Actuarial Governance, Ethics and Future Practice

Introduction and context

Actuaries work at the boundary between financial mathematics, uncertainty, institutional decision-making and public interest. Their models influence premiums, reserves, solvency capital, pensions, investments and claims decisions. Consequently, technical competence alone is insufficient. Actuarial work must also be governed, communicated and performed ethically.

Actuarial governance is the system through which actuarial assumptions, models, judgments, data, advice and professional responsibilities are directed, reviewed and controlled. It connects the individual actuary to the insurer’s board, senior management, risk function, auditor, supervisor, policyholders and wider society.

The central challenge is that actuarial conclusions are rarely mechanically determined. Future mortality, morbidity, inflation, lapse rates, catastrophe losses and investment returns are uncertain. Professional judgment is therefore unavoidable. Governance does not eliminate judgment; it makes judgment transparent, challengeable and accountable.

Learning objectives

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

1. Explain the meaning and structure of actuarial governance.
2. Distinguish professional ethics, professional standards, regulation and internal controls.
3. Apply actuarial principles to pricing, reserving and capital decisions without confusing their financial components.
4. Identify conflicts of interest, bias, model risk and communication failures.
5. Relate actuarial practice to recognised frameworks, including IAA standards, professional codes, Solvency II, ORSA and IFRS 17.
6. Evaluate emerging issues involving artificial intelligence, climate risk, data governance and sustainability.
7. Design a proportionate governance process for an actuarial model or opinion.

1. The architecture of actuarial governance

A sound governance system has several mutually reinforcing layers:

Governance should operate across the entire actuarial cycle:

1. define the purpose and decision to be supported;
2. specify data, assumptions and methodology;
3. build or select the model;
4. independently review and validate it;
5. communicate results, uncertainty and limitations;
6. obtain appropriate approval;
7. monitor outcomes and revise when conditions change.

A model that is mathematically sophisticated but poorly documented is weakly governed. Conversely, extensive documentation cannot compensate for invalid assumptions or biased data.

Roles and accountability

The board is responsible for overall risk appetite and oversight, although it does not normally perform the actuary’s technical work. Senior management implements strategy. The appointed or responsible actuary provides professional advice within the applicable legal framework. The risk function challenges assumptions and aggregation. Internal audit evaluates controls independently. The supervisor assesses whether the insurer is sound and policyholders are protected.

The three-lines model is useful but not absolute. Actuaries may work in the first line, such as pricing or reserving, or in a second-line risk role. Their professional obligations remain applicable in either position. A reporting line must not prevent an actuary from raising a material concern.

2. Ethics and professional judgment

Ethics concerns what a professional ought to do, not merely what is technically possible or legally permissible. A lawful decision can still be misleading, unfair or inconsistent with professional integrity.

Core ethical principles commonly found in actuarial codes include:

• integrity and honesty;
• competence and continuing professional development;
• impartiality and freedom from inappropriate influence;
• confidentiality and responsible information use;
• clear, fair and timely communication;
• compliance with applicable standards and law;
• responsibility to the public interest.

The public-interest dimension is particularly important in insurance. The immediate client may be an insurer, but the consequences of advice may fall on policyholders, claimants, pensioners and taxpayers.

An ethical decision framework

When faced with a difficult instruction, an actuary can proceed as follows:

1. Identify affected stakeholders, including those without contractual power.
2. Clarify the decision, the applicable law and professional standards.
3. Separate facts, assumptions, judgments and value choices.
4. Identify conflicts, incentives and possible harms.
5. Consider reasonable alternatives and their distributional effects.
6. Document the analysis and seek appropriate peer or governance review.
7. Communicate the conclusion, uncertainty and limitations.
8. Escalate or withdraw if the work cannot be performed competently or independently.

“Client preference” is not an adequate reason for selecting an unsupported assumption. Nor is commercial pressure a justification for omitting a material limitation.

Conflicts of interest

Conflicts may be actual, potential or perceived. Examples include remuneration linked to a target result, advising both parties to a transaction, using confidential information for another client, or approving a model that the actuary designed without independent review.

A proportionate response may include disclosure, consent, segregation of duties, independent review or withdrawal. Disclosure alone is not sufficient where impartiality cannot reasonably be preserved.

