本研究使用台灣經濟新報資料庫(TEJ)的資料,利用Merton (1974)的選擇權評價法與Altman(1968)的信用評分法計算各公司在2001年至2007年的違約距離,並使用波動度模型更準確地對選擇權評價法所需要的參數做預測,最後以排序法,ROC 曲線與AUC值來判 斷各模型在估計違約機率的能力。本研究結果顯示(1)GARCH(1,1)下的股價報酬率標準差所導出的DD值,其違約預測能力優於考慮一階自我相關AR(1)或歷史波動度模型所導出DD值的違約預測能力。(2) 考慮一階自我相關GARCH(1,1)下的的股價報酬率標準差所導出DD值,其違約預測能力優於信用評分法。
To calculate the default probability of Taiwan stock market from 2001 to 2007, we conduct Merton (1974) and Altman’s Z-score. In order to use the parameters correctly, we use GARCH process to predict those. We test the power of default prediction by the Receiver Operating Characteristic (ROC) curves and Area under Curve (AUC). The findings of this study are as follows: 1. The GARCH (1, 1) model is superior to the AR(1) model and historical volatility model. 2. The GARCH (1, 1) model is superior to the Altman’s Z-score model.