Use this guide to connect actuarial assumptions, calculations and financial decisions associated with the requested CAS Exam 7 title. Begin with claims data and reserving methods, then study uncertainty, valuation, risk management and capital. Each concept includes an original worked example and a specific error to avoid. Examples use simplified monetary units and state the assumptions needed for their conclusions.
Unpaid claims estimation
1. Incremental and cumulative development data
Incremental claims record activity during a development period; cumulative claims total activity through that period. A development triangle organizes comparable origin periods at successive ages. Convert between the two before applying a method, and distinguish an unobserved future cell from an observed zero.
Worked example: Payments of 40, 25 and 15 produce cumulative payments of 40, 65 and 80. The third incremental payment is 80 − 65 = 15.
Mistake to avoid: Treating blank future cells as zero payments distorts development patterns.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
2. Ultimate claims and unpaid liability components
Ultimate claims equal amounts already paid plus amounts still unpaid. When reported incurred claims equal paid claims plus case reserves, the difference between estimated ultimate and reported incurred is an aggregate IBNR provision, potentially including development on known claims. State how the data define IBNR.
Worked example: Ultimate claims of 150, paid claims of 90 and case reserves of 40 imply unpaid claims of 60 and aggregate IBNR of 20.
Mistake to avoid: Adding case reserves again after calculating ultimate minus paid double-counts them.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
3. Reconciling claims data before modeling
Reconcile triangle totals to an independently controlled extract or ledger on the same valuation date and basis. Check currency, recoveries, claim expenses, reopened claims and transaction cutoffs. A discrepancy is a data question before it becomes a development-factor question; otherwise a model can project an extraction error.
Worked example: Paid claims of 72 plus case reserves of 28 give incurred claims of 100. An incurred extract showing 97 requires investigation of the missing 3.
Mistake to avoid: Changing selected factors to conceal an unexplained reconciliation difference.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
4. Segmentation and credible development patterns
Group claims with reasonably similar reporting, settlement and coverage characteristics. Excessive aggregation can hide changing business mix, while excessive segmentation produces unstable estimates. Compare patterns and volumes together; a segment should have both a meaningful operational distinction and enough information for the chosen method.
Worked example: Fleet liability claims settle more slowly than private motor claims. A growing fleet share can increase an aggregate development factor even when both segment patterns remain unchanged.
Mistake to avoid: Interpreting a mixture change as deterioration within every segment.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
5. Volume-weighted development factors
A volume-weighted age-to-age factor divides the sum of later cumulative amounts by the sum of corresponding earlier amounts. It weights individual ratios by their earlier claim amounts. Use matched observations and inspect influential origin periods; the arithmetic average answers a different weighting question.
Worked example: Two periods develop from 100 to 120 and from 200 to 300. The volume-weighted factor is 420 ÷ 300 = 1.40; the simple average is 1.35.
Mistake to avoid: Averaging ratios while describing the result as volume-weighted.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
6. Chain-ladder projection to ultimate
Chain ladder multiplies the latest cumulative amount by the remaining selected development factors. Its central assumption is that historical development relationships remain applicable to the projected cohorts. Separate the cumulative development factor from individual age-to-age factors, and calculate unpaid claims using the appropriate paid basis.
Worked example: Latest paid claims of 90, a remaining factor of 1.50 and a tail factor of 1.10 give ultimate claims of 148.50 and unpaid claims of 58.50.
Mistake to avoid: Adding successive development factors instead of multiplying them.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
7. Tail development beyond observed ages
A tail factor represents development after the last modeled age. Select it using mature experience, suitable external comparisons or a defensible extrapolation, then assess sensitivity. Small percentage changes can materially affect long-tail liabilities because the factor applies to a substantial cumulative amount.
Worked example: Claims developed to the last modeled age total 200. A tail factor of 1.03 produces ultimate claims of 206, adding 6 beyond that age.
Mistake to avoid: Assuming the oldest observed development age necessarily represents final settlement.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
8. Bornhuetter-Ferguson estimation
Bornhuetter-Ferguson combines observed claims with an independently selected expected ultimate amount for the portion not yet emerged. On a paid basis, unpaid claims equal expected ultimate multiplied by the unpaid proportion. The expected ultimate assumption and development pattern both matter, especially for immature cohorts.
