary supplies exposure and event totals, not a verified detailed model of surrender by moneyness or charge expiry. The difference among variable, fixed-rate deferred, and registered indexed-linked contracts should be retained in every cross-product inference.

Historical comparison is possible only with attention to cohort and observation changes. The 2019–2021 variable-annuity experience study includes different contributors and coverage rates from the recent study. Comparing its more than 500,000 surrender events to the 2022–2024 count without denominator and portfolio composition adjustment is insufficient to conclude a temporal rise in individual propensity. The correct comparison reweights or stratifies exposure by contract duration, age, rider type, guarantee status, account value, calendar period, and qualifying status.

Model specification, exposure, and selection

For contract ii exposed during interval (t,t+Δt](t,t+\Delta t], define a cause-specific surrender hazard λiS(t∣Xi(t))\lambda_i^S(t\mid X_i(t)) and distinct hazards for death, annuitization, and other exit. With piecewise constant hazards, the probability of any termination within the interval equals $1-\exp[-\Delta t(\lambda^S+\lambda^D+\lambda^A+\cdots)]$. The surrender probability is the share of that total associated with the surrender cause, not simply the total lapse probability. For a cause-specific intensity fixed over the interval,

Pi(S first in Δt∣Xi)=λiSλiS+λiD+λiA(1−e−Δt(λiS+λiD+λiA)).P_i(S\text{ first in }\Delta t\mid X_i)= \frac{\lambda_i^S}{\lambda_i^S+\lambda_i^D+\lambda_i^A} \left(1-e^{-\Delta t(\lambda_i^S+\lambda_i^D+\lambda_i^A)}\right).

The formula assumes constant cause-specific intensities during the interval and noninformative measurement conditional on the modeled state. It must be replaced or adapted when recording is interval-censored, partial withdrawals are recurrent, or health and information states evolve.

A practical intensity specification may include contract duration, calendar period, age, surrender-charge expiry proximity, credited-minus-market rate spread, insurer-specific MVA, account-to-benefit-base ratio, GLWB activation, tax qualification, prior withdrawals, distribution channel, and macroeconomic state. Functional forms might be stratified hazards, logit models, generalized additive models, or survival models with time-varying covariates. Interactions are essential: a large interest-rate advantage outside the contract can still be offset by a valuable lifetime guarantee, whereas a zero charge may trigger exit among holders with a less valuable guarantee.

Identification requires careful selection accounting. Data supplied by insurers are shaped by the companies' product offerings, distribution networks, and sales vintages. More responsive contracts may be withdrawn sooner, changing the composition of surviving business. Portfolios in force at study inception are left-truncated, new issues enter during the study, death competes with surrender, and administrative definitions can differ. Policyholder-specific health or precautionary needs are often unobserved. Observational coefficients should be presented as associations unless an identified causal research design is implemented.

Competing exercise assumptions and welfare

Three stylized behavioral rules illustrate the difference between option value and projected cash flows. Under never surrender, the policy remains in force except for other contract events; an economic surrender right is ignored. Under perfect exercise, the holder takes every financially favorable action within the model's state space. Under empirical dynamic behavior, probabilities respond to contract economics and observed behavioral covariates, with estimated dispersion and uncertainty. The first can understate exercise cost for a valuable option. The second can overstate actual insurer outflows by assuming attention, liquidity, and foresight that real holders may not possess. The third is sensitive to data history, regime changes, and extrapolation beyond observed states.

These assumptions also define different risk interpretations. A calibrated average surrender rate might give approximately correct central outflows for an unchanged historical portfolio but fail after abrupt rate changes when many holders have newly in-the-money replacement opportunities. Conversely, fitting a steep rate-spread sensitivity to a short, rising-rate sample can exaggerate surrender during other regimes. Rational exercise is useful as an upper benchmark for contractual financial flexibility, subject to its valuation assumptions, while actual cash-flow modeling requires empirically constrained behavior.

