Advanced Long Term Actuarial Mathematics

Advanced Long-Term Actuarial Mathematics

When to Use

• Build and interpret survival models: \(l_x\), \(q_x\), \(\mu_x\), select vs ultimate mortality
• Value life insurance and annuity benefits (term, whole life, endowment, temporary/whole life annuities)
• Derive premiums and reserves via equivalence principle, Thiele, prospective vs retrospective views
• Model multiple decrements and Markov/multi-state paths (disability, pension, active/retired)
• Apply yield curves and discounting for long-dated liability cash flows
• Frame mortality improvement and longevity risk at a technical level
• Outline profit testing and emerging cost for life products (conceptual)
• Estimate and graduate mortality tables; document data and smoothing choices
• Connect math to pricing, reserving, and pension liability measurement; hand execution to actuarial-analyst

When NOT to Use

• P&C frequency-severity, aggregate loss, credibility, or short-tail reserving math → advanced-short-term-actuarial-mathematics
• Triangle workbooks, exhibits, statutory tie-outs, or model run packs only → actuarial-analyst
• Appointed actuary opinions, regulatory sign-off, or enterprise capital policy → actuary, appointed-chief-actuary
• Enterprise assumption governance, assumption papers, and change control → assumption-setting
• ALM duration matching, LDI hedge design, and investment policy depth → asset-liability-management
• Life/health product mechanics, underwriting, or claims without contingency math focus → life-health-insurance
• Pension ERISA, PBGC, fiduciary, or plan design narrative without contingency formulas → pension-retirement-funds
• Exam cram or past-exam solutions as the sole deliverable (support professional application; exam study is secondary)
• General data science, ML pipelines, or quant research without life-contingency framing → data-scientist, quantitative-researcher
• Credential pathway and exam strategy only → associate-actuary

Related skills

Core Workflows

1. Problem framing (ALTAM-aligned)

Before deriving formulas:

1. Horizon — Long-duration; valuation vs pricing vs funding purpose 2. State space — Single life, joint life, or multi-state (active/disabled/retired/dead) 3. Benefit/premium basis — Net vs gross; continuous vs discrete; payment timing 4. Mortality basis — Ultimate vs select; table vintage; improvement scale (if any) 5. Interest/discount — Flat rate vs term structure; real vs nominal 6. Deliverable — Formula spec, APV tables, reserve recursion, sensitivity—not filing sign-off 7. Peer execution — Route spreadsheet builds and filing exhibits to actuarial-analyst

See `references/altam_scope_and_principles.md`.

2. Survival models and mortality

1. Define life table functions: \(l_x\), \(d_x\), \(q_x\), \(p_x\), \(\mu_x\) 2. Distinguish ultimate vs select mortality and duration since selection 3. State independence assumptions for joint lives unless modeling dependence 4. Document radix, age basis, and table source (regulatory, industry, experience) 5. Link force of mortality to discrete \(q_x\) under stated assumptions

See `references/survival_models_and_mortality.md`.

3. Life insurance and annuities

1. Map benefit types: level term, whole life, endowment, deferred benefits 2. Write actuarial present value (APV) with correct survival and payment timing 3. Cover annuities: temporary, whole life, due vs immediate, varying payments 4. Address joint-life and last-survivor structures at formula level 5. Flag riders and options for product context → life-health-insurance

See `references/life_insurance_and_annuities.md`.

4. Premiums and reserves

1. Apply equivalence principle for net/gross premiums; define expense and profit loadings 2. Build prospective and retrospective reserves; reconcile where standard 3. Use Thiele's equation for continuous reserves; discrete analogs where needed 4. Distinguish net level premium vs gross premium reserves 5. Summarize deficiency and emerging cost implications (conceptual)

See `references/premiums_and_reserves.md`.

5. Multiple decrement and state models

1. Set up multiple decrement table: dependent vs independent decrements 2. Compute probabilities of single and combined decrements; associated single decrement rates 3. Extend to Markov multi-state models (transition intensities, Kolmogorov forward equations) 4. Apply to disability, pension, and long-term care state spaces (math only) 5. Hand pension plan law and governance narrative to pension-retirement-funds

See `references/multiple_decrement_and_state_models.md`.

6. Longevity, improvement, estimation, and discounting

1. Frame mortality improvement scales and longevity risk (framework-level) 2. Describe estimation from experience; graduation and smoothing methods 3. Apply yield curves to long cash flows; note ALM interaction at high level → asset-liability-management 4. Outline profit testing cash flows and sensitivity to mortality and interest 5. Package assumptions for actuary review; governance → assumption-setting

See `references/longevity_improvement_and_estimation.md`.

Deliverable standards

Label output as technical modeling support, not actuarial opinion, legal advice, or filed regulatory submission.

