公司破產預警模式之建立－－以台灣上市公司為例

A Study of Bankruptcy Forecasting Models: Evidence from Taiwan

本研究的目的是為突顯台灣企業破產的財務特徵，建立具有較高準確度的破產預測模型。利用單變數分析主要是分析74組配對的正常公司與破產危機公司，從1990年至2012年所有破產前三年的產業關聯財務比率（Industry relative Financial Ratio, IRR）的結果，顯示這具有很高的鑑別效度，可以用來區分危機或破產公司和正常公司。隨後，建立我們所提出的IRR logistic迴歸模型，本研究預測分析模型顯示，優於Altman的Z分數模型，在破產前一年，具有較高的預測準確度。實證研究結果顯示，我們所提出的破產模型比Altman的Z分數模型，更適合評估台灣的破產公司。

The purpose of this study is to highlight the financial characteristics of bankrupt companies in Taiwan, and to build bankruptcy forecasting models with higher forecasting accuracy. Univariate analysis results for all three years prior to bankruptcy for 74 equally matched bankrupt and nonbankrupt public companies from 1990 to 2012 show that the debt to assets industry-relative ratio (IRR) possesses the highest discriminatory power for differentiating between bankrupt and nonbankrupt companies. Subsequent forecasting analysis of model building shows that our proposed IRR logistic regression models outperform Altman's Z-score models, and possess high classification forecasting accuracy one year before bankruptcy. Ultimately, the empirical results suggest that our proposed bankruptcy models are more suitable for assessing the likelihood of firm bankruptcy in Taiwan than Altman's Z-score models.