Worked example: Paid claims are 40, expected ultimate is 120 and the expected paid proportion is 25%. Estimated unpaid claims are 120 × 75% = 90; ultimate is 130.
Mistake to avoid: Multiplying observed paid claims by the unpaid proportion.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
9. Cape Cod calibration
A basic Cape Cod approach estimates an expected loss ratio from claims divided by development-adjusted premium, then uses that ratio in a Bornhuetter-Ferguson calculation. Premium must be comparable across cohorts. Rate changes, exposure shifts and claim trends can otherwise make the pooled ratio misleading.
Worked example: Paid claims of 180 divided by adjusted premium of 240 give a 75% ratio. Another cohort has premium 100, paid claims 10 and 20% paid emergence; its unpaid estimate is 60.
Mistake to avoid: Pooling unadjusted premiums across materially different rate levels.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
10. Frequency-severity estimation
Estimate unpaid claims through the expected number of future payments or claims and their expected severity. Define the count carefully: outstanding reported claims, unreported claims and reopened claims may need separate assumptions. Severity must correspond to the same count definition and payment basis.
Worked example: Thirty outstanding reported claims and twenty expected unreported claims each have expected remaining cost of 2. The estimated unpaid amount is 50 × 2 = 100.
Mistake to avoid: Multiplying outstanding claim counts by ultimate severity that includes payments already made.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
11. Operational changes in paid and incurred patterns
Settlement speed and case-reserving practice can change development without changing underlying ultimate costs. Compare paid development, incurred development, closure rates and case reserves together. A faster payment process may invalidate historical paid factors; stronger case reserves may similarly distort incurred factors.
Worked example: Payments accelerate while incurred claims and estimated ultimate costs remain stable. Applying old paid factors to the larger paid amount would overstate ultimate claims if the acceleration persists.
Mistake to avoid: Treating every increase in early paid claims as higher ultimate severity.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
12. Reserve rollforwards and estimate changes
For a fixed set of claims, opening unpaid liability minus payments plus changes in estimated ultimate equals closing unpaid liability. This reconciliation separates cash settlement from revisions to expected total cost. Keep new origin periods, transfers and other scope changes outside the comparison or identify them explicitly.
Worked example: Opening unpaid claims are 70, payments are 35 and closing unpaid claims are 50. The increase in estimated ultimate is 50 + 35 − 70 = 15.
Mistake to avoid: Calling the entire decline in unpaid reserves favorable development without accounting for payments.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
Stochastic reserving and statistical models
13. Process, parameter and model uncertainty
Process uncertainty concerns random future outcomes even with known parameters. Parameter uncertainty concerns estimated quantities such as development factors. Model uncertainty concerns the adequacy of the chosen structure. A predictive distribution should identify which sources it includes; changing the model can matter more than refining parameter estimates.
Worked example: Independent process variance of 16 and parameter variance of 9 give predictive variance of 25 and standard deviation of 5, before any allowance for model uncertainty.
Mistake to avoid: Presenting process variability alone as total reserve uncertainty.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
14. Mack-style chain-ladder uncertainty
The Mack framework quantifies chain-ladder uncertainty using assumptions about conditional development means, variances and independence between origin periods. It does not require a complete distributional specification. Its results remain conditional on the development structure; common calendar shocks can undermine the independence assumption.
Worked example: A judicial change affects several origin periods simultaneously. A calculation assuming independent origin periods can miss the shared source of reserve variation.
Mistake to avoid: Interpreting a distribution-free moment framework as protection against incorrect development assumptions.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
15. Overdispersed Poisson reserving models
An overdispersed Poisson model can relate expected incremental claims to origin and development effects, often through a log link. Its variance is proportional to its mean through a dispersion parameter. This is a mean-variance modeling framework, not a claim that monetary payments are literal Poisson counts.
Worked example: A fitted cell has expected incremental claims of 80 and dispersion of 3. Its modeled variance is 3 × 80 = 240.
