CRO Quantitative Risk Modeling & Analytics Flashcards

Backtesting a VaR model involves:

Answer: Comparing the model's historical loss predictions against actual observed losses

Backtesting validates a VaR model by comparing its predicted loss thresholds against realized portfolio losses over a historical period to assess predictive accuracy.

A CRO wants to quantify the risk that two risks materialize simultaneously. Which technique is MOST appropriate?

Answer: Copula models to capture joint probability distributions and tail dependence

Copula models capture the dependency structure between two or more risk variables, including tail dependence where multiple risks spike simultaneously during crises.

Risk-adjusted return on capital (RAROC) is calculated as:

Answer: Net revenue minus expected losses divided by economic capital

RAROC measures profitability relative to the risk taken by dividing risk-adjusted net revenue (after expected losses) by the economic capital required to support that risk.

In scenario analysis, a 'reverse stress test' is specifically designed to:

Answer: Identify scenarios that would cause the organization to fail, then assess their plausibility

Reverse stress testing starts with a failure outcome (e.g., insolvency) and works backward to identify what scenarios could realistically cause that outcome.

Which of the following is a key challenge when applying machine learning models in enterprise risk management?

Answer: Explainability — regulators and auditors require that model decisions can be clearly justified

Regulators increasingly require model explainability (especially under SR 11-7 and GDPR), making black-box machine learning models challenging to validate and defend to supervisors.

When a quantitative risk model consistently underestimates losses during periods of market stress, this is MOST likely caused by:

Answer: Assuming linear correlations that break down during crises due to contagion effects

In crisis periods, correlations between assets spike non-linearly due to contagion, causing models calibrated on normal-period linear correlations to drastically underestimate joint losses.