Insurance by BogdanovaThe Encyclopedia of Insurance and Risk Science

RESEARCH DOSSIER

Life Insurance Premium and Reserve Mathematics

Abstract. Life insurance premium and reserve mathematics translates contingent life benefits, premium-payment conditions, expenses, mortality, and the time value of money into coherent contract values. This monograph develops annual discrete-time life-contingency valuation from first principles and relates premium equivalence to prospective and retrospective reserves, recursive funding identities, gross premium expenses, and sensitivity to actuarial assumptions. A stochastic future-loss formulation identifies precisely when an expected-value price corresponds to an initial reserve of zero. A separate cohort-accounting derivation proves the equality of prospective and retrospective reserves under a common technical basis and exposes the role of survival conditioning. The research distinguishes the mathematical policy value from contractual surrender benefits and from measurements required for financial reporting or solvency supervision. A fully specified twenty-year term-life illustration, computed independently by forward and backward methods, demonstrates the net and gross premiums, reserve development, and the effects of changing mortality and discount rates. Comparative interpretation draws on scholarly life-contingency literature and directly checked sources from the Society of Actuaries, Actuarial Standards Board, International Actuarial Association, NAIC, IFRS Foundation, and EIOPA. The resulting synthesis identifies the central operational requirement: the same observable contract can carry different legitimate valuation amounts because cash-flow boundaries, assumptions, timing, risk margins, and accounting purposes differ. The mathematical identities remain exact within their stated models; actual adequacy depends on the quality and governance of the valuation basis.

Keywords: Life insurance; actuarial mathematics; life contingencies; net premium; gross premium; equivalence principle; expected present value; term insurance; whole life insurance; endowment; premium annuity; prospective policy value; retrospective reserve; reserve recursion; mortality; interest rate; expense loading; Thiele differential equation; principle-based reserving; IFRS 17; Solvency II; actuarial model validation.

Introduction

A life insurance contract creates cash flows whose occurrence and timing depend on survival, death, and sometimes the exercise of contractual options. Premiums finance those cash flows before their actual realization. Reserves express the amount assigned at a specified valuation date to the remaining obligations and financing. Their relationship can be derived mathematically once the contract, payment calendar, mortality law, discount basis, and expense definitions have been fixed. These quantities are conditional on survival and on the information available at the valuation date.

Life-contingency mathematics has a long documentary history. Edmond Halley's 1693 Breslau mortality investigation is an early foundational attempt to turn mortality observations into quantitative survival and annuity analysis. Contemporary formal treatment appears in Dickson, Hardy, and Waters (2020), whose Actuarial Mathematics for Life Contingent Risks connects premiums, policy values, multiple-state modeling, investment features, and stochastic longevity. The modern challenge is to preserve the precision of these mathematical objects while recognizing changes in life portfolios, product design, financial assumptions, and international reporting institutions.

The principal claim of this dossier is that premium equivalence, reserve recursion, and the prospective–retrospective identity are manifestations of one internally consistent conditional cash-flow account. They can be derived from a common expected present-value equation. A change in cash-flow definition, information set, valuation basis, or timing changes the object that is calculated. This observation is developed here as an analytical synthesis of established actuarial mathematics and existing institutional measurement rules; it is not proposed as a new actuarial theory.

The subject boundary is the mathematical valuation of life insurance premiums and reserves. Insurance demand, companywide capital allocation, underwriting fairness, non-life loss development, and detailed financial-statement implementation are treated only where they clarify the cash-flow identities. The work addresses a hypothetical contract rather than a customer's insurance or investment choice.

Research Scope, Notation, and Method

Contract states and time conventions

Consider one person aged x at policy issue and a contract with n annual insurance periods. The annual effective valuation interest rate is i, assumed greater than minus one, and the annual discount factor is v = 1/(1+i). Define q_{x+j} as the probability of death between the j-th and (j+1)-th policy anniversaries, conditional on survival to the former anniversary; p_{x+j}=1-q_{x+j}. For a fixed selected insured-life mortality basis, the probability of survival for t complete policy years is

tpx=∏j=0t−1px+j,0px=1.{}_tp_x=\prod_{j=0}^{t-1}p_{x+j},\qquad {}_0p_x=1.

The symbol q is a conditional probability rather than a population death count. Its dependence on age, duration since underwriting, sex or other modeled classification, calendar year, and experience selection must be recorded separately from the notation. The Society of Actuaries' 2015 Valuation Basic Tables illustrate why life-insurance assumptions may be stratified by smoker status, age definition, and relative-risk classes. Those tables do not automatically represent the mortality of an arbitrary portfolio, country, or future underwriting cohort.

For the central model, the death benefit S is paid at the end of the policy year of death, provided death occurs during the n covered years. A level premium is due at the beginning of each policy year if the insured is then alive. There is no lapse, surrender, dividend, investment option, tax, inflation indexing, or reinsurance; these omissions are deliberate experimental controls. No premium is due after death or after the n-th premium date. We first use a constant annual interest rate and a fixed issue mortality basis. Additional assumptions are introduced explicitly when needed.

