The Application of Value-at-Risk Model and Stress Testing Approaches-the case for Financial Tsunami

[[abstract]]本論文針對2008年的金融海嘯為例，以及亞洲金融風暴、無壓力事 件發生時期的比較，來探討風險值模型與壓力測試的效用。本研究在 投資組合中先利用四種風險值模型：變異數-共變異數法、歷史模擬 法、蒙地卡羅模擬法以及極值理論搭配上T-copula關聯結構，再根據 正確性、損失幅度及保守性進行回顧測試評估模型績效，選擇出最佳 的風險值估計模型。接著歸納出壓力情境和兩種壓力測試法下之檢 定，檢驗在金融海嘯中，市場遭受衝擊發生損失時，壓力測試值的表 現是否能彌補風險值的不足。 實證結果顯示，在95%信心水準下，亞洲金融風暴融時期與無壓力 事件發生時期之壓力值約可覆蓋一半的風險值失敗天數，而金融海嘯 期間下，兩種壓力值仍然無法覆蓋其大幅損失程度。[[abstract]]This paper takes Financial Tsunami in 2008, Asia Financial Crisis and none of pressure events periods as examples to compare the performance of Value-at-Risk(VaR) model and Stress testing approaches. First, we use four VaR methods：the variance-covariance method, the historical simulation method, Monte Carlo simulation method and T-copula method to estimate the VaR values. Then, we employ accuracy, expected shortfall and conservatism in the backtesting of VaR models as criteria to choose the best one. Last, we find the pressure events and examine whether stress test value can make up the deficient of VaR models when the market suffered huge loss in Financial Tsunami? The results reveal that stress value can cover about half of the exceptions in VaR models in Asia Financial Crisis and none of pressure events period with 95% confident level. However, both of stress values still can’t avoid huge loss effectively in Financial Tsunami