7. Actuarial Foundations, Pricing, and Reserving

Introduction and Context

Insurance converts uncertain future losses into a more predictable financial arrangement. Policyholders pay premiums in advance or periodically, while insurers promise to pay benefits or indemnities when specified events occur. The actuarial function provides the quantitative foundation for this exchange. It estimates the frequency and severity of insured events, discounts future cash flows, allows for expenses and capital requirements, and determines whether premiums and reserves are adequate.

Actuarial work is not limited to calculating a premium. It connects probability, economics, finance, statistics, demography, claims management, regulation, and business strategy. In life insurance, mortality, longevity, morbidity, lapses, investment returns, and expenses are central. In health insurance, incidence, treatment costs, provider behaviour, medical inflation, and utilization patterns matter. In general insurance, claims frequency, claims severity, catastrophe exposure, inflation, legal developments, and claims settlement practices are particularly important.

Pricing concerns the amount charged for future insurance protection. Reserving concerns the amount that must be held to meet obligations arising from policies already written or claims already incurred. A product may be profitable on a pricing basis but still require substantial reserves because benefits will be paid many years later. Conversely, inadequate pricing may be concealed temporarily by favourable claims experience or investment income.

This chapter develops the principal actuarial concepts used in life, health, and general insurance. It also considers credibility, experience rating, technical provisions, sensitivity analysis, professional judgement, model risk, and the interpretation of actuarial reports.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the role of probability, mortality and morbidity tables, and time value of money in insurance.

2. Distinguish frequency, severity, aggregate loss, expected loss, and risk measures.

3. Construct insurance premiums by allowing for claims, expenses, profit, contingencies, and capital.

4. Describe credibility theory and experience rating and identify suitable applications.

5. Explain the purpose and methods of life, health, and general insurance reserving.

6. Distinguish case reserves, incurred-but-not-reported reserves, and premium-related provisions.

7. Apply sensitivity analysis to actuarial assumptions and interpret its results.

8. Evaluate actuarial reports, including their assumptions, limitations, governance, and uncertainty.

The Actuarial View of Insurance Risk

Random variables and probability

An insured loss can be represented by a random variable. Let X denote the monetary amount of a claim from one policy during a specified period. X may be zero if no claim occurs. Its expected value, E(X), is the probability-weighted average of all possible outcomes. Although the expected value is useful for pricing, it does not describe the entire risk. Two portfolios may have the same expected loss but very different variability, skewness, or catastrophe potential.

For a portfolio of n policies, aggregate loss is commonly written as:

S = X1 + X2 + ... + XN

where N is the number of claims and Xi is the amount of the ith claim. N is the frequency variable, while Xi is the severity variable. If frequency and severity are independent, the expected aggregate loss is:

E(S) = E(N) × E(X).

This relationship is fundamental but should not be applied mechanically. Dependence may arise from weather events, epidemics, inflation, common medical providers, economic conditions, or changes in legal rules. Catastrophe risks are especially unlikely to satisfy simple independence assumptions.

Frequency and severity models

Frequency models estimate how often claims occur. Poisson, binomial, negative binomial, and zero-inflated models are frequently used, depending on the characteristics of the data. The Poisson model assumes that events occur independently at a stable average rate. The negative binomial model allows greater dispersion and is often more suitable where observed variation exceeds the Poisson assumption.

Severity models estimate claim amounts. Lognormal, gamma, Pareto, Weibull, and related distributions may be considered. Heavy-tailed distributions are important in liability, catastrophe, and large commercial claims because a small number of claims may account for a large share of total costs. Deductibles, policy limits, coinsurance, and reinsurance transform the original severity distribution. Pricing must therefore model the insured layer rather than merely the reported gross loss.

The aggregate-loss distribution supports decisions about premiums, reinsurance, capital, solvency, and risk appetite. Expected loss alone is insufficient where management must protect against adverse but plausible outcomes.

Mortality, Morbidity, and Life-Contingent Cash Flows

Mortality tables

A mortality table records the survival and death experience of a defined population. Important functions include lx, the number surviving to age x; dx, the number dying between ages x and x+1; and qx, the probability that a person aged x dies before reaching age x+1. The one-year survival probability is px = 1 − qx.

