EBA欧洲银行-ORA_CP10r_10页_208kb

CEBS Guidelines Summary: Comments on AMA and IRB Approaches

Core Content

This document presents a detailed critique of the CEBS (Committee of European Banking Supervisors) Guidelines issued on January 20, 2006, regarding the implementation, validation, and assessment of Advanced Measurement Approaches (AMA) and Internal Ratings Based (IRB) methods. The comments are from OpRisk Advisory, a firm specializing in operational risk management. The key areas of concern include the definition of risk, expected loss, external data usage, and confidence intervals.

Main Issues and Concerns

1. Definition of Risk

• Problem: The CEBS Guidelines do not explicitly define the term "risk."
• Impact: This leads to inconsistent interpretations across the banking industry.
• Three Interpretations:
  • Definition I: Risk is an incident (e.g., fraud).
  • Definition II: Risk is a measure of uncertainty, specifically negative variance from the mean.
  • Definition III: Risk is the product of likelihood and impact (common in traditional ORM).
• Conclusion: Definition III is flawed and misaligned with Basel II principles. It treats risk as an incident, not a measure, and leads to inaccurate risk assessments.

2. Definition of Expected Loss (EL)

• Problem: The CEBS Guidelines use median instead of mean to define EL.
• Impact: This creates confusion and distorts the concept of unexpected loss (UL), which is the true measure of risk.
• Example: If two businesses have the same median loss, but one has a higher mean loss due to skewness, the former appears more profitable under median-based pricing.
• Conclusion: Using the mean is critical for accurate product pricing and risk-adjusted profitability. Misusing the median leads to systemic risk and poor investment decisions.

3. External Data Uses

• Problem: The CEBS Guidelines allow banks to selectively incorporate external data after scaling and adjusting for controls.
• Impact: This practice is unscientific and can lead to results that vary by a factor of 1000.
• Key Point: External loss data must be used in the context of its distribution, not as isolated data points.
• Conclusion: Banks should adopt objective and scientific methods for using external data. Manual selection and scaling of data points is impractical and leads to manipulation.

4. Confidence Intervals

• Problem: The CEBS Guidelines recommend confidence intervals based on mechanical aspects of the model.
• Impact: This underestimates uncertainty and does not reflect expert judgment.
• Recommendation: Confidence intervals should be based on variations in assumptions and weights derived from expert opinion.
• Conclusion: Confidence intervals should be stress-tested by regulators to ensure scientific validity and transparency.

Key Recommendations

• Define Risk Properly: Risk should be defined as negative variance from the mean, not as a product of likelihood and impact.
• Use Mean for Expected Loss: EL should be based on the arithmetic mean, not the median, to ensure accurate pricing and profitability calculations.
• Adopt Scientific External Data Methods: Banks should use objective methods for integrating external data into their models, rather than manual selection and scaling.
• Include Expert-Based Confidence Intervals: Confidence intervals should reflect expert judgment and be stress-tested by regulators to ensure realistic uncertainty ranges.

Conclusion

The CEBS Guidelines, while well-intentioned, contain flawed definitions and inadequate methodologies that risk undermining the effectiveness of Basel II. If these issues are not addressed, banks may continue to view ORM as a meaningless compliance exercise, rather than a valuable tool for improving risk management and business decision-making. The document emphasizes that accurate and scientific ORM practices are essential for the success of Basel II and the long-term stability of the banking industry.