2012-05-14-奥纬咨询-Lessons_Learned_from_Internal_Model_Developement_10页_310kb

Financial Services - Lessons Learned: Internal Model Development

• Balancing Act: Model developers face conflicting demands from Risk Functions (accuracy/consistency), Business (avoiding capital constraints), and Model Validation (regulatory compliance). The model must be Realistic, Understandable, Sufficient, and Practical.
• Development Process: Establish the correct team and structure early, covering stakeholders from the beginning. Define clear principles and use these consistently.
• Scope and Prioritization: Focus resources on high-impact risks/items using metrics like existing model outputs, sensitive asset values, and technical provisions. Less material items can simplify modeling.
• Data and Calibration: Choose appropriate data sources robustly and apply sound calibration principles. Avoid over-fitting data and ensure "explanation plausibility." Strive for robustness through alternative approaches (data/methods) and expert review.
• Calibration Balance: Avoid overly complex or exotic distributions. Use understandable methodologies and balance fit/simplicity, leading with the balance between accuracy and manageability.
• Expert Judgment: Incorporate expert judgment to sanity-check results, especially when statistical methods or data are limited. Ensure capital requirements are not overly conservative, factoring in appropriate business levers (shocks, events, risk mitigants).
• Model Limitations & Recalibration: Acknowledge and understand model limitations. Define triggers for recalibration (sensitivity analysis, recap). Complement models with deterministic stress testing.
• Conclusion: No "cookie-cutter" approach exists; each project is unique. Follow an iterative, principle-driven process, accepting necessary trade-offs, especially in regulatory communication and capital optimization.