Mortality tables may be based on population data or insured lives. Insured-life tables can differ from population tables because of underwriting, socioeconomic selection, product design, and policyholder behaviour. Tables may also include mortality improvement assumptions, which are particularly significant for annuities and long-term protection products.

Morbidity tables

Morbidity tables measure the incidence, duration, recovery, or termination of illness and disability. They are used in health insurance, critical illness, income protection, long-term care, and disability products. A morbidity assumption may concern the probability of entering a claim state, the probability of recovery, the duration of a claim, or the transition from one health state to another.

Health insurance often requires more than a simple annual incidence rate. Costs depend on diagnosis, treatment intensity, provider networks, age, geographic location, medical technology, admission rates, length of stay, and prescription patterns. Consequently, morbidity models are frequently linked to utilization and cost models.

Life-contingent valuation

Life insurance benefits are contingent on survival or death. The present value of a future payment is obtained by combining a discount factor with the probability that the payment will occur. If a benefit of B is payable at the end of one year on death, its expected present value is approximately:

B × v × qx,

where v = 1/(1+i), i is the effective annual interest rate, and qx is the probability of death during the year. Multi-year benefits require the relevant survival and death probabilities for each future period.

An annuity payable while a person survives has a different structure. Each payment is multiplied by the probability of survival to the payment date and by the corresponding discount factor. The actuarial present value therefore reflects both financial uncertainty and demographic uncertainty.

Time Value of Money and Assumptions

Insurers receive premiums before many benefits are paid. Money available today can earn investment income, so future cash flows must be discounted. The discount rate may be a risk-free rate, a prescribed regulatory rate, an investment-based rate subject to constraints, or a rate consistent with the valuation framework.

The choice of discount rate has major consequences. A higher rate reduces the present value of long-term liabilities, while a lower rate increases it. Long-duration products such as annuities and whole-life insurance are particularly sensitive to interest rates. Discounting should not be used to conceal uncertainty or to justify assumptions unsupported by assets and investment strategy.

Other important assumptions include mortality, morbidity, lapse, expenses, inflation, claim settlement speed, recoveries, taxes, and future business volumes. Assumptions may be deterministic, stochastic, best-estimate, prudent, or prescribed, depending on the purpose of the calculation.

Premium Construction

A premium is normally developed in stages rather than selected as a single arbitrary amount. The central components are expected claims, expenses, risk or contingency margins, cost of capital, profit, and sometimes taxes or levies.

The pure risk premium, or expected loss cost, is the expected value of insured claims per exposure unit. A simplified gross premium may be expressed as:

Gross premium = expected claims + claim expenses + acquisition expenses + administration expenses + risk margin + profit allowance + taxes and charges.

If expenses are expressed as proportions of premium, an algebraic loading method may be needed. For example, if expected claims are 60 monetary units, fixed expenses are 10, and variable expenses are 20% of premium, the premium before profit and other charges is:

P = 60 + 10 + 0.20P,

so P = 87.50. The calculation illustrates why a percentage loading cannot simply be added to claims without considering its denominator.

Life insurance premiums

Life premiums may be level, stepped, single, or flexible. A level premium is designed so that payments are relatively stable even though the underlying mortality cost generally increases with age. Early premiums may exceed the current cost of insurance, producing a reserve that supports later benefits. A stepped premium more closely follows changing risk costs but may become expensive at older ages.

The equivalence principle sets the expected present value of premiums equal to the expected present value of benefits and expenses, subject to the chosen basis. Commercial pricing may then add profit, capital costs, distribution costs, and risk margins. Product design must consider guarantees, options, surrender values, bonuses, and policyholder behaviour.

Health insurance pricing

Health premiums depend on expected utilization multiplied by expected cost per service, adjusted for policy terms and expenses. Medical trend includes price inflation, changes in treatment patterns, technology, provider charges, and changes in the mix of insured members. Selection is important because individuals with higher expected healthcare needs may be more likely to purchase or retain coverage.

Risk adjustment, deductibles, copayments, benefit limits, network design, preauthorization, and disease-management programmes affect both claim costs and behaviour. Pricing models should distinguish genuine risk differences from variables that are legally or ethically inappropriate to use.