Assignment type matrix

When to load references

• Scope, ALTAM alignment, principles → references/altam_scope_and_principles.md
• Survival models and mortality → references/survival_models_and_mortality.md
• Life insurance and annuities → references/life_insurance_and_annuities.md
• Premiums and reserves → references/premiums_and_reserves.md
• Multiple decrement and state models → references/multiple_decrement_and_state_models.md
• Longevity, improvement, estimation, discounting → references/longevity_improvement_and_estimation.md

ALTAM scope and principles

Table of contents

1. Purpose and alignment 2. Long-term vs short-term 3. Core building blocks 4. Modeling principles 5. Boundaries with peer skills 6. Ethics and reliance

Purpose and alignment

This reference supports advanced long-term actuarial mathematics in the spirit of SOA ALTAM (Advanced Long-Term Actuarial Models): stochastic and deterministic models for life contingencies, annuities, and pension-adjacent liabilities where survival, decrements, and long-dated cash flows drive outcomes.

The skill is tool-agnostic and concept-first. Implementation in R, Python, Excel, Prophet, or vendor valuation systems belongs in actuarial-analyst unless the user explicitly wants formulas and interpretation only.

Long-term vs short-term

Route P&C aggregate-loss work to advanced-short-term-actuarial-mathematics.

Core building blocks

Standard life contingency setup:

• Survival — \(l_x\), \(q_x\), \(p_x\), \(\mu_x\); select vs ultimate tables
• Benefits — Death, survival, annuity payments with defined timing (due vs immediate)
• Discounting — Interest \(i\) or force \(\delta\); term structures for long liabilities
• Premiums/reserves — Equivalence principle; prospective and retrospective balances
• Decrements — Death, lapse, disability, retirement in single or multi-state form

Joint lives and last-survivor benefits extend the state space; Markov models generalize decrements.

Modeling principles

1. Define the estimand — APV, net premium, reserve, funded status contribution—not notation alone 2. State payment timing — Continuous vs discrete; due vs immediate; fractional year conventions 3. Document mortality basis — Table name, year, select period, improvement scale, credibility of experience 4. Interest and inflation — Nominal vs real; flat vs curve; alignment with ALM when relevant 5. Prospective purpose — Pricing vs valuation vs funding; net vs gross; statutory vs economic (overview) 6. Reproducibility — Assumption set ID, table version, and rounding rules for published factors

Boundaries with peer skills

Ethics and reliance

• Present results as technical modeling pending actuary review
• Do not substitute for appointed actuary statements, rate filing adequacy, or pension actuarial opinion
• Cite table vintage, improvement uncertainty, and small experience in limitations
• Exam preparation may use this material but should not reduce deliverables to memorized exam templates without business context

Life insurance and annuities

Table of contents

1. Actuarial present value (APV) 2. Life insurance benefits 3. Annuities 4. Varying benefits and expenses 5. Joint-life and group extensions

Actuarial present value (APV)

Actuarial present value of a payment stream:

\[ \text{APV} = \mathbb{E}\left[\sum \text{payments} \times \text{discount} \times \text{survival/decrement indicators}\right] \]

Standard commutation notation (ultimate mortality, interest \(i\)):

State whether functions are due (\(D_x\) at start of year) or immediate.

Life insurance benefits

Common unit benefits (pay 1 on event):

Deferred insurance multiplies by \({}_m p_x\) and discount for deferral period \(m\).

Annuities

Continuous annuities use \(\bar{A}\), \(\bar{a}\) with \(\delta\) instead of \(i\).

Temporary annuity certain adds non-life contingency layer when guaranteed periods apply.

Varying benefits and expenses

• Level vs increasing benefits (e.g., \( (1+j)^t \) — document inflation linkage)
• Expense charges as percent of premium or per policy; route gross premium work to premiums_and_reserves.md
• Refund or return of premium features alter death benefit path

Product wording and rider mechanics → life-health-insurance.

Joint-life and group extensions

• Joint-life annuity — pay while both alive; APV uses joint survival
• Last survivor — pay until last death; requires joint decrement logic
• Group certificates often use salary scales and service — pension state models in multiple_decrement_and_state_models.md

Do not present APV factors as pricing without expense, persistency, and capital margins.

Longevity improvement, estimation, and discounting

Table of contents

1. Mortality improvement frameworks 2. Longevity risk 3. Estimation from experience 4. Graduation and smoothing 5. Yield curves and long cash flows 6. Profit testing overview

Mortality improvement frameworks

Mortality improvement — systematic decline in mortality rates over calendar time.

Common approaches (conceptual):

Distinguish:

• Base table (cross-section at valuation date)
• Projection (future calendar years)
• Long-term rate vs short-term improvement

Longevity risk

Longevity risk — risk that realized mortality is lower than assumed (payments last longer).

