Operational Risk - Model Risk

Source :

1. incorrect model specificiation
2. incorrect model application
3. implementation risk
4. calibration error
5. programming errors
6. data errors

Quantifying model risk :

• single unknown parameter
• two unknown parameter
• unknown correlations
• mixing parameter and distribution risk

Managing model risk :

• 瞭解 model risk 是不可避免的
• 分析 model performance 對主要假設的敏感性
• 用已知方法驗證 model 針對簡單問題的正確性
• 選用可以提供有用結果且最簡單的 model
• place more confidence in a model been backtested and stress tested
• 注意小問題
• 定時評估更新模型

Independent risk oversight (IRO) :

• unit head reports directly to CEO
• sit on BoD to communicate performance and strategy

Model risk implication :

• 如果金融商品的正確價格不管任何時候都能取得，則 model risk 不存在
• 金融商品不用太複雜也會有 model risk exposure
• 在金融機構 trading book 或 banking book 上的資產應被用不同方式分析

Model risk under EMH :

• Source of risk
• not having sophisticated or realistic enough model

Model risk outside EMH :

• factors limiting arbitrages
• 時間限制 (limitation on timeframes)
• regulatory and risk management constraints
• 流動性差 (poor liquidity)
• 資訊成本高 (high information costs)
• changes in pricing relationships
• 市場流動性需求改變
• 風險趨避改變
• change in accepted wisdom
• source of model risk
• shift from "finding the best model" to "how today's pricing methodologies may change in the future"

Model risk manager :

• Primary objectives
• Identify model currently used to arrive at observed market prices
• Identify how pricing methodologies used today may change in the future
• Key tasks
• Gather market intelligence
• Reserve engineering of prices