CAIA Level II common mistakes
CAIA Level II · Understand the traps and how to avoid them.
Writing a target allocation as a bare number (e.g., 'raise private equity to 20%') without stating the funding source, the multi-year deployment schedule, and how the portfolio stays inside its liquidity and risk budget during the J-curve.
Level I trained candidates to compute a single optimal weight; Level II constructed-response items reward the full decision chain, and candidates reproduce the one number they memorized.
What to do instead
Use a fixed answer template for every allocation question: objective → binding constraints (liquidity, horizon, regulatory, tax) → sizing → funding source → implementation/pacing → monitoring and rebalancing trigger. Never end on the weight.
Treating percentage of capital as percentage of risk - assuming a 15% allocation to a 2x-volatility asset contributes 15% of portfolio risk - and not testing what happens when correlations converge in a stress event.
Capital weights are visible and easy; marginal contribution to risk requires the covariance matrix and candidates skip it under time pressure.
What to do instead
Compute contribution to risk as w_i × (Σw)_i / σ_p for each sleeve and report it alongside capital weights; then re-run the risk budget with all pairwise correlations set to 0.8 and state whether the allocation still holds.
Ranking private equity funds by headline gross IRR and selecting the highest number without converting to net-of-fee terms or comparing against a public-market equivalent.
Gross IRR is the figure quoted in marketing materials and is cognitively 'sticky'; candidates also forget that IRR embeds an implicit reinvestment assumption at the IRR itself.
What to do instead
Always present net IRR alongside DPI, TVPI and RVPI, compute a public-market equivalent (KS-PME) using the same cash-flow dates, and compare within vintage year rather than across vintages.
Treating committed capital as invested capital when building a pacing plan, then concluding the program has no near-term liquidity drain.
The commitment is the salient number in the fund documents, while the unfunded liability and the J-curve are timing concepts that must be modelled explicitly.
What to do instead
Build a commitment schedule showing expected annual takedown percentages, management fee drag on committed capital, expected distribution timing, net cash flow per year and the unfunded ratio; state each assumption so the grader can follow it.
Reporting a real estate deal's return as a cap rate or cash-on-cash yield and labelling it the IRR, and presenting the leveraged return without the debt terms behind it.
Cap rate is the first number in every offering memorandum, and leverage mechanics (amortization, DSCR covenants, floating-rate resets) are tedious to lay out.
What to do instead
Decompose return into income return + appreciation + leverage contribution − capital expenditure/reserves, state the loan terms and DSCR, and show exit value sensitivity to a ±100bp exit cap rate move. Label each metric explicitly.
Evaluating a hedge fund using the reported Sharpe ratio and standard deviation at face value when the fund holds illiquid or appraisal-priced positions.
Serial correlation in smoothed marks understates volatility and inflates Sharpe; the reported statistics look clean and candidates have no reason to adjust them.
What to do instead
Un-smooth the return series (Getmansky–Lo–Makarov or a first-order autocorrelation adjustment), recompute volatility, Sharpe and maximum drawdown on the adjusted series, and report the fund's beta to equities specifically during the worst equity deciles.
Claiming managed futures provide unconditional 'crisis alpha' diversification and quoting a single full-period correlation figure to support it.
The full-period correlation is low and the crisis-alpha narrative is memorable, so the conditional nature of the benefit never surfaces in the answer.
What to do instead
Compute correlation and beta conditional on the worst equity-market deciles, discuss dependence on trend persistence, whipsaw losses in V-shaped reversals, and roll/carry costs; state explicitly the regimes in which the diversification fails.
Conflating spread duration with interest-rate duration and applying only one of them to a credit scenario, or reporting a bond's expected return with no expected-loss term.
Both are called 'duration' and both scale a yield change, so candidates collapse them into a single sensitivity.
What to do instead
Write the decomposition explicitly: ΔP/P ≈ −D_rate × Δy_risk-free − D_spread × Δspread + convexity, then add expected loss = probability of default × LGD × (1 − recovery), and show each term separately in the scenario.
Answering an ethics item by declaring the conduct 'a violation' or 'unethical' without naming the specific Standard involved, the duty it creates, and the concrete action required of the candidate (disclose, refuse, escalate, or withdraw).
The moral judgment is easy and feels like the answer; the standards-based reasoning and prescribed remedy are what actually earn marks.
What to do instead
Use a fixed four-part structure: identify the Standard → state the duty owed and to whom → state the specific required action → state escalation/remediation and any documentation needed. Never stop at the verdict.
Answering emerging-topic questions on digital assets, tokenized real assets or carbon markets with generic filler ('high volatility, evolving regulation') instead of naming the specific structural issue the scenario turns on.
These topics feel unfamiliar, so candidates default to risk boilerplate rather than applying a framework that transfers from traditional alternatives.
What to do instead
Anchor every emerging-topic answer to four transferable dimensions: valuation approach (no cash flow → cost of carry or network value), custody and key control (qualified custodian, key management), operational/legal structure (transfer agent, jurisdiction, settlement finality), and liquidity/exit. Name the dimension the scenario is actually testing.