Expected-value operators

For the n-year term insurance, the actuarial present value of a unit death benefit is

Ax:n‾∣1=∑j=0n−1vj+1 jpx qx+j.A_{x:\overline n|}^{1} =\sum_{j=0}^{n-1}v^{j+1}\,{}_jp_x\,q_{x+j}.

For the unit annual premium annuity-due, payable for at most n years while alive,

a¨x:n‾∣=∑j=0n−1vj jpx.\ddot a_{x:\overline n|} =\sum_{j=0}^{n-1}v^j\,{}_jp_x.

We will abbreviate these two quantities as A and a when the contract and assumptions are clear. Both are nonnegative; the premium denominator a is positive for a policy with a premium due at issue. The death-benefit timing is visible in the exponent j+1 and the premium timing in the exponent j. Changing either exponent changes the valuation.

The general formulation is a random future loss L_0 = PV(future benefits and expenses) minus PV(future premiums). The policy's expected loss is E[L_0], conditional on the applicable issue information. An equivalence premium sets E[L_0]=0 on the chosen technical basis. Expected loss, observed profit, statutory reserves, and required solvency capital are distinct measurements. The International Actuarial Association's international actuarial practice standards provide an international reference for professional methods and judgments; a jurisdiction's adoption and binding requirements must be assessed independently.

Research design and limitations

The analysis uses symbolic derivations, survival-conditioned present values, an independent forward cohort calculation, backward recursion, and explicit numerical perturbations of interest and mortality. It then compares the resulting mathematical values with narrowly defined institutional accounting and prudential measurements. It does not estimate any mortality distribution from raw data, infer causal effects, calculate company statutory liabilities, or assume universal adoption of any international standard.

A reference is used for a proposition only within the scope its original issuer supports. Official mortality-table pages establish table design and data provenance; actuarial standards establish professional guidance within their scope; accounting and supervisory texts establish formal reporting definitions. Scholarly textbook metadata and chapter coverage support the mathematical context; equations below are also derived directly and tested numerically. Source-to-claim inspection and the limitations of accessible full text are recorded separately in the accompanying evidence ledger.

Life Insurance Premium Principles

A level net premium for term insurance

Let the death benefit be S. Under the annual assumptions above, the expected present value of benefits equals S A. The expected present value of a level annual net premium P equals P a. The equivalence principle therefore gives

P=SAa=∑j=0n−1S vj+1 jpxqx+j∑j=0n−1vj jpx.P=\frac{S A}{a} =\frac{\sum_{j=0}^{n-1}S\,v^{j+1}\,{}_jp_x q_{x+j}} {\sum_{j=0}^{n-1}v^j\,{}_jp_x}.

The premium is a price per premium due date, conditional on survival at that date. Dividing the sum of expected benefits by n instead of a would ignore discounting and the death-contingent termination of premium payments. Likewise, computing all future death benefits from unconditional q values instead of survival-weighted probabilities would count mutually inconsistent death events.

For a single premium at issue, the equivalence price becomes S A because the premium payment itself is certain at issue. For a premium payable over m years, with m at most n, replace a by the m-payment annuity-due. A level premium can therefore rise if the premium-paying period shortens while the benefit period remains constant. The statement depends on holding the benefit basis and discount assumptions fixed.

Whole life, endowment, and survival benefits

If coverage lasts for the insured's entire remaining lifetime within an appropriately terminating life table, the corresponding whole-life functions are

Ax=∑j≥0vj+1jpxqx+j,a¨x=∑j≥0vjjpx,Px=SAxa¨x.A_x=\sum_{j\ge0}v^{j+1}{}_jp_xq_{x+j},\qquad \ddot a_x=\sum_{j\ge0}v^j{}_jp_x,\qquad P_x=S\frac{A_x}{\ddot a_x}.

These series require the usual convergence and limiting-survival conditions. For a pure endowment that pays S at time n if the person survives, the single-premium value is S v^n{}_np_x. For an n-year endowment combining a death benefit during the term and a survival payment at maturity, the benefit present value is

S(Ax:n‾∣1+vnnpx).S\left(A_{x:\overline n|}^{1}+v^n{}_np_x\right).

The survival payment is a separate event from death during the term. It must be added only once and discounted to the contractual maturity. The cash-flow vocabulary in the Cambridge life-contingencies text includes term, whole-life, endowment, annuities, and policy values, which constitute distinct actuarial objects under a common valuation framework.

The stochastic future loss and pricing risk

Let K_x denote the curtate remaining lifetime: K_x=j when death occurs in policy year j+1. Under the n-year term policy,

L0=S vKx+11{Kx<n}−P∑j=0n−1vj1{Kx≥j}.L_0=S\,v^{K_x+1}\mathbf1_{\{K_x<n\}} -P\sum_{j=0}^{n-1}v^j\mathbf1_{\{K_x\ge j\}}.