Mistake to avoid: Using Poisson variance equal to the mean after estimating material overdispersion.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
16. Log links and exposure offsets
A log link makes explanatory effects multiplicative on the response mean. A log-exposure offset fixes exposure proportionality rather than estimating its coefficient. Distinguish counts, rates and severities before adding an offset; their denominators and interpretations differ.
Worked example: A count model predicts 20 claims for 100 exposure units. With unchanged risk characteristics, an exposure offset predicts 40 for 200 units. A coefficient of 0.20 implies a multiplier of approximately 1.221.
Mistake to avoid: Reading a log-link coefficient of 0.20 as an additive increase of 0.20 claims.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
17. Calendar effects and identifiability
Calendar effects can capture inflation, legal changes or operational shifts affecting several cohorts at once. Origin, development and calendar coordinates are structurally related, so unrestricted effects cannot all be uniquely estimated. Identification constraints or a reduced structure are needed, and their substantive interpretation deserves scrutiny.
Worked example: Elevated payments appear along one calendar diagonal across several origin periods. This supports investigating a shared calendar influence instead of assigning a separate severity increase to each cohort.
Mistake to avoid: Estimating unrestricted origin, development and calendar trends as independently identifiable effects.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
18. Residual diagnostics
Residuals compare observed and fitted values on a scale appropriate to the model. Examine patterns by origin, development and calendar period, rather than relying only on an overall fit statistic. Clusters of similarly signed residuals suggest systematic structure the model has missed.
Worked example: Under a Poisson variance assumption, an observation of 100 with fitted mean 64 has Pearson residual (100 − 64) ÷ √64 = 4.5.
Mistake to avoid: Comparing raw residual sizes across cells with very different expected variability.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
19. Bootstrap predictive reserving
A residual bootstrap can generate pseudo-history, refit the model and project future claims. Adding future process draws produces a predictive distribution; refitting alone mainly reflects estimation uncertainty. Residual adjustment, influential cells and structural dependence affect whether resampling represents the actual risk.
Worked example: Refitted models produce future means of 90 and 110. Simulating future payments around each mean adds variability that a distribution containing only those two means would omit.
Mistake to avoid: Calling a bootstrap of fitted reserve means a complete predictive distribution.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
20. Model selection and predictive validation
Compare models using consistent likelihood definitions, fit diagnostics and genuinely withheld observations. Information criteria penalize unnecessary complexity but do not prove that extrapolation is reliable. Historical valuation backtests are useful when each fit uses only information available at that valuation.
Worked example: Comparable models have AIC values of 240 and 230. The second is preferred by AIC, but poor prediction of withheld calendar periods remains evidence against relying on it.
Mistake to avoid: Using information from later valuations when reconstructing an earlier forecasting exercise.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
21. Aggregating reserve uncertainty
The variance of aggregate reserves includes component variances and twice each pairwise covariance. Dependence can arise from inflation, shared case-reserving changes or common modeling assumptions. Calculate uncertainty on a consistent basis; adding standard deviations or assuming independence produces different answers.
Worked example: Two reserve estimates have variances of 25 and 36 and covariance of 12. Aggregate variance is 25 + 36 + 2 × 12 = 85.
Mistake to avoid: Discarding covariance simply because reserves are reported in separate business segments.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
22. Confidence intervals and predictive intervals
A confidence interval for an expected reserve concerns uncertainty in its estimated mean. A predictive interval concerns a future realized amount and also includes process variability under the model. Specify the target quantity and probability interpretation before comparing interval widths or communicating uncertainty.
Worked example: A model gives a mean-estimation interval of 90–110 and a predictive interval of 60–150 at the same nominal level. The wider interval addresses future payment outcomes.
Mistake to avoid: Using a confidence interval for the mean to describe the range of future claims.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
23. Bayesian updating and credibility
Bayesian estimation combines a prior distribution with a likelihood to produce a posterior. In a normal model with known variances, precision determines the weights. Prior relevance matters: a precise prior from a different risk environment can dominate the calculation while weakening its practical usefulness.
Worked example: A prior mean of 100 with variance 25 and an independent observation of 120 with variance 100 produce posterior mean 104: the observation receives 20% weight.