General insurance pricing

General insurance rates may be developed using exposure-based methods, loss-cost approaches, burning-cost methods, generalized linear models, or catastrophe models. Rating factors can include location, construction, vehicle characteristics, occupation, prior claims, limits, deductibles, and security controls.

A technical price should be distinguished from a market price. The technical price reflects estimated risk and required return; the market price also reflects competition, capacity, strategic objectives, customer relationships, and regulatory constraints. Persistent divergence between price and technical cost requires explicit management rather than accidental cross-subsidization.

Credibility and Experience Rating

A new or small portfolio may have little reliable loss experience. A large portfolio may have substantial experience but still be affected by random fluctuation. Credibility theory combines collective information, such as industry or portfolio data, with the individual experience of the risk.

A simple credibility premium is:

P = ZP_e + (1 − Z)P_c,

where Pe is the experience-based estimate, Pc is the collective or manual premium, and Z is the credibility factor between zero and one. Z generally increases with exposure volume, data quality, stability, and the relevance of the experience.

Limited-fluctuation credibility asks whether the observed number of claims is sufficient for the actual experience to fall within a desired tolerance of the underlying expected value with a specified probability. Bayesian credibility treats prior collective information and observed experience as components of a posterior estimate.

Experience rating is common in group health, workers’ compensation, commercial motor, and other products where the policyholder’s own experience is informative. It must be adjusted for changes in exposure, benefit design, claims handling, inflation, large losses, and credibility. A reduction in claims following a risk-management intervention may justify a lower rate, but the actuary must determine whether the improvement is persistent and causally related to the intervention.

Reserving and Technical Provisions

Pricing looks forward from the point at which coverage is offered. Reserving looks backward and forward simultaneously: it considers obligations arising from past underwriting and estimates the future payments required to discharge them.

Life reserves

For a life policy, the prospective reserve is commonly the present value of future benefits and expenses minus the present value of future net premiums. The reserve is therefore not simply the amount of premium collected. It depends on the valuation assumptions, policy duration, guarantees, expenses, surrender behaviour, and investment basis.

A retrospective reserve may be calculated from accumulated past premiums and investment returns less accumulated benefits and expenses. Under consistent assumptions, prospective and retrospective methods should reconcile, subject to differences in timing and valuation conventions.

Health and general insurance reserves

Claim reserves include amounts for reported claims that have not yet been paid and claims incurred but not reported (IBNR). A case reserve is established for an individual reported claim. IBNR captures claims that have occurred but are not yet reported, as well as development on known claims that has not been fully recognized.

Premium-related provisions cover future claims and expenses associated with unexpired coverage. Depending on the accounting or regulatory framework, these may be described as unearned premium provisions, liability for remaining coverage, premium deficiency reserves, or loss-component provisions.

Reserving methods

The chain-ladder method projects historical development patterns from reported or paid claims. It is useful where past development is reasonably stable, but it can be misleading after changes in claims handling, settlement practices, policy limits, or exposure mix.

The Bornhuetter–Ferguson method combines an a priori expected loss ratio with observed emergence. It is often useful for immature accident years or portfolios with limited credible experience. The expected-loss method relies heavily on the initial expected loss ratio and is therefore appropriate only when that assumption is well supported.

Other approaches include frequency-severity reserving, loss-ratio methods, generalized linear models, stochastic reserving, and scenario-based approaches. Stochastic models produce a distribution of possible reserve outcomes rather than a single estimate, supporting risk margins and capital analysis.

Technical provisions generally comprise a best-estimate liability and an additional margin for uncertainty, depending on the applicable regulatory or accounting framework. The exact terminology varies across jurisdictions, but the underlying principle is that reported liabilities should be sufficient, neither deliberately optimistic nor unnecessarily excessive.

Sensitivity, Scenario, and Capital Analysis

A point estimate can create false confidence. Sensitivity analysis changes one assumption at a time—for example, mortality, morbidity, lapse, inflation, discount rate, or claim settlement speed—to show the effect on premiums, reserves, profit, or solvency. Scenario analysis changes several related assumptions to represent a coherent event, such as a severe epidemic, a prolonged inflation shock, or a major catastrophe.

