Claims, Fraud and Underwriting Feedback

Introduction and Context

Underwriting does not end when a policy is issued. The ultimate test of an underwriting decision is revealed through claims: what happened, why it happened, how much it cost, how quickly it developed, whether coverage responded as intended, and whether similar losses are likely to recur. Claims therefore provide the most important empirical feedback for underwriting management.

A mature insurer treats claims information as a strategic learning system rather than merely an operational record. Claims trends influence risk selection, rating, deductibles, limits, exclusions, policy conditions, reinsurance structures and renewal decisions. Reserving information reveals the probable final cost of claims before all payments are made. Litigation developments can alter the interpretation of coverage and generate systemic exposure. Fraud indicators help distinguish genuine deterioration in a portfolio from opportunistic or organized behavior. Root-cause analysis enables underwriters to correct the underlying risk rather than simply react to loss frequency.

The central management challenge is to convert imperfect, delayed and sometimes biased claims evidence into disciplined underwriting action. A high claim ratio may indicate inadequate pricing, poor risk selection, inflation, catastrophe accumulation, claims leakage, fraud, or a temporary shock. These explanations require different responses. A sound feedback process therefore combines quantitative analysis with investigation, legal interpretation, operational review and professional judgment.

Learning Objectives

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

1. Explain how claims data supports underwriting feedback and portfolio governance.
2. Distinguish reported, paid, incurred and ultimate claims measures.
3. Interpret frequency, severity, loss ratio, development and reserving information.
4. Identify fraud indicators while respecting fairness, privacy and due process.
5. Apply root-cause analysis to recurrent or severe losses.
6. Assess how litigation and wording developments affect underwriting decisions.
7. Translate claims intelligence into pricing, policy wording, risk selection and renewal strategy.
8. Evaluate the limitations of claims-based feedback and design appropriate controls.

1. Claims as an Underwriting Feedback System

A claim is both a financial event and an information event. It provides evidence about the insured, the hazard, the policy contract, the distribution channel, the claims process and the assumptions used in underwriting. Claims feedback is effective only when information moves systematically from claims operations to underwriting, actuarial, legal, risk engineering, compliance and senior management.

A useful feedback loop contains six stages:

1. Capture: record exposure, policy, event, claimant, cause, coverage and financial information.
2. Validate: check data quality, coverage, duplication, coding and reserve reasonableness.
3. Analyse: examine frequency, severity, development, concentration, causation and trends.
4. Interpret: distinguish random variation from structural deterioration.
5. Decide: modify selection, price, terms, limits, deductibles, controls or renewal posture.
6. Monitor: test whether the intervention produced the intended result.

The loop must operate at several levels. Individual claims may reveal a defective process or unusual hazard. Account-level analysis may identify a poorly managed insured. Segment-level analysis may show adverse selection or inadequate rates. Portfolio-level analysis may expose catastrophe accumulation, social inflation or an emerging liability trend.

Claims data is not automatically underwriting truth

Claims records are shaped by reporting practices, claim-handling decisions, policy limits, legal representation, settlement incentives, inflation and changes in coverage. A fall in paid claims may reflect slower settlement rather than better experience. A rise in reported claims may result from improved reporting or a new notification requirement. Underwriters must therefore ask what the data measures, what it omits and what institutional process produced it.

2. Core Claims Measures

Frequency and severity

Claim frequency measures how often claims occur relative to exposure. A general form is:

Exposure units may be vehicle-years, insured values, payroll, revenue, employee-years, policy-years or another measure appropriate to the class. Frequency is meaningful only when the exposure denominator is consistent over time and comparable across segments.

Claim severity measures the average cost of claims:

The numerator must specify whether it represents paid, incurred or ultimate cost. Severity is often highly skewed because a small number of large claims may account for a substantial proportion of total loss.

For a portfolio with reasonably stable exposure, expected loss cost can be represented as:

This relationship assumes that frequency and severity are defined consistently and that their product is being used as an expected-value approximation. It does not imply that every policy has the same risk, nor does it capture dependence, catastrophe clustering or policy expenses.

Paid, incurred and ultimate losses

Paid loss is the amount already disbursed. Reported loss or case-incurred loss generally combines paid amounts with the case reserve established by the claims handler. Incurred loss is commonly expressed as:

The formula applies to a claim or group of claims when the case reserve is defined as the outstanding amount estimated by the claims function. It excludes or includes claim expenses depending on the insurer’s accounting convention; the convention must be stated explicitly.