Taking expectations recovers S A-P a. Although E[L_0] equals zero at the net equivalence premium, the random loss is generally not identically zero. Its variance depends on the joint variability of benefit timing and premium cessation. A collection of policies can diversify a component of independent mortality volatility, while parameter uncertainty, contagious risk, concentration, and interest risk may remain common across policies. Thus the equivalence premium is an actuarial starting point and does not itself specify a competitive retail price or solvency margin.

A portfolio-percentile approach instead selects a premium that meets a probability statement for a defined aggregate loss or profit random variable. An expected-profit approach adds an explicit target contribution. These approaches answer different questions; neither follows solely from the equation E[L_0]=0. They require a defined portfolio, financial constraints, and assumptions about dependence and expenses.

Timing, selection, and model boundaries

In a model with monthly premiums, the unit premium annuity must contain the correct monthly conditional survival weights and discount factors. A mortality rate q for an entire year cannot simply be divided by twelve without defining a within-year mortality convention. A constant-force assumption yields a different fractional-age expression than a uniform distribution of deaths. Benefits paid on death rather than at the end of the year also require a corresponding continuous-time present value.

Life insurance pricing may distinguish select mortality soon after underwriting from ultimate mortality, account for policyholder lapse and reinstatement, or project improvement by calendar year. The Society of Actuaries' individual life mortality research directory documents evolving datasets and studies, including material dated 2026. Those studies motivate careful basis selection; their existence alone does not validate any particular assumption used by the numerical illustration below.

Expense and Gross Premium Structure

Classifying premium cash flows

The net premium above finances only the stated benefits on the defined basis. A gross premium also addresses acquisition and maintenance expenses, premium-related costs, claim-handling expenses, and potentially profit, taxes, reinsurance, and capital costs. These terms must be attached to actual payment triggers. Acquisition cost at issue has a different time profile from a maintenance cost at each premium date, and a claim expense is contingent on death in the insured period.

Define α as a fixed acquisition expense at issue, c as a fixed expense each year a premium is due, β as the fraction of gross premium G consumed by collection or proportional administration, and K as an expense paid with each covered death claim. Under the same n-year term structure, a zero-expected-profit gross equivalence premium satisfies

(1−β)G a=(S+K)A+α+c a,(1-\beta)G\,a=(S+K)A+\alpha+c\,a,

hence

G=(S+K)A+α+c a(1−β)a,0≤β<1.G=\frac{(S+K)A+\alpha+c\,a}{(1-\beta)a}, \qquad 0\le\beta<1.

The acquisition expense α is certain at issue. The term c a accounts for future maintenance only on premium dates while the policy remains in force. The term K A attaches claim expenses to the modeled death event. If premiums are waived after a specified event or maintenance expenses continue after premium cessation, these conventions must be altered. The gross formula is not a universal tariff: it is the exact solution to the expressly stipulated model.

Profit, contingency margins, and capital

If m is an additional target fraction of the gross premium devoted to expected profit, and β+m is below one, the simple level-loading equation becomes

Gtarget=(S+K)A+α+c a(1−β−m)a.G_{\mathrm{target}} =\frac{(S+K)A+\alpha+c\,a}{(1-\beta-m)a}.

This is an internal expected-value allocation, not a risk-capital rule. It does not establish the distribution of the insurer's profit, the capital supporting adverse scenarios, or the contractual service margin under financial reporting. A capital charge may be modeled as a separate cash-flow stream, or as an amount tied to capital held through time; folding it into an undifferentiated percentage can obscure the source of expense and the required return.

The economics of first-year acquisition spending are especially important. A level-premium term policy can have a negative prospective gross value immediately after issue when acquisition expenditure exceeds early premium contributions, even when the gross premium exactly balances all expected lifetime costs on the original pricing basis. This value is conditional on the valuation instant; it is not a claim that a statutory insurer may report a negative required reserve or distribute capital accordingly.

Premium decomposition and consistency

One may write the gross premium as a collection of benefit, expense, and target-return components divided by a premium-paying annuity. Such decomposition is useful for diagnostics but is not fully additive if expense terms depend nonlinearly on the gross premium, policyholder behavior, inflation, or premium changes. A lapse-sensitive expense model is endogenous: premium level affects lapse, which affects premium duration, which changes the premium's own denominator. Solving the resulting fixed-point or simulation problem requires a stated behavior model and numerical method.

For professional data governance, the Actuarial Standards Board's ASOP 23 on data quality addresses appropriate data selection, review, use, and disclosures. A technically flawless algebraic solution remains contingent on whether inputs reflect the insured class, currency, claim settlement practice, policy term, and expense scale being valued.

