Actuarial Pricing and Reserving

Introduction and Context

Health insurance pricing and reserving translate uncertain future healthcare costs into financially adequate, fair and governable obligations. Pricing determines the premium required for a future period; reserving determines the liability for insured events that have already occurred or for coverage that has already been issued. In managed care, these tasks are complicated by provider contracts, utilization management, benefit design, medical inflation, coding changes, delayed claims reporting, selection, changes in treatment patterns and policyholder behaviour.

The actuarial objective is not simply to predict average claims. It is to estimate the distribution of possible outcomes and ensure that premiums and technical provisions remain adequate under plausible adverse experience. A sound process therefore combines morbidity analysis, claims development, credibility, trend analysis, risk adjustment, experience rating, scenario testing and governance.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Distinguish incidence, prevalence, frequency, severity and morbidity measures.
2. Construct a basic prospective health insurance premium.
3. Apply credibility, trend, risk adjustment and experience-rating concepts.
4. Explain incurred, reported and paid claim reserves and technical provisions.
5. Interpret sensitivity and scenario analysis for pricing adequacy.
6. Identify professional, ethical, regulatory and data-governance risks.

1. The Actuarial Cost of Health Risk

A health insurance portfolio consists of people exposed to uncertain events such as hospitalization, outpatient treatment, chronic disease management, disability or long-term care. Actuarial analysis separates the risk into several related dimensions.

1.1 Morbidity

Morbidity describes the frequency and nature of illness or healthcare need in a population. It may be measured through diagnoses, episodes, admissions, days of care, procedures, prescriptions or expenditures. Morbidity is broader than mortality because the same disease may produce very different levels of resource use.

1.2 Incidence and prevalence

Incidence measures new cases during a specified period among people at risk. A simple incidence rate is:

Prevalence measures existing cases at a point in time or over an interval:

Incidence is especially useful for estimating new episodes, while prevalence is important for chronic disease, capitation and long-term care. Incidence and prevalence should not be treated as interchangeable. For a stable population with long average disease duration, prevalence may substantially exceed annual incidence.

1.3 Frequency and severity

Frequency is the number of claims or service events per exposure unit. Severity is the cost per claim or event. If aggregate claims are represented by S = ∑_i=1^N X_i, where N is claim count and X_i is the cost of claim i, then the expected aggregate cost is commonly decomposed as:

This identity assumes that claim count and individual claim cost are independent, or that the expression is being used as an approximation. More generally:

where X̄ is average cost per claim conditional on the observed claim count. Frequency may be measured per member, member-month, admission or episode. The denominator must be stated because changing exposure units changes the numerical result.

1.4 Utilization and unit cost

Health expenditure can also be decomposed into exposure, utilization and price:

For a hospital benefit, utilization may be admissions per 1,000 member-years and unit cost may be cost per admission. This decomposition helps management identify whether deterioration arises from more services, more expensive services, a change in case mix or a combination.

2. Premium Construction

A premium has several layers. The claims cost is the expected value of covered benefits. The risk margin compensates for uncertainty. Expenses, commissions, taxes, cost-of-capital requirements and profit or contribution margins are then incorporated according to the product and regulatory framework.

A simplified prospective premium per exposure unit is:

The formula assumes that variable expenses, taxes and target margin are expressed as proportions of gross premium. Fixed expenses are expressed per exposure unit. If no such deductions apply, the denominator is one. All components must use the same period, population, benefit definition and monetary unit.

The expected claims estimate may be written as:

For multiple benefits:

The benefits may include inpatient care, outpatient care, pharmacy, maternity, mental health and emergency transport. Correlation between benefits matters when estimating the risk margin even if expected claims are calculated benefit by benefit.

2.1 Worked pricing example

An insurer expects 10,000 member-years of exposure. Annual expected utilization is 0.18 admissions per member-year and average admissible cost is CU 4,000 per admission. Expected claims are therefore:

Suppose the insurer adds a risk margin of 4% of expected claims and fixed expenses of CU 600,000. Variable expenses are 10% of gross premium, taxes are 2%, and the target margin is 5%.

Risk margin equals CU 288,000. The numerator is therefore CU 8,088,000. The denominator is:

The required gross premium is:

The premium per member-year is approximately CU 974.46. The result is valid only if the expense and margin ratios are genuinely percentages of gross premium and if the exposure and claims assumptions refer to the same coverage period.

