CrTAS common mistakes
CrTAS · Understand the traps and how to avoid them.
Computing VaR with a 1-day horizon and then scaling to 10 days by multiplying by 10 instead of by the square root of 10, or scaling when returns are not i.i.d. normal.
Candidates memorize 'square root of time' but forget it requires i.i.d. returns and zero mean; they also confuse the 99% one-tailed z of 2.326 with the 95% value of 1.645 when converting volatility to VaR.
What to do instead
Write the scaling assumption down before applying it: VaR_10d = VaR_1d × sqrt(10) only under i.i.d. and no autocorrelation. For each question, state the confidence level, horizon, and whether the z is one- or two-tailed before plugging numbers in.
Treating expected shortfall as a multiple of VaR (e.g., 'ES ≈ 1.5 × VaR') or reporting ES at a different confidence level than the VaR it accompanies.
Candidates memorize a rough multiplier from a normal distribution and apply it to fat-tailed or skewed portfolios, and they miss that ES must be at the same confidence level and horizon as the VaR it complements.
What to do instead
Compute ES as the average of losses beyond the VaR threshold, or use the closed-form normal result ES = σ·φ(z)/(1−α). Check that confidence level and horizon match the VaR before comparing them.
Confusing PD, LGD, and EAD in the expected loss formula, or applying the wrong exposure measure (drawn vs. undrawn, CCF applied to the wrong base).
Under time pressure, candidates plug numbers into EL = PD × LGD × EAD without checking units (PD as a decimal, not percent) or forget that the undrawn portion is multiplied by the credit conversion factor before entering EAD.
What to do instead
Write EL = PD × LGD × EAD at the top of scratch work and label each input with its units. For commitments, compute EAD = drawn + CCF × undrawn explicitly before multiplying.
Netting counterparty exposure across the wrong set of trades, or ignoring collateral when computing current exposure versus potential future exposure.
Candidates mix netting sets with legal entities and forget that netting only applies within an enforceable master netting agreement, and they treat collateral as reducing PFE rather than current exposure.
What to do instead
Before netting, ask: same legal entity, same netting agreement, enforceable in the relevant jurisdiction? Then compute exposure = max(MTM − collateral, 0) + PFE and keep PFE gross of collateral unless the question states otherwise.
Mixing up the liquidity coverage ratio and net stable funding ratio numerators and denominators, or classifying a funding source as stable when it is not.
Both ratios use 'high-quality liquid assets' and 'stable funding' language, and candidates apply retail deposit runoff assumptions without checking whether deposits are insured, operational, or from a financial institution.
What to do instead
Memorize the structures separately: LCR = HQLA / net cash outflows over 30 days; NSFR = available stable funding / required stable funding over one year. Build a short table of runoff and ASF/RSF factors from the question's own data before calculating.
Treating duration gap and repricing gap as interchangeable, or ignoring the equity effect of a rate shock on the economic value of equity.
Candidates learn both gaps as 'rate sensitivity' measures and apply a duration gap formula to a repricing-gap question, or forget that a positive duration gap means liabilities reprice faster and equity falls when rates rise.
What to do instead
Classify the question first: earnings-at-risk (repricing gap, NII) versus economic value of equity (duration gap). State the sign convention explicitly: ΔEVE ≈ −D_gap × A × Δr / (1+r), and check the direction of the equity change against the rate move.
Answering operational-risk questions by naming a tool (RCSA, KRI, loss data) without linking it to the risk event type or the control objective the scenario describes.
Candidates recognize the vocabulary and pick the most familiar acronym rather than reading what the scenario asks: identification, assessment, monitoring, or mitigation.
What to do instead
For each scenario, label the step in the risk-management cycle (identify, assess, monitor, control, report) and the Basel event category (internal fraud, external fraud, employment practices, clients/products, damage to physical assets, business disruption, execution/delivery). Choose the tool that matches the step.
Confusing model validation with model development and assuming a validated model is a compliant model, or missing that backtesting and benchmarking are separate exercises.
Candidates conflate the three lines of defence and assume validation is a one-time sign-off rather than an independent challenge, and they treat backtesting (comparing predictions with outcomes) and benchmarking (comparing with an alternative model) as the same test.
What to do instead
Map each activity to its owner: development (first line), independent validation and ongoing monitoring (second line), internal audit (third line). For each question, ask whether it tests predictive accuracy (backtest) or relative performance (benchmark).
Mixing up settlement risk, Herstatt risk, and counterparty credit risk, or ignoring the timing of payment versus delivery in a foreign-exchange transaction.
Candidates treat all 'the other side might not pay' risks as the same and miss that settlement risk arises when one leg is paid before the other is received, which is distinct from pre-settlement credit exposure.
What to do instead
Define each term in one line before answering: settlement (Herstatt) risk is principal-at-risk during the payment-versus-payment window; pre-settlement risk is replacement cost before maturity. Then identify which window the scenario describes.
Applying a stress scenario as a single shock to one risk factor while ignoring second-order effects (correlation breakdown, funding contagion, collateral calls) and not reconciling the stressed capital need to the regulatory buffer stack.
Candidates treat stress testing as a sensitivity exercise and forget that severe scenarios combine market, credit, and liquidity stresses, and they mis-stack the capital buffers (CCoB, CCyB, G-SIB, systemic risk) when computing the total requirement.
What to do instead
For every scenario, list the transmission channels: P&L, capital, liquidity, and collateral. Then build the buffer stack in order: minimum requirement + capital conservation buffer + countercyclical buffer + systemic buffers, and state which buffer is being breached and what restriction follows.