IFoA SP7 cheat sheet

Indemnity principle: the insured is restored to pre-loss financial position, not better off; valued policies (agreed value) are the exception and invite moral hazard.

Property covers: fire, business interruption (BI), engineering/construction, household, commercial property; perils may be named or all-risks with exclusions.

Liability covers: employers', public/products, professional indemnity, directors' and officers'; occurrence vs claims-made trigger is the key distinction.

Motor: compulsory third-party bodily injury in most jurisdictions; own-damage and theft are optional; no-claims discount (NCD) is a credibility-weighted experience rating.

Marine/aviation/transport and energy are specialty lines with high severity, low frequency, and heavy reliance on reinsurance and survey data.

Health and personal accident: short-tail, high-frequency, morbidity-driven; often community-rated with regulatory constraints on underwriting.

Market cycle: soft markets erode rates and widen terms; hard markets follow large losses or capital shocks; cycle length varies by line and jurisdiction.

Distribution channels (brokers, agents, direct, MGA/MGU, bancassurance) change acquisition cost, data quality and underwriting control.

Regulatory environment: licensing, conduct rules, policyholder protection schemes, compulsory covers and price regulation shape product design.

Pure premium = expected frequency x expected severity; loaded premium adds expenses, profit, contingency and reinsurance cost.

Burning cost = incurred claims (net of recoveries) / exposure, used for short-tail and reinsurance pricing.

Credibility: Z = sqrt(n / n0) or Z = n / (n + k); complement of credibility applied to a broader class or prior.

Frequency-severity models: Poisson or negative binomial for counts; lognormal/gamma/Pareto for severity; aggregate loss via Panjer recursion or simulation.

GLM pricing: link function g(mu) = X*beta; Tweedie with power 1<p<2 handles zero-inflated pure premium; offset for log exposure.

Risk factors must be statistically significant, legally permissible, and not unfairly discriminatory; proxy discrimination risk from correlated variables.

Large loss and catastrophe loading: separate attritional from large and CAT losses; CAT models give exceedance probability (EP) curves.

Underwriting cycle management: technical vs market price; combined ratio = loss ratio + expense ratio; underwriting result = premium minus claims and expenses.

Exposure bases: payroll, turnover, vehicle-years, property sum insured, or written/earned premium; choice affects trend and on-level factors.

Chain-ladder: development factor f_k = sum(C_{i,k+1}) / sum(C_{i,k}); ultimate = latest cumulative x product of remaining factors.

Bornhuetter-Ferguson: ultimate = reported + expected unreported, where expected unreported = prior ultimate x (1 - 1/percent developed).

Mack chain-ladder gives standard error of reserves without distributional assumption; bootstrap (ODP) gives full predictive distribution.

IBNR vs IBNER: IBNR covers claims not yet reported; IBNER is development on reported claims; together they form the outstanding claims reserve.

Case reserves are individual estimates; they can be inadequate (under-reserved) or redundant (over-reserved); test via development on reported counts.

Tail factors extend beyond the observed triangle; critical for liability, professional indemnity and latent claims with long reporting lags.

Discounting: reserves discounted for solvency and accounts; rate choice (risk-free, locked-in) materially changes liability value.

Claims inflation: separate from economic inflation; social, judicial and medical inflation often exceed CPI; adjust triangles before projection.

Uncertainty: process variance (random) vs parameter/estimation variance (model); both must be combined in a predictive distribution.

Proportional: quota share (fixed % of premium and losses) and surplus (share of sums insured above retention).

Non-proportional: excess of loss (per risk or per occurrence) and aggregate stop-loss; recoveries triggered when loss exceeds attachment.

Reinstatements: additional cover after a loss, often paid pro rata; free or paid reinstatements affect the effective rate on line.

Rate on line = reinsurance premium / limit; expected loss ratio = expected recoveries / premium.

Burning cost and exposure rating: experience-based vs exposure-based methods; blend with credibility.

Alternative risk transfer: cat bonds, ILWs, sidecars, captives and parametric covers; basis risk is the key drawback of parametric triggers.

Facultative vs treaty: facultative is per-risk, treaty is portfolio-wide; treaties may include hours clauses and event definitions.

Solvency credit: reinsurance reduces required capital but introduces counterparty credit risk and recovery lag.

Solvency II structure: SCR = min capital to survive a 1-in-200 one-year event; MCR is a lower floor; own funds split into tiers.

SCR modules: non-life underwriting, market, counterparty default, life, health; aggregation via correlation matrix; operational risk added.

Risk margin = cost of capital (6% CoC under Solvency II) on projected SCR run-off; increases with duration.