Mistake to avoid: Choosing credibility weights from sample size alone without considering variability and comparability.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
24. Monte Carlo precision in tail estimates
Simulation estimates introduce numerical uncertainty in addition to modeled insurance uncertainty. Tail quantities usually require more simulations than central quantities because fewer draws inform them. Compare independent runs or suitable error estimates, and distinguish numerical stability from confidence in the underlying model.
Worked example: With 10,000 draws, roughly 100 observations lie above a continuous distribution's 99th percentile. That limited tail sample can make an estimated tail average unstable.
Mistake to avoid: Assuming a smooth-looking simulated histogram establishes accurate extreme-tail estimates.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
Insurance valuation and financial analysis
25. Reserve adequacy and equity value
Understated claim liabilities overstate net assets when other balance-sheet items are unchanged. A valuation should assess reserve adequacy before interpreting reported equity or profitability. The effect on economic value can also depend on payment timing, taxes and recoverability, which must be modeled separately when relevant.
Worked example: Assets and other liabilities remain fixed. Increasing claim liabilities by 12 reduces equity from 80 to 68 before any tax effects.
Mistake to avoid: Treating reported book equity as reliable without examining the liabilities supporting it.
Reference: Casualty Actuarial Society
26. Accounting and economic measurement bases
Accounting measurements follow an applicable reporting framework; economic measurements reflect an explicitly chosen valuation basis. Differences can arise from discounting, risk allowances and asset measurement. Reconcile each difference rather than assuming one balance sheet can be converted into another with a single universal adjustment.
Worked example: An illustrative nominal claim estimate is 100 and its discounted cash-flow value is 92. The difference of 8 reflects timing and the selected discount basis, not automatically profit.
Mistake to avoid: Applying an economic discount adjustment as though every accounting framework requires it.
Reference: Casualty Actuarial Society
27. Discounting liability cash flows
Present value depends on both expected payment amounts and their timing. Discount each cash flow using a rate consistent with its period and valuation basis. Separate changes in expected claims from changes in discount assumptions so their financial effects remain interpretable.
Worked example: Expected payments of 55 in one year and 60.50 in two years, discounted at an illustrative annual 10%, have present value 50 + 50 = 100.
Mistake to avoid: Discounting the entire unpaid amount for one period when payments occur at several dates.
Reference: Casualty Actuarial Society
28. Discounted equity cash flows
An equity valuation discounts cash flows available to shareholders using a consistent required return on equity. Define distributions after necessary reinvestment and capital retention. Explicit forecast cash flows and any terminal value must use the same valuation date and ownership basis.
Worked example: Equity cash flows of 12 after one year and 13.20 after two years contribute 12 ÷ 1.10 + 13.20 ÷ 1.10² = 21.82 to value at a 10% return.
Mistake to avoid: Calling the explicit forecast contribution the full company value when later cash flows remain.
Reference: Casualty Actuarial Society
29. Earnings, retained capital and distributions
Accounting earnings do not necessarily equal cash available for shareholders. Growth, claim payments and capital needs can require retention. For an insurer, establish the cash-flow and financing definitions carefully before importing valuation formulas designed for nonfinancial businesses.
Worked example: An insurer earns 30 but must retain 18 to support its modeled business and capital needs. With no other adjustments, distributable cash is 12.
Mistake to avoid: Discounting all earnings as immediate dividends while also assuming capital-intensive growth.
Reference: Casualty Actuarial Society
30. Terminal growth valuation
A constant-growth terminal value equals the next period's sustainable equity cash flow divided by the required return minus growth. The required return must exceed growth. Sustainable distributions must reflect the capital needed to support that growth; the formula does not justify an arbitrary assumption.
Worked example: At the terminal date, next year's distributable cash is 11, required return is 10% and perpetual growth is 2%. Terminal value is 11 ÷ 0.08 = 137.50.
Mistake to avoid: Using the final forecast year's cash flow without advancing it to the next year.
Reference: Casualty Actuarial Society
31. Adjusted net asset valuation
An adjusted net asset approach values assets and liabilities on a specified basis and subtracts liabilities from assets. Examine asset impairment, reserve adequacy and items absent from reported balances. This approach describes existing net resources and may not capture the value of profitable future business.
