Pricing, Rating and Underwriting Profitability

Introduction

Insurance pricing is the process of converting uncertainty about future losses into a premium that is adequate, competitive, equitable and commercially sustainable. Underwriting management extends beyond deciding whether to accept a risk. It requires the insurer to determine the amount to charge, the terms and conditions to impose, the capital required to support the business, and the degree to which the portfolio contributes to long-term profitability.

A premium must perform several functions simultaneously. It must fund expected claims, meet acquisition and administration expenses, compensate for uncertainty and catastrophe potential, support regulatory and capital requirements, and provide an acceptable return to owners. Pricing therefore combines actuarial analysis, underwriting judgment, portfolio management, finance, regulation and market strategy.

The central distinction in this chapter is between expected cost and risk-adjusted price. Expected cost is the statistically anticipated claim cost. The final premium also reflects expenses, uncertainty, capital, reinsurance, taxes, market conditions and the insurer’s strategic objectives.

Learning Objectives

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

1. Explain exposure measures, loss frequency, severity and loss costs.
2. Construct a basic insurance rate and distinguish premium, rate and exposure base.
3. Apply expense provisions, credibility, experience rating and schedule rating.
4. Explain catastrophe loads, reinsurance effects and risk margins.
5. Calculate and interpret underwriting profitability measures.
6. Evaluate pricing decisions using managerial, regulatory and ethical perspectives.
7. Identify limitations of models and the governance requirements for responsible pricing.

1. The Architecture of an Insurance Price

An insurance price is usually expressed as a premium for a specified period and exposure. A rate is the price per unit of exposure, while the premium is the rate multiplied by the insured exposure, subject to minimum premiums, deductibles, limits and other adjustments.

Examples of exposure bases include:

The exposure base should be related to the mechanism generating losses. Payroll may be appropriate for workers’ compensation because workplace activity and wages are associated with claims, whereas building value may be more relevant for property damage.

A simplified pricing identity is:

This identity is conceptual. In practice, taxes, brokerage, reinsurance cost, policy fees, investment income, premium discounts and capital costs may be modelled separately. The components must be defined consistently: for example, whether claims include allocated loss-adjustment expenses and whether expenses are measured as a percentage of premium or as a fixed amount per policy.

Pure premium and loss cost

The pure premium, often called the loss cost, is expected claim cost per exposure unit. If the expected aggregate claims for a portfolio are 2,400,000 monetary units and the portfolio contains 60,000 exposure units, the pure premium is 40 monetary units per exposure unit.

The formula assumes homogeneous exposure units or an adjustment that makes them sufficiently comparable. It does not include insurer expenses, profit, taxes or contingencies unless explicitly defined to do so.

Frequency and severity

Loss frequency measures how often claims occur:

Loss severity measures the average cost per claim:

When the same period, claim definition and exposure base are used, the loss cost is the product of frequency and severity:

This relationship follows because claim count divided by exposure multiplied by claim cost divided by claim count equals claim cost divided by exposure. It is valid when frequency and severity are measured consistently and the number of claims is non-zero. If no claims occur in the observation period, severity cannot be calculated directly, although the loss cost may be zero in the observed data.

Suppose 1,000 vehicle-years produce 50 claims with total incurred losses of 150,000 monetary units. Frequency is 0.05 claims per vehicle-year, severity is 3,000 monetary units per claim, and loss cost is 150 monetary units per vehicle-year.

2. Data Used in Pricing

Pricing data may be derived from internal experience, industry loss costs, market databases, catastrophe models, engineering studies or expert judgment. The main data adjustments are as follows:

Development

Claims reported or paid at the valuation date may not represent their ultimate cost. Incurred but not reported claims, reopened claims, litigation and late development must be considered. Actuaries use loss development factors to estimate ultimate claims.

If reported claims at an early development age are 800,000 and the selected cumulative development factor is 1.25, estimated ultimate claims are:

Thus, ultimate claims are 1,000,000 monetary units. The factor is appropriate only for the relevant line, accident period, development age and claims basis.

Trending

Historical losses must often be adjusted to the conditions expected during the future rating period. Trend may reflect inflation, medical-cost escalation, wage growth, legal change, vehicle repair costs, claims frequency changes and social inflation.

A simple annual trend adjustment is:

where t is the number of years between the historical midpoint and the prospective period. This assumes a constant compound trend. Real applications may use separate frequency and severity trends:

The use of trends must avoid double counting. For example, a severity trend should not include an inflation adjustment already embedded in a development factor.

