[[alternative]]Efficient Simulation and Approximation of the Default Probability of Credit Risk Risk under Normal Copula Model

計畫編號：NSC101-2410-H032-033 研究期間：201208~201307 研究經費：488,000[[abstract]]本研究考慮在信用風險的Normal copula模型下，採用靈活且彈性高的蒙地卡羅模擬法計算破產機率，並探討利用重點抽樣法降低蒙地卡羅模擬法的計算成本高的問題。採用最新的中偏差法計算出最佳效率的重點抽樣法，提高蒙地卡羅模擬法的計算效率。 實證方面，從 Datastream 及台灣經濟新報資料庫蒐集一些國內外股票上市公司的資料，以常用的Z分數、ZETA等建議的風險因子變數，作為Normal copula模型的控制風險因子，並估計其參數。利用蒙地卡羅模擬法程式來估算破產機率並與實務常用的KMV模型、信用矩陣模型做比較，並用CreditRisk+模型估計風險資本，提供信用風險管理者做為決策參考的依據。[[abstract]]In this research plan, we consider using flexible and elastic Monte Carlo simulation to evaluate the default probability of Credit Risk under the Normal copula model, then use importance sampling to reduce the high computational cost problem of Monte Carlo simulation method. Apply the latest moderate deviation method to calculate the optima efficiency of the importance sampling method, and improve the computational efficiency of Monte Carlo simulation. Empirically, we can collect some prices on domestic and foreign stocks from Datastream and TEJ databases. Use the risk factors of Z-score and ZETA model to regard as the risk factors of Normal copula model. Then, estimate the unknown parameters of the Normal copula model. Finally, run the Monte Carlo simulation program to evaluate the default probability of Credit Risk and compare with popular KMV, RiskMatrics model and employ CreditRisk+ model to estimate risk capital.[[sponsorship]]行政院國家科學委員