CORES Loss Data Collection & Analysis 5 — Questions and Answers
Question 1: A firm recovers a portion of an operational loss through legal action three years after the original event. How should this recovery be treated in capital modeling?
- Reduce the original loss event's gross amount retroactively
- Record as a separate positive cash flow with no link to the original loss
- Record as a recovery against the original event, reducing net loss but preserving gross loss history (Correct answer)
- Exclude from the loss database as it falls outside the modeling window
Correct answer: Record as a recovery against the original event, reducing net loss but preserving gross loss history
Recoveries should be recorded against the original loss event to maintain accurate net loss figures while preserving gross loss data for frequency/severity analysis.
Question 2: What is the primary challenge of using Extreme Value Theory (EVT) to model operational risk loss severity?
- EVT requires normally distributed data
- EVT is only applicable to market risk
- EVT requires a sufficiently large sample of tail observations that may not exist in internal data (Correct answer)
- EVT cannot model losses above a certain threshold
Correct answer: EVT requires a sufficiently large sample of tail observations that may not exist in internal data
EVT's accuracy depends on having enough extreme observations to fit a reliable tail distribution, which is difficult given the rarity of catastrophic operational losses.
Question 3: In the context of loss data collection, what does 'gross loss' specifically exclude?
- Insurance recoveries and proceeds from litigation settlements (Correct answer)
- Losses from failed transactions
- Losses attributed to third-party vendors
- Losses below the minimum reporting threshold
Correct answer: Insurance recoveries and proceeds from litigation settlements
Gross loss is the total financial impact before subtracting any insurance recoveries, indemnifications, or other compensations received.
Question 4: Which control is MOST effective at preventing loss events from being misclassified into the wrong Basel II Level 1 event type?
- Automated system flags based on transaction amount
- Mandatory second-line-of-defense review of event type assignments (Correct answer)
- Annual training on Basel event type definitions for all staff
- Requiring business lines to self-certify the event type
Correct answer: Mandatory second-line-of-defense review of event type assignments
Independent second-line review ensures business line self-reporting does not introduce systematic classification bias driven by incentives or misunderstanding.
Question 5: A financial institution's operational loss database shows a high frequency of low-severity losses but very few high-severity losses. Which statistical distribution property does this most likely indicate?
- Thin-tailed (platykurtic) severity distribution
- Heavy-tailed (leptokurtic) severity distribution (Correct answer)
- Symmetric normal distribution of loss severity
- Uniform distribution of loss events
Correct answer: Heavy-tailed (leptokurtic) severity distribution
Operational risk severity distributions are typically heavy-tailed, meaning severe events are rarer but far larger than a normal distribution would predict.
Question 6: When performing a loss data collection program gap analysis, which finding would most urgently require remediation before using the data in an AMA capital model?
- Some events are missing secondary business line codes
- The database contains only 3 years of history for a key business unit (Correct answer)
- Event descriptions average fewer than 50 characters
- Recovery amounts are captured quarterly rather than monthly
Correct answer: The database contains only 3 years of history for a key business unit
Only 3 years of data is insufficient to credibly model tail risk and fails to meet the Basel minimum of 5 years (3 years during initial implementation) of quality data.
Question 7: Which of the following best describes the concept of 'expected loss' (EL) in the context of an operational risk loss distribution?
- The single largest loss observed in the historical dataset
- The average annual loss that is anticipated based on historical frequency and severity (Correct answer)
- The loss amount at the 99.9th percentile confidence interval
- The minimum loss threshold required for regulatory reporting
Correct answer: The average annual loss that is anticipated based on historical frequency and severity
Expected loss represents the mean of the loss distribution, reflecting what a firm anticipates losing on average per year from operational risk events.
CORES Loss Data Collection & Analysis 5
A firm recovers a portion of an operational loss through legal action three years after the original event.
How should this recovery be treated in capital modeling?