Exposure and mix changes

A portfolio may change because it contains larger buildings, younger drivers, different industries or new geographical concentrations. Historical results must be adjusted for changes in exposure mix. Aggregate improvement in loss ratio can be misleading if it is caused only by a change in business composition.

On-leveling and rate changes

Historical premiums may be adjusted to the current rate level before a loss ratio is calculated. Otherwise, old premiums may make current performance appear artificially strong or weak. The adjustment must reflect the effective dates and earning patterns of rate changes.

3. Basic Rating Methods

Manual or class rating

Under manual rating, risks are grouped into classes with common characteristics. A class rate is multiplied by the exposure and modified by approved factors. The method is efficient and transparent, but broad classes can create cross-subsidies and may not reflect individual risk accurately.

Judgment rating

For unusual risks, credible historical data may not exist. An underwriter may use engineering information, contractual conditions, occupancy, construction, management quality and comparable risks to determine a premium. Judgment rating is necessary for large or novel risks, but it should be documented and subject to governance controls.

Experience rating

Experience rating adjusts a manual premium using the insured’s own historical claims. It is most useful where the insured has sufficient exposure and stable operations. A simple credibility-weighted approach is:

where Z is the credibility assigned to the insured experience, between 0 and 1. The formula assumes the two loss-cost estimates are expressed on the same basis and that credibility is selected using an appropriate actuarial method.

If an insured’s loss cost is 120, the manual loss cost is 100 and credibility is 0.60, the indicated loss cost is 112 monetary units. The remaining 40% is supported by broader manual data because the insured’s own experience is not fully reliable.

Credibility

Credibility is not a subjective statement that data are “good.” It is a quantitative or structured assessment of how much reliance should be placed on observed experience. Greater credibility normally results from larger exposure, more claims, longer observation periods, consistent coverage and stable risk characteristics.

The classical limited-fluctuation approach may define full credibility as the exposure needed to achieve a selected probability that observed results lie within a specified tolerance of the expected value. A common approximation for claim counts is:

where N is the required expected claim count, z is the standard-normal value associated with the selected confidence level, and k is the permitted relative error. This approximation assumes independent claims and a Poisson-like count process. It is not automatically suitable for heavy-tailed losses, correlated catastrophes or changing portfolios.

Schedule rating

Schedule rating starts with a base rate and applies debits or credits for individually assessed characteristics. In commercial property, factors may include construction, occupancy, protection, exposure to hazards, management controls and physical condition.

Schedule rating is valuable when qualitative risk information has predictive value but is not fully represented in the statistical class plan. Its weakness is inconsistency: unsupported judgment can create arbitrary or discriminatory outcomes. Rating plans should therefore specify factors, permitted ranges, evidence requirements and review procedures.

4. Expense Provisions and Rate Indication

Expenses may include commissions, brokerage, underwriting, administration, claims handling, premium taxes, assessment charges, technology costs and general overhead. Some expenses vary with premium, while others vary with policy count or exposure.

Let:

• L = expected loss and loss-adjustment expense ratio;
• V = variable expense ratio, expressed as a proportion of premium;
• F = fixed expense per exposure unit;
• Q = exposure units;
• P = premium per exposure unit;
• R = target underwriting profit and contingency provision as a proportion of premium.

A rate indication that includes fixed expenses is:

Solving for P gives:

The second expression applies when L is a ratio of premium rather than a monetary amount per exposure unit. Care is essential because a loss cost per exposure unit cannot be inserted directly into a formula requiring a loss ratio.

A more operational form begins with a loss cost C per exposure unit:

This assumes that C and F are both monetary amounts per exposure unit, while V and R are percentages of premium. The denominator must be positive. If expected loss, variable expense and target profit provisions consume all or more than the premium, the indicated rate is undefined or economically unsustainable.

Worked example: indicated premium

An insurer estimates expected claims of 420 monetary units per exposure unit. Fixed expenses are 60 per unit, variable expenses are 22% of premium, and the target profit and contingency provision is 5% of premium.

The indicated premium is approximately 657.53 monetary units per exposure unit. At that premium, claims are 63.9% of premium, fixed expenses are 9.1%, variable expenses are 22%, and the profit provision is 5%, subject to rounding.