Best Estimate Liabilities (BEL) plus risk margin = technical provisions; BEL is discounted expected cash flows.

Available vs eligible own funds: surplus assets over liabilities; tiering limits how much lower-quality capital counts.

ORSA: insurer's own assessment of risk and solvency, forward-looking, approved by the board.

IFRS 17: contractual service margin (CSM) released over coverage; risk adjustment for non-financial risk; onerous contracts recognised immediately.

Capital modelling: copula or aggregate simulation; diversification benefit from imperfect correlation; allocation via Euler or Shapley.

Regulatory capital vs economic capital: regulatory is formula-driven and comparable; economic reflects firm's own risk profile.

Duration gap: mismatch between asset and liability durations creates interest rate sensitivity; immunisation sets PV and duration equal.

Short-tail lines (motor, property) suit short-duration bonds; long-tail lines (liability, PI) benefit from longer duration and inflation-linked assets.

Liquidity: claims can be lumpy after a catastrophe; hold liquid assets or committed facilities to avoid forced selling.

Credit risk: spread widening increases unrealised losses and may trigger capital charges; default risk in reinsurance recoveries.

Currency matching: match assets to liabilities by currency to avoid FX translation risk; hedging via forwards or swaps.

Return on capital: investment return contributes to overall profitability; risk-adjusted return on capital (RAROC) compares lines.

Regulatory constraints: localisation rules, admissible asset lists, concentration limits and matching requirements.

ALM in a hard market: rising rates reduce bond values but increase reinvestment yield; effect on reserves and capital must be modelled jointly.

ERM framework: risk identification, assessment, treatment, monitoring and reporting; embedded in strategy and decision-making.

Risk appetite statement: qualitative and quantitative tolerances; links to limits, capital and underwriting authority.

Three lines model: business owns risk; risk/compliance oversight; internal audit provides independent assurance.

Board responsibilities: approve risk appetite, ORSA, capital plan; senior managers execute; actuarial function provides technical advice.

Risk register: likelihood x impact scoring; heat maps; emerging risks (cyber, climate, pandemic) tracked separately.

Control environment: segregation of duties, underwriting authority limits, claims triage, fraud detection, data governance.

Stress and scenario testing: reverse stress tests identify scenarios that breach solvency; used in ORSA and capital planning.

Conduct risk: treating customers fairly, product governance, complaints handling; failures create reputational and financial loss.

Data quality: completeness, accuracy, consistency, timeliness; reconcile to financial ledgers and policy admin systems.

Granularity: policy-level vs aggregate; more granularity improves segmentation but increases noise and overfitting risk.

Missing data: imputation (mean, regression, multiple imputation); document assumptions and test sensitivity.

Model validation: back-testing, out-of-sample testing, benchmarking against simpler models, expert review.

Overfitting: too many parameters relative to data; use cross-validation, regularisation (ridge/lasso) and parsimony.

GLM diagnostics: residual plots, deviance, AIC/BIC; check link function and distributional fit.

Machine learning: gradient boosting and random forests can outperform GLMs but reduce interpretability; regulators may require explainability.

Judgement: model output is an input, not the answer; document overrides, rationale and sensitivity to key assumptions.

Emerging data: telematics, IoT and satellite imagery change risk assessment; privacy and consent constraints apply.

Actuaries' Code: integrity, competence and care, compliance, and communication; apply professional judgement and be transparent.

Conflicts of interest: identify, disclose and manage; independence of the actuarial function is a regulatory expectation.

Peer review: independent check of assumptions, methods and results before advice is relied upon.

Communication: tailor to audience (board, regulator, client); state purpose, scope, data, methods, assumptions, uncertainty and limitations.

Whistleblowing: report material concerns (e.g. inadequate reserves) through proper channels, including the regulator if needed.

Technical provisions opinions and reserve reports: state reliance on data and others, and any limitations on scope.

Ethical dilemmas: pressure to smooth results, understate reserves or misprice; document concerns and escalate.

Continuing professional development: maintain competence in relevant practice areas; record CPD activity.

Reinsurance market capacity: global capital and cat bond issuance influence pricing and terms in specialty lines.

Lloyd's model: syndicates, managing agents and members; capital backing and central fund provide policyholder security.

Mutuals vs stock insurers: mutuals focus on member value; stock insurers on shareholder return; governance and capital differ.

Distribution and commission: broker commission and MGA profit share affect net premium and underwriting discipline.

Catastrophe exposure: concentration in coastal, flood and wildfire zones drives accumulation management and cat modelling.