Prospective Policy Values

The reserve as a conditional future-value difference

Fix an integer anniversary t between zero and n. Let the insured be alive immediately before any premium due at time t. Define

kpx+t=∏r=tt+k−1px+r,0px+t=1.{}_kp_{x+t}=\prod_{r=t}^{t+k-1}p_{x+r},\qquad {}_0p_{x+t}=1.

For t less than n, the unit benefit and premium values of the remaining term are

At=∑j=tn−1vj−t+1(∏r=tj−1px+r)qx+j,A_t=\sum_{j=t}^{n-1}v^{j-t+1} \left(\prod_{r=t}^{j-1}p_{x+r}\right)q_{x+j},
at=∑j=tn−1vj−t(∏r=tj−1px+r).a_t=\sum_{j=t}^{n-1}v^{j-t} \left(\prod_{r=t}^{j-1}p_{x+r}\right).

The net prospective policy value immediately before the premium at t is

Vt=SAt−Pat.V_t=S A_t-Pa_t.

At issue V_0=0 when P is calculated by the same net equivalence basis. At maturity for a pure term insurance V_n=0, after the last covered period and the payment of any valid terminal claims. A pure endowment or an endowment at its benefit-payment boundary instead requires its explicit maturity cash flow. Treating every policy as having a terminal reserve of zero just before maturity would be incorrect.

Prospective valuation is conditional valuation, not the difference between unconditional totals remaining from issue. Dividing by the probability of survival to duration t, where appropriate, converts issue-date cohort expectations into a value per surviving contract. This distinction is central to exact retrospective equivalence.

Before-premium and after-premium values

The time instant must be attached to every reported value. With a level premium P due at anniversary t,

Vtafter premium=Vt+P,V_{t}^{\mathrm{after\ premium}}=V_t+P,

when the after-premium value is the amount retained by the insurer immediately after collecting that premium and before paying any other simultaneous cost. If expenses are deducted at that instant, the net financing position changes accordingly. A positive difference of P between the two instants is a cash-receipt convention; it is not new earned investment income. Consistent valuation of an end-of-year benefit also requires the one-year discount before it contributes to the beginning-of-year reserve.

A prospective gross reserve with expenses

The contractual gross premium G and the insurer's future actual expenses produce a gross prospective value. For a surviving policy at time t before premium,

Wt=(S+K)At+(c−(1−β)G)at+α1{t=0},W_t=(S+K)A_t+(c-(1-\beta)G)a_t+\alpha\mathbf1_{\{t=0\}},

in the stipulated no-lapse model. Thus W_0=0 when the gross premium satisfies its own equivalence equation and no target profit is included. For later anniversaries α is a historical cost and does not appear in future cash flows. A gross value at t greater than zero represents positive expected future net cash outflow; a negative amount means expected future premium inflows exceed modeled outflows. Neither sign alone determines a regulatory reserve floor.

When mortality, interest, or expenses are updated, one must decide which items are held fixed. Recalculating G is a repricing exercise; retaining the original contractual G and recomputing W_t is a current-assumption value exercise. Confusing those experiments hides the economic consequences of an assumption change.

Policy values, surrender values, and accounting liabilities

A contractual cash surrender value is determined by policy terms and applicable law. It need not equal V_t or W_t. A company's asset share, typically reflecting accumulated experience and allocated expenses for the policy or cohort, is another distinct amount. A supervisory best estimate, a statutory reserve, and a financial reporting liability each follow a defined measurement rule. Mathematical policy-value identities help implement these measurements, but do not dictate all institutional additions, floors, margins, or grouping requirements.

Under IFRS 17 Insurance Contracts, applicable accounting measurement involves estimates of fulfillment cash flows, discounting, an adjustment for nonfinancial risk, and a contractual service margin for relevant groups of contracts. IFRS 17 applies to annual reporting periods beginning on or after January 1, 2023, subject to the standard's transition requirements. A net individual-policy reserve calculated from original mortality and interest does not become an IFRS 17 balance-sheet measurement merely by attaching an accounting label to it. Reporting contract groups, boundary rules, actual assumptions, and service-margin recognition matter.

Retrospective Reserves

The surviving-cohort construction

A retrospective reserve begins with amounts financed in the past, accumulated to the valuation date, and allocates the resulting expected fund to surviving contracts. Consider one unit of initial issued business. Let l_j={}jp_x denote the expected survival proportion at the beginning of policy year j; hence l_0=1 and l{j+1}=l_j p_{x+j}. These are expected proportions under the specified mortality basis, not realized fractional persons.

At time t, for t greater than zero, the accumulated net-premium fund before expenses is

Ft=∑j=0t−1ljP(1+i)t−j−∑j=0t−1ljqx+jS(1+i)t−j−1.F_t=\sum_{j=0}^{t-1}l_jP(1+i)^{t-j} -\sum_{j=0}^{t-1}l_jq_{x+j}S(1+i)^{t-j-1}.

