CIMA common mistakes
CIMA · Understand the traps and how to avoid them.
Computes z-scores or confidence intervals from mean and standard deviation for any return series - including hedge fund, private equity, or option-overlay returns - and states a probability of a loss exceeding X% without ever checking skewness or kurtosis.
Introductory quantitative material drills the normal distribution, so candidates reach for the two-moment shortcut whenever they see a mean and a standard deviation. Fat tails and negative skew are invisible in those two statistics, and exam vignettes rarely flag them explicitly.
What to do instead
Before quoting any normal-based probability, run the four-moment check. If skew is negative or excess kurtosis is positive, say so and switch to downside deviation, Sortino, or empirical/bootstrapped percentiles. State the distributional assumption explicitly as part of the answer rather than leaving it implicit.
Treats duration as 'years until the bond pays back' and selects the longer-maturity bond as the higher-duration bond without adjusting for coupon or yield level.
Duration is a time-like word and maturity is the most visible field on a bond quote. The present-value-weighted definition is the part candidates skip under time pressure, so they fall back on maturity as a proxy.
What to do instead
Anchor two definitions: Macaulay duration = present-value-weighted average time to cash flows; modified duration = Macaulay/(1+y/k) = first-order price sensitivity. Then apply the sign rules: at equal maturity, lower coupon and lower yield mean higher duration; zero-coupon duration equals maturity.
Discounts FCFF at the cost of equity, or FCFE at WACC, or adds net borrowings to an FCFF-derived value, when producing an equity value.
Both DCF variants are presented on the same page with near-identical numerators, so the cash flow and discount rate get matched by habit rather than by the claim each represents.
What to do instead
Pair claims with claimants before computing anything: FCFF → WACC → enterprise value → subtract net debt → equity value; FCFE → cost of equity → equity value directly. Write the pair down as a first step so the discount rate cannot drift.
Labels a bias from the client's narrative rather than from its mechanism - e.g. calls a client who will not sell an inherited concentrated position 'loss aversion' when the driver is endowment/status quo, or calls a client who chased last year's top-performing fund 'herding' when it is recency.
Biases are taught through vivid, overlapping vignettes. Candidates pattern-match on the surface details of the story (a stock, a loss, a crowd) instead of on what is actually driving the behaviour.
What to do instead
Ask two diagnostic questions: (a) Is the belief factually wrong (cognitive) or is a feeling driving it (emotional)? (b) What is the client protecting - a prior commitment (endowment, status quo), a self-image (overconfidence, self-attribution), or a recent salient outcome (recency, availability)? Name the mechanism first, then attach the label.
Recommends 'educate the client about the data' as the remedy for every identified bias, including clearly emotional ones such as regret aversion or loss aversion.
Education feels like the professional and safe answer, and the implication of the cognitive/emotional split for remedies is the least-drilled part of the behavioural material.
What to do instead
Apply the split explicitly: cognitive biases are corrected - education, better information, a more rigorous decision process, second opinions. Emotional biases are adapted to - change the portfolio or the process around the client (automatic rebalancing, pre-commitment rules, limited access to volatile positions) rather than arguing them out of a feeling.
Estimates portfolio risk by averaging or summing the individual asset standard deviations, omitting the covariance terms entirely.
The weighted-average formula is correct for expected return, so it gets copied across to risk. The double summation is tedious and error-prone under exam time pressure, so it is quietly dropped.
What to do instead
Write the general form first: variance = ΣΣ wi wj σi σj ρij. Then do the two-asset case symbolically: σp² = w1²σ1² + w2²σ2² + 2w1w2σ1σ2ρ12. Use the boundary check: at ρ = +1 risk equals the weighted average; any ρ < 1 pulls portfolio risk strictly below it.
Computes Sharpe and Treynor ratios with the risk denominators swapped (β under excess return for Sharpe, σ for Treynor), or asserts that one measure always dominates the other.
Both are excess return over a risk measure, the two symbols blur under recall pressure, and the 'which measure when' rule is a single sentence that is easy to skip over in review.
What to do instead
Let the question choose the denominator. Total risk (σ) applies when the portfolio is the investor's entire holding → Sharpe. Systematic risk (β) applies when the portfolio is one sleeve inside an already diversified whole → Treynor. Rank single-manager, standalone mandates on Sharpe; rank managers to be added to a diversified portfolio on Treynor or Jensen's alpha - and be ready to explain that the rankings can legitimately disagree when σ and β rank differently.
Reports the money-weighted (IRR) return as the manager's performance for a period in which the client made large contributions or withdrawals.
The IRR is the figure printed on the client statement and the most mechanically familiar to compute. The TWR/MWR distinction is conceptual rather than procedural, so it gets bypassed.
What to do instead
Apply the rule: manager evaluation → time-weighted return, which neutralises client-controlled cash flow timing; client experience and goal achievement → money-weighted return. When flows are large relative to assets, state why the two diverge - a timing effect - and which question each number answers.
Links sub-period returns by adding or averaging them, or annualises by multiplying a quarterly return by four and by dividing a multi-year cumulative return by the number of years.
Arithmetic is the default mental operation, and simple division feels like it produces an 'average annual return'. The compounding alternative requires an exponent or a root, which is slower and more error-prone by hand.
What to do instead
Default to compounding and roots: cumulative = Π(1 + rt) − 1; annualised = (1 + cumulative)^(1/n) − 1; sub-annual annualisation = (1 + r)^k − 1, never r × k. Use the built-in sanity check that the geometric mean is always less than or equal to the arithmetic mean, and investigate whenever a computed result violates it.
Records a single 'risk tolerance' score from the questionnaire and builds the asset allocation directly from it, without separately documenting risk capacity, required return, or the five IPS constraints.
Willingness is what the questionnaire measures and what clients talk about openly. Ability to bear loss comes from the balance sheet and required return comes from a goal projection, both of which take real analytical work that candidates try to shortcut.
What to do instead
Quantify all three separately: attitude (willingness), capacity (ability to absorb loss without derailing the plan), and required return (need to meet stated goals). The binding constraint is the minimum of the three, and any gap between willingness and ability must be documented and resolved with the client. Write out liquidity, time horizon, tax, legal/regulatory, and unique circumstances before proposing any allocation.