Ultimate loss is the estimated final cost after all claims in the cohort have been settled and all relevant development has occurred. It may be estimated using actuarial methods such as chain-ladder, Bornhuetter–Ferguson, frequency-severity models or individual claim assessments. Ultimate loss is an estimate, not an observed fact at early development stages.

Loss ratio

The loss ratio compares claims cost with premium:

Claims cost may be paid, incurred or ultimate, and the choice materially affects interpretation. Earned premium is premium attributable to the period of risk, not necessarily the cash received during that period. A loss ratio above 100% means claims cost exceeds earned premium before considering operating expenses, acquisition costs, reinsurance and capital costs.

A simplified underwriting result can be expressed as:

The identity assumes consistent accounting bases and excludes investment income, taxes and other non-underwriting items. For managerial analysis, insurers often use an expense ratio and a combined ratio:

A combined ratio below 100% indicates an underwriting profit on the stated basis; it does not necessarily imply total company profitability.

Worked example: interpreting portfolio deterioration

An insurer writes 20,000 policy-years. It records 1,000 claims and incurred claims of CU 7.5 million. Earned premium is CU 10 million and underwriting expenses are CU 2.2 million.

Frequency = 1,000 ÷ 20,000 = 0.05 claims per policy-year, or 5 claims per 100 policy-years.

Average severity = CU 7,500,000 ÷ 1,000 = CU 7,500.

Expected incurred loss cost per policy-year = 0.05 × CU 7,500 = CU 375.

Loss ratio = CU 7,500,000 ÷ CU 10,000,000 × 100 = 75%.

Expense ratio = CU 2,200,000 ÷ CU 10,000,000 × 100 = 22%.

Combined ratio = 75% + 22% = 97%.

The portfolio appears marginally profitable before other adjustments. However, management should investigate whether the result includes catastrophe claims, reserve strengthening, large losses, inflation, or a change in exposure mix. The average severity may conceal a heavy-tailed distribution, and the apparent profit may not be sustainable.

3. Claims Development and Reserving Insights

Claims mature over time. Property claims may settle quickly, whereas bodily injury, professional liability, workers’ compensation and environmental claims may develop for many years. Comparing immature accident-year claims with mature years without adjustment can produce misleading conclusions.

A claims development triangle arranges claims by accident year or underwriting year and development age. Each cell may contain paid or incurred cumulative claims. The diagonal contains recent, immature observations; older rows show how losses developed toward ultimate values.

A simplified chain-ladder approach calculates development factors. If cumulative claims at development age 12 months are CU 4 million and at 24 months are CU 5 million, the observed age-to-age factor is:

Thus, the factor is 5 ÷ 4 = 1.25. Applying such factors requires assumptions of reasonably stable development patterns, credible historical data and comparable claims-handling practices. It can be unreliable after a major change in policy terms, claims operations, inflation, legal environment or portfolio composition.

Reserving insights are valuable to underwriters for several reasons:

• Persistent adverse development may indicate underpricing or inadequate selection.
• Faster-than-expected settlement may indicate a change in claims strategy rather than lower ultimate cost.
• Increasing case reserves may signal more severe claims, legal escalation or conservative reserving.
• Large incurred-but-not-reported exposure may reveal reporting delays or latent injury patterns.
• Reserve strengthening concentrated in a segment may justify underwriting action before annual results visibly deteriorate.

Underwriters should distinguish reserve adequacy from risk quality. A well-reserved poor portfolio may show a stable incurred ratio, while an inadequately reserved portfolio may appear profitable until later development. Underwriting committees should therefore review paid, incurred and ultimate views together.

Reserving governance

Claims and actuarial functions should document assumptions, segment definitions, changes in case-reserving practice and the treatment of large or catastrophe claims. Independent actuarial review, reserve monitoring and reconciliation to financial reporting reduce the danger that underwriting decisions rely on unstable numbers. Underwriters should not direct reserves to improve reported performance; claims estimates must remain independent and evidence-based.

4. Claims Trends and Portfolio Diagnosis

A trend is meaningful only after controlling for exposure, mix and external conditions. Useful dimensions include:

• accident year and report year;
• product, coverage and policy form;
• geography and catastrophe zone;
• distribution channel and producing agent;
• industry, occupation or vehicle type;
• claim cause and loss location;
• limit, deductible and attachment point;
• claim size bands and development age;
• litigated versus non-litigated claims;
• new business versus renewal business.