3. Recognised standards and regulatory frameworks

Specific requirements vary by jurisdiction, so an actuary must identify the governing framework before commencing work. Important international and national reference points include the following.

International Actuarial Association standards

The International Actuarial Association (IAA) issues International Standards of Actuarial Practice (ISAPs). ISAPs are model standards rather than automatically binding law in every country. They provide internationally consistent guidance on matters such as general actuarial practice, data quality, assumptions, risk classification, social security and insurance contracts. A national association or regulator may adopt, modify or replace them.

An actuary should therefore distinguish:

• an IAA model standard;
• a locally adopted actuarial standard;
• a regulatory rule;
• non-mandatory professional guidance.

The distinction matters because the source determines enforceability and the applicable compliance test.

Codes and disciplinary systems

National actuarial bodies, such as the Institute and Faculty of Actuaries, the Society of Actuaries, the Casualty Actuarial Society and equivalent bodies in other jurisdictions, publish codes of conduct and standards. These generally address competence, communication, conflicts, confidentiality and compliance. Disciplinary procedures give the code practical force.

Solvency regulation and ORSA

Under the European Union’s Solvency II framework, insurers are subject to risk-sensitive solvency requirements, governance requirements and supervisory reporting. The Own Risk and Solvency Assessment (ORSA) is a forward-looking process through which an insurer assesses its own risk profile, solvency needs and ability to remain within risk appetite. ORSA is not simply a regulatory form; it should connect strategy, stress testing, capital management and board oversight.

Other jurisdictions use different capital regimes, including risk-based capital systems and principles-based frameworks. The governing principle is similar: capital assessment should reflect material risks, uncertainty and the insurer’s business model.

IFRS 17

IFRS 17 is an accounting standard for insurance contracts. It requires disciplined measurement, grouping, assumptions, discounting and recognition of insurance service results. Its requirements differ from solvency measurement and regulatory reporting. An actuary must not assume that an IFRS 17 liability is automatically the same as a prudential technical provision.

Technical actuarial standards

The UK Technical Actuarial Standards, for example, provide principles for technical actuarial work and emphasize communication, data, assumptions, models and materiality. Other jurisdictions have comparable standards. A defensible actuarial report should identify the standards relied upon and explain any departures or limitations.

4. Pricing, profit and transparent financial definitions

A common governance failure occurs when financial quantities are defined inconsistently. In particular, “premium before profit” must not be used while the margin variable already includes profit.

Let:

• P = written premium or premium charged;
• E[L] = expected loss cost, in currency units;
• C = expenses and other explicitly identified costs, in currency units;
• K = required compensation for capital, uncertainty or risk margin, in currency units;
• Π = target profit, in currency units.

If the premium is defined as the total premium charged, including the target profit, the additive formulation is:

Alternatively, if a single margin M is defined to include both capital compensation and target profit, then:

These two formulations are equivalent only when M is explicitly defined as K + Π. If M means risk margin only, profit must be added separately. The terms must not be mixed.

A multiplicative expense-and-profit loading is a different convention. If B = E[L] + K is the technical cost base, expenses are e of premium and target profit is r of premium, then:

This formula assumes that both expense and profit percentages are proportions of premium, that e + r < 1, and that the cost base is correctly specified. It is not equivalent to simply multiplying the loss cost by 1 + e + r.

Worked example: reconciling premium definitions

Suppose expected losses are 600,000 currency units, expenses are 120,000, capital compensation is 30,000 and target profit is 50,000. Under the additive approach:

P = 600,000 + 120,000 + 30,000 + 50,000 = 800,000.

If M is defined as the combined capital-and-profit margin, then M = 30,000 + 50,000 = 80,000, and P = 600,000 + 120,000 + 80,000 = 800,000.

The interpretation is different if “premium before profit” is required. In that case, define:

P₀ = E[L] + C + K = 750,000,

and total premium after profit is P = P₀ + Π = 800,000. Thus, P₀ and P must not be labelled interchangeably.

Pricing governance also requires monitoring adequacy, fairness, affordability, claims experience, expense allocation, reinsurance, concentration and regulatory restrictions. A technically adequate average premium may still create unfair outcomes if rating variables are proxies for protected characteristics or if vulnerable customers are systematically disadvantaged.