Premiums collected at the beginning of year j accumulate for t-j years. Death benefits paid at the end of year j accumulate for t-j-1 years. The retrospective reserve per surviving contract is

Rt=Ftlt,lt>0,R0=0.R_t=\frac{F_t}{l_t},\qquad l_t>0, \quad R_0=0.

The denominator matters. Without division by l_t, F_t is the aggregate expected fund attributable to one original policy at issue, whereas V_t is per policy alive at duration t. Comparing the two unadjusted values would mix distinct populations.

The derivation can also be expressed with expected present values. Multiplying F_t by v^t and recalling l_t={}_tp_x gives

vtltRt=P∑j=0t−1vjlj−S∑j=0t−1vj+1ljqx+j.v^t l_t R_t = P\sum_{j=0}^{t-1}v^jl_j -S\sum_{j=0}^{t-1}v^{j+1}l_jq_{x+j}.

The result is the issue-date present value of past net premiums minus past benefit claims, accumulated and divided by the expected surviving exposure. It is thus a mathematically defined expected-fund value, not a report of actual investment assets.

Proof of prospective–retrospective equivalence

Premium equivalence at issue states

P∑j=0n−1vjlj=S∑j=0n−1vj+1ljqx+j.P\sum_{j=0}^{n-1}v^jl_j =S\sum_{j=0}^{n-1}v^{j+1}l_jq_{x+j}.

Split both sums into periods before t and periods from t onward. Move the past net cash-flow balance to the left and the future net benefit balance to the right. The common identity becomes

P∑j=0t−1vjlj−S∑j=0t−1vj+1ljqx+j=S∑j=tn−1vj+1ljqx+j−P∑j=tn−1vjlj.P\sum_{j=0}^{t-1}v^jl_j -S\sum_{j=0}^{t-1}v^{j+1}l_jq_{x+j} = S\sum_{j=t}^{n-1}v^{j+1}l_jq_{x+j} -P\sum_{j=t}^{n-1}v^jl_j.

Dividing by v^t l_t converts the right side into S A_t-Pa_t and the left side into F_t/l_t. Therefore

Rt=Vt\boxed{R_t=V_t}

for every duration at which l_t is positive, provided the premium was calculated on the same mortality, interest, contract, payment, and benefit basis and no unrecorded cash flows intervene. The identity is accounting algebra applied to a consistent expected-value model; it requires neither a claim that actual experience equals expected experience nor an additional independent principle.

This proof also makes the main failure modes transparent. If the issue premium is recalculated on one set of mortality assumptions and future benefits are reserved on another, the issue equivalence equation no longer cancels the historical and prospective segments. If realized investment returns replace the assumed i in the historical fund, the retrospective experience account differs from the prospective technical reserve. If actual surrenders, dividends, commissions, taxes, or reinsurance are introduced on one side alone, the cash-flow sets are no longer identical.

Gross retrospective funding

With acquisition expense α, proportional premium expense βG, maintenance expense c, and death claim expense K, define the gross fund

Ft(g)=−α(1+i)t+∑j=0t−1lj((1−β)G−c)(1+i)t−j−∑j=0t−1ljqx+j(S+K)(1+i)t−j−1.F_t^{(g)} =-\alpha(1+i)^t +\sum_{j=0}^{t-1}l_j\left((1-\beta)G-c\right)(1+i)^{t-j} -\sum_{j=0}^{t-1}l_jq_{x+j}(S+K)(1+i)^{t-j-1}.

The expected gross retrospective reserve is F_t^{(g)}/l_t. Under zero-profit gross equivalence, it equals W_t on the same basis. The initial acquisition cost is accumulated as a time-zero outflow. Removing that cost retrospectively while retaining it in the issue gross premium equation would produce an apparent but artificial inconsistency.

Actual cohort experience and attribution

The expected-fund identity is exact at the level of the assumed mortality law. An observed cohort has an integer number of deaths, actual premium collections, actual expenses, timing differences, and realized asset returns. An empirical asset share may be constructed by summing those actual cash flows and allocating resulting assets to surviving contracts. It can differ materially from both R_t and V_t without demonstrating an error in either formula. Actuarial analysis should identify the portion of the difference attributable to mortality, investment return, persistency, expense overruns, and model basis changes rather than using the single term reserve indiscriminately.

Reserve Recursions

A one-year funding identity

The conditional prospective net value at anniversary t satisfies

Vt+P=v(qx+tS+px+tVt+1).V_t+P=v\left(q_{x+t}S+p_{x+t}V_{t+1}\right).

The left side is resources available immediately after collecting the premium and before the year's investment accumulation. The right side is the discounted expected funding requirement: a death claim at the end of the year or the surviving policy's reserve at the next anniversary. Rearranging gives

Vt+1=(Vt+P)(1+i)−qx+tSpx+t,px+t>0.V_{t+1}= \frac{(V_t+P)(1+i)-q_{x+t}S}{p_{x+t}}, \qquad p_{x+t}>0.