Regulatory change: IFRS 17, Solvency II reform and climate disclosure requirements reshape reporting and capital.

Claims environment: legal costs, litigation trends and fraud rates affect severity and reserving.

Competition and pricing: price comparison websites and direct channels compress margins and increase price sensitivity.

Experience rating: manual rate adjusted by individual loss experience; credibility weighting determines weight on own experience.

Retrospective rating: premium depends on actual losses within a range; used for large commercial risks.

Deductibles and limits: change frequency and severity; model via truncated/limited distributions and deductible factors.

Catastrophe pricing: exposure-based using EP curves; return period and probable maximum loss (PML) inform pricing.

Segmentation: balance homogeneity, credibility, cost and regulatory acceptability; too fine splits create unstable rates.

Rate monitoring: on-level factors adjust historical premiums to current rate level; needed for trend analysis.

Expense loading: fixed vs variable expenses; commission, admin, claims handling and premium taxes.

Profit and contingency load: target return on capital, cost of capital and uncertainty margin.

Underwriting authority: limits by size, class and risk type; escalation to senior underwriters for complex risks.

Triangle diagnostics: check for changes in case reserving, mix of business, settlement rates and claims handling.

Averages: simple average, volume-weighted average and average of ratios; volume-weighted is standard chain-ladder.

Expected loss ratio method: ultimate = exposure x expected loss ratio; suitable for immature years with little data.

Cape Cod method: expected loss ratio applied to exposure, adjusted by actual development; blends BF and chain-ladder.

Case outstanding development: project case reserves separately from IBNR; useful when case adequacy changes.

Frequency-severity reserving: project reported counts and average severities separately; captures changes in mix.

Large losses: separate large claims from attritional; model large claims individually or via a severity distribution.

Reserve risk: variability of ultimate; model via stochastic methods and quantify at a chosen confidence level.

Actuarial judgement: select ultimate by combining methods with weights; document rationale and sensitivity.

Treaty terms: retention, limit, co-participation, exclusions, hours clause, reinstatements and profit commission.

Proportional treaty: ceding commission covers acquisition costs; sliding scale shares profit with reinsurer.

Excess of loss layers: attachment point and limit define layer; expected loss to layer from severity distribution.

Aggregate stop-loss: protects against accumulation of losses over a period; trigger on aggregate ratio.

Catastrophe cover: per-event excess of loss; event definition and hours clause critical for accumulation events.

Reinsurance credit risk: collateral, trust funds, letters of credit mitigate counterparty risk.

Retrocession: reinsurance of reinsurance; adds layers of counterparty risk.

Captive reinsurance: parent-owned insurer; regulatory and tax considerations; capital efficiency.

Parametric triggers: payout based on index (wind speed, earthquake magnitude); fast settlement but basis risk.

Solvency II balance sheet: assets at market value; technical provisions = BEL + risk margin; own funds = assets minus liabilities.

SCR standard formula: modular approach with correlation matrix; undertaking-specific parameters possible with approval.

Internal model: full or partial; requires use test, statistical quality, calibration and validation standards.

MCR: linear formula based on premiums, claims and technical provisions; triggers regulatory intervention.

Capital tiers: Tier 1 (equity, retained earnings), Tier 2 (subordinated debt), Tier 3 (deferred tax assets); limits by tier.

IFRS 17 groups: portfolios split into onerous, no-significant-possibility-of-loss and others; CSM released over coverage period.

Risk adjustment: compensation for non-financial risk; confidence level or cost of capital approach.

Dividend and capital management: maintain buffer above SCR; consider volatility, growth and rating agency requirements.

Group supervision: solo vs group solvency; intra-group transactions and fungibility of capital.

Cash flow matching: match asset cash flows to expected claim payments; exact matching rare due to uncertainty.

Duration and convexity: duration measures first-order sensitivity; convexity captures curvature; both matter for large rate moves.

Credit spread risk: widening spreads reduce asset values and increase capital charges; default risk in corporate bonds.

Inflation-linked assets: index-linked bonds match liability inflation exposure; limited supply in some markets.

Equity and property: higher expected return but volatile; capital charges and liquidity constraints.

Strategic asset allocation: set by risk appetite, liability profile and capital; reviewed regularly.

Tactical asset allocation: short-term deviations from strategy; governed by limits and delegated authority.

Derivatives: interest rate swaps, FX forwards and options for hedging; counterparty and basis risk.

Liquidity management: hold cash and high-quality liquid assets; stress test for catastrophe claim surges.

Risk taxonomy: insurance, market, credit, liquidity, operational, strategic, reputational and emerging risks.

