違約機率之實證研究-離散時間危險模型與Merton模型

Abstract

本文針對台灣上市上櫃公司的資料，以因素分析法選取變數的組合，另外再根據股價資料，別分建立離散時間危險模型與Merton模型，進而估計公司發生違約的機率值，並比較兩種模型的預測準確度。本研究將變數分為四個因子，依序為財務結構、償債能力、經營效能與獲利能力。衡量模型預測準確度的方法有K-S檢定、ROC曲線與AUC值的分析。實證結果發現，兩模型皆可以有效地將違約公司與正常公司區分開來，且離散時間危險模型預測違約機率的表現能力比Merton模型還要來的好。

Based on the data of Taiwan corporations trading in TSE and OTC，this study used factor analysis to choose variables and according to stock data to construct financial distress prediction models, such as discrete-time survival model and Merton’ s model and then estimated the default probability when company went into bankruptcy. Furthermore, I compared the accuracy of two models. This study classified the variables into four categories, which are financial structure、ability to pay、efficiency of administration and ability to profit. The methods used in analyzing the Models’ prediction accuracy are K-S test、ROC curve and AUC. The empirical results showed that these two models can both validate the distribution of independent variables of non-default group differ from that of default group and the discrete-time survival model actually predict the default probability better than Merton’s model. Keywords: Merton；Discrete-Time Survival；ROC Curve；AUC；Factor Analysis