Worked example: Assets of 250 and liabilities of 190 imply net assets of 60. Recognizing an asset impairment of 5 reduces adjusted net assets to 55.
Mistake to avoid: Adding an adjustment already reflected in the asset valuation.
Reference: Casualty Actuarial Society
32. Price-to-book comparisons
Price-to-book compares equity market value with book equity. Interpret the multiple alongside reserve quality, asset risk, expected returns and growth requirements. Comparable insurers can deserve different multiples because book value quality and future profitability differ.
Worked example: Market equity value of 120 divided by book equity of 80 gives a multiple of 1.50. Another insurer at 1.20 is not necessarily cheaper if its book liabilities are understated.
Mistake to avoid: Applying a peer multiple without checking whether accounting and reserve bases are comparable.
Reference: Casualty Actuarial Society
33. Underwriting performance and total returns
On a consistently defined earned-premium basis, the combined ratio compares claims and underwriting expenses with premium. Total shareholder profitability also depends on investment results, financing, taxes and equity employed. A favorable underwriting ratio alone does not determine return on equity.
Worked example: Earned premium is 200, claims are 120 and underwriting expenses are 50. The illustrative combined ratio is 85%, and underwriting profit is 30 before other items.
Mistake to avoid: Equating a 15% underwriting margin with a 15% return on equity.
Reference: Casualty Actuarial Society
34. Reinsurance recoverability in valuation
Gross claim obligations and expected reinsurance recoveries are separate economic quantities. Assess recovery amounts, timing and counterparty performance rather than treating nominal recoverables as certain cash. Use consistent gross and net bases throughout the valuation.
Worked example: Gross unpaid claims of 100 and nominal recoverables of 30 imply net claims of 70. If expected recoveries are only 25, the economic net amount becomes 75.
Mistake to avoid: Subtracting full nominal recoverables while separately counting the same counterparty impairment again.
Reference: Casualty Actuarial Society
Enterprise risk management
35. Risk capacity, appetite and tolerances
Risk capacity describes the ability to bear risk; appetite expresses the risk an organization chooses to accept. Tolerances translate appetite into operational limits. Define the metric, aggregation basis and time horizon before comparing a proposed exposure with any of these boundaries.
Worked example: An organization has capacity of 100, aggregate appetite of 70 and a business-unit tolerance of 45. A unit exposure of 50 breaches its tolerance even though it is below aggregate appetite.
Mistake to avoid: Treating unused organization-wide capacity as permission to exceed a unit limit.
Reference: Casualty Actuarial Society
36. Risk registers linked to objectives
A useful risk register links each risk to an objective, cause, consequence, owner and response. Broad categories organize discussion but should not replace concrete exposure descriptions. Distinguish the initiating event from its financial effects so overlapping entries do not inflate aggregated risk.
Worked example: Unexpected claims inflation threatens reserve adequacy. The entry identifies affected portfolios, an owner, inflation sensitivities and monitoring indicators rather than simply recording “financial risk.”
Mistake to avoid: Counting one inflation event separately under every consequence without recognizing their shared cause.
Reference: Casualty Actuarial Society
37. Inherent and residual risk
Inherent risk describes exposure before specified controls; residual risk describes exposure after them. Evaluate control effectiveness and cost using a consistent scenario basis. Expected loss is useful, but a control's value can also arise from reducing severe outcomes or maintaining risk within tolerances.
Worked example: An event with probability 10% and loss 50 has expected loss 5. A control reducing probability to 4% leaves expected loss 2; including control cost 1 gives expected cost 3.
Mistake to avoid: Assuming a documented control eliminates risk without evidence of its effectiveness.
Reference: Casualty Actuarial Society
38. Leading and lagging risk indicators
Leading indicators signal changing exposure before the main financial effect; lagging indicators describe outcomes already observed. An indicator needs a definition, owner and escalation response. A movement should trigger investigation rather than automatically prove a loss or establish its cause.
Worked example: Overdue recoverables increase from 5 to 12 before recorded credit losses rise. This can signal collection or counterparty concerns that merit examination.
Mistake to avoid: Treating an indicator threshold as a mechanically calculated loss estimate.
