FRM Part II common mistakes
FRM Part II · Understand the traps and how to avoid them.
When converting a VaR from one confidence level and horizon to another, the candidate scales only by the square root of time and keeps the original z-score (e.g., takes a 1-day 95% VaR, multiplies by √10, and reports it as the 10-day 99% VaR).
The candidate has memorized VaR = z·σ·√t as a single formula rather than as two separate transformations, and forgets that the multiplier changes with the confidence level (1.645 vs 2.326). The √t rule itself also silently assumes i.i.d. normal returns, an assumption the question may be testing.
What to do instead
Before computing, write the target pair (α, horizon) on the page. Convert in two explicit steps: scale volatility by √t, then apply the z-score that matches the stated α. If returns are autocorrelated or fat-tailed, flag that √t scaling understates long-horizon risk rather than applying it silently.
Applies a delta-normal (linear) VaR to a book containing short options or other clearly nonlinear exposures and reports the number as if it were valid, without noting the direction of the bias.
The item stem supplies a delta and a notional, and multiplying them is the fastest path to an answer. Gamma, and therefore the convexity of the loss profile, never enters the calculation, even though short-gamma positions lose far more than delta predicts in a tail move.
What to do instead
Before computing, classify each position as linear or nonlinear. For any net short-gamma book, state that delta-normal VaR understates tail loss and use full revaluation or add the second-order term ½·Γ·(ΔS)². Read the last line of the question: it often asks for the direction of bias, not the number.
Plugs the observed equity volatility straight into the Merton distance-to-default or EDF calculation as if it were the asset volatility.
Equity is a call option on the firm's assets, so equity volatility and asset volatility are related, not equal. The candidate treats the two as interchangeable because both appear in the same model, and the resulting distance-to-default is wrong.
What to do instead
Recall that σ_E·E = N(d1)·σ_A·A. Either solve the two-equation system (equity value as a call plus the volatility relationship) simultaneously, or use the asset volatility the question supplies. Sanity-check that asset volatility must be lower than equity volatility for a levered firm.
Converts a default probability across horizons by proportional scaling: multiplies an annual PD by the fraction of a year, or annualizes a quarterly PD by multiplying by four, and uses the result as if it were a cumulative PD.
The candidate treats PD as a linear rate rather than a survival-based quantity, and does not distinguish the hazard rate (marginal, conditional) from the cumulative default probability. Both errors produce plausible-looking but wrong numbers.
What to do instead
Work in hazard rates: λ = −ln(1 − PD_annual); cumulative PD(t) = 1 − e^(−λt); marginal (period) PD over period t = e^(−λ(t−1)) − e^(−λt). First decide whether the question asks for a conditional, marginal, or cumulative figure, and label your answer accordingly.
In an LDA calculation, takes the 99%/one-year quantile of the aggregate loss distribution as the operational risk capital figure, or adds expected loss on top of the quantile rather than netting it out.
The candidate carries over the market-risk convention (99%, 10 days) into operational risk, and confuses expected loss with unexpected loss. Regulatory operational risk capital is calibrated at 99.9% over a one-year horizon and covers unexpected loss only.
What to do instead
Keep a one-line convention table per risk type (market: 99%/10-day or ES at 97.5%; operational: 99.9%/1-year, UL only). For LDA, capital = 99.9% quantile of aggregate losses minus expected annual loss; state explicitly which term the question is asking for.
Computes the Standardised Measurement Approach Business Indicator Component by applying the 12%, 15%, 18% coefficients to the entire Business Indicator instead of only the portion falling in each bucket, and omits the cap applied to the interest component of BI.
The marginal (tiered) structure of the BIC is easy to flatten into a single average rate under time pressure, and the ILDC cap of 2.25% of interest-earning assets is a small detail buried in the BI definition.
What to do instead
Write out the components before computing: BI = ILDC + SC + FC, where ILDC = min(|interest income − interest expense|, 2.25% × interest-earning assets) + dividend income; SC = max(other operating income, other operating expense); FC = |trading book net P&L| + |banking book net P&L|. Then apply BIC coefficients marginally: 12% on the first €1bn, 15% on €1bn–€30bn, 18% above €30bn. Then add the loss component (15 × ten-year average annual operational losses).
Mixes up LCR and NSFR inputs: places wholesale funding or loans on the wrong side of the ratio, or attaches a one-year horizon to the LCR.
Both ratios are 'liquidity ratios ≥100%' with a stock-over-flow structure, so the templates look interchangeable in memory. The distinguishing features (HQLA and 30-day net outflows vs available and required stable funding over one year) get lost.
What to do instead
Anchor two lines: LCR = HQLA / net cash outflows over 30 calendar days, with inflows capped at 75% of outflows; NSFR = available stable funding / required stable funding over a one-year horizon, with ASF and RSF factors assigned by liability and asset type and maturity. Sketch which side each balance-sheet item belongs to before touching the numbers.
Builds the maturity ladder using contractual maturities only - treating non-maturity core deposits as maturing overnight and committed but undrawn credit lines as having no cash-flow consequence - and then reports the resulting short-term gap as the funding need.
A static contractual ladder is mechanical and quick, whereas behavioural assumptions require judgment. The candidate also ignores the stress scenario the question specifies, in which previously stable deposits run and undrawn commitments get drawn.
What to do instead
First identify whether the question says 'contractual' or 'behavioural'. Under behavioural assumptions, assign core non-maturity deposits a stable expected maturity and treat a stated run-off percentage of them as an outflow. In stress, committed lines convert into outflows even though nothing has been drawn. State the sign convention (+ = surplus) before computing the cumulative gap.
Takes a hedge fund or private-asset index's reported mean and standard deviation at face value and computes a Sharpe ratio, beta, or volatility directly from the published series.
Reported returns for illiquid assets are smoothed by appraisal-based pricing, producing positive serial correlation; backfill and survivorship bias further flatter the series. The candidate treats the reported statistics as if they came from a liquid, complete sample.
What to do instead
Check the series for first-order autocorrelation and state the direction of each bias: serial correlation understates volatility and beta, overstates the Sharpe ratio and reported alpha, and understates drawdowns. De-smooth the returns using the observed autocorrelation before computing risk statistics, and add back an estimate of survivorship and backfill effects.
Treats climate risk as a standalone new risk category that should generate its own capital charge, instead of as a driver that transmits into credit, market, liquidity and operational risk.
Headlines frame climate as a distinct 'new' risk, and the candidate answers with a generic statement about climate being unmodelled rather than tracing the transmission channels: physical risk and transition risk flowing into asset valuations, default probabilities, collateral values, and liability exposures.
What to do instead
For any climate-risk item, name the transmission channel and the standard risk it lands in: physical risk → collateral and PD deterioration (credit), stranded assets → repricing (market), insurance and liability claims and litigation (operational), deposit and funding flight (liquidity). Only after mapping it to a known risk category should you discuss measurement or capital treatment.