Sources:

• Process risk — random fluctuation in small portfolios
• Level risk — wrong average mortality level
• Trend risk — improvement faster than assumed
• Basis risk — mismatch between hedge and liability cohort

Mitigation (overview only): reinsurance, longevity swaps, pension risk transfer → pension-retirement-funds; ALM for interest-longevity interaction → asset-liability-management.

Estimation from experience

Workflow for experience analysis (math framing; execution in actuarial-analyst):

1. Define exposure (policy years, central exposed to risk) 2. Stratify age, gender, duration, product, underwriting class 3. Compute \( \hat{q}_x = \text{deaths} / \text{exposure} \) with credibility checks 4. Compare to industry or regulatory standard table (A/E ratios) 5. Propose standard table or blend; submit to assumption-setting

Address censoring (lapses, end of study) and large claims separately for health riders.

Graduation and smoothing

Graduation — smooth raw \(\hat{q}_x\) to produce publishable rates.

Checks after graduation:

• Smoothness — no jagged runs in \(q_x\)
• Adherence — weighted fit to raw experience
• Tail — plausible at high ages; avoid zero denominators

Yield curves and long cash flows

Long-dated liabilities require discounting aligned with cash-flow timing:

1. Map liability cash flows by duration bucket 2. Select spot or par curve; document liquidity premium if any 3. Compute duration and convexity for ALM dialogue → asset-liability-management 4. For pensions, note segment rates (IRC) vs market curves (economic)

Nominal vs real — link benefit indexation (COLA) to inflation assumptions in assumption-setting.

Profit testing overview

Profit testing for life products (conceptual steps):

1. Projection — premiums, claims, expenses, reserves, investment income by duration 2. Scenario grid — mortality, lapse, interest, expense shocks 3. Metrics — PV profit, profit margin, IRR, break-even year 4. Regulatory capital — cite need for actuary; do not fabricate RBC formulas without source

Emerging cost ties actual experience to pricing assumptions over time—pair with premiums_and_reserves.md.

Hand model implementation and sensitivity exhibits to actuarial-analyst; pricing sign-off to actuary.

Multiple decrement and state models

Table of contents

1. Multiple decrement tables 2. Associated single decrements 3. Markov multi-state models 4. Pension and disability applications 5. Implementation notes

Multiple decrement tables

Multiple decrement model: several exit causes (death, lapse, retirement, disability) in the same interval.

Notation (example causes \(j=1,\ldots,m\)):

• \(q_x^{(j)}\) — probability of decrement \(j\) between \(x\) and \(x+1\)
• \(q_x^{(\tau)}\) — total decrement probability: \(q_x^{(\tau)} = \sum_j q_x^{(j)}\) (if mutually exclusive)
• \(l_x^{(\tau)}\) — survivors in active state before decrements

\[ l_{x+1}^{(\tau)} = l_x^{(\tau)} (1 - q_x^{(\tau)}) \]

Dependent decrements — all causes compete in the same interval (standard for pension active populations).

Independent decrements — assume each cause operates on a hypothetical population; use only when theory or regulation requires.

Associated single decrements

Associated single decrement rate \(q_x^{'(j)}\): rate for cause \(j\) if it were the only decrement.

Common relationship (Makeham-style independence assumption):

\[ 1 - q_x^{(\tau)} = \prod_j (1 - q_x^{'(j)}) \]

Document when converting experience by cause to pricing tables.

Markov multi-state models

States \(S = \{1,\ldots,K\}\); transition intensity \(\mu_{ij}(t)\) from \(i\) to \(j\).

Kolmogorov forward equations for probabilities \(P_{ij}(s,t)\):

\[ \frac{\partial}{\partial t} P_{ij}(s,t) = \sum_k P_{ik}(s,t)\,\mu_{kj}(t) - \mu_{ij}(t)\,P_{ij}(s,t) \]

Actuarial present values weight cash flows by state occupancy probabilities.

Typical state spaces:

Pension and disability applications

Pension liability measurement (conceptual):

1. Project benefit accrual by status and service 2. Apply decrements (termination, mortality, disability, retirement) 3. Discount with segment rates or curve (overview) — detail in pension-retirement-funds 4. Sum APV of projected benefits

Disability income — payment while disabled; recovery to active; waiver of premium as rider math.

Do not conflate funding vs accounting vs economic measurement bases.