Stress testing examines adverse but plausible conditions. Reverse stress testing begins with a specified failure outcome, such as breach of a solvency threshold, and identifies the circumstances that could produce it. These techniques help boards and management understand risk concentration and the adequacy of capital.

The most sensitive assumption is not necessarily the most uncertain one. A small change in an assumption with high financial leverage may matter more than a large change in a relatively immaterial assumption. Results should therefore be presented with explanations of materiality, correlation, plausibility, and management action.

Reading and Challenging an Actuarial Report

An actuarial report should identify its purpose, valuation date, scope, data sources, methods, assumptions, material judgements, results, and limitations. A reader should ask:

• Are the data complete, accurate, and consistent with policy and claims records?
• Have exposure changes, inflation, claims handling, and benefit changes been allowed for?
• Are assumptions supported by experience, external evidence, or documented judgement?
• Is the method appropriate for the maturity and stability of the portfolio?
• Are uncertainty, adverse development, and model risk disclosed?
• Are results reconciled to accounting records and prior reports?

An actuarial opinion is not a guarantee. It is a professional conclusion based on specified information, assumptions, methods, and materiality thresholds. The report should distinguish facts, assumptions, estimates, and recommendations. Management remains responsible for decisions, governance, controls, and the consequences of using the results.

Illustrative Case Study: Pricing and Reserving for a Group Health Portfolio

Background

Insurer Meridian provides annual medical insurance to mid-sized employers. The portfolio has grown rapidly, but its claims data are only moderately credible because several large groups joined during the year. The insurer is preparing renewal rates and year-end technical provisions.

Situation and problem

The observed claims ratio is 68%, compared with the original expected ratio of 62%. However, medical inflation is estimated at 8%, one large claim distorted the year, and a new hospital network has changed treatment patterns. Management wants to apply the observed ratio directly to all renewals and use a chain-ladder reserve based on paid claims.

Analysis

The actuary first separates ordinary claims from the exceptional large claim and adjusts the experience for exposure growth, benefit changes, and the timing of earned premium. A frequency-severity analysis indicates that utilization has increased moderately, while the average cost per admission has risen materially. The new network has shortened settlement delays but increased the average treatment cost. Consequently, paid development patterns from the previous year are not fully representative.

Credibility analysis gives 60% weight to the portfolio’s adjusted experience and 40% to an external collective benchmark. The projected claims ratio is then trended for medical inflation and adjusted for the new benefit design. For reserving, the actuary uses a Bornhuetter–Ferguson approach for the immature accident period and a development method for mature periods. A separate allowance is made for IBNR and for claims reported but not yet fully adjudicated.

Sensitivity analysis shows that a two-percentage-point increase in medical trend has a larger effect on the reserve than removal of the exceptional claim. A stress scenario involving increased hospital utilization would reduce the portfolio’s solvency margin substantially.

Approach and solution

The insurer adopts risk-adjusted renewal rates rather than applying the raw observed claims ratio. Groups with credible and stable experience receive experience-rated adjustments, while smaller groups are priced using a credibility-weighted manual rate. Provider contracts are reviewed, claims notifications are monitored, and a utilization-management programme is introduced. The reserve is supported by a documented method selection memorandum and reconciled to claims and finance data.

Outcome

The final pricing indication is higher than the initial management proposal but lower than a simple extrapolation of the raw claims ratio. The reserve increases because the original paid-development assumption understated the effect of changed settlement and provider patterns. Management accepts the rate indication and establishes quarterly monitoring of trend, utilization, large claims, and reserve development.

Learning points

The case demonstrates that actuarial estimates require adjustment for data maturity, structural change, credibility, and claims processes. A single observed ratio is not automatically a reliable price. Similarly, a reserving method should reflect the nature of the available data rather than convenience alone.

Professional, Ethical, and Governance Considerations

Actuaries must act with competence, integrity, objectivity, and due care. They should disclose material limitations, avoid unsupported precision, and document significant judgement. Conflicts may arise when commercial targets encourage optimistic assumptions or when the same individual is pressured to approve both pricing and reporting results without adequate challenge.