This forward recursion is exactly the corresponding retrospective expected-fund update. It is numerically convenient when V_0 and a constant pricing basis are known. Backward recursion is convenient when a terminal boundary value is known. Both must agree with direct cash-flow summation to within numerical precision. A difference larger than justified floating-point error points to a timing, boundary, mortality, cash-flow, or implementation discrepancy.

If p_{x+t}=0, conditioning on survival beyond that point is undefined. Such an endpoint cannot be handled by division by zero. A calculation must end with the last feasible duration or use an alternative state representation.

Backward policy valuation

With a term insurance and terminal value V_n=0, the recursion is applied backward:

Vt=v(qx+tS+px+tVt+1)−P,t=n−1,…,0.V_t=v\left(q_{x+t}S+p_{x+t}V_{t+1}\right)-P, \quad t=n-1,\ldots,0.

In particular the net equivalent premium is recovered as the value that makes the backward-calculated V_0 equal to zero. This provides an implementation cross-check independent of any accumulation formula. If a maturity survival benefit is payable at time n, the boundary is changed to reflect its payment instant. If a policy has a surrender option, payment guarantee, bonus, premium waiver, or several insured lives, the state space and boundary payments must be enlarged. An algebraic two-state recursion cannot value features whose cash flows are absent from its states.

Gross premium recursion

The analogous gross recursion includes premium collection costs and maintenance expenses:

Wt+(1−β)G−c−α1{t=0}=v(qx+t(S+K)+px+tWt+1).W_t+(1-\beta)G-c-\alpha\mathbf1_{\{t=0\}} =v\left(q_{x+t}(S+K)+p_{x+t}W_{t+1}\right).

This equation assumes acquisition expenses α are paid immediately at t=0; for t greater than zero, there is no new acquisition charge. In the presence of profit loading, this recursion still values the contractual gross cash flows and expenses, but the initial boundary generally differs from zero because anticipated profit is not modeled as a benefit or an actual expense. A target profit should not be counted as a claim obligation merely to force W_0 to zero.

Continuous-time limiting form

Suppose premium inflow π(t) and maintenance outflow e(t) occur continuously while the policy remains active, the force of interest is δ(t), the mortality intensity is μ_{x+t}, and the death benefit S(t) is paid instantly on death. For a differentiable prospective value V(t), a small-interval survival/death balance yields the classical Thiele equation

dV(t)dt=δ(t)V(t)+π(t)−e(t)−μx+t(S(t)−V(t)).\frac{dV(t)}{dt} =\delta(t)V(t)+\pi(t)-e(t) -\mu_{x+t}\bigl(S(t)-V(t)\bigr).

The term μ(S-V) is the mortality-weighted net amount at risk: on death the insurer pays S and releases the reserve otherwise required for a surviving policy. In the annual discrete equation the analogous funding term contains qS+pV_{t+1}. The continuous equation is not obtained by merely substituting q for μ and i for δ; its state transition, premium intensity, and payment timing must match the limiting model.

For a multiple-state insurance product, the continuous reserve is indexed by the occupied state and transition intensities connect the states. A death-only model is therefore a special case of broader actuarial state modeling. This relationship is included in the life-contingencies textbook architecture, which treats both policy values and multiple-state models.

Validation controls and numerical stability

An implementable validation suite should test: issue-equivalence residual, agreement of direct present values with recursion, prospective–retrospective reconciliation, terminal boundary, sign and range of survival probabilities, premium timing, claim timing, mortality-unit scaling, and stability under plausible perturbations. Numerical overflow can occur when survival probabilities become extremely small and retrospective reserves divide by l_t; backward conditioning and scaled survival representations may be more stable near extreme ages. Very small p can also amplify rounding in the forward recursion. Exact identities must be checked where defined, while practical tolerances depend on machine precision and contract scale.

The U.S. Actuarial Standards Board's ASOP 41 on actuarial communications supports documentation of assumptions, methods, and actuarial findings. A reproducible calculation should specify whether reserve values are pre-premium or post-premium, whether nominal or real cash flows are used, and which profit and expense terms have been included.

Interest and Mortality Sensitivity

Fully specified twenty-year illustration

The following illustration is constructed solely for mathematical verification. It is not fitted to any country's mortality table and should not be used as an insured-life rate recommendation. A single insured aged x has a twenty-year term contract with S=100,000 currency units payable at the end of the covered year of death. Level premiums are due at the start of each covered year while alive. The effective annual valuation interest rate is 3.5 percent, and the conditional probability of death in year j, j=0,...,19, is

qx+j=0.00150+0.00012j.q_{x+j}=0.00150+0.00012j.