3. Trend, Seasonality and Benefit Changes

Historical claims cannot be transferred mechanically into future prices. A trend assumption typically reflects several components:

• medical price inflation;
• utilization change;
• technology and treatment intensity;
• provider reimbursement changes;
• demographic and morbidity change;
• benefit, network or cost-sharing changes; and
• coding, data and claims-administration changes.

If annual trend is t and the projection period is k years, the compounded projection factor is:

For monthly projections, a consistent monthly rate should be used, or the annual rate should be converted according to the chosen compounding convention. A claim cost of CU 1,000 projected for two years at 6% annual trend becomes CU 1,123.60, not CU 1,120, because compounding is applied.

Trend must be separated from changes in exposure and benefit design. If a deductible increases, observed claims may fall even though underlying medical need has not changed. If a network expands, utilization may rise while average negotiated price falls. Seasonality is also important: respiratory illness, births and elective procedures may cluster in particular months.

4. Credibility and Experience Rating

A small portfolio may have highly volatile observed claims. Credibility theory combines its own experience with a broader manual or benchmark rate. A basic limited-fluctuation approach is:

Here Z lies between zero and one. A higher Z is appropriate when the portfolio has greater volume, stability and relevance. The formula assumes that the manual estimate is a reasonable estimate for the portion of risk not supported by credible experience.

A common conceptual credibility criterion is that exposure must be sufficient for the observed result to fall within a selected tolerance of the true expected result with a selected probability. Exact methods depend on the frequency distribution, severity distribution and regulatory practice. Credibility is not simply a reward for large claims; it is a measure of statistical reliability.

Experience rating

Experience rating adjusts the premium using the group’s own historical claims, subject to adjustments for trend, benefit changes, exposure changes, large claims, credibility and risk pooling. A simplified approach is:

Adjustments may include:

• completion of claims to ultimate cost;
• removal or capping of unusually large claims;
• trend to the rating period;
• changes in membership demographics;
• changes in benefits and provider networks; and
• risk-adjustment transfers.

A purely retrospective approach can be unfair where a group has experienced random catastrophic claims. Conversely, ignoring credible adverse experience can produce inadequate premiums. Sound practice uses explicit credibility and pooling rules rather than informal negotiation.

5. Risk Adjustment and Selection

Risk adjustment transfers or redistributes funds to account for differences in expected cost that are not fully reflected in premiums. Its purposes include reducing incentives to select healthier members, supporting competition on efficiency rather than risk avoidance, and improving affordability in regulated markets.

Risk-adjustment models may use age, sex where legally permitted, diagnoses, pharmacy data, prior utilization, disability status or socioeconomic variables. A risk score is often interpreted relative to a population mean of one. If a member group has a risk score of 1.20 and the standard expected cost is CU 2,500 per member-year, a mechanical expected-cost estimate is:

This is not automatically the final premium. It may require geographic, network, benefit, credibility and regulatory adjustments.

Risk adjustment has limitations. Diagnosis-based models may reward coding intensity, fail to capture unmet need, or reproduce inequities in the data. Variables such as income, ethnicity or disability require careful legal and ethical review. Models should be monitored for calibration, discrimination, gaming and changes in clinical coding.

6. Reserving and Technical Provisions

Pricing concerns future coverage; reserving concerns obligations arising from past coverage or incurred events. Major health insurance reserve components include:

1. Case reserves: estimates for reported claims not yet settled.
2. IBNR reserves: incurred but not reported claims.
3. IBNER reserves: incurred but not enough reported, where reported claims require further development.
4. Unexpired risk or premium deficiency provisions: additional amounts required when future costs under existing coverage exceed the related unearned premium and other expected inflows.
5. Unearned premium reserves: premium relating to coverage not yet provided, where applicable.
6. Claims handling expense provisions: direct and indirect costs of settling claims, depending on the applicable accounting framework.

The exact classification and valuation rules depend on jurisdiction and the applicable insurance accounting standard. Actuaries must distinguish actuarial best estimates from prudential margins or risk adjustments required by regulation.