The social value of liquidity depends on heterogeneity. A holder with sufficient outside liquid savings may place little incremental value on the surrender right. Another may need emergency cash despite giving up a valuable longevity guarantee. An observed surrender therefore neither proves that the guarantee was valueless nor proves irrationality. The research must preserve the distinction between revealed contract action, possible information constraints, and the underlying household welfare function.

Liquidity, Asset-Liability Management, and Hedge Costs

Funding of surrender outflows

Surrender is both an option exercise and a cash-flow event on the insurer's balance sheet. At time tt, let NtN_t be the exposed contract population and let LitL_{it} be a full-surrender indicator. Gross cash redemption requirements over a period are

RtS=∑i=1NtLitSit+∑i=1NtWit,R_t^S=\sum_{i=1}^{N_t} L_{it}S_{it} +\sum_{i=1}^{N_t} W_{it},

where WitW_{it} is the net cash paid for any allowed partial withdrawals that are not already included in full surrender. The separation avoids counting the same contract's final payout twice. Total liquidity need also includes benefit payments, debt service, collateral and derivative variation margin, expenses, taxes, and new-business cash receipts. A stable estimate of aggregate mean redemptions is insufficient for controlling a tail event involving simultaneous surrender and asset-market deterioration.

Funding sources include short-duration cash assets, coupon income, maturing bonds, marketable securities, committed liquidity lines, and cash received from new business, each subject to availability and contractual restrictions. An insurer may also hedge the market exposures of the liability rather than hold assets that exactly mature with surrender payments. Surrender uncertainty changes the effective duration of liabilities: when interest rates rise, the risk of early cash outflow can shorten liability duration at the same time that bond prices fall. A static duration match constructed under a low-surrender assumption can deteriorate endogenously under rate shocks.

The ASOP No. 7 cash-flow standard, revised in December 2025 and effective June 1, 2026, provides the professional United States setting for life and health asset-liability cash-flow projections. Its applicability depends on the assignment. For annuity surrender models, the relevant methodological principle is to consider the contractual cash flows and market/behavioral interactions of the insurer and the portfolio rather than evaluate an isolated surrender amount as if its financing were frictionless.

A surrender charge may recover some frictional funding cost, but the size of that charge and actual realized costs need not match. Fees that are financially valuable to the insurer can reduce the holder's exercise payoff and redistribute costs between exiting and continuing policyholders. Pricing must separate these streams to avoid treating fees simultaneously as a cost reduction and a separate hedge benefit.

Correlated exercise and market stress

Consider a high-rate shock that changes both contractholder incentives and the market prices of fixed-income assets. Let DtD_t be the number of full-surrender events, Sˉt\bar S_t their weighted mean cash value, and PtAP_t^A the market price of assets sold to fund those outflows. At an aggregate level, gross surrendered cash approximately equals DtSˉtD_t\bar S_t; any realized asset-sale loss is determined by the gap between the assets' previous carrying or modeled values and their prices upon sale, according to the relevant valuation basis. The joint distribution of DtD_t and PtAP_t^A is central. An assumption that exercise is independent of rates and asset conditions can understate combined stress even if the unconditional surrender count is correctly estimated.

The Solvency II delegated regulation requires consideration of interest and exercise interactions. Its policyholder behavior provision directs undertakings to examine historical and prospective behavior, the benefit of exercising contractual options, economic conditions, management actions, and other relevant drivers. Independence of these factors requires empirical justification. This regulatory requirement closely matches the scientific conclusion that behavior and financial markets should be modeled jointly, while the standard's precise legal reach remains the EU solvency framework.

The Solvency II lapse-risk provision addresses lapse up, lapse down, and mass lapse stresses in the standard formula. Such capital stresses serve a prudential loss-absorption purpose; they are not interchangeable with a commercial pricing assumption or a central behavioral forecast. Mass lapse impacts depend on which policies surrender, the effect on liabilities and assets, and the structure of contractual guarantees. EIOPA's clarification concerning lapse stress illustrates the importance of exact termination definitions and policy-by-policy stress effects. A mass lapse is a portfolio scenario requiring funding and capital analysis, not a claim that all holders will rationally choose the same action.