Implementation notes

1. Conservation — decrements sum to at most total exit probability 2. Timing — decrements at start vs end of interval 3. Selection — disability rates often depend on duration in state 4. Data — sparse causes need graduation or pooling 5. Peer skills — plan design narrative → pension-retirement-funds; tables → actuarial-analyst

Premiums and reserves

Table of contents

1. Equivalence principle 2. Net and gross premiums 3. Reserve definitions 4. Thiele's equation 5. Deficiency and emerging cost

Equivalence principle

Equivalence principle: actuarial present value of benefits = actuarial present value of premiums (at issue, for net premium).

\[ \text{APV}(\text{benefits}) = \text{APV}(\text{premiums}) \]

Solve for level net premium \(P\) over \(m\) payments on an \(n\)-year policy:

\[ P \cdot \ddot{a}_{x:\overline{m\|}} = \text{APV}(\text{benefits}) \]

Document funding period vs benefit period when they differ.

Net and gross premiums

Expense formats:

• Percent of premium (renewal commission analog)
• Percent of sum insured per year
• Fixed per policy per year

Gross premium equivalence: \(\text{APV}(\text{benefits}+\text{expenses}) = \text{APV}(\text{gross premiums})\).

Hand expense studies and experience allocation to actuarial-analyst.

Reserve definitions

At duration \(t\) (prospective, net premium reserve):

\[ {}_t V = \text{APV}(\text{future benefits}) - \text{APV}(\text{future net premiums}) \]

Retrospective form (when applicable):

\[ {}_t V = \text{APV}(\text{past premiums}) - \text{APV}(\text{past benefits}) \quad \text{(with adjustments)} \]

Gross premium reserve uses gross premium income and includes unearned expense allowances per convention.

Prospective vs retrospective reconciliation is a standard control—state assumptions if they differ.

Thiele's equation

Continuous Thiele reserve dynamics:

\[ \frac{d}{dt} {}_t V = \delta \, {}_t V + P_t - \mu_{x+t} \,(1 - {}_t V) - \text{(other decrements)} \]

Interpretation:

• \(\delta \, {}_t V\) — interest on reserve
• \(P_t\) — premium income rate
• \(\mu_{x+t}(1 - {}_t V)\) — death benefit outgo net of reserve released

Discrete analogs step reserves year-by-year with \(q_x\), \(p_x\), and expense cash flows.

Use Thiele for sensitivity and profit emergence; discrete recursions for valuation tables.

Deficiency and emerging cost

Deficiency (conceptual): prospective reserve negative at some duration under net premium basis—signals inadequate premium or high early benefits.

Emerging cost / profit testing:

1. Project cash flows by policy year (premiums, claims, expenses, investment) 2. Discount at earned rate or hurdle rate 3. Allocate profit by duration; test IRR and NPV under scenarios

Link assumption changes to assumption-setting; full model builds to actuarial-analyst.

Survival models and mortality

Table of contents

1. Life table notation 2. Force of mortality 3. Select vs ultimate mortality 4. Joint-life structures 5. Practical checks

Life table notation

Given radix \(l_0\) (often \(l_0 = 100{,}000\)):

Curtate vs complete life assumptions affect fractional-year benefits; state which convention applies.

Force of mortality

Force of mortality at age \(x\):

\[ \mu_x = \lim_{t \to 0} \frac{{}_t q_x}{t} \]

Under constant force between integer ages: \({}_t p_x = e^{-\int_0^t \mu_{x+s}\,ds}\).

Discrete link (common approximation):

\[ q_x \approx 1 - e^{-\mu_x}, \quad \mu_x \approx -\ln(1 - q_x) \]

Use consistent conversion when moving between continuous and discrete models.

Select vs ultimate mortality

• Ultimate table: mortality depends on attained age only
• Select table: lower mortality shortly after underwriting (select period), then merges to ultimate

Document:

1. Select period length and basis (e.g., duration since issue) 2. Ultimate rates used after select wears off 3. Whether aggregate or individual underwriting applies

Misapplying select tables (e.g., on renewals without selection) biases pricing and reserves.

Joint-life structures

For lives \((x)\) and \((y)\) with independence (unless modeling dependence explicitly):

• Joint survival \({}_n p_{xy} = {}_n p_x \cdot {}_n p_y\)
• Last survivor probabilities combine survival and single-death events
• First death benefits use \({}_n q_{xy}\) or complementary structures

State correlation or common shock if independence is relaxed (escalate to specialist models).

Practical checks

1. Monotonicity — \(l_x\) non-increasing; \(q_x \in [0,1]\) 2. Closure — \(l_{x+1} = l_x (1 - q_x)\) within rounding 3. Radix and age — Issue age vs attained age; birthday conventions 4. Gender/smoker — Segment tables; do not blend without explicit rules 5. Improvement — Distinguish base table from projection scale (see longevity_improvement_and_estimation.md)

Hand experience studies and credibility-weighted table construction execution to actuarial-analyst; assumption approval to assumption-setting.