6.1 Paid and incurred development

Claims are often organized into development triangles by accident month or quarter and development age. A paid triangle records cumulative payments; an incurred triangle records paid amounts plus case reserves. Development factors estimate movement from an observed development age to ultimate cost.

If cumulative incurred claims at 12 months are CU 5 million and the selected 12-to-ultimate factor is 1.25, estimated ultimate claims are:

The implied reserve at that valuation date is CU 1,250,000, before considering separate expenses or other adjustments. The method assumes that the selected development factor is appropriate for the cohort and that the data are comparable over time.

6.2 Bornhuetter–Ferguson method

When recent experience is immature, a Bornhuetter–Ferguson approach combines an a priori expected loss with observed emergence. If expected ultimate claims are CU 10 million and the expected proportion reported by the valuation date is 40%, the expected reported amount is CU 4 million. If actual reported claims are CU 3 million, the estimated ultimate is:

The method is less sensitive than a pure chain-ladder method to random early experience, but it depends on a credible a priori expected loss and an appropriate reporting pattern.

6.3 Discounting and risk adjustment

Some frameworks require or permit discounting future cash flows. If a payment of C_t is expected at time t and the effective annual discount rate is r, present value is:

The rate, payment timing, inflation assumptions and currency must be consistent. A risk adjustment for non-financial risk is conceptually distinct from discounting: discounting reflects timing of money, while the risk adjustment reflects uncertainty about amount and timing.

7. Pricing Adequacy and Monitoring

Pricing adequacy is assessed by comparing actual and expected experience, not merely by observing whether claims exceed premiums in one month. Useful indicators include:

• loss ratio = incurred claims ÷ earned premium;
• claims per member-month;
• admission frequency per 1,000 member-years;
• average cost per admission;
• completion factor and reserve development;
• persistency and lapses;
• risk-adjusted morbidity ratio; and
• actual-to-expected ratios by segment.

For a loss ratio, incurred claims should be defined consistently with earned premium and should include an appropriate reserve estimate. A high loss ratio may result from underpricing, adverse selection, random fluctuation, reserve strengthening or a temporary trend shock.

Sensitivity and scenario analysis

Sensitivity analysis changes one assumption at a time. Scenario analysis changes several related assumptions to represent a coherent future state. For example, a respiratory epidemic scenario might combine higher admission frequency, longer hospital stays, delayed elective care and increased provider prices.

Suppose baseline claims are CU 100 million. A frequency increase of 8% and severity increase of 5% produce:

Claims increase by 13.4%, not 13%, because the effects compound. Management should assess the impact on loss ratio, solvency capital, liquidity, reinsurance recoveries, provider capacity and customer affordability.

A useful sensitivity table is:

The table assumes other variables remain fixed. In reality, frequency and severity may be correlated, and policyholder or provider responses may alter the result.

8. Illustrative Case: Meridian Managed Health Plan

Background

Meridian is a fictional regional managed care plan with 80,000 member-years. It offers inpatient, outpatient and pharmacy benefits to employer groups. Its previous pricing assumed stable utilization, but recent data show increased emergency admissions and higher specialty-drug expenditure.

Problem

The latest reported loss ratio is 86%, compared with a target of 78%. However, the year is only 75% developed. Management must decide whether to increase premiums, strengthen reserves, revise provider contracts or redesign benefits.

Analysis

The actuarial team first reconciles enrollment and claims data, separates medical claims from administration costs, and constructs paid and incurred triangles. It identifies:

• emergency admission frequency 9% above expectation;
• average inpatient severity 4% above expectation;
• specialty-drug cost 15% above expectation;
• membership risk score 1.06 rather than the assumed 1.00;
• a one-time data-coding change affecting diagnosis prevalence; and
• significant credibility differences between large employer groups and small groups.

The team trends credible claims, adjusts for the coding change, applies risk adjustment and completes immature claims. It then develops three scenarios:

Approach and solution

Meridian adopts a segmented response. Large groups receive experience-rated indications with explicit credibility and pooling limits. Small groups receive a community or manual rate adjusted for risk. The plan negotiates specialty-drug contracts, expands prior authorization with clinical safeguards, and introduces case management for high-cost admissions. It does not use diagnosis data without reviewing coding quality and fairness.