Policyholder behavior can also reduce some liabilities on surrender by extinguishing guarantees that otherwise would have continued. Consequently the direction of the insurer's net economic loss from lapse can differ across blocks of business. Surrender of a deeply valuable lifetime guarantee may improve a particular liability position, while surrender of a fixed-rate account backed by depressed bonds may crystallize loss or force sales. Effective stress modeling aggregates contingent asset and liability changes by policy rather than assuming a universal sign.

Dynamic and static hedging limitations

Financial option risk can be hedged with duration instruments, interest-rate derivatives, equity options, and dynamic rebalancing, depending on product design. A fixed deferred annuity with guaranteed surrender value can expose the insurer to a combination of interest-rate options and withdrawal behavior; a variable annuity with GLWB requires market, longevity, fee, and behavioral sensitivity. Full replication may be impossible when policyholder actions depend on non-traded income needs or health information, when mortality is imperfectly hedgeable, or when markets become illiquid during stress.

The variable-annuity valuation research emphasizes coherent treatment of contract features in valuation. A practical hedge must additionally consider basis mismatch between modeled and actual indices, rebalancing delays, transaction spreads, collateral demands, derivative counterparty risk, asset credit migration, and path dependence. Hedging an option under perfect exercise can differ materially from hedging an empirically behaviorally weighted liability, because the delta and gamma around the exercise boundary change.

Let V(θ)V(\theta) be modeled liability value, with state θ\theta containing rates, volatility, account returns, mortality, and surrender behavior parameters. A local first-order sensitivity to a driver θj\theta_j is ∂V/∂θj\partial V/\partial\theta_j. When the exercise boundary moves discontinuously, second-order or regime-change effects can dominate finite local sensitivities. A hedge calibrated to the value of ∂V/∂r\partial V/\partial r under a fixed lapse rate may fail in a sharp rise in rates if ∂λS/∂r\partial \lambda^S/\partial r also increases. Joint repricing and liquidity stress must therefore supplement conventional Greeks.

Costs also include the option to retain a contract. A holder who keeps a high guaranteed crediting rate when market rates fall can impose an opportunity cost on the insurer, while a holder who surrenders when rates rise may do the same through forced asset sales. This two-sided selection makes some long-dated guarantees resemble an asymmetric funding option, with its realized direction determined by contract terms, market state, and behavioral constraints.

Accounting and supervisory treatment

IFRS 17 measures insurance contract groups using expected fulfilment cash flows and an explicit treatment of nonfinancial risk, within a broader recognition and measurement architecture. The IFRS 17 standard overview provides the formal institutional context. The text of paragraph B62, as reproduced in the endorsed IFRS 17 regulatory text, explicitly requires estimates of how holders exercise options including surrender, and an allowance in the nonfinancial risk adjustment for potential deviations of actual behavior from expected behavior. This is an accounting measurement requirement, not a rule that every contract must be priced to the perfect exercise boundary.

An IFRS reporting estimate and an insurer's internal economic option value can diverge because of measurement objectives, discounting, grouping, risk-adjustment requirements, and the distinction between experience and replicable traded-market components. Cash-flow estimates must not double count the value of surrender if expected redemptions are already represented in simulated contract outcomes. Likewise, guaranteed benefits extinguished on surrender must be removed coherently rather than retained as liabilities after redemption.

Under Solvency II, technical provisions, asset valuation, and capital stress operate within a different purpose: policyholder protection through prudential solvency. The EIOPA guidance on technical provisions supports governance of technical provision assumptions and valuation methodology. Evidence of consumer advantage, macroeconomic conditions, and insurer actions should inform a prospective surrender assumption in the appropriate legal scope. The resulting risk capital stress is conceptually distinct from the premium charged for surrender flexibility.

The United States uses state-level insurance law, statutory accounting, product and reserving requirements, and professional actuarial standards. The NAIC model-law collection documents model regulation as a harmonization instrument, while enacted rules vary by jurisdiction. A model regulation must not be described as uniformly effective law in every state. Contract disclosures and surrender-value guarantees therefore require analysis of the actual approved contract form and applicable state law. United States actuarial practice, IFRS reporting in adopting jurisdictions, and the EU prudential framework are different institutional layers; a scientific comparison should identify the purpose of each before aligning numerical results.

