Best Practices for Reviewing Actuarial Assumptions

Actuarial assumptions influence pricing, reserving, capital, financial reporting and strategic decisions across an insurer. A well-designed peer review tests whether those assumptions are reasonable, appropriately supported and applied consistently. It also helps management distinguish between genuine changes in underlying risk and movements caused by data quality, model structure or expert judgement.

For Australian insurers, the process sits within a particularly demanding environment. AASB 17 has increased the focus on transparent estimates and contractual service results, while APRA expects sound actuarial governance and documentation. Market conditions add further pressure: claims inflation, severe weather, reinsurance costs, wage movements and changing customer behaviour can all make historical experience a less reliable guide to future outcomes.

Establish The Purpose And Scope

A peer review should begin with a written statement of purpose. It should specify whether the review covers premium liabilities, outstanding claims liabilities, life insurance best estimate liabilities, risk adjustment inputs, capital projections or a broader set of financial assumptions. The reviewer should understand how the work will be used, who will rely on it and which decisions depend on the outcome.

Scope should reflect materiality and risk rather than follow a fixed checklist. A small assumption with little financial effect may require a desktop assessment, while a high-impact assumption with limited data may need detailed replication and sensitivity testing. The review should identify relevant portfolios, valuation dates, lines of business, models, data sources and interfaces with finance, risk and underwriting.

The terms of reference should also define independence. A reviewer who helped develop the original assumption may provide valuable context, but the challenge process needs enough separation to be credible. In a smaller Australian insurer, complete organisational separation may be impractical. Clear disclosure of the reviewer’s involvement, an independent sign-off and review by the appointed actuary or board committee can provide useful safeguards.

The review should align with the insurer’s governance framework and reporting calendar. For an APRA-regulated entity, this includes considering the expectations of CPS 320 and the evidence needed for internal committees, auditors and regulators. Agreeing deliverables early prevents a rushed review when the annual accounts are already approaching approval.

Understand Data, Segmentation And Experience

Assumptions are only as credible as the data supporting them. The reviewer should trace key figures from source systems through extraction, transformation and modelling. This includes checking claims development triangles, policy counts, exposure measures, premium records, benefit payments, lapses, mortality, expenses and recoveries. Reconciliations to the general ledger and regulatory returns can reveal unexplained differences before they become embedded in the valuation.

Segmentation deserves close attention. A broad class may combine risks with very different claim frequencies, severities, policy terms or customer profiles. Australian motor portfolios, for example, may show different experience across metropolitan Sydney, regional New South Wales and remote areas. Home insurance data may also need to distinguish flood, cyclone, bushfire and storm exposure rather than relying on a single national average.

The reviewer should investigate changes in the composition of the portfolio. A reduction in claims could reflect tighter underwriting, fewer policies, higher deductibles or a shift towards lower-risk customers. Conversely, rising claim costs may be linked to repair supply chains, labour shortages or imported parts rather than a permanent deterioration in risk. Discussions with claims, underwriting and operations teams help identify these explanations.

Data limitations should be documented rather than hidden behind a precise-looking result. Missing fields, inconsistent coding, short experience periods and changes in claims handling can materially affect the selected assumption. Where the evidence is weak, the reviewer should assess whether a proxy, external benchmark, credibility adjustment or explicit margin is appropriate.

Challenge Methods And Expert Judgement

A strong review examines the method used to derive each assumption. For experience-based assumptions, this may involve checking exposure definitions, weighting periods, outlier treatment, development patterns, trend selections and credibility procedures. For economic assumptions, it may include testing discount rates, inflation, salary growth, foreign exchange and investment returns against reliable external information.

The reviewer should reproduce enough of the analysis to understand how results were generated. Full model rebuilding is not always necessary, but key calculations, formulas, parameter selections and controls should be independently tested. Reperformance can uncover errors that a high-level discussion misses, particularly where spreadsheets, manual adjustments and data exports connect several systems.