The reserve committee selects a blended development approach: chain-ladder for mature cohorts and Bornhuetter–Ferguson for the latest quarters. It documents assumptions, performs a reserve sensitivity range and establishes quarterly monitoring of emergency admissions, specialty-drug costs and provider payment changes.

Outcome

The plan implements a measured premium adjustment rather than reacting to the preliminary 86% loss ratio. The final indication reflects completed claims, risk mix, benefit changes and credible experience. Governance reports show the base and adverse scenarios separately, enabling management to assess affordability and capital resilience.

Learning points

The case demonstrates that pricing adequacy is a process rather than a single ratio. Data validation, development of immature claims, credibility, risk adjustment, clinical interpretation and governance must operate together. Utilization management can reduce cost, but it must not create unsafe barriers to medically necessary care.

9. Governance, Ethics and Regulation

Actuarial work in health insurance affects access to treatment and financial protection. Key governance principles include:

• document data sources, exclusions and transformations;
• maintain version control for assumptions and models;
• independently review material pricing and reserving judgments;
• test models for calibration, stability and disparate impact;
• disclose material uncertainty to boards and regulators;
• protect personal health information through access controls and minimization;
• separate clinical necessity decisions from financial incentives where appropriate; and
• provide appeal and correction mechanisms for inaccurate risk data.

Regulators may prescribe minimum capital, premium review, loss-ratio standards, community-rating rules, reserve methods, actuarial opinions or technical-provision requirements. Compliance does not eliminate professional judgment. An actuarial estimate should be transparent about uncertainty rather than presenting a false level of precision.

10. Technology, AI and Contemporary Developments

Modern insurers use claims warehouses, electronic health records, pharmacy data, predictive models and automated anomaly detection. Machine learning may improve prediction of admissions or high-cost claims, but predictive accuracy is not the same as causal validity or fairness. Models trained on historical utilization may learn barriers to care rather than underlying health need.

Good practice includes out-of-time validation, subgroup performance testing, drift monitoring, explainability appropriate to the decision, human review and strict governance of protected health information. AI should support—not silently replace—actuarial judgment, clinical review and accountability.

Emerging pressures include gene and specialty therapies, virtual care, value-based provider contracts, climate-related health events, ageing populations, antimicrobial resistance and changing post-pandemic utilization. These developments increase the need for dynamic scenario analysis and close collaboration among actuaries, clinicians, finance professionals, data scientists and regulators.

Chapter Summary

Actuarial pricing estimates the prospective cost of health benefits and adds risk, expense, tax and margin components under a consistent premium formula. Morbidity may be described through incidence and prevalence, while claims cost is commonly decomposed into frequency and severity. Trend must reflect medical prices, utilization, technology, demographics, provider payment and benefit changes. Credibility and experience rating balance portfolio data with broader information, and risk adjustment addresses differences in expected morbidity and selection incentives.

Reserving estimates obligations arising from past coverage. Case reserves, IBNR, IBNER, unearned premium and unexpired-risk provisions require careful definitions and appropriate development methods. Chain-ladder methods rely on development patterns; Bornhuetter–Ferguson methods combine observed emergence with an a priori expected loss. Pricing adequacy requires actual-to-expected monitoring, scenario analysis and sensitivity testing. Ethical and regulatory governance is essential because actuarial decisions influence affordability, access, solvency and the fair treatment of members.

Key Terms

Actuarial pricing: Estimation of premiums needed to fund expected claims, expenses, risk and required returns.

Morbidity: Frequency and nature of illness or healthcare need in a population.

Incidence: Occurrence of new cases during a specified period.

Prevalence: Proportion of a population with an existing condition.

Frequency: Number of claims or service events per exposure unit.

Severity: Cost per claim or service event.

Exposure: Unit measuring the population at risk, such as member-months.

Medical trend: Change in healthcare cost caused by price, utilization, technology and related factors.

Credibility: Degree to which observed experience is considered statistically reliable.

Experience rating: Premium determination using a group’s adjusted historical experience.

Manual rate: Benchmark rate derived from broader portfolio or population data.

Risk adjustment: Modification or transfer of funds for differences in expected risk.

Case reserve: Estimate for a reported claim not yet fully settled.

IBNR: Incurred but not reported claims reserve.

IBNER: Reserve for additional development on claims already reported.