Scientific Validity, Sensitivity, and International Comparison

Assumption governance and data provenance

An evidence-based surrender model has five data layers. First is the contract layer: issue date, approved form, surrender schedule, guaranteed cash value, market-value adjustment, partial withdrawal allowance, fees, benefit bases, and state/jurisdiction. Second is the market layer: interest-rate term structure, credit spreads, index paths, implied or realized volatility, and asset-liability position. Third is the behavior layer: full surrender, withdrawals, annuitization, rider activation, deposits, and exchanges. Fourth is the exposure layer: entry/exit dates, actual time at risk, mortality, censoring, and missing records. Fifth is the governance layer: assumption version, calibration sample, model owner, validation tests, override decisions, and effective dates.

A source-to-claim audit must keep denominators distinct. SOA reports refer variously to contract-count exposure, contract-value exposure, new premium share, and surrender events. Dividing event totals by count exposure provides an unadjusted event-to-exposure quotient, not necessarily an age-adjusted annual hazard or a population-causal parameter. For the 2022–2024 variable-annuity report, the crude ratio of approximately 620,000 events to 11.5 million contract-count surrender exposures is about 5.4%. The report's values are approximate, and its exposure aggregation requires the study's own definition; no claim is made here that the calculated quotient is an independent insurer-inforce annual rate. A uniform hazard fitted to that quotient would ignore charge-expiry concentration, GLWB activation, observation-period censoring, and firm-level effects.

The ASOP No. 56 modeling standard supplies a relevant governance reference for model design, validation, documentation, and communication, subject to actuarial assignment and applicability. Governance requires independent replication of formula outputs, tests of data transformations, and documentation of material uncertainty. A defensible model can reveal its unidentifiable parameters instead of replacing the missing observations with an unexplained industry average.

Validation should distinguish within-sample fit, out-of-time performance, and behavior under previously unobserved rate regimes. Backtests can compare predicted and actual full-surrender counts by policy vintage, surrender-charge duration, guarantee status, age, and year. Calibration tests should include observed-versus-expected curves, likelihood measures, survival calibration, and financial sensitivity. Synthetic scenarios can test code consistency, but cannot establish empirical accuracy. External validation requires independent portfolio data or a justified transportability analysis.

Stress design and parameter sensitivity

A release-quality valuation should expose at least the following dimensions in explicit scenarios: parallel and nonparallel yield-curve shifts; rate-spread competition; high and low equity returns; changes in surrender-charge or MVA effectiveness; GLWB in-the-money levels; annuity-owner age; emergency withdrawal shocks; unusually correlated surrenders; stressed collateral and asset-sale haircuts; increased mortality or longevity; and governance overrides. This is an analysis scope, not a prescribed regulatory stress grid.

One simple behavior sensitivity takes the form

logit⁡(piS)=αd(i)+βr(rexternal−rcredit)+βc ci+βg gi+βa ai+βq qi,\operatorname{logit}(p_i^S)=\alpha_{d(i)}+\beta_r(r_{\mathrm{external}}-r_{\mathrm{credit}}) +\beta_c\,c_i+\beta_g\,g_i+\beta_a\,a_i+\beta_q\,q_i,

where αd(i)\alpha_{d(i)} is a contract-duration effect, cic_i captures the charge or MVA, gig_i a guarantee moneyness measure, aia_i attained age, and qiq_i qualification or withdrawal-state indicators. Coefficients are unknown until estimated and validated; signs cannot be set universally from generic theory. In particular a guarantee base that appears favorable by one accounting comparison may have different remaining lifetime value once survival, fees, and exercise restrictions are considered.

Stress combinations are more informative than one-at-a-time changes. Raising rates affects the economic continuation value and potentially surrender hazard; worsening liquidity can raise forced-sale costs; an adverse equity return can increase guaranteed living-benefit moneyness and reduce surrender propensity in contracts with valuable riders. Because drivers can offset, a globally fixed “lapse shock increases liabilities” rule is inadequate outside an explicitly defined product and regulatory capital calculation.