5. Catastrophe Loads and Reinsurance

Ordinary claims may be estimated from historical frequency and severity, but catastrophe losses are infrequent, severe and often correlated across policyholders. Historical data alone may therefore be inadequate. Catastrophe pricing can use event catalogues, hazard maps, exposure data, vulnerability functions, financial terms and stochastic simulation.

A catastrophe load may be expressed as an expected annual catastrophe loss per exposure unit. A risk margin may then be added for uncertainty, capital usage and model error. These are conceptually different:

• Expected catastrophe loss compensates for the modeled mean loss.
• Volatility or risk margin compensates for uncertainty around that mean.
• Reinsurance cost pays for risk transfer and may include reinsurer expenses and capital costs.

A catastrophe model does not predict the exact event that will occur. It estimates a distribution of possible losses. A key measure is the annual exceedance probability, which represents the chance that loss exceeds a specified amount in a year. Model outputs must be tested against exposure data, engineering knowledge and historical experience.

Reinsurance affects pricing in at least three ways. It transfers part of the loss distribution, creates a direct cost, and may reduce the capital required to support the portfolio. A policy premium that ignores reinsurance cost may overstate underwriting profitability. Conversely, a reinsurance cost should not be added without considering the part of the loss it replaces.

6. Profitability Metrics

Loss ratio

The ratio is meaningful only when numerator and denominator use compatible periods, valuation dates and definitions. A low loss ratio is not necessarily evidence of good performance if expenses are excessive or reserves are understated.

Expense ratio

The premium basis must be stated. Acquisition expenses may be associated with written premium, while an analysis of earned underwriting results may use earned premium.

Combined ratio

A combined ratio below 100% indicates underwriting gain before items excluded from the ratio, such as investment income, taxes or certain corporate costs. A ratio above 100% indicates underwriting loss on that basis.

If the loss ratio is 64% and the expense ratio is 31%, the combined ratio is 95%, representing a 5% underwriting margin on the relevant premium base.

Underwriting margin and return on equity

Return on equity is broader:

ROE includes investment results, capital structure, taxes and non-underwriting items. It should not be inferred from the combined ratio alone.

Economic value and risk-adjusted performance

Managers may evaluate economic profit as profit after charging for the capital committed to the risk:

The cost of capital is not a universal percentage. It depends on the insurer’s capital framework, required solvency margin, diversification, reinsurance and investor expectations. Risk-adjusted metrics are especially important when two portfolios have similar expected margins but very different volatility.

7. A Practical Pricing Workflow

A disciplined workflow normally includes:

1. Define coverage, limits, deductibles, exclusions and policy period.
2. Select the exposure base and verify its measurement.
3. Collect claims, exposure, premium and underwriting data.
4. Adjust claims for development, trend, mix and rate level.
5. Estimate frequency, severity and expected loss cost.
6. Add catastrophe loss, reinsurance and uncertainty provisions where relevant.
7. Allow for variable and fixed expenses.
8. Apply credibility, experience or schedule adjustments.
9. Test affordability, competitiveness and sensitivity.
10. Review legal, regulatory, fairness and governance requirements.
11. Monitor actual results and update assumptions.

Sensitivity analysis should examine variables such as claim inflation, frequency deterioration, catastrophe loss, policy persistency, expense overruns and reinsurance renewal cost. Scenario analysis is preferable to reliance on a single point estimate.

8. Applied Case: Commercial Property Portfolio

The following is an illustrative case, not a report of an actual insurer.

Background

Harbor Shield Insurance is pricing a portfolio of 2,000 small commercial properties. The portfolio contains offices, shops and small warehouses. Historical claims are incomplete for the most recent year, and a new flood map indicates greater geographic concentration than previously assumed.

Problem

The existing premium is based on a broad class rate. Management observes that the combined ratio has deteriorated from 92% to 104%. It must determine whether the deterioration reflects inadequate pricing, adverse claims development, expense growth or a change in portfolio mix.

Analysis

The pricing team first develops the latest claims to ultimate cost and separates ordinary claims from catastrophe-related flood losses. It then adjusts historical losses for building-cost inflation and applies current exposure values. Frequency is stable, but severity has increased because construction and contents replacement costs have risen. The team also finds that higher-risk warehouses represent a larger share of the current portfolio.

The proposed prospective loss cost is 520 monetary units per property-year. Expected catastrophe loss is 45 per property-year, and fixed expenses are 70. Variable expenses are 20% of premium. A 4% risk and profit provision is required.