Reference: Casualty Actuarial Society
39. Stress tests and reverse stress tests
A stress test calculates the effects of a specified adverse scenario. A reverse stress test starts with an unacceptable outcome and investigates scenarios capable of causing it. Preserve plausible relationships between assumptions; combining individually severe shocks does not automatically produce a coherent scenario.
Worked example: Equity starts at 80 and the minimum acceptable modeled level is 50. Investment losses of 20 plus a claims increase of 15 reduce equity to 45.
Mistake to avoid: Describing an arbitrary severe scenario as a probability-calibrated forecast.
Reference: Casualty Actuarial Society
40. Common causes across risk categories
Enterprise analysis examines events that affect multiple functions simultaneously. A shared cause can connect underwriting losses, asset losses, operational disruption and liquidity needs. Map these channels before estimating dependence so apparently separate risk categories do not conceal concentrated exposure.
Worked example: A regional catastrophe generates property claims, disrupts local operations and weakens investments exposed to the same region. Separate departmental limits can miss the combined concentration.
Mistake to avoid: Assuming organizational separation establishes statistical independence.
Reference: Casualty Actuarial Society
41. Risk ownership and independent challenge
Risk owners manage exposures and implement responses; independent challenge examines assumptions, measurements and adherence to agreed limits. Effective governance makes both responsibilities explicit. A model is more useful when someone can explain its assumptions and someone else can question them without owning the commercial decision.
Worked example: An underwriting team proposes growth, actuarial analysis quantifies reserve implications, and an independent risk function challenges concentration assumptions before the designated decision-maker approves the plan.
Mistake to avoid: Assigning everyone responsibility while leaving no identifiable owner for action.
Reference: Casualty Actuarial Society
42. Choosing and monitoring risk responses
Risk responses include avoiding, reducing, transferring and retaining exposure. Compare their effects on expected cost, tail risk, operational feasibility and residual exposure. Transfer can create counterparty or basis risk. Monitor the implemented response because its effectiveness can change as the underlying business changes.
Worked example: A transfer costs 4 and reduces expected retained loss by 3. It increases expected cost by 1 but may still be justified by a necessary reduction in severe losses.
Mistake to avoid: Rejecting or accepting a response using expected cost alone when tail constraints also matter.
Reference: Casualty Actuarial Society
Risk measures and capital allocation
43. Expected loss and variance
Expected loss is the probability-weighted average outcome. Variance measures squared deviations from that mean and therefore uses squared monetary units. Neither describes every feature of a loss distribution; two distributions with the same mean and variance can have different tail behavior.
Worked example: Loss is 0 with probability 80% and 10 with probability 20%. Mean loss is 2, variance is 0.8 × 4 + 0.2 × 64 = 16, and standard deviation is 4.
Mistake to avoid: Reporting variance as though it were a monetary loss amount.
Reference: Casualty Actuarial Society
44. Value at Risk as a quantile
Value at Risk at probability p is the smallest loss level whose cumulative probability is at least p. It identifies a quantile, not the average loss beyond that quantile. Specify the loss definition, time horizon and probability level when interpreting it.
Worked example: Losses are 0 with probability 90%, 20 with probability 8% and 100 with probability 2%. The 95% Value at Risk is 20 because cumulative probability reaches 98% there.
Mistake to avoid: Interpreting Value at Risk as the maximum possible loss.
Reference: Casualty Actuarial Society
45. Tail Value at Risk with discrete outcomes
Tail Value at Risk averages the worst specified fraction of the loss distribution. With discrete outcomes, include only the required fraction of probability mass at the cutoff. This quantile-average definition avoids ambiguity when many outcomes equal Value at Risk.
Worked example: For losses 0, 20 and 100 with probabilities 90%, 8% and 2%, the worst 5% contains all 2% at 100 and 3% at 20. Its average is 52.
Mistake to avoid: Averaging every outcome at or above the quantile when that includes more than the intended tail.
Reference: Casualty Actuarial Society
46. Subadditivity and diversification
A subadditive risk measure does not assign the combined portfolio more risk than the sum of component risks. Value at Risk can fail this property, particularly for discontinuous loss distributions. Diversification claims therefore depend on both the joint distribution and the chosen risk measure.