Governance should include clear ownership of assumptions, independent review for material models, version control, data validation, experience monitoring, and escalation of adverse results. Regulatory frameworks may prescribe valuation bases, capital requirements, governance standards, and actuarial opinions. Compliance with a prescribed formula does not eliminate the need for professional judgement.

Fairness is also important. Pricing variables should have a defensible relationship to risk and should comply with applicable anti-discrimination and consumer-protection requirements. In health and life insurance, sensitive personal data require strict confidentiality, lawful processing, appropriate access controls, and careful explanation of automated decisions.

Technology, Analytics, and Emerging Practice

Modern actuarial work increasingly uses cloud computing, automated data pipelines, generalized linear models, machine learning, telematics, electronic health records, and real-time claims analytics. These tools can improve segmentation and identify fraud or emerging trends, but predictive performance does not by itself establish fairness, causality, or regulatory acceptability.

Machine-learning models may be difficult to interpret, unstable when the environment changes, and vulnerable to data leakage or biased historical decisions. Model governance should therefore address explainability, validation, drift monitoring, cybersecurity, privacy, and human oversight. Artificial intelligence should support—not replace—professional accountability.

Emerging practice also includes climate-risk modelling, pandemic scenarios, longevity risk analysis, embedded insurance, usage-based insurance, parametric covers, and more frequent or continuous pricing. These developments increase the value of actuarial analytics while making assumption governance and model limitations more important.

Best Practices

Sound actuarial practice includes the following:

• Define the purpose, valuation basis, and decision to be supported before modelling.
• Reconcile exposure, premium, claims, and financial data before analysis.
• Separate frequency, severity, timing, and dependency assumptions.
• Adjust historical experience for changes in mix, inflation, benefits, and operations.
• Use credibility rather than relying mechanically on either internal or external data.
• Select reserving methods according to portfolio maturity and structural stability.
• Present central estimates together with ranges, sensitivities, and key risks.
• Validate models independently and monitor actual outcomes against assumptions.
• Document judgement, limitations, overrides, and changes from prior analyses.
• Ensure that actuarial conclusions are communicated clearly to non-specialist decision-makers.

Chapter Summary

Actuarial foundations provide the quantitative basis for insurance pricing, reserving, capital management, and solvency assessment. Probability models distinguish claim frequency from severity and support analysis of aggregate loss. Mortality and morbidity tables quantify life-contingent risks, while discounting converts future benefits and expenses into present values. Premiums must allow for expected claims, expenses, risk, capital, profit, and applicable charges.

Credibility and experience rating combine portfolio-specific evidence with broader information. Reserving estimates obligations arising from past business through case reserves, IBNR, premium provisions, and technical provisions. Chain-ladder, Bornhuetter–Ferguson, expected-loss, frequency-severity, and stochastic methods each have appropriate uses and limitations. Sensitivity and stress testing reveal the effect of assumption uncertainty. Finally, actuarial reports must be read critically, with attention to data quality, model appropriateness, governance, ethics, regulation, and professional judgement.

Key Terms

Probability: A numerical representation of the likelihood of an event.

Random variable: A variable whose value depends on the outcome of an uncertain event.

Frequency: The number or rate of claims occurring in a specified period.

Severity: The monetary amount of an individual claim.

Aggregate loss: The total value of all claims in a portfolio or period.

Expected loss: The probability-weighted average cost of insured losses.

Mortality table: A table describing survival and death probabilities by age.

Morbidity table: A table describing illness, disability, recovery, or health-state transitions.

Discount factor: A factor used to convert a future cash flow into present value.

Pure risk premium: The expected claim cost before broader expense and profit loadings.

Credibility: The degree of reliance placed on an entity’s own experience.

Experience rating: Pricing that reflects the insured’s historical claims experience.

Case reserve: An estimate for a reported claim that remains unpaid or unsettled.

IBNR: Claims incurred but not yet reported, including related future development.

Technical provision: An insurer’s valuation of obligations arising from insurance contracts.

Chain-ladder method: A reserving method that projects claims using historical development patterns.

Bornhuetter–Ferguson method: A reserving method combining expected losses with observed emergence.

Sensitivity analysis: Analysis of the effect of changing one or more assumptions on results.

Stress testing: Evaluation of outcomes under adverse but plausible conditions.