5. Reserving, capital and model governance

A reserve is an estimate of future payments and related expenses, conditional on information available at the valuation date. It is not a guaranteed future cash flow. Governance should distinguish best estimate, risk adjustment, prudential margin, accounting measurement and solvency capital.

For a set of future benefit or claim payments X_t, a simplified present value is:

where i_t is the effective discount rate applicable to payment time t. If a single constant annual effective rate i is appropriate, the special case is:

The constant-rate formula must not be used when a term structure is materially relevant without justification.

Worked example: present value and sensitivity

Suppose expected payments are 100,000 at the end of year 1, 120,000 at the end of year 2 and 150,000 at the end of year 3. At a constant annual effective discount rate of 5%:

PV = 100,000/1.05 + 120,000/1.05² + 150,000/1.05³.

The components are approximately 95,238, 108,844 and 129,576, giving:

PV ≈ 333,658.

At 6%, the corresponding value is approximately 94,340 + 106,787 + 125,890 = 327,017. The difference, about 6,641, illustrates interest-rate sensitivity. A governance report should disclose the rate, timing convention, payment assumptions and sensitivity rather than presenting a single number without context.

Model governance requires an inventory of models, ownership, approval thresholds, validation, change control, access control, independent review and retirement procedures. Validation should examine conceptual soundness, data, coding, calibration, performance, stability, limitations and use. Back-testing compares predicted outcomes with subsequent experience, while stress testing examines adverse but plausible scenarios.

Materiality should be assessed quantitatively and qualitatively. A small numerical error may be material if it affects a regulatory breach, vulnerable policyholders or a key management decision.

6. Data governance, artificial intelligence and explainability

Actuarial work increasingly uses granular data, machine learning and automated decision systems. These tools may improve prediction, but they introduce risks:

• incomplete, inaccurate or non-representative data;
• historical discrimination embedded in labels;
• proxy variables for legally protected characteristics;
• opaque model outputs;
• data leakage and cyber risk;
• instability under changing economic or climate conditions;
• automation bias, where users accept an output without challenge.

A governance framework should document data lineage, lawful basis for use, quality checks, feature selection, fairness testing, performance by subgroup, human oversight and appeal procedures. Explainability should be proportionate to the decision’s impact. A high-impact claims or pricing decision generally requires more than a generic statement that a model is “accurate.”

The actuary remains accountable for professional work even when software is supplied by a vendor. Vendor reliance requires due diligence, contractual access to documentation, validation evidence, security assessment and contingency planning.

7. Climate, sustainability and systemic risk

Climate change affects physical hazards, liability, mortality, morbidity, asset values, inflation, reinsurance capacity and claims settlement. It creates both acute risks, such as floods and storms, and transition risks arising from policy, technology and consumer behaviour.

Traditional models based solely on historical averages may be inadequate where the risk environment is non-stationary. Actuaries should consider scenario analysis, multiple time horizons, correlation uncertainty and model limitations. Climate scenarios are not forecasts; they are conditional explorations of possible pathways.

Systemic risks also challenge diversification assumptions. Pandemic risk, cyber accumulation, supply-chain disruption and sovereign stress may cause losses across lines that were historically treated as weakly correlated. Capital models should test dependency assumptions rather than relying uncritically on historical correlations.

8. Communication and actuarial opinions

An actuarial opinion should enable a reasonably informed user to understand:

• the purpose and scope of the work;
• the data used and significant data limitations;
• methods and assumptions;
• material judgments and alternatives considered;
• uncertainty, sensitivities and scenario results;
• reliance on other experts or models;
• applicable standards and any departures;
• the conclusion and its intended use.

False precision is a communication error. Reporting a reserve of 333,658 currency units does not imply that the estimate is accurate to one currency unit. Rounding and ranges may communicate uncertainty more honestly, provided they are supported by the analysis.

Technical language should be adapted to the audience without removing material qualifications. A board presentation may use a concise dashboard, but underlying papers should preserve sufficient detail for challenge and audit.

9. Illustrative case: automated motor insurance pricing

Background. An insurer introduces a machine-learning model for motor premiums. The model uses claims history, location, mileage, vehicle characteristics and a credit-derived indicator. It improves aggregate predictive accuracy by 8% compared with the existing tariff.

Problem. After implementation, consumer representatives report that premiums have increased sharply for residents of several low-income districts. The data team argues that the model is neutral because it does not include income or ethnicity directly.