Risk appetite metrics: capital adequacy ratio, earnings volatility, combined ratio, concentration limits.

Risk committees: board risk committee, audit committee, underwriting committee; clear terms of reference.

Actuarial function: coordinate technical provisions, opine on underwriting and reinsurance, contribute to risk management.

Internal controls: underwriting authority, claims authority, IT access controls, reconciliation and audit trails.

Outsourcing: due diligence, contract terms, monitoring and contingency plans; regulator notification.

Business continuity: disaster recovery, cyber resilience, pandemic planning.

Culture: tone from the top, speak-up channels, incentives aligned with risk appetite.

Climate risk: physical and transition risk; scenario analysis and disclosure requirements.

Data sources: policy admin, claims, finance, external (credit, weather, geospatial); reconcile and document.

Data transformations: exposure adjustments, on-leveling, trend, development; document each step.

Outliers: identify and treat (cap, remove, investigate); document treatment and impact.

Model selection: compare AIC/BIC, deviance, cross-validation; parsimony and interpretability.

Regularisation: ridge (L2), lasso (L1) and elastic net for high-dimensional data; shrink coefficients.

Machine learning: train/test split, cross-validation, hyperparameter tuning; monitor for drift.

Explainability: SHAP, partial dependence, feature importance; regulators may require interpretable models.

Model risk: governance, validation, documentation, version control and periodic review.

Judgement: expert opinion supplements data; document assumptions and test sensitivity.

Actuaries' Code principles: integrity, competence and care, compliance, communication.

Professional judgement: apply skill and care; document reasoning and limitations.

Independence: avoid conflicts; disclose any relationships that could influence advice.

Confidentiality: protect client information; disclose only with consent or legal requirement.

Peer review: independent review of significant advice before reliance.

Reporting: state scope, data, methods, assumptions, results, uncertainty and limitations.

Regulatory reporting: actuarial opinions on technical provisions; comply with local requirements.

Ethics: honesty, fairness, transparency; escalate concerns about inadequate reserves or mispricing.

CPD: maintain and record continuing professional development.

Product lifecycle: design, pricing, underwriting, claims, review and withdrawal; product governance throughout.

Terms and conditions: exclusions, warranties, conditions precedent; clarity reduces disputes.

Regulatory conduct: fair treatment of customers, transparent pricing, complaints handling.

Market concentration: large players vs niche specialists; MGA growth changes distribution.

Reinsurance cycles: capacity constraints after large losses; pricing and terms tighten.

Climate change: increased frequency/severity of weather events; underwriting and pricing response.

Cyber risk: growing line; accumulation and systemic risk concerns; limited historical data.

Pandemic risk: business interruption and event cancellation; coverage disputes and exclusions.

Technology: automation, AI underwriting and claims; data privacy and bias concerns.

Loss ratio = incurred claims / earned premium; expense ratio = expenses / written premium; combined ratio = sum.

Pure premium method: expected losses / exposure; load for expenses and profit.

Frequency-severity: separate models for claim counts and average claim size; combine for pure premium.

GLM: distribution (Poisson, gamma, Tweedie), link (log), offset (log exposure); interpret coefficients as multiplicative.

Credibility: limited fluctuation and greatest accuracy; Bayesian methods for small data.

Catastrophe models: hazard, vulnerability, financial modules; output EP curves and PML.

Reinsurance pricing: burning cost, exposure rating, blended; load for expenses and profit.

Rate adequacy: compare indicated vs charged rate; monitor loss ratio trend.

Segmentation: balance homogeneity, credibility and regulatory acceptability.

Chain-ladder assumptions: stable development pattern, no calendar-year effects; test via diagnostics.

Bornhuetter-Ferguson: robust for immature years; relies on prior loss ratio.

Mack model: distribution-free standard error; assumes independence across accident years.

Bootstrap: resample residuals to generate predictive distribution; ODP scale parameter.

Tail factor: estimate from benchmark or curve fit; critical for long-tail lines.

Discounting: use risk-free rate; consider tax and inflation.

Case reserving: individual estimates; monitor adequacy via case outstanding development.

Large losses: separate treatment; model individually or via severity distribution.

Uncertainty: process and parameter variance; combine for predictive distribution.

Quota share: cedes fixed proportion; reduces volatility and capital; ceding commission covers expenses.

Surplus: cedes proportion above retention; proportional to sum insured.

Excess of loss: per risk or per occurrence; attachment and limit define layer.

Aggregate stop-loss: protects aggregate loss ratio; trigger on cumulative losses.

Reinstatements: free or paid; affect effective cost.

