IAI SP7 common mistakes
IAI SP7 · Understand the traps and how to avoid them.
Selecting a single chain-ladder result for the latest accident years and presenting it as the reserve estimate, without computing a Bornhuetter–Ferguson or expected-loss benchmark for immature years.
Candidates often memorise the mechanics of the chain-ladder method and treat it as the default answer, overlooking that it gives excessive weight to thin, immature data when case reserves or reported counts are unreliable.
What to do instead
Compute both a development-based estimate and an expected-loss benchmark, compare the implied ultimate loss ratios, and document why one method is preferred for each accident year. Explain the selection using case reserve adequacy, large-loss treatment and changes in settlement patterns.
Applying a paid development pattern to incurred data, or mixing paid and incurred development factors without reconciling the implied case reserve estimate.
Paid and incurred methods look interchangeable in exam conditions, and candidates often focus on producing a number rather than checking whether the data and method are aligned.
What to do instead
Project paid and incurred data separately, reconcile the difference to the current case reserve, and test whether the implied case reserve adequacy is credible. If mixing factors, show the reconciliation and explain the assumption.
Using historical development factors without adjusting for claims inflation, or applying a single average factor across accident years with different settlement regimes and court award trends.
Candidates reproduce the arithmetic from the development triangle but forget that development factors embed historical inflation, superimpose inflation, and changes in claims handling.
What to do instead
Restate historical payments to current cost levels, use accident-year-specific inflation assumptions where material, and test sensitivity to superimposed inflation. Explain how inflation affects both the central estimate and the tail.
Reporting IBNR as a single point estimate and omitting a range, the tail beyond the observed development period, and the volatility from large claims or latent exposures.
Exam answers often reward a tidy number, so candidates stop at the central estimate and do not demonstrate an understanding of reserving uncertainty.
What to do instead
Present the central estimate alongside a range from stochastic or scenario analysis, explicitly quantify tail and large-claim volatility, and state the limitations of the chosen range. Link the uncertainty discussion to reserve risk capital.
Treating the SCR as a single formula output and failing to reconcile the internal model's risk modules to the regulatory capital requirement or to explain the diversification benefit.
Candidates memorise the list of risk modules but do not practise the aggregation step, so they cannot explain how module-level capital becomes a company-level SCR.
What to do instead
Show the module-level capital figures, the correlation assumptions or copula used, the diversification credit, and a reconciliation to the regulatory return. Explain why diversification is not a free lunch and how it is constrained.
Stating risk appetite only in qualitative terms (for example, 'we are conservative') without linking it to capital metrics such as probability of ruin, target rating, dividend capacity or SCR coverage.
Risk appetite is often learned as a governance concept rather than a quantitative framework, and candidates do not connect it to the reserving and capital numbers in the question.
What to do instead
Quantify the tolerance: state the target SCR coverage, acceptable probability of ruin, rating agency capital target and dividend constraints. Show how the risk appetite drives the choice of capital model, reinsurance structure and reserve margin.
Modelling reinsurance recoveries at face value without allowing for counterparty default, disputes, reinstatement premiums or the timing of recoveries.
Candidates focus on the gross-to-net arithmetic and treat reinsurance as a pure risk transfer, forgetting that recoveries are credit exposures and may be delayed or disputed.
What to do instead
Apply a credit risk adjustment for counterparty default, model reinstatement premiums and aggregate limits explicitly, and reflect recovery timing in the capital and reserve calculations. Explain the effect on both best-estimate reserves and the SCR.
Assuming every reinsurance treaty provides a full capital credit without testing whether it transfers significant insurance risk, including deposit accounting, experience accounts, side letters or financing elements.
Candidates equate the existence of a reinsurance contract with risk transfer and do not consider that some structures are financing arrangements that should not reduce regulatory capital.
What to do instead
Test each treaty for meaningful risk transfer, identify financing or deposit components, and adjust the capital credit accordingly. Explain how the accounting and regulatory treatment differs from the legal form.
Answering a reserving or capital question without citing the applicable IRDAI solvency regime, the Indian GAAP or IFRS 17 presentation, and the relevant IAI actuarial standards.
Candidates learn the technical methods in isolation and do not integrate the regulatory and professional framework, so the answer lacks jurisdictional context.
What to do instead
State the applicable regulatory basis explicitly, explain the key differences between the reporting bases where relevant, and reference the professional standards governing the work. Show how the regulatory regime changes the reserve or capital conclusion.
Writing a board or audit committee note that presents technical results without explaining the key judgments, data limitations, method choices and sensitivity to a non-technical audience.
Candidates demonstrate technical competence but not communication competence, and they assume the reader shares their technical background.
What to do instead
Structure the note with purpose, data quality, methods and assumptions, results, uncertainty and recommendations. Translate technical terms into plain language, highlight the three or four judgments that most affect the result, and state what would change the conclusion.