Approach and solution

Total expected cost per property-year before variable expenses is 635:

The indicated premium is:

The insurer does not apply the same final premium to every property. It develops territorial and construction factors, uses schedule credits for verified flood protection, and applies a deductible structure that reflects the insured’s ability to retain small losses. Experience rating is used only for larger accounts with sufficiently credible data.

Before implementation, management tests a severe flood scenario, reviews whether the flood factor is legally permissible and checks that data do not act as an unjustified proxy for protected characteristics. It also communicates the basis of material price changes to intermediaries and policyholders.

Outcome

The revised plan identifies that the prior combined ratio understated prospective risk because catastrophe exposure and property mix had changed. The new rate indication supports adequacy, while segmentation and risk-control credits reduce unnecessary cross-subsidy. Results will be monitored quarterly using earned premium, ultimate losses, catastrophe emergence and retention by segment.

Learning points

1. Historical loss ratios must be adjusted before they are used for prospective pricing.
2. Catastrophe exposure cannot be treated as ordinary independent claims.
3. A technically adequate portfolio rate may still require individual risk modification.
4. Pricing governance includes fairness, documentation and communication, not merely arithmetic.

9. Limitations, Governance and Ethics

Pricing models are subject to sampling error, parameter uncertainty, model misspecification, changing behaviour and data quality problems. A model may be statistically accurate for historical data but unsuitable after a legal, technological or environmental change. Correlation can invalidate assumptions of independent claims, particularly in cyber, pandemic, flood and catastrophe portfolios.

Governance should include data lineage, version control, peer review, validation, outcome monitoring, authority limits and escalation procedures. Material assumptions should be documented, including exposure definitions, claims valuation, trend selection, credibility, catastrophe treatment and expense allocation.

Regulatory requirements commonly address rate adequacy, fairness, transparency, solvency and prohibited discrimination, although specific rules vary by jurisdiction. An insurer should not use a technically predictive variable without examining whether it produces unjustified adverse effects or conflicts with law and public policy. Pricing segmentation must be supported by legitimate insurance rationale and appropriate controls.

Artificial intelligence can assist with data cleaning, feature discovery, claims triage and scenario analysis. It does not eliminate actuarial or underwriting responsibility. AI systems require explainability appropriate to the decision, bias testing, drift monitoring, cybersecurity controls and human oversight. A model that improves predictive accuracy but cannot be governed or fairly explained may be unsuitable for deployment.

10. Best Practices for Underwriting Managers

• Match the exposure measure to the loss mechanism.
• Separate frequency, severity, catastrophe and expense assumptions.
• Use ultimate rather than immature claims where appropriate.
• Avoid double counting trend, development or reinsurance effects.
• Distinguish monetary amounts per exposure from ratios of premium.
• Use credibility methods proportionate to data quality and risk complexity.
• Document judgmental schedule credits and debits.
• Test rates under adverse scenarios, not only expected conditions.
• Monitor both portfolio results and segment-level cross-subsidies.
• Link pricing decisions to capital, reinsurance and risk appetite.
• Review fairness, regulatory compliance and customer communication.
• Recalibrate models when claims, exposure or external conditions change.

Chapter Summary

Insurance pricing converts exposure and loss information into a premium that must fund expected claims, expenses, uncertainty and an acceptable return. Frequency and severity combine to produce loss cost, but historical observations require adjustment for development, trend, exposure mix and rate level. Manual, judgment, experience and schedule rating methods serve different data and risk conditions. Credibility determines how much reliance should be placed on insured-specific experience.

Expense provisions require careful distinction between fixed costs per exposure and variable costs as a percentage of premium. Catastrophe loads and reinsurance must reflect correlated, high-severity risks rather than ordinary claims assumptions. Loss ratio, expense ratio, combined ratio, underwriting margin and risk-adjusted profit provide complementary views of performance. Effective underwriting management combines quantitative discipline with governance, fairness, scenario testing, technology oversight and continuous monitoring.

Key Terms

Exposure measure: A unit used to quantify the amount of insurance risk subject to pricing.

Rate: Price charged per unit of exposure.

Premium: Amount paid for insurance coverage.

Loss cost: Expected claims cost per exposure unit before insurer expenses and profit provisions.

Frequency: Number of claims divided by exposure units.

Severity: Average claim cost.

Pure premium: Another term for loss cost.

Development: Change in estimated claim cost as claims mature.

Trend: Expected change in frequency or severity over time.