Thus the first-year q is 0.00150 and the twentieth-year q is 0.00378. Each rate lies strictly between zero and one. There are no lapses, taxes, dividends, surrender options, mortality improvements, or claims after the twentieth covered year. For gross pricing only, set α=200 at issue, c=35 per premium date, β=0.05 of gross premium, and K=1,000 at a covered death. There is no additional target profit loading.

Direct discounted summation yields A=0.0348095 and a=14.4458589, using the definitions above. The resulting net premium is P=240.9652375 currency units per premium date. Under the given expense model, the zero-profit gross premium is G=307.5996716. Currency units and payment timing apply uniformly to every number in the tables.

Reserve trajectory and identity test

The following table reports pre-premium net prospective V_t, independent forward retrospective R_t, and pre-premium gross W_t. Figures are rounded to four decimals for display.

Duration t Net prospective V_t Net retrospective R_t Gross W_t
0 0.0000 0.0000 0.0000
1 99.5483 99.5483 -92.4162
5 411.2553 411.2553 253.0241
10 582.1433 582.1433 470.2933
15 465.0965 465.0965 405.4752
19 124.2522 124.2522 111.6499
20 0.0000 0.0000 0.0000

The negative gross value at year one records the prospective recovery of the initial acquisition expense through later premium flows after that initial expense has already been paid. It has no direct interpretation as an authorized negative statutory reserve. In this illustration, the net value rises in the early middle durations because early premiums exceed contemporaneous expected death costs, then declines as the remaining premium-paying and coverage terms shorten.

The complete twenty-one-point net prospective and retrospective arrays agreed to a maximum absolute difference of approximately 6.82×10^-13 currency units using double-precision arithmetic. The initial gross expected-loss residual was approximately 9.09×10^-13, and both types of term reserve reached zero at duration twenty. These are calculation residuals for the hypothetical model, not measures of model error relative to actual insurance experience.

Interest sensitivity: repricing versus fixed-contract reserves

With mortality held at its base values, changing the valuation interest rate changes both discounting and the level premium required to reprice a newly issued policy. To measure the economic effect on an existing contract, the right comparison instead retains the original premium P=240.9652375 and revalues the remaining future cash flows at duration ten.

Annual interest rate Newly repriced net P Duration-ten V_10 using original P
2.0% 250.1096 673.3164
3.5% 240.9652 582.1433
5.0% 232.2720 502.3852

In this contract and range of assumptions, higher discount rates reduce the level net premium and the fixed-premium duration-ten prospective value. That sign is obtained from this cash-flow pattern and should not be asserted as a universal theorem for every life product, especially products with material investment guarantees, options, inflation-linked benefits, or changing premium schedules. For a value of the form V_t(i)=PV_i(benefits)-P PV_i(premiums), the derivative with respect to i is the difference of two duration-weighted derivatives; its sign depends on their timing. Repricing and existing-policy valuation must be labeled separately to avoid treating a newly computed premium as a retroactive contractual cash flow.

Mortality sensitivity

Now fix the valuation rate at 3.5 percent and multiply every annual q in the illustration by a scalar k, with k equal to 0.75, 1, or 1.25. This deliberately simplified perturbation preserves valid probability ranges and holds expense assumptions and the original premium schedule fixed for the prospective reserve comparison.

Mortality multiplier k Newly repriced net P Duration-ten V_10 using original P
0.75 180.9016 -75.1598
1.00 240.9652 582.1433
1.25 300.9108 1234.9794

The direction is intuitive for this pure death-benefit term contract: higher modeled mortality raises death-cost exposure and the required premium. At the lower mortality assumption the original contractual premium is high enough relative to modeled remaining death benefits that the technical prospective net value becomes negative. This is a conditional mathematical value and is not a claim about permissible financial-statement presentation. In other products, mortality improvements can increase the present value of annuity or survival guarantees; the sign of mortality sensitivity therefore depends on which contingent event creates the contractual benefit.

A scalar shock to all q values is a diagnostic, not an empirical mortality forecast. An actual mortality analysis may model insured selection, duration, medical developments, mortality improvement by calendar year, underwriting class, geographic heterogeneity, and statistical credibility. The Society of Actuaries' mortality experience and improvement studies provide appropriate starting points for investigating those dimensions, while source selection must be matched to product and jurisdiction.

Assumption governance and statistical uncertainty

Mortality rates estimated from past data have sampling and specification uncertainty. Rates observed in a population may not directly estimate insured-life mortality because underwriting, socioeconomic composition, exposure definitions, and contract durations differ. A large dataset can still be biased if its populations, definitions, or calendar periods are poorly matched to the contractual risk. The official 2015 valuation table documentation describes multiple alternative table classifications and subsequent updates; it does not license a one-size-fits-all forecast.

Discounting creates a separate basis question. A locked-in technical rate used for pricing, an expected asset return used for an internal projection, and a regulatory market-consistent curve are not automatically interchangeable. For a variable term structure, replace v^j by the appropriate discount factor D(0,j) and use D(t,u) for future amounts valued at t. If discount rates are stochastic and correlated with benefits, in-force behavior, or options, a deterministic discount factor applied to unconditional expected cash flows may not capture the joint distribution. Model design must preserve the dependence structure that is material to the product.