A practical diagnostic separates four questions:

1. Is frequency changing? This may indicate behavior, controls, reporting or selection problems.
2. Is severity changing? Possible causes include inflation, repair costs, medical costs, liability awards or limits.
3. Is mix changing? Growth in a hazardous segment can raise portfolio losses even if segment rates are unchanged.
4. Is development changing? Delayed reporting or litigation may make current results look artificially favorable.

Trend analysis should use both absolute and relative measures. A small segment can have a volatile loss ratio because one claim dominates the denominator. Credibility, exposure-weighted analysis, large-loss thresholds and confidence intervals may be needed before taking action. Quantitative significance is not the same as managerial significance: one severe event may justify risk-control intervention even when statistical evidence is limited.

5. Fraud: Detection, Investigation and Control

Insurance fraud involves deliberate deception or misrepresentation intended to obtain an improper benefit. It may occur at application, policy servicing, claim reporting, settlement or provider billing stages. Fraud can be opportunistic, organized, internal or facilitated by external networks.

Common indicators include:

• material inconsistencies between application, incident report and supporting documents;
• loss shortly after inception or shortly before expiry;
• repeated claims involving the same claimant, address, repairer, medical provider or legal representative;
• suspiciously identical wording or invoices across unrelated claims;
• unusual escalation of claimed value after initial notification;
• inability to produce ownership, maintenance or purchase evidence;
• signs that the loss occurred before the policy began;
• excessive pressure for immediate settlement;
• unusual relationships among insureds, witnesses and service providers;
• claims patterns that differ sharply from comparable exposure.

An indicator is not proof of fraud. A sound fraud management process uses triage, verification, specialist investigation and documented decisions. Automated scoring can prioritize cases, but it should not determine denial without human review and procedural safeguards.

Fraud and underwriting feedback

Fraud findings should be coded separately from ordinary claims causes. Otherwise, an underwriter may incorrectly conclude that a class has deteriorating physical risk when the primary issue is organized fraud. Feedback may lead to stronger identity verification, proof-of-loss requirements, vendor controls, payment restrictions, revised warranties, improved wording or enhanced pre-bind screening.

Fraud controls must be proportionate. Excessive suspicion can delay legitimate claims, damage customer trust and create regulatory exposure. Insurers should provide clear reasons for adverse decisions, maintain evidence trails and apply investigation standards consistently.

6. Root-Cause Analysis of Losses

Root-cause analysis asks why a loss occurred and why existing controls failed to prevent, detect or limit it. The immediate cause is not always the underlying cause. For example, a warehouse fire may be immediately caused by an electrical fault, while deeper causes include poor maintenance, combustible storage, inadequate inspection and weak management accountability.

Useful techniques include:

• Five Whys: repeatedly ask why the event occurred until controllable systemic causes are identified.
• Fishbone analysis: organize causes under categories such as people, process, equipment, environment, management and materials.
• Fault-tree analysis: begin with an undesired event and logically map combinations of failures.
• Bow-tie analysis: connect threats to a top event and then to consequences, with preventive and mitigative barriers.
• Pareto analysis: rank causes or loss types to identify a relatively small number of dominant contributors.

Root-cause findings should be translated into underwriting variables. A repeated machinery loss may justify inspection requirements, maintenance warranties, higher deductibles, engineering recommendations or a revised appetite statement. A series of professional liability claims may require client-selection criteria, contract review, cyber controls or a sublimit for a newly exposed service.

The analysis should distinguish controllable and uncontrollable causes. Weather may not be controllable, but location, construction, drainage, warning systems and business continuity planning may be. The objective is not to assign blame; it is to improve risk control and contract design.

7. Litigation, Social Inflation and Policy Wording

Litigation affects both the amount and timing of claims. Legal costs, expanded theories of liability, collective actions, third-party litigation funding, broad interpretations of coverage and increased damages can produce social inflation. The term should be used carefully because it may combine several distinct influences: legal doctrine, claimant behavior, jury awards, regulatory developments, public expectations and defense costs.

Underwriters should monitor:

• new court decisions affecting coverage interpretation;
• changes in statutory liability or limitation periods;
• developments in professional, cyber, product and directors’ liability;
• trends in defense costs and settlement values;
• litigation frequency by jurisdiction;
• aggregation and occurrence interpretation;
• allocation among policies and insurers;
• the enforceability of exclusions, warranties and conditions.