Worked example: Each of two independent risks loses 50 with probability 3%. Each has 95% Value at Risk of zero, but their combined zero-loss probability is 97%² = 94.09%, giving combined Value at Risk of 50.
Mistake to avoid: Assuming every quantile-based capital measure always rewards diversification.
Reference: Casualty Actuarial Society
47. Correlation-based capital aggregation
Under a specified square-root aggregation model, two capital amounts combine as the square root of their squared amounts plus twice their product times correlation. This structure resembles standard-deviation aggregation. Its suitability depends on distributional and calibration assumptions; it is not a universal rule for insurance capital.
Worked example: Capital amounts of 3 and 4 with correlation 0.5 aggregate to √(9 + 16 + 12) = √37, approximately 6.08.
Mistake to avoid: Applying ordinary correlations mechanically to unrelated tail-capital measures.
Reference: Casualty Actuarial Society
48. Euler allocation and marginal contribution
For a differentiable, degree-one homogeneous capital function, Euler allocation multiplies each exposure by its marginal effect on total capital. The allocations sum to portfolio capital. They reflect diversification under the chosen model and can differ substantially from standalone capital.
Worked example: For K = √(x₁² + x₂²), exposures 3 and 4 give K = 5. Euler allocations are 9 ÷ 5 = 1.8 and 16 ÷ 5 = 3.2.
Mistake to avoid: Using derivatives alone as allocations without multiplying by the corresponding exposures.
Reference: Casualty Actuarial Society
49. Proportional allocation versus risk contribution
Proportional allocation distributes total capital using selected weights, such as standalone capital. It is transparent and fully allocates the total, but it need not represent each segment's marginal contribution or diversification benefit. State why the weights suit the intended decision.
Worked example: Total capital is 6 and standalone weights are 4 and 8. Proportional allocations are 6 × 4/12 = 2 and 6 × 8/12 = 4.
Mistake to avoid: Describing a convenient proportional allocation as a model-derived marginal risk contribution.
Reference: Casualty Actuarial Society
50. Risk-adjusted return on capital
Risk-adjusted return on capital compares consistently defined profit with consistently defined capital. Align the time horizon, expenses, expected losses and capital basis before comparing businesses. A higher ratio can be informative, but it does not establish that all available business should be accepted.
Worked example: Business A earns 6 on allocated capital of 40, giving 15%. Business B earns 8 on capital of 80, giving 10%, despite its higher absolute profit.
Mistake to avoid: Comparing one segment's gross margin with another segment's profit after expenses.
Reference: Casualty Actuarial Society
51. Expected costs and risk loads
A risk load adds compensation for uncertainty to expected costs under a stated pricing principle. Distinguish the expected-loss component from the uncertainty component to avoid counting the mean twice. A mathematical loading formula is an analytical choice, not an automatically applicable market or regulatory requirement.
Worked example: Expected loss is 2 and a selected tail measure is 8. A loading of 25% of their difference is 1.5, producing a loss-plus-risk amount of 3.5 before expenses.
Mistake to avoid: Adding a tail measure that already contains expected loss without recognizing the overlap.
Reference: Casualty Actuarial Society
52. Reinsurance layers and retained losses
For a simplified covered loss, a layer's recovery is the amount above attachment, capped at the layer limit. Retained loss equals gross loss minus recovery. Actual analysis must respect contractual definitions, exclusions, aggregation and recoverability; the simple formula only applies within its stated assumptions.
Worked example: A layer attaches at 40 and has limit 30. A covered loss of 60 produces recovery 20 and retention 40; a loss of 100 produces recovery 30 and retention 70.
Mistake to avoid: Applying the limit to the entire loss instead of the amount above attachment.
Reference: Casualty Actuarial Society
Economic capital and asset-liability management
53. Economic capital for unexpected loss
Economic capital measures resources needed for a defined adverse outcome under a chosen risk model. Specify the horizon, loss basis and whether expected loss is already funded. A centered unexpected-loss convention differs from measuring the full loss quantile; neither should be used without explaining the balance-sheet treatment.
Worked example: A selected loss quantile is 30 and expected loss of 8 is already provided for. Under a quantile-minus-mean convention, economic capital is 22.