The international measurement boundary

The U.S. National Association of Insurance Commissioners describes principle-based reserving as replacing purely static formula assumptions with a framework that considers insurer experience and future economic conditions while retaining required minima. Its valuation manual archive lists annual editions, including the 2026 edition in effect for the research cutoff. VM-20 is relevant to individual life products within its specified scope and uses statutory concepts, including net premium, deterministic, and stochastic reserve requirements, that go beyond this dossier's single deterministic net premium formula. The Actuarial Standards Board's adopted ASOP 52 on life reserves addresses professional work under the NAIC Valuation Manual. These are U.S.-specific systems and must be read with their governing rules and current versions.

Under the European Union's Solvency II Article 77, technical provisions are generally the sum of a best estimate and a risk margin. The best estimate is a probability-weighted present value of future cash flows using the relevant risk-free term structure. The prescribed assumptions, cash-flow scope, and separate risk-margin computation differ from a historical issue-date net premium policy value. The direct reference to future settlement cash flows underscores the need to model expenses and policyholder behavior when material.

The IFRS Foundation's IFRS 17 measurement description locates financial reporting within grouped insurance contracts, current fulfillment cash-flow estimates, explicit nonfinancial risk adjustment, and a contractual service margin when applicable. The International Actuarial Association separately identifies ISAP 4 for IFRS 17 among its model standards. Neither measurement is the issue-equivalence calculation in another name. Comparing reported values requires a bridge that reconciles cash-flow definitions, grouping, assumptions, risk adjustments, discount curves, and profit-recognition rules.

Research synthesis and reproducibility

Three conclusions follow from the derivations and the worked checks. First, a level premium is a discounted, survival-weighted cash-flow ratio, so premium frequency and mortality exposure must be aligned. Second, a prospective reserve, an expected retrospective fund, and a reserve recursion are equivalent descriptions of the same stipulated technical valuation only when they contain the same cash flows on the same basis. Third, real-world reserve numbers require an explicit purpose label. Pricing, portfolio experience attribution, solvency supervision, and financial reporting can assign different values to the same contract without violating the underlying mathematics.

Reproduction needs only the stated q_j sequence, n, S, i, and expense assumptions. Generate survival weights iteratively as l_0=1 and l_{j+1}=l_j(1-q_j). Sum A and a. Calculate P and G from the closed forms. For every t, sum prospective benefits and premiums using conditional survival l_j/l_t. Separately initialize R_0=0 and apply the forward recursion. Check the gross prospective sum and the gross recursion. Repeat the pricing calculation under each changed interest or mortality basis; for reserve sensitivity retain the base P. Store calculations in adequate floating-point precision and round only for presentation.

This is a complete theoretical and numerical demonstration of a narrowly specified product. Applied implementation would also require experience calibration, lapse and expense behavior, parameter uncertainty, realistic discount curves, asset-liability and capital models, statutory contract boundaries, and independent governance appropriate to the undertaking.

Conclusion

Life insurance premium and reserve mathematics rests on a single disciplined treatment of contingent cash flows. Discount factors assign time value, survival and death probabilities assign event weights, policy provisions establish which amounts are paid, and the premium schedule establishes the financing stream. On a fixed technical basis the net equivalence premium balances expected present values at issue. Prospective reserves apply that same comparison to the remaining contingent cash flows. Retrospective reserves allocate the accumulated expected past funding balance to policies surviving to the valuation date. The two constructions are equal under common assumptions, and the reserve recursion expresses their common one-period cash-flow identity.

Expense loadings make the distinction between net and gross mathematics explicit. Acquisition, maintenance, proportional collection, and death-claim expenses have different triggers and payment dates. A gross premium must reflect those cash-flow definitions. A target profit loading can be included in pricing while remaining conceptually separate from a genuine future claim or operating expense.

The twenty-year demonstration verifies the main identities numerically and shows how changes in mortality or discount rates alter both newly priced premiums and the value of an existing policy with contractual premiums held fixed. Its results are reproducible and intentionally illustrative. Their generalizable content lies in the form of the cash-flow and sensitivity analysis, not in the hypothetical numerical rates.

International frameworks demonstrate the additional discipline needed to move from a theoretical individual policy value to a formally reported reserve or liability. U.S. principle-based reserves, Solvency II technical provisions, and IFRS 17 insurance contract liabilities incorporate institutional scope, risk treatment, assumption selection, grouping, and recognition requirements. The coherent scientific interpretation of a life reserve therefore begins by declaring its valuation purpose, contract boundary, contingent cash flows, assumption basis, and reporting instant. Once those definitions are fixed, both the mathematics and its limitations become testable.

References