Policy wording is a risk-control instrument. Ambiguous language may be interpreted against the drafter in some legal systems, while an exclusion that is technically broad may be ineffective if it conflicts with the policy’s reasonable expectations or mandatory law. Wording changes should be reviewed by legal, claims, actuarial and underwriting specialists. They should be tested against realistic claim scenarios rather than drafted in isolation.

Feedback from litigated claims may lead to clearer definitions, reporting clauses, consent provisions, claims cooperation duties, dispute-resolution clauses, sublimits, exclusions or endorsements. However, wording must remain transparent and legally enforceable. A contract designed solely to avoid payment may create reputational, regulatory and conduct risk.

8. Translating Feedback into Underwriting Decisions

Claims intelligence can influence four principal decisions.

Risk selection

The insurer may revise appetite, referral rules, minimum control standards, geographic limits, industry prohibitions or authority thresholds. Selection should consider prospective risk, not simply punish a policyholder for historical claims. Underwriters should investigate whether the insured corrected the cause and whether the claim was exceptional.

Pricing

Pricing responses may include rate changes, changes to deductibles, risk-adjusted fees, experience rating, catastrophe loadings, increased margins for uncertainty or revised reinsurance costs. Price increases should reflect expected loss, expenses, capital and uncertainty. Claims data must be adjusted for exposure changes, inflation, development, large losses and coverage changes before being used as a rate indication.

Policy wording and structure

Feedback can support revised definitions, exclusions, sublimits, waiting periods, warranties, conditions precedent, deductibles, coinsurance, attachment points or claims-made reporting provisions. Each change should be assessed for legal validity, customer impact, operational feasibility and consistency with the intended risk transfer.

Renewal strategy

Renewal decisions should be evidence-based. Possible responses include continuation, targeted remediation, revised terms, increased deductible, co-insurance, risk engineering, reduced limits, non-renewal or referral to senior authority. A renewal meeting should address both the loss outcome and management response. A high-loss account with excellent corrective action may be preferable to a low-loss account with deteriorating controls.

A useful governance tool is an underwriting feedback register:

9. Technology, Analytics and Artificial Intelligence

Modern insurers use claims warehouses, geospatial tools, text analytics, graph analysis and predictive models. Natural-language processing can identify emerging causes in adjuster notes and court judgments. Image analytics may support damage assessment. Network analysis can reveal relationships among claimants, providers, repairers and addresses. Machine learning can prioritize claims for investigation or estimate severity.

These technologies require governance. Important controls include:

• documented purpose and model scope;
• representative and high-quality training data;
• validation, calibration and drift monitoring;
• explainability appropriate to the decision;
• human oversight and appeal mechanisms;
• privacy, security and retention controls;
• testing for unfair discrimination and proxy bias;
• separation of fraud suspicion from final adverse action.

Historical claims data may encode past bias, inconsistent investigation or unequal access to legal support. A model that predicts claims cost accurately on average may still produce unfair outcomes for protected groups or vulnerable customers. Governance should therefore assess both predictive performance and conduct impact.

10. Illustrative Case: Commercial Property Renewal

Background

This illustrative case concerns Meridian Insurance, which insures a regional food-processing company. The account has been profitable for four years but produced three water-damage claims and one electrical fire in the latest two years. The insured requests renewal at unchanged terms.

Problem

The claims department reports incurred losses of CU 1.8 million, while paid losses are only CU 0.9 million. Two claims remain open, and one has been litigated because the insured disputes a coverage limitation. The underwriter initially considers non-renewal because the recent incurred loss ratio is high.

Analysis

A cross-functional review separates the evidence. Root-cause analysis shows that the water losses arose from inadequate inspection of roof drainage and aging sprinkler valves. The electrical fire followed an overloaded production line. Risk engineering confirms that the insured has since installed monitoring devices, replaced valves and adopted a documented maintenance schedule. The litigated claim resulted partly from ambiguous wording about stock deterioration following interruption of refrigeration.

Claims analytics also show that comparable food-processing accounts have experienced similar water losses during severe storms. The account’s frequency is elevated, but its corrected controls are materially better than those of peer accounts. Reserve review indicates that the CU 0.9 million outstanding amount is plausible, but ultimate cost remains uncertain because of the litigation.