Credibility: Degree of reliance placed on observed experience.

Experience rating: Adjustment based on the insured’s own claims experience.

Schedule rating: Individual debits or credits based on assessed risk characteristics.

Catastrophe load: Provision for expected catastrophe loss and, where specified, related uncertainty.

Variable expense: Expense that changes in proportion to premium.

Fixed expense: Expense allocated per policy or exposure unit rather than as a premium percentage.

Loss ratio: Claims and loss-adjustment expenses divided by premium.

Expense ratio: Underwriting expenses divided by premium.

Combined ratio: Loss ratio plus expense ratio.

Underwriting margin: Underwriting profit expressed relative to premium.

Model risk: Risk that a model is inappropriate, incorrectly implemented or misused.

Review Questions

Short Answer Questions

1. Distinguish a rate from a premium and provide one example of an exposure base.
2. Define loss frequency and loss severity.
3. Why are historical claims developed to ultimate cost?
4. What is the purpose of credibility in experience rating?
5. What does a combined ratio below 100% generally indicate?

Long Answer Questions

1. Explain how an insurer converts historical claims data into a prospective loss cost.
2. Compare manual rating, judgment rating, experience rating and schedule rating.
3. Explain the treatment of fixed expenses, variable expenses and profit provisions in a rate indication.
4. Discuss why catastrophe risks require methods beyond ordinary frequency–severity analysis.
5. Evaluate the governance and ethical issues involved in data-driven insurance pricing.

Analytical / Application Questions

1. An insurer estimates a loss cost of 300 monetary units per exposure, fixed expenses of 50, variable expenses of 25% of premium and a profit provision of 5%. Calculate the indicated premium per exposure unit.
2. A portfolio has 80 claims, 4,000 exposure units and total incurred claims of 360,000 monetary units. Calculate frequency, severity and loss cost, and explain the assumptions behind the calculation.
3. An insured has an experience loss cost of 150, a manual loss cost of 120 and credibility of 0.40. Calculate the indicated loss cost and explain why the manual component remains important.

Multiple Choice Questions with Answers

Question 1

Which measure represents expected claims cost per exposure unit before insurer expenses and profit provisions?

A. Loss cost

B. Expense ratio

C. Combined ratio

D. Underwriting margin

Correct Answer: A. Loss cost

Question 2

A portfolio has 40 claims from 2,000 exposure units. What is its claim frequency?

A. 0.02

B. 0.20

C. 50

D. 2,000

Correct Answer: A. 0.02

Question 3

If total claims are 240,000 monetary units and there are 60 claims, what is average severity?

A. 4,000

B. 3,000

C. 240

D. 0.025

Correct Answer: A. 4,000

Question 4

Which method gives greater weight to an insured’s own claims when sufficient credible experience exists?

A. Experience rating

B. Pure manual rating

C. Flat taxation

D. Unrelated market indexing

Correct Answer: A. Experience rating

Question 5

An insured loss cost is 150, the manual loss cost is 100, and credibility is 0.40. What is the indicated loss cost?

A. 120

B. 130

C. 140

D. 150

Correct Answer: B. 130

Question 6

If the loss ratio is 62% and the expense ratio is 33%, what is the combined ratio?

A. 29%

B. 62%

C. 95%

D. 105%

Correct Answer: C. 95%

Question 7

Which item is normally a variable expense in a rate indication?

A. A commission calculated as a percentage of premium

B. An annual office lease allocated per branch

C. A fixed policy-record storage charge

D. A salaried executive cost that does not vary with premium

Correct Answer: A. A commission calculated as a percentage of premium

Question 8

Why are catastrophe loads treated separately in many pricing analyses?

A. Catastrophe losses are often correlated and highly severe

B. Catastrophe losses never affect premiums

C. Catastrophe claims are always independent

D. Catastrophe losses have no uncertainty

Correct Answer: A. Catastrophe losses are often correlated and highly severe

Question 9

A combined ratio above 100% generally indicates:

A. An underwriting loss before items outside the ratio

B. A guaranteed investment loss

C. That claims are zero

D. That the expense ratio is zero

Correct Answer: A. An underwriting loss before items outside the ratio

Question 10

Which control is most appropriate for an AI-assisted pricing model?

A. Model validation, bias testing and drift monitoring

B. Removing all human review

C. Using undocumented variables without testing

D. Ignoring changes in claims behaviour

Correct Answer: A. Model validation, bias testing and drift monitoring