Scenario weighting also requires clarity. Probabilities fitted from historical behavior under the physical measure are appropriate to certain real-world cash-flow forecasts. No-arbitrage pricing of hedgeable cash flows uses a calibrated risk-neutral construction. Prudential capital applies the regime's prescribed or approved stressed valuation assumptions. Decision-theoretic household modeling uses personal subjective and objective risks. Mixing these measures silently can yield a mathematically precise number with no coherent economic meaning.

Jurisdictional distinctions and effective dates

The European Union's Solvency II Article 26 stipulates behavior-sensitive technical provision assumptions; lapse risk in the standard formula has its own scenario construction. Insurance contract groups applying IFRS 17 have a distinct accounting requirement to reflect expected option exercise, with paragraph B62 making surrender an explicit illustration. In the United States, state-level annuity contract standards, valuation frameworks, and consumer disclosures coexist with actuarial professional guidance. A calculation made under one of these systems should be labeled by purpose and legal applicability.

The EU review of Solvency II has transitional timing. EIOPA's July 2026 implementation announcement identifies regulatory implementation work associated with the revised framework and its 2027 applicability. A rule scheduled for January 2027 should not be described as binding in November 2026 merely because preparatory guidance is publicly available. Any later publication version of this research must repeat a dated legal applicability review rather than treat this historical cutoff as eternally current.

Accounting, solvency, and consumer protection also differ in how they measure financial and nonfinancial risk, when assumptions may change, what disclosures are required, and how benefit payments are legally defined. A cross-jurisdictional comparison should therefore use a common cash-flow object, map the applicable contract and reporting regime, and reconcile differences explicitly. This preserves meaningful international scientific comparison without pretending that one annuity formula is universally legislated.

Reproducibility, limitations, and review

The main arithmetic calculation is deliberately small enough to reproduce from the printed inputs. The exact low-rate continuation is $102.9/1.03=99.9029126214; the high-rate continuation is \102.9/1.12=91.875.Undera5. Under a 5% charge, the initial increment is \0.99\times0.40\times3.125/1.03=1.2014563107. Under a zero charge, it is \0.99[0.60(100-99.9029126214)+0.40(100-91.875)]/1.03=3.1797766048.The10. The 10% charge yields zero because the two continuation values exceed90$. Values are in identical nominal units, rounded only for display.

The illustration establishes state-contingent option logic but excludes endogenous product replacement rates, stochastic mortality, full yield curve calibration, portfolio market liquidity, withdrawals prior to exercise, income taxes, additional surrender periods, and guarantee state transitions. None of these omissions invalidates the arithmetic; each limits its scope. An expanded replicable research design would calibrate an arbitrage-free scenario generator, estimate cause-specific behavior with observed annuity contract exposures, simulate the insurer's financial and nonfinancial contingent obligations, and compare hedge error across competing assumptions.

The direct public reports and official institutional pages support the claims assigned to them. Publisher abstracts support limited identification of financial models and conclusions described in those abstracts. Restricted commercial experience dashboards, issuer-specific full contract forms, and underlying SOA microdata have not been obtained. This manuscript does not describe unaccessed data as inspected, nor does it infer precise national surrender coefficients from reported aggregate counts. No external peer review has been commissioned or claimed. The original contribution is a transparent source-based integration rather than a new named actuarial framework or empirically demonstrated universal law.

Integrated Findings and Research Agenda

Reconciliation of the valuation objects

The research establishes four analytically useful quantities within existing actuarial economics:

  1. The contractually payable surrender or withdrawal amount depends on the elected date, contract state, adjustments, and legal conditions.
  2. The market-consistent incremental option value expresses the value of the holder's allowable exercise flexibility under specified asset dynamics and valuation measure.
  3. The projected cash outflow and liability effect depend on actual or modeled holder behavior, benefit extinction, guarantee state transitions, and the insurer's accounting or valuation objective.
  4. The household liquidity value depends on consumption needs, outside assets, health, taxes, bequests, information, and the future benefits lost on exercise.