Analysis. The actuarial governance committee identifies several concerns. Location may be a proxy for protected characteristics. Historical claims may reflect unequal access to repair services or differential reporting, rather than pure risk. The credit-derived indicator may create affordability and fairness concerns. Aggregate accuracy does not establish fairness across subgroups. There is also insufficient documentation of feature construction and model drift monitoring.

Approach. The responsible actuary requires:

1. a data lineage and legal-use review;
2. subgroup performance and outcome testing;
3. proxy and correlation analysis;
4. removal or modification of variables that cannot be justified;
5. a comparison with a constrained or interpretable model;
6. premium-change limits and customer impact analysis;
7. independent validation and documented board approval;
8. ongoing monitoring, appeals and a model-change trigger.

The insurer also separates the technical risk estimate from the commercial decision about affordability and distribution. The final pricing structure explicitly documents the premium definition, expense treatment, risk margin and target profit.

Outcome. The revised model retains much of the predictive improvement but uses fewer opaque variables, applies fairness controls and introduces monitoring. The board receives both accuracy and distributional-impact metrics. The model is approved for a limited period subject to review.

Learning points. Predictive performance is not the sole criterion for actuarial acceptability. Governance must address legality, fairness, explainability, accountability, uncertainty and the interests of affected policyholders.

10. Future practice and professional capability

Future actuaries will need a combined profile of mathematical, technological, regulatory and interpersonal competence. Important capabilities include:

• stochastic modelling and uncertainty communication;
• data engineering and programming literacy;
• model validation and algorithmic accountability;
• understanding of accounting, capital and enterprise risk;
• climate and sustainability analysis;
• behavioural and distributional analysis;
• negotiation and board-level communication;
• ethical reasoning and professional courage.

Technology will automate some calculations but increase the value of problem formulation, assumption selection, validation and judgment. Generative AI may assist with coding, documentation and scenario exploration, but outputs require verification. Confidential client data should not be entered into an unauthorised system. An AI-generated explanation is not evidence of model validity.

Best practice is therefore not to resist innovation, but to place innovation inside a controlled framework: define permitted uses, protect data, retain human accountability, test outputs, record versions and provide a route for challenge.

Chapter Summary

Actuarial governance integrates law, professional standards, corporate oversight, data and model controls, judgment and monitoring. Ethical actuarial practice requires integrity, competence, impartiality, confidentiality, clear communication and attention to the public interest. IAA ISAPs, national codes, local standards, Solvency II, ORSA and IFRS 17 illustrate the need to distinguish international guidance, enforceable regulation, accounting measurement and solvency assessment.

Financial definitions must be internally consistent. If M includes capital compensation and profit, the premium equation must state that explicitly; if premium is before profit, profit must be added separately. Reserving and capital models require transparent assumptions, validation, sensitivity analysis and appropriate communication. Artificial intelligence, climate change and systemic risk make data governance, fairness, explainability and scenario analysis increasingly important. The future actuary is both a quantitative analyst and a professional steward of trust.

Key Terms

Actuarial governance: The structures and processes used to direct, control and review actuarial work.

Actuarial opinion: A professional conclusion on a defined actuarial matter, supported by methods, assumptions and disclosures.

Best estimate: An estimate of expected future cash flows using appropriate information and assumptions.

Code of conduct: Binding or enforceable professional principles governing members’ behaviour.

Conflict of interest: A circumstance in which personal or competing duties may impair impartiality.

Data lineage: The documented history of data sources, transformations and uses.

IFRS 17: An accounting standard governing measurement and reporting of insurance contracts.

ISAP: An International Standard of Actuarial Practice issued by the IAA.

Model risk: The possibility that a model or its use produces materially misleading results.

ORSA: Own Risk and Solvency Assessment, a forward-looking insurer assessment of risks and solvency needs.

Professional judgment: Reasoned expert choice where evidence does not determine one automatic answer.

Risk adjustment: An amount reflecting uncertainty in non-financial insurance cash flows, where applicable.

Risk margin: Compensation or provision for uncertainty, defined according to the relevant framework.

Solvency II: The European Union’s risk-based insurance supervisory framework.

Stress test: Analysis of outcomes under adverse specified conditions.