Model risk: The risk that a model, its data, assumptions, or implementation produces misleading results.

Review Questions

Short Answer Questions

1. Distinguish claim frequency from claim severity.

2. Why are mortality and morbidity tables important in insurance valuation?

3. What is the purpose of discounting future insurance cash flows?

4. Define credibility in the context of insurance pricing.

5. What is an IBNR reserve?

Long Answer Questions

1. Explain the principal components of a gross insurance premium and show how variable expenses affect premium construction.

2. Discuss the use of mortality, morbidity, lapse, and interest assumptions in valuing a life insurance product.

3. Compare the chain-ladder, expected-loss, and Bornhuetter–Ferguson reserving methods.

4. Explain how credibility theory supports experience rating in a heterogeneous insurance portfolio.

5. Discuss the governance, ethical, and professional requirements involved in preparing an actuarial report.

Analytical / Application Questions

1. An insurer expects annual claims of 75 monetary units, fixed expenses of 12, and variable expenses equal to 15% of premium. Calculate the premium before profit and taxes, and explain the calculation.

2. A health portfolio’s claims ratio has increased, but its provider network, benefit design, and settlement speed have also changed. Identify the adjustments required before using the experience for pricing and reserving.

3. A life insurer reports that a lower discount rate, improved longevity, and higher lapse rates all affect its annuity reserves. Explain the likely direction of each effect and recommend appropriate sensitivity tests.

22. Multiple Choice Questions with Answers

Question 1

In a frequency-severity model, what does the frequency variable represent?

A. The monetary amount of each claim

B. The number of claims in a specified period

C. The investment return on premiums

D. The duration of a policy

Correct Answer: B. The number of claims in a specified period

Question 2

If claim frequency and severity are independent, aggregate expected loss is generally calculated as:

A. E(N) + E(X)

B. E(N) − E(X)

C. E(N) × E(X)

D. E(N) ÷ E(X)

Correct Answer: C. E(N) × E(X)

Question 3

A mortality table primarily provides information about:

A. Claim settlement expenses

B. Survival and death probabilities by age

C. Hospital provider charges

D. Reinsurance commissions

Correct Answer: B. Survival and death probabilities by age

Question 4

What is the main effect of reducing the discount rate on a long-term insurance liability, other assumptions remaining constant?

A. It generally increases the present value of the liability

B. It eliminates mortality risk

C. It reduces the number of claims

D. It converts a reserve into a premium

Correct Answer: A. It generally increases the present value of the liability

Question 5

A credibility-weighted premium combines:

A. Only the largest claims with the smallest claims

B. Experience-based information with collective information

C. Premium income with investment income only

D. Reported claims with taxes only

Correct Answer: B. Experience-based information with collective information

Question 6

Which reserve is intended to reflect claims that have occurred but have not yet been reported?

A. Acquisition expense reserve

B. IBNR reserve

C. Surrender value reserve

D. Contingency premium

Correct Answer: B. IBNR reserve

Question 7

Which reserving method combines an a priori expected loss estimate with observed claims emergence?

A. Chain-ladder method

B. Bornhuetter–Ferguson method

C. Simple average method

D. Payback method

Correct Answer: B. Bornhuetter–Ferguson method

Question 8

A premium includes expected claims, expenses, risk margin, and profit. Why cannot a percentage expense loading always be added directly to expected claims?

A. Because percentage expenses may be calculated as a proportion of the final premium

B. Because claims never vary from year to year

C. Because expenses are unrelated to insurance operations

D. Because discounting is prohibited in pricing

Correct Answer: A. Because percentage expenses may be calculated as a proportion of the final premium

Question 9

Which statement best describes sensitivity analysis?

A. It proves that an actuarial estimate is certain

B. It changes assumptions to examine the effect on results

C. It replaces all professional judgement with automation

D. It measures only past claim frequency

Correct Answer: B. It changes assumptions to examine the effect on results

Question 10

Why may a machine-learning pricing model require additional governance?

A. It cannot process any insurance data

B. It may be difficult to interpret, biased, unstable, or affected by data leakage

C. It always produces the same result as a manual model

D. It removes the need for regulatory compliance

Correct Answer: B. It may be difficult to interpret, biased, unstable, or affected by data leakage