These quantities often move together but are not identical. Surrender charges change immediate proceeds and exercise boundaries; guarantees change continuation values; interest-rate shocks change both the attractiveness of exits and the value of supporting assets. A defensible comparison expresses all four explicitly and documents the basis of conversion between them.

The conceptual separation has a direct model-control consequence. A behavior-adjusted liability projection cannot simply add the full perfect-exercise option premium to an already modeled expected cash-surrender cost. That may double count the same contractual flexibility. Likewise, a cash-flow forecast with a low average surrender assumption cannot alone establish that the written exercise right has little economic value: it may fail in precisely those market states when redemptions become concentrated and expensive. Both errors emerge from confusing model outputs that answer different questions.

Policyholder protection and insurer resilience

Liquidity provisions contribute to consumer flexibility, especially when long-duration retirement assets must meet unexpected cash needs. The value of this flexibility should be displayed alongside the cost of losing income and investment guarantees, surrender charges, market-value adjustments, replacement charges, and taxation. A holder-friendly product comparison must identify the actual transferable rights, not merely the advertised accumulation balance.

For insurers, long-duration protection requires consistent investment, behavior, and contract modeling. The greatest liquidity problem may arise when redemption demands and asset market losses are positively linked. The greatest valuation uncertainty may arise where product guarantees dissuade surrender in states that a simple rate-sensitive lapse model would characterize as exit opportunities. The direction and magnitude must be investigated product by product and cohort by cohort.

The 2022–2025 United States experience summaries offer a strong descriptive case for segmentation by charge expiry, guarantee status, and product type. They do not deliver an internationally transferable surrender hazard. A future empirical international comparison should begin with harmonized contract and event definitions, documented rate and tax regimes, and appropriately adjusted exposure denominators.

Empirical and methodological research priorities

A first research priority is a public, de-identified cross-product survival-analysis dataset connecting policy features, guarantee moneyness, contract duration, economic state, and voluntary surrender versus other competing exits. A second is joint validation of behavior and asset-liability cash-flow models under observed rapid changes in interest rates. A third is a comparison of hedges calibrated to perfect exercise, empirically estimated exercise, and robust uncertainty sets. A fourth is household-level analysis of emergency liquidity, commutation, annuitization timing, and the welfare loss of surrendered guarantees. A fifth is systematic mapping of state, national, and international contract obligations to model events, including effective-date versioning.

The methods are established: contingent-claims pricing, optimal stopping, stochastic control, competing-risk survival models, scenario simulation, and asset-liability stress testing. Advancing this field requires more credible data linkage, explicitly specified contractual rights, and faithful reconciliation of valuation objectives. New theory is unnecessary to justify these concrete research steps.

Conclusion

Annuity surrender rights are economically material contractual options embedded in long-duration insurance arrangements. Their value depends on the right to choose the timing and form of liquidity, the contractual surrender amount, the market state, competing guarantees, mortality, and the decision behavior of the policyholder. The cash paid when a contract is surrendered, the economic value of being able to surrender, and the insurer's anticipated cash outflows are separate quantities requiring explicit reconciliation.

The scholarly literature on participating insurance, American-style surrender, and guaranteed withdrawals explains the option structure. Official experience studies show that actual exercise is conditional on product type, surrender-charge expiry, lifetime withdrawal guarantees, and policyholder characteristics. Regulatory and accounting sources require behaviorally informed valuation for their respective purposes, while insurance asset-liability management translates correlated redemptions into funding and hedge risks. The synthetic example demonstrates that a contractual charge can change the optimal exercise state and substantially reduce the modeled option increment.

The resulting scientific conclusion is a unified practice of contract-first valuation, coherent measure selection, behavior-sensitive cash-flow projection, transparent liquidity stress testing, and source-documented model validation. This conclusion is an original analytical synthesis of existing actuarial, financial, and insurance research. It does not constitute a new authorial theory, a market quote, a jurisdiction-specific contractual interpretation, or personalized retirement or insurance advice.

References