CORES CORES Model Risk & Quantitative Methods 2 — Questions and Answers
Question 1: What is the main limitation of relying solely on internal loss data for operational risk capital modeling?
- Internal data is too expensive to collect
- Internal loss history may be too short or lack sufficient large-loss observations, underestimating tail risk (Correct answer)
- Regulators prohibit the use of internal loss data
- Internal data always overestimates capital requirements
Correct answer: Internal loss history may be too short or lack sufficient large-loss observations, underestimating tail risk
Internal loss histories are often short and may not include extreme tail events, so relying solely on them can underestimate the true severity of rare, high-impact losses.
Question 2: How does the use of external loss data address the limitations of internal data in operational risk modeling?
- External data replaces internal data entirely in all models
- External data supplements internal data to improve the estimation of low-frequency, high-severity loss distributions (Correct answer)
- External data is used only for regulatory reporting, not modeling
- External data eliminates the need for scenario analysis
Correct answer: External data supplements internal data to improve the estimation of low-frequency, high-severity loss distributions
External loss data provides observations of rare, large-scale loss events at peer institutions, which helps calibrate the tail of loss distributions that internal data cannot capture.
Question 3: What is a Monte Carlo simulation used for in operational risk quantification?
- To physically simulate office environments for business continuity testing
- To generate thousands of random loss scenarios by sampling from frequency and severity distributions to estimate aggregate loss distributions (Correct answer)
- To audit vendor contracts for financial terms
- To automatically approve risk models for regulatory submission
Correct answer: To generate thousands of random loss scenarios by sampling from frequency and severity distributions to estimate aggregate loss distributions
Monte Carlo simulation uses repeated random sampling from loss frequency and severity distributions to build an aggregate loss distribution and estimate percentile-based capital figures.
Question 4: Which of the following is a key governance control for managing model risk in a financial institution?
- Requiring all models to be built by the same team
- Maintaining a comprehensive model inventory with periodic validation and performance monitoring (Correct answer)
- Using only vendor-supplied models without internal review
- Restricting model use to the IT department
Correct answer: Maintaining a comprehensive model inventory with periodic validation and performance monitoring
A centralized model inventory that tracks all models in use, their purpose, validation status, and ongoing performance is a foundational model risk governance control.
Question 5: What does 'backtesting' a risk model involve?
- Rebuilding the model from scratch each quarter
- Comparing model predictions or outputs against actual historical outcomes to assess accuracy (Correct answer)
- Testing the model on data from a competitor institution
- Requiring the model to pass a physical hardware performance test
Correct answer: Comparing model predictions or outputs against actual historical outcomes to assess accuracy
Backtesting compares a model's predicted outcomes against realized historical results to evaluate whether the model's assumptions and outputs are accurate.
Question 6: In operational risk, what does 'fat-tail' or 'heavy-tail' distribution refer to?
- A loss distribution where most losses cluster near the mean with very few extremes
- A loss distribution with a higher-than-normal probability of extreme, large-magnitude losses (Correct answer)
- A standard normal distribution used in credit risk
- A distribution exclusively used in market risk modeling
Correct answer: A loss distribution with a higher-than-normal probability of extreme, large-magnitude losses
Fat-tail distributions assign higher probability to extreme outcomes than a normal distribution, which is critical for operational risk modeling because large, rare losses are common in practice.
CORES CORES Model Risk & Quantitative Methods 2
What is the main limitation of relying solely on internal loss data for operational risk capital modeling?