Approach and solution

Meridian renews, but not on unchanged terms. It requires verified maintenance records, introduces a water-damage deductible, applies a sublimit and endorsement for stock deterioration, and adjusts the rate to reflect corrected but still material exposure. The insurer also requires electrical thermal imaging and quarterly compliance evidence. Legal and claims teams revise the wording to clarify causation and documentation requirements.

Outcome

At the next renewal, no major water loss occurs. Minor incidents are reported early, and the new monitoring system produces evidence supporting rapid adjustment. The account remains profitable, while the insured retains coverage aligned with its operational needs.

Learning points

1. Incurred losses must be interpreted with reserve and litigation uncertainty.
2. Corrective action can be more informative than historical claims alone.
3. Wording ambiguity can create claims cost even when physical risk is understood.
4. Renewal strategy can combine risk improvement, pricing and contract redesign.
5. Claims feedback is strongest when claims, actuarial, legal and engineering functions collaborate.

11. Challenges and Limitations

Claims feedback faces several structural limitations. Long-tail claims create delayed evidence. Low-frequency, high-severity events produce statistical volatility. Historical data may not represent future climate, technology, regulation or customer behavior. Portfolio changes can destroy comparability. Reserving assumptions can mask deterioration. Fraud detection may generate false positives. Litigation can change the meaning of historical claims after the data has been used for pricing.

There is also a danger of feedback loops. If an insurer increases price or restricts terms for a segment, the healthiest customers may leave while higher-risk customers remain. The resulting claims experience may then appear to validate the original action. Management should monitor retention, mix and new-business quality after major underwriting changes.

Another limitation is organizational: claims teams may code causes differently, underwriters may lack access to detailed notes, and commercial pressure may discourage escalation. Governance therefore requires common taxonomies, data standards, audit trails and incentives that reward sustainable underwriting rather than short-term premium growth.

12. Best-Practice Governance Framework

An effective claims-to-underwriting framework should include:

1. A defined claims taxonomy covering cause, coverage, severity, litigation, fraud and recovery.
2. Consistent exposure and premium bases for frequency and loss-ratio analysis.
3. Regular claims triangles and reserve-development reviews.
4. Large-loss and catastrophe reporting separate from attritional experience.
5. Formal referral triggers for deterioration, litigation or fraud concentration.
6. Root-cause investigations for recurrent or severe losses.
7. Legal review of wording and claims precedents.
8. Documented renewal actions and account-level remediation plans.
9. Model-risk, privacy and fairness controls for analytics.
10. Independent assurance through audit, actuarial review and governance committees.

The strongest organizations close the loop by measuring whether underwriting changes actually improve outcomes. They compare post-intervention frequency, severity, claim development, customer outcomes and retention with the original objectives. Feedback is not complete until learning changes behavior and the result is evaluated.

Chapter Summary

Claims provide the principal empirical feedback for underwriting management, but they must be interpreted in context. Frequency, severity, loss ratio, paid losses, incurred losses and ultimate losses answer different questions. Development triangles and reserve analysis help reveal emerging deterioration before claims are fully settled. Claims trends should be segmented by exposure, product, geography, cause, development age and litigation status.

Fraud indicators support investigation and control, but suspicion is not proof and automated tools require human oversight. Root-cause analysis moves beyond the immediate cause of loss to identify failures in people, process, equipment, management and controls. Litigation and social inflation may affect both claim cost and policy interpretation, making legal and wording feedback essential.

Underwriting feedback should lead to disciplined changes in risk selection, pricing, policy wording, deductibles, limits, engineering requirements and renewal strategy. The process must respect actuarial independence, customer fairness, privacy, regulatory obligations and model governance. The best feedback systems are cross-functional, evidence-based and monitored after implementation.

Key Terms

Claims development: Change in the estimated or paid cost of claims as they mature.

Claim frequency: Number of claims relative to an appropriate exposure base.

Claim severity: Average cost per claim.

Case reserve: Claim amount estimated by the claims handler for an outstanding claim.

Combined ratio: Loss ratio plus expense ratio.

Earned premium: Premium allocated to the period for which insurance protection is provided.

Fraud indicator: Fact or pattern that warrants further verification for possible deception.

Incurred loss: Paid loss plus case outstanding reserve, subject to the insurer’s accounting convention.

Loss ratio: Claims cost divided by earned premium.

Loss triangle: Arrangement of claims data by accident or underwriting period and development age.

Ultimate loss: Estimated final cost of a claim or claims cohort.

Litigation severity: The cost impact associated with defended, disputed or litigated claims.

