CAS Exam 9 common mistakes
CAS Exam 9 · Understand the traps and how to avoid them.
Treating the variance of a portfolio as the sum of individual asset variances, or forgetting the covariance/correlation terms when combining two risky assets in the CAS-style portfolio math problem.
Candidates memorize the single-asset variance formula and carry it over to multi-asset problems, especially under time pressure when the two-asset case looks deceptively like a weighted average. The covariance term is the whole point of diversification, but it is the piece most easily dropped.
What to do instead
Before computing, write the full variance expression for the portfolio explicitly: w1²σ1² + w2²σ2² + 2w1w2ρσ1σ2. Substitute numbers only after the symbolic line is complete. Do at least five drill problems that vary ρ from −1 to +1 and confirm that ρ = +1 reproduces the weighted-average σ and ρ = −1 gives the minimum-variance combination. If the answer does not change when ρ changes, the covariance term was lost.
Computing a risk measure (VaR, TVaR, expected shortfall) for the wrong horizon or confidence level, or reporting a loss quantile as if it were a probability, because the problem statement embeds the parameter in a sentence rather than a table.
Exam questions often state the confidence level in prose ('the 99th percentile of the loss distribution') while the intermediate distribution is given as a probability of exceeding a threshold. Candidates flip between the tail probability and the confidence level, or use an annual figure for a one-day horizon.
What to do instead
Circle the horizon and the confidence level in the question stem before touching the distribution, and rewrite them in a one-line header ('99%, one year, loss basis'). Convert to a consistent convention immediately: if the table gives P(L > x), the 99% VaR is the x where that probability equals 0.01. Then state the answer as a loss amount, not a probability, and sanity-check magnitude against the mean and standard deviation of the loss distribution.
Applying the CAPM or a single-factor beta to a project or division without adjusting for the leverage or business risk of the comparison firm, and then using the resulting cost of equity in a WACC calculation.
Candidates find an industry beta, plug it straight into the CAPM, and move on. They forget that an observed equity beta is levered and must be unlevered with the comparables' debt-to-equity ratio, then relevered at the target capital structure. The error survives to the final answer and is invisible unless the grader checks the intermediate beta.
What to do instead
Build a fixed four-step template for every cost-of-capital question: (1) unlever each comparable's beta with βu = βL / [1 + (1 − t)(D/E)]; (2) average the unlevered betas; (3) relever at the target D/E; (4) run CAPM, then WACC. Write the D/E used at each step next to the number. If the question gives no target structure, say so explicitly rather than silently reusing the comparable's leverage.
Confusing the risk-management frameworks and their components - for example, mixing up the steps of the risk-management process, the roles of the board versus management, or the definitions of risk appetite, risk tolerance, and risk capacity - and answering with a plausible-sounding but misattributed list.
The governance and framework material is prose-heavy and reads similarly across sources, so candidates skim it and rely on general business intuition. Under exam pressure they produce a generic list that omits the specific element the question is testing, such as the distinction between risk appetite (willingness to accept) and risk capacity (ability to bear).
What to do instead
Build a one-page comparison table with one row per framework and columns for purpose, components, owner, and output. For each term, write a one-sentence definition in your own words plus a concrete insurance example (e.g., capacity constrained by surplus, appetite set by the board). Practice free-response answers that must name the framework's stages in order, and self-grade against the table rather than against memory.
Answering ERM questions with generic corporate-governance language instead of the specific ERM elements being tested - such as the distinction between risk identification, risk quantification, risk prioritization, and risk monitoring, or the role of a risk owner versus a risk committee.
ERM is the most conceptual domain, and candidates substitute buzzwords ('tone at the top', 'risk culture') for the precise mechanism the question asks about. This earns partial credit at best on free-response items that require a named component and an explanation of how it operates.
What to do instead
For each ERM component, prepare a three-part note: what it is, who performs it, and what artifact it produces (risk register, heat map, capital model output, board report). When practicing, force yourself to answer 'which specific element and what does it do' before writing prose. Avoid any sentence that could appear unchanged in a different domain's answer.
Mishandling correlation and dependence in aggregation - assuming independence when combining risks, or treating a copula or correlation matrix as if it implied linear correlation in the tails - and then reporting an aggregate capital or VaR figure that is too low.
Independence is the default mental model and the easiest arithmetic. Candidates also conflate 'correlation' with 'dependence' and apply a single correlation parameter to a tail aggregation problem where the dependence structure matters most. The resulting understatement is directionally wrong in exactly the scenario the exam is testing.
What to do instead
Before aggregating, write down the dependence assumption and its justification. When the question specifies a copula or scenario, compute the aggregate under at least two dependence assumptions (independence and perfect dependence) to bracket the answer, then state which is appropriate and why. For tail measures, note explicitly that a linear correlation of zero does not imply tail independence.
Miscalculating the cost of capital or capital adequacy by mixing book and market values, or by double-counting the cost of debt in the WACC weights, so the final rate or ratio is internally inconsistent.
Insurance company financial analysis questions supply balance-sheet data in book terms and market data for equity, and candidates grab whichever number is nearby. WACC weights must be market-value based for the target structure, and the after-tax cost of debt must use the marginal tax rate, not the effective rate.
What to do instead
Write the WACC formula with labeled inputs before substituting: market value of equity, market value of debt, after-tax cost of debt, cost of equity. Flag every input as book or market and convert where required. Check that the weights sum to 1.00 and that the resulting WACC lies between the after-tax cost of debt and the cost of equity; if it does not, an input or weight is wrong.
Ignoring the time value and the timing mismatch in reserving or capital questions - discounting losses at the wrong rate, using undiscounted cash flows in a present-value comparison, or failing to align the duration of assets and liabilities when assessing interest-rate risk.
Candidates focus on the loss distribution and treat the discounting as an afterthought, or they apply a risk-free rate when the question specifies a risk-adjusted rate. Duration mismatch questions are missed because the candidate computes duration for each side but never compares them.
What to do instead
For every cash-flow problem, draw a small timeline with the discount rate written above the arrow and the period index below it. Compute the present value of assets and liabilities separately, then compare both the present values and the durations. State the sign of any surplus or gap explicitly. If a question gives both a risk-free and a risk-adjusted rate, write one sentence on which applies and why before calculating.
Producing a numerically correct free-response answer with no stated assumptions, no intermediate steps, and no interpretation, losing method and communication points even when the final number matches.
Candidates practice multiple-choice style thinking and treat free-response as a place to deposit a number. CAS free-response grading rewards demonstrated reasoning, explicit assumptions, and a concluding sentence that interprets the result in the business context of the question.
What to do instead
Adopt a fixed answer skeleton for every free-response item: restate the question's key parameters, list assumptions, show the formula, show substituted values, give the result with units, then write one sentence on what it means for the insurer. Practice under timed conditions and grade your own answers against that skeleton. If a required input is missing, state the assumption you are making rather than leaving a blank.
Allocating study time evenly across domains and skipping the quantitative portfolio and capital-modeling problems because they feel slow, then running out of time or leaving blanks on the highest-value computational items.
Conceptual reading feels productive and is easier to schedule, so quantitative practice gets deferred until late. On exam day, unfamiliar computational setups consume disproportionate time, and candidates abandon questions rather than capturing partial credit.
What to do instead
Build a practice log by domain and question type, and weight sessions toward the quantitative items you have not yet solved cold. Time-box each problem (roughly the exam's per-point pace) and record actual versus target time. For any problem you cannot finish, write the setup and first two steps anyway, since partial credit is available. Review errors by cause (setup, arithmetic, interpretation) rather than by topic alone.