Technical provision: Recognized liability for expected insurance obligations under the applicable framework.

Chain-ladder method: Reserving method based on historical claims-development factors.

Bornhuetter–Ferguson method: Reserving method combining observed claims with an a priori expected loss.

Loss ratio: Incurred claims divided by earned premium.

Scenario analysis: Evaluation of a coherent set of changed assumptions representing a possible future state.

Review Questions

5 Short Answer Questions

1. Distinguish incidence from prevalence.
2. Why must frequency and severity use clearly defined exposure units?
3. What is the purpose of credibility weighting?
4. Distinguish IBNR from a case reserve.
5. Why are scenario analysis and sensitivity analysis both useful?

5 Long Answer Questions

1. Explain the components of a prospective health insurance premium and the assumptions required for a gross-premium formula.
2. Discuss how medical trend should be developed and monitored in a managed care plan.
3. Compare chain-ladder and Bornhuetter–Ferguson reserving methods, including their principal assumptions.
4. Explain how risk adjustment can reduce selection incentives while creating ethical and data-governance risks.
5. Describe a governance framework for actuarial pricing and reserving in health insurance.

3 Analytical / Application Questions

1. A portfolio has 20,000 member-years, expected frequency of 0.25 admissions per member-year and severity of CU 3,200. Calculate expected claims and explain two limitations of the estimate.
2. A plan’s baseline claims are CU 75 million. Frequency is expected to rise by 7% and severity by 6%. Calculate projected claims and interpret the result.
3. Reported claims are CU 4 million, expected ultimate claims are CU 12 million and the expected reported proportion is 50%. Apply the Bornhuetter–Ferguson logic and discuss why the estimate differs from simply doubling reported claims.

Multiple Choice Questions with Answers

Question 1

Which measure describes new cases occurring during a specified period?

A. Prevalence

B. Incidence

C. Severity

D. Persistency

Correct Answer: B. Incidence

Question 2

If expected frequency is 0.20 claims per member-year and expected severity is CU 5,000, expected cost per member-year is:

A. CU 1,000

B. CU 5,200

C. CU 25,000

D. CU 200

Correct Answer: A. CU 1,000

Question 3

In the formula Z × Observed + (1 − Z) × Manual, Z represents:

A. Medical inflation

B. A tax ratio

C. Credibility weight

D. Claim severity

Correct Answer: C. Credibility weight

Question 4

Which reserve concerns claims that have occurred but have not yet been reported?

A. Case reserve

B. IBNR reserve

C. Unearned premium reserve

D. Capital reserve

Correct Answer: B. IBNR reserve

Question 5

Claims of CU 100 million increase by 8% in frequency and 5% in severity. The projected claims are:

A. CU 108.0 million

B. CU 105.0 million

C. CU 113.4 million

D. CU 113.0 million

Correct Answer: C. CU 113.4 million

Question 6

A risk score of 1.20 applied to a standard expected cost of CU 2,500 produces an unadjusted expected cost of:

A. CU 2,080

B. CU 2,500

C. CU 2,700

D. CU 3,000

Correct Answer: D. CU 3,000

Question 7

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

A. Bornhuetter–Ferguson

B. Simple average

C. Expense loading

D. Community rating

Correct Answer: A. Bornhuetter–Ferguson

Question 8

A loss ratio is generally calculated as:

A. Earned premium divided by incurred claims

B. Incurred claims divided by earned premium

C. Paid claims divided by exposure

D. Premium divided by membership

Correct Answer: B. Incurred claims divided by earned premium

Question 9

Which statement best distinguishes sensitivity analysis from scenario analysis?

A. Sensitivity analysis changes one assumption, while scenario analysis changes a coherent set of assumptions

B. Sensitivity analysis is used only for reserves

C. Scenario analysis never includes adverse assumptions

D. They are mathematically identical in all applications

Correct Answer: A. Sensitivity analysis changes one assumption, while scenario analysis changes a coherent set of assumptions

Question 10

Why is diagnosis-based risk adjustment subject to governance concerns?

A. It cannot use any healthcare data

B. It may reflect coding intensity or reproduce inequities in historical data

C. It always produces lower premiums

D. It eliminates the need for actuarial judgment

Correct Answer: B. It may reflect coding intensity or reproduce inequities in historical data