Mistake to avoid: Subtracting expected loss when it has not been funded elsewhere.
Reference: Casualty Actuarial Society
54. One-year development and full runoff risk
One-year reserve risk concerns payments during the year plus the revised value of remaining obligations. Full runoff risk concerns the eventual total settlement outcome. The two answer different questions and require different models; a one-year reassessment is not simply one year's cash payments.
Worked example: Opening liability is 100, first-year payments are 30 and closing liability is 80. One-year deterioration is 10. If eventual total payments are 130, full runoff deterioration is 30.
Mistake to avoid: Measuring one-year reserve risk using paid claims alone.
Reference: Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
55. Joint asset and liability stresses
Economic surplus equals assets minus liabilities on a consistent valuation basis. Stress both sides together because the same event can alter investment values and claim obligations. Preserve dependence and avoid applying incompatible valuation assumptions to assets and liabilities.
Worked example: Assets of 180 and liabilities of 140 give surplus 40. An asset loss of 10 and liability increase of 15 reduce surplus to 15, a deterioration of 25.
Mistake to avoid: Stress-testing assets while holding economically sensitive liabilities fixed without justification.
Reference: Casualty Actuarial Society
56. Cash-flow matching and liquidity
Cash-flow matching aligns available asset proceeds with expected liability payments by date and currency. A portfolio can have adequate total value yet lack cash when needed. Matching expected payments reduces timing risk but does not eliminate unexpected claims, asset defaults or operational access constraints.
Worked example: Assets provide 10 next month while expected claim payments are 15. The immediate funding gap is 5 even if another asset matures with ample proceeds later.
Mistake to avoid: Using total asset value as proof that near-term payments can be met.
Reference: Casualty Actuarial Society
57. Duration and surplus sensitivity
Modified duration approximates the percentage value change from a small parallel yield shift. Surplus sensitivity depends on the monetary duration of both assets and liabilities, not just their durations separately. Keep liability cash-flow and discount assumptions explicit.
Worked example: Assets are 120 with duration 4; liabilities are 100 with duration 6. A one-percentage-point yield rise changes assets by about −4.8 and liabilities by −6, increasing surplus by about 1.2.
Mistake to avoid: Comparing asset and liability durations without weighting them by their values.
Reference: Casualty Actuarial Society
58. Convexity beyond linear duration
Convexity adds a second-order term to the duration approximation of value changes. For fixed cash flows with positive convexity, it improves the estimate for larger yield moves. Optionality, changing cash flows and nonparallel yield movements can require more detailed valuation rather than a single approximation.
Worked example: Duration is 5, convexity is 30 and yields rise by 0.02. Approximate value change is −5 × 0.02 + 0.5 × 30 × 0.02² = −9.4%.
Mistake to avoid: Entering a 2% yield change as 2 instead of 0.02.
Reference: Casualty Actuarial Society
59. Inflation and liability hedge basis risk
A hedge can track a different index from the costs driving claims. General inflation, wage inflation and repair-cost inflation need not move together. Evaluate the relationship between asset payoffs and actual liability drivers, including lags and coverage limits, rather than relying on a shared label.
Worked example: Expected claim costs of 100 rise by 8% to 108, while a fixed asset payment remains 100. The resulting gap is 8 despite unchanged nominal asset proceeds.
Mistake to avoid: Assuming any inflation-sensitive asset exactly offsets claims inflation.
Reference: Casualty Actuarial Society
60. Contingent liquidity under stressed sales
Liquidity planning considers cash, reliable inflows and proceeds obtainable from asset sales under stress. Market value is not necessarily immediately realizable value. Apply explicit timing and haircut assumptions, and examine whether asset-sale pressure coincides with unusually large claim payments.
Worked example: Cash is 12 and assets worth 30 can be sold with a modeled 20% haircut, producing 24. Available funds are 36 against payments of 40, leaving a gap of 4.
Mistake to avoid: Counting full market values and stressed sale proceeds as two separate liquidity resources.
Reference: Casualty Actuarial Society
Sources
Source context:
- Exam 7 - Advanced Estimation of Claims Liabilities | Casualty Actuarial Society
- Casualty Actuarial Society