Expert judgement needs the same discipline as statistical analysis. Judgement may be necessary where data is sparse, a product has changed, a regulatory intervention has occurred or a major catastrophe has distorted experience. The rationale should explain the evidence considered, the alternatives rejected, the people consulted and the expected effect on liabilities or capital.

Useful challenge questions focus on direction, scale and persistence. Is the selected trend consistent with observed claims and relevant economic indicators? Has a one-off event been treated as recurring? Is an allowance already captured elsewhere in the model? Would another reasonable actuary reach a materially different outcome? The reviewer should record responses and unresolved disagreements rather than smoothing them out in the final report.

For emerging work patterns and insurance needs, assumptions may need to reflect changing occupation and income patterns. A discussion of gig economy insurance can help teams consider how platform work, irregular earnings and labour shortages may affect exposure, affordability and claims behaviour.

Test Uncertainty, Interactions And Outcomes

Sensitivity testing should be proportionate to the assumption’s importance. One-way sensitivities show the effect of changing a single input, while scenario analysis explores combinations such as higher claims inflation, slower settlement, increased catastrophe frequency and weaker premium growth. Reverse stress testing can identify the movement needed to threaten a capital target, earnings plan or risk appetite limit.

Interactions between assumptions deserve special care. A claims inflation change may affect payment patterns, settlement costs, expense provisions and discounting. A lapse assumption can influence both future premiums and the mix of policyholders who remain insured. Reviewing each input in isolation can therefore understate the uncertainty of the overall result.

The review should compare selected assumptions with actual outcomes as they emerge. Back-testing may reveal persistent bias, delayed recognition of trends or excessive reliance on old data. It should be performed with care: an unfavourable result does not automatically prove that the original assumption was unreasonable, because random variation and new events can produce short-term differences.

For Australian portfolios, scenario design should reflect local risk drivers. Cyclone and flood events in Queensland, bushfire conditions in parts of New South Wales and Victoria, and repair-cost pressures in major cities may have different timing and severity patterns. A useful scenario is grounded in the portfolio’s actual exposure and reinsurance structure, rather than copied from a generic international template.

Results should be translated into business language. Senior management needs to know the estimated financial effect, the key uncertainties, the likelihood of change and the decisions that may be affected. A technical appendix can preserve detail, while an executive summary should make clear whether the assumption is supported, supported with limitations, or requires remediation.

Document Findings And Embed Governance

A peer review report should provide a traceable record of what was examined and why. It should describe the scope, data reviewed, methods tested, limitations, key findings, materiality assessment, sensitivities and recommended actions. Each finding should identify an owner and a target date, with priority based on financial impact, control weakness and regulatory significance.

The report should distinguish between observations and formal recommendations. An observation may suggest improved documentation, while a recommendation may require revised assumptions, additional controls or a repeat analysis. Clear wording helps committees understand whether they are being asked to note an issue, approve a change or accept a documented risk.

Governance does not finish when the report is issued. The accountable actuary should track remediation, retain evidence and confirm whether changes were incorporated into valuation, pricing, capital or financial reporting processes. Material disagreements should be escalated through the appropriate actuarial, risk, finance or board committee rather than left in working papers.

Peer review quality also improves when professionals compare practices across the market. Events such as the IASA Conference programme bring together insurance finance, accounting, technology, risk and actuarial perspectives, creating useful opportunities to discuss controls, reporting expectations and practical approaches with industry colleagues.

Practical Actions For A Reliable Review

A consistent process makes it easier to repeat the work each valuation cycle and to demonstrate that challenge is meaningful. The following actions provide a practical foundation:

An effective review is proportionate, sceptical and constructive. It does not aim to replace the responsible actuary’s judgement or produce false precision. Its purpose is to show that the judgement was formed from relevant evidence, challenged by an appropriately independent professional and communicated clearly to decision-makers.

Use these practices to strengthen the next valuation cycle, improve conversations between actuarial, finance, risk and operations teams, and create a review file that stands up to scrutiny. In a market where assumptions can shift quickly, disciplined peer review gives Australian insurers a clearer basis for reporting, planning and responding to risk.