Rate on line: premium / limit; compare across layers.

Burning cost: historical recoveries / exposure; load for trend and expenses.

Exposure rating: uses exposure curves to estimate expected loss to layer.

Alternative risk transfer: cat bonds, ILWs, captives; basis risk for parametric.

Solvency II: SCR 1-in-200 one-year VaR; MCR lower floor.

Own funds: available and eligible; tiering limits.

Risk margin: 6% cost of capital on projected SCR.

BEL: discounted expected cash flows; includes expenses and reinsurance.

Standard formula: modular with correlation matrix; simplifications available.

Internal model: use test, statistical quality, calibration, validation.

ORSA: forward-looking assessment; board approval.

IFRS 17: CSM, risk adjustment, onerous contracts.

Capital allocation: Euler, Shapley; diversification benefit.

Duration gap: asset duration minus liability duration; immunisation sets equal.

Convexity: second-order sensitivity; matters for large rate moves.

Liquidity: hold liquid assets for catastrophe claims.

Credit risk: spread widening and default; capital charges.

Currency matching: match assets and liabilities by currency.

Inflation: index-linked bonds match liability inflation.

Strategic asset allocation: set by risk appetite and liability profile.

Derivatives: hedging; counterparty and basis risk.

Stress testing: scenario analysis for market and liquidity risk.

ERM: identify, assess, treat, monitor, report.

Risk appetite: qualitative and quantitative; links to limits.

Three lines: business, risk/compliance, internal audit.

Board: approve risk appetite, ORSA, capital plan.

Risk register: likelihood x impact; emerging risks.

Controls: segregation of duties, authority limits, fraud detection.

Stress testing: reverse stress tests; ORSA.

Conduct risk: fair treatment, product governance.

Culture: tone from top, speak-up channels.

Data quality: completeness, accuracy, consistency, timeliness.

Granularity: policy-level vs aggregate; overfitting risk.

Missing data: imputation; document assumptions.

Model validation: back-testing, out-of-sample, benchmarking.

Overfitting: cross-validation, regularisation, parsimony.

GLM diagnostics: residuals, deviance, AIC/BIC.

Machine learning: gradient boosting, random forests; explainability.

Judgement: document overrides and sensitivity.

Emerging data: telematics, IoT; privacy constraints.

Actuaries' Code: integrity, competence, compliance, communication.

Conflicts of interest: identify, disclose, manage.

Peer review: independent check.

Communication: tailor to audience; state scope, data, methods, assumptions, uncertainty, limitations.

Whistleblowing: report material concerns.

Reserve reports: state reliance and limitations.

Ethical dilemmas: document and escalate.

CPD: maintain competence.

Product design: cover, limits, deductibles, exclusions, pricing.

Underwriting: risk selection, pricing, terms.

Claims: triage, investigation, settlement.

Distribution: brokers, agents, direct, MGA.

Regulation: licensing, conduct, capital.

Market cycle: soft/hard; capacity.

Catastrophe exposure: concentration, modelling.

Climate: physical and transition risk.

Technology: automation, AI, data privacy.

Pure premium = frequency x severity.

Burning cost = incurred claims / exposure.

Credibility: Z = sqrt(n/n0) or n/(n+k).

GLM: link, distribution, offset.

Tweedie: 1<p<2 for zero-inflated.

Risk factors: significant, permissible, not unfairly discriminatory.

Large loss and CAT loading.

Combined ratio = loss ratio + expense ratio.

Exposure bases: payroll, turnover, vehicle-years.

Chain-ladder: development factors; ultimate.

BF: reported + expected unreported.

Mack: standard error; distribution-free.

Bootstrap: predictive distribution.

IBNR vs IBNER.

Case reserves: individual estimates.

Tail factors: long-tail lines.

Discounting: risk-free rate.

Claims inflation: social, judicial, medical.

Proportional: quota share, surplus.

Non-proportional: excess of loss, stop-loss.

Reinstatements: free or paid.

Rate on line = premium / limit.

Burning cost and exposure rating.

Alternative risk transfer: cat bonds, ILWs, captives.

Facultative vs treaty.

Solvency credit: reduces capital; counterparty risk.

Hours clause: event definition.

SCR: 1-in-200 one-year VaR.

MCR: lower floor.

Risk margin: 6% CoC.

BEL: discounted expected cash flows.

Own funds: available vs eligible; tiers.

ORSA: forward-looking; board approval.

IFRS 17: CSM, risk adjustment, onerous contracts.

Capital modelling: copula, simulation; diversification.

Regulatory vs economic capital.