常數彈性變異數過程下的結構式模型破產預測分析

Abstract

本篇論文使用結構式模型來衡量信用風險，假設公司的資產價值分配及不同破產界線假設下，觀察是否能夠準確預測出公司破產的發生，以及探討不同的分配是否更能描述真實公司價值，並進行實證分析。在金融風暴期間，我們分別針對金融公司未來三個月、六個月及一年預測違約機率來比較模型的優劣。在布朗運動假設下，使用內生界限架構下的模型較能準確的預測出公司違約的發生，在常數彈性變異數過程下，使用歐式買權架構下的模型較能準確的預測出公司未來的違約。

In this paper we measure the credit risk under structural model. We include two different processes under three different structural models and compare which model has best ability to predict default probability if the firm is going to bankruptcy. During financial crisis period, we estimate the default probability of financial companies in future three months, six months, and one year. In our empirical result, we conclude that when we use the endogenous barrier framework under geometric Brownian motion (GBM), it will have more power to predict default than other models. On the other hand, we use the European option framework under Constant Elasticity of Variance (CEV) process has more powerful to predict default than other models. In summary, using the endogenous barrier framework under GBM has the most powerful prediction on default.