Target profit: The explicitly defined profit amount included in a pricing decision.

Technical standard: A professional requirement governing the performance or communication of technical work.

Validation: Independent assessment of a model’s conceptual and operational adequacy.

Review Questions

Short Answer Questions

1. What is actuarial governance, and why is it necessary?
2. Distinguish an IAA model standard from a national regulatory requirement.
3. Why must an actuary disclose material assumptions and limitations?
4. What is model risk?
5. State two ethical risks associated with automated insurance pricing.

Long Answer Questions

1. Explain the relationship between professional ethics, actuarial standards, regulation and corporate governance.
2. Discuss the roles of the board, responsible actuary, risk function, internal audit and supervisor.
3. Explain how an insurer should govern a reserving model from development through monitoring.
4. Compare solvency measurement, IFRS 17 measurement and actuarial best-estimate analysis.
5. Evaluate the implications of climate change and systemic risk for future actuarial practice.

Analytical / Application Questions

1. An insurer defines a margin as including both capital compensation and profit but adds profit again in its premium formula. Diagnose the error and restate a consistent formula.
2. Using the cash flows in the worked reserving example, explain why the reserve changes when the discount rate increases and identify two governance disclosures required.
3. Design a governance response for an AI claims model that is accurate overall but produces materially different outcomes across demographic groups.

Multiple Choice Questions with Answers

Question 1

In the additive premium formulation, if M is defined as including both capital compensation and target profit, which equation is consistent?

A. P = E[L] + C + M

B. P = E[L] + C + M + Π

C. P = E[L] + M − C

D. P = E[L] − C + Π

Correct Answer: A. P = E[L] + C + M

Question 2

Using the worked example, what is the approximate present value of the three payments discounted at 5% annually?

A. 327,017

B. 333,658

C. 350,000

D. 370,000

Correct Answer: B. 333,658

Question 3

Which statement best describes an IAA ISAP?

A. It is automatically binding law in every jurisdiction.

B. It is a model standard that may be adopted or modified locally.

C. It is an insurer’s internal pricing manual.

D. It is an accounting standard replacing IFRS 17.

Correct Answer: B. It is a model standard that may be adopted or modified locally.

Question 4

The primary purpose of ORSA is to:

A. Set individual policy premiums.

B. Provide a historical claims database.

C. Assess the insurer’s own risks, solvency needs and future resilience.

D. Replace the external audit.

Correct Answer: C. Assess the insurer’s own risks, solvency needs and future resilience.

Question 5

If expected losses are 600,000, expenses 120,000, capital compensation 30,000 and target profit 50,000, what is the total additive premium?

A. 680,000

B. 750,000

C. 780,000

D. 800,000

Correct Answer: D. 800,000

Question 6

Which action is most appropriate when an actuary discovers a material conflict of interest?

A. Ignore it if the result appears reasonable.

B. Disclose and manage it proportionately, or withdraw if impartiality cannot be preserved.

C. Transfer all responsibility to software.

D. Delete the relevant documentation.

Correct Answer: B. Disclose and manage it proportionately, or withdraw if impartiality cannot be preserved.

Question 7

Why is aggregate predictive accuracy insufficient to approve an AI insurance model?

A. Accuracy is never relevant to actuarial work.

B. The model may produce unfair or materially different outcomes across groups.

C. Machine-learning models cannot process insurance data.

D. Regulation requires every model to be manual.

Correct Answer: B. The model may produce unfair or materially different outcomes across groups.

Question 8

Which item is most directly associated with model validation?

A. Testing conceptual soundness, data, code, calibration and performance.

B. Increasing the target profit without analysis.

C. Removing all sensitivity tests.

D. Allowing the model owner to approve it without review.

Correct Answer: A. Testing conceptual soundness, data, code, calibration and performance.

Question 9

If a premium before profit is P₀ = 750,000 and target profit is 50,000, what is the total premium?

A. 700,000

B. 750,000

C. 800,000

D. 37,500,000

Correct Answer: C. 800,000

Question 10

Which statement about climate scenarios is correct?

A. They are guaranteed forecasts.

B. They explore conditional possible pathways and are not necessarily forecasts.

C. They eliminate parameter uncertainty.

D. They apply only to life insurance.

Correct Answer: B. They explore conditional possible pathways and are not necessarily forecasts.