運用財務比率及公司治理變數建構財務危機預警模型之研究:單期與多期資料之比較

由於全球企業之財務危機事件陸續爆發，進而影響至世界各國經濟而導致破產、倒閉以及改變經營型態。而台灣也在本土型金融風暴後，接連傳出掏空、違約交割，甚至是企業之管理階層過度運用財務槓桿或非法操作造成周轉不靈，最後導致財務危機。過去在財務危機之探討中較多是以單期資料來進行研究，因此本研究欲以單期與多期之資料來做比較，初步將單期資料透過常態性檢定、平均數檢定以及共線性檢定找出在所有財務比率以及公司治理變數中，有顯著效果較明顯之24個變數，並運用羅吉斯迴歸，以最後的9個較顯著自變數建立單期資料(危機前一年)之模型1，並計算預測分類準確率。同樣對多期資料進行檢定後，將19個顯著效果較明顯之財務比率、公司治理變數來運用於羅吉斯迴歸中，以最後的7個顯著自變數建立多期資料(危機前三年至前一年)之模型2，並計算預測分類準確率。實證結果發現，危機前一年中有五個變數在訓練組和測試組有較顯著的影響與解釋力，分別為財務比率變數中的負債比率、現金流量比率，公司治理變數中的外部個人持股、有息負債利率以及必要控制持股。其中負債比率、外部個人持股、有息負債比率和必要控制股數皆與財務危機之發生呈正向關係，現金流量比率和財務危機之發生呈負向關係。危機前三年至前一年中有三個變數在訓練組和測試組有較顯著的影響與解釋力，分別為淨值週轉率(次)、總資產週轉率(次)以及當季季底PB，且與財務危機之發生皆呈負向關係。在危機前一年之訓練組及測試組的預測準確率可達87.3%和85.4%，在危機前三年至前一年之訓練組及測試組的預測準確率可達91.4%和98.8%，故本研究模型可用於台灣上市櫃公司來檢驗財務危機之預測準確率，作為風險防範之參考。

關鍵字

財務危機；財務比率；公司治理變數；羅吉斯迴歸

並列摘要

Due to the occurrence of financial crisis in global enterprises, it affected the world’s economies and led to fraudulent bankruptcy and changes in business models. Also, the occurrence of emptying and credit default swap after the local financial crisis in the Taiwan. And even the excessive use the financial leverage or illegal operations by the management of the company, resulting in a financial crisis. In the past, almost discussion and research on single-period data. Therefore, this research is divided into single-period and multi-period data to compare. First, using the Normality Test, T-Test and Collinearity Test on single-period data (one year before financial crisis). Then, find the 24 significant variables of the financial ratio and corporate governance variables. After using these significant variables on Logistic Regression Model, build the single-period data’s model 1 with remaining 9 significant variables and calculate the predictive accuracy rate. Similarly, in the multi-period data (one to three year before financial crisis), using the remaining 19 significant variables of the financial ratio and corporate governance variables on Logistic Regression Model. Then, build the model 2 with 7 significant variables and calculate the predictive accuracy rate. The empirical results show that the five variables in single-period data have the significant level and explanatory power in the training group and the test group, which are the the debt ratio and cash flow ratio of the financial ratio, external individual shareholding, liability with interest ratio and necessary controlling shareholding of the corporate governance variables. Among them, the debt ratio external individual shareholding, liability with interest ratio and necessary controlling shareholding have the positively related with the occurrence of financial crisis. The cash flow ratio has the negatively related with the occurrence of financial crisis. There have three variables in muti-period data have the significant level and explanatory power in the training group and the test group, which are the net turnover rate (times), the total asset turnover rate (times) and the price book ratio. All of them are negatively related with the occurrence of the financial crisis. In single-period data, the predictive accuracy of the training group and test group could be approximately up to 87.3% and 85.4%. In muti-period data, the predictive accuracy of the training group and test group could be approximately up to 91.4% and 98.8%. As a result, the research models are useful for banks to prediction accuracy if the companies have the possibilities of the occurrence of financial crisis, it could be the prevention.