企業危機預測---結合Logit模型與Merton模型

Abstract

在近三十年來發展的財務危機預警模型中，以會計資訊為基礎的logit模型，由於近來頻傳會計報表不實的事件，導致其預警能力備受質疑；而以市場股價資訊為基礎的Merton模型，其對財務危機的預警能力亦遭受到市場不效率及資產價值為常態分配的批評。 本研究嘗試結合上述二類財務危機預警模型，將Merton模型計算出的違約機率引進logit模型中，透過會計資訊與市場資訊的結合，彌補二類模型在企業破產預測上的不足，以期達到更高的預測正確率。 研究結果顯示，利用Merton模型計算出的違約機率在logit模型中不僅能顯著區分危機企業與正常企業，同時，引進市場資訊後的logit模型也由於擴大了可用的資訊集合，而使得預測正確率能有效提升。

Due to the decreasing transparency of the financial reports, traditional logit models, usually employing accounting numbers as predicting variables, are facing questions of their effectiveness. However, the market value based corporate credit model, such as Merton type models, are also encountering problems of market inefficiency and inappropriate assumptions of value distribution. The purpose of this study is, therefore, to improve traditional failure prediction models by proposing an integrated model that incorporates both accounting and market credit information. We introduce the variable, Expected Default Frequency (EDF) generated from Merton model, as a predicting variable into traditional logit model. The empirical results show that Expected Default Frequency variable is not only significant in the model but also have higher failure predictive power than the traditional model.