Root-cause analysis: Structured investigation of underlying causes and control failures.

Social inflation: Growth in liability costs associated with legal, social and behavioral developments beyond ordinary economic inflation.

Underwriting feedback loop: Process by which claims evidence is analysed and used to modify future underwriting.

Risk selection: Decision about which risks an insurer accepts and on what conditions.

Reserve development: Change in estimated claim liabilities over time.

Special Investigation Unit: Specialized function investigating suspected insurance fraud or abuse.

Model drift: Decline in model performance because relationships in new data differ from those in development data.

Review Questions

Short Answer Questions

1. Distinguish paid, incurred and ultimate loss.
2. Why must claims trends be adjusted for exposure and portfolio mix?
3. State two appropriate uses and one limitation of fraud indicators.
4. What is the purpose of root-cause analysis in underwriting?
5. How can litigation influence policy wording?

Long Answer Questions

1. Explain how a claims feedback loop can influence risk selection, pricing, wording and renewal strategy.
2. Discuss the use of claims development and reserving information in underwriting governance.
3. Critically evaluate the relationship between claims frequency, severity and expected loss cost.
4. Explain how social inflation and litigation developments can create underwriting risk.
5. Describe governance requirements for using artificial intelligence in claims fraud detection.

Analytical / Application Questions

1. A portfolio has stable frequency but sharply increasing severity. Analyse at least five possible explanations and recommend the data needed before changing rates.
2. An account has poor historical claims but has implemented verified engineering improvements. Design a renewal strategy that balances historical evidence and prospective risk.
3. Develop a root-cause and underwriting-response plan for repeated commercial property water losses, including claims, engineering, pricing and wording actions.

Multiple Choice Questions with Answers

Question 1

Which measure combines claim frequency and average severity to estimate expected loss cost per exposure unit?

A. Claim frequency plus severity

B. Claim frequency multiplied by average severity

C. Earned premium divided by claims

D. Paid loss minus case reserve

Correct Answer: B. Claim frequency multiplied by average severity

Question 2

Incurred loss is generally calculated as:

A. Earned premium plus expenses

B. Ultimate loss minus premium

C. Paid loss plus case outstanding reserve

D. Case reserve minus paid loss

Correct Answer: C. Paid loss plus case outstanding reserve

Question 3

A loss ratio of 75% and an expense ratio of 22% produce a combined ratio of:

A. 53%

B. 75%

C. 97%

D. 122%

Correct Answer: C. 97%

Question 4

Which statement about a fraud indicator is most accurate?

A. It is conclusive proof of fraud

B. It should automatically result in claim denial

C. It warrants verification and possible investigation

D. It is relevant only after litigation begins

Correct Answer: C. It warrants verification and possible investigation

Question 5

What is the principal purpose of root-cause analysis?

A. To assign blame to the claimant

B. To identify underlying causes and control failures

C. To eliminate the need for reserves

D. To replace policy wording review

Correct Answer: B. To identify underlying causes and control failures

Question 6

Why can comparing immature accident-year claims with mature years be misleading?

A. Immature claims have already reached ultimate value

B. Claims may develop and increase over time

C. Mature claims cannot be segmented

D. Earned premium is never available for mature years

Correct Answer: B. Claims may develop and increase over time

Question 7

Which underwriting action most directly responds to repeated maintenance-related losses?

A. Ignore the losses if premium is growing

B. Require risk improvements and verify compliance

C. Remove all claims reporting requirements

D. Replace earned premium with written premium

Correct Answer: B. Require risk improvements and verify compliance

Question 8

Which is an important governance control for AI-based fraud detection?

A. Use the model without human review

B. Avoid monitoring after implementation

C. Test for bias and provide appropriate human oversight

D. Delete all model documentation

Correct Answer: C. Test for bias and provide appropriate human oversight

Question 9

If cumulative claims are CU 4 million at 12 months and CU 5 million at 24 months, the observed development factor is:

A. 0.80

B. 1.00

C. 1.25

D. 9.00

Correct Answer: C. 1.25

Question 10

Why should policy wording changes be reviewed by claims and legal specialists?

A. To ensure the wording reflects claims realities and is legally enforceable

B. To guarantee that no claim can ever be paid

C. To eliminate the need for underwriting judgment

D. To ensure all policies use identical exclusions

Correct Answer: A. To ensure the wording reflects claims realities and is legally enforceable