本研究以台灣1996 年至2008年之曆年制上市櫃公司中曾經發生財務危機之公司為樣本,而配對樣本公司之選取,則以相同產業且公司資產總額近似之公司為配對公司。本研究採取1:4的比率來降低財務危機公司過度抽樣的缺點,最後本研究的研究樣本是危機公司190家,健全公司760家。本研究經由文獻探討參照公開說明書所使用的會計資訊後,選取23項變數,進行獨立樣本T檢定,檢測上述23項變數的平均數在財務危機公司與配對之財務正常公司有顯著差異的14項變數,納入Logictic迴歸中進行下一步分析,實證結果發現:在財務危機發生前一年(T-1),股價標準差與董監質押比率對財務危機的發生有顯著正向影響,而淨營運資金對資產總額比值、淨值市價對數值-即公司規模及現金流量比率對財務危機的發生有顯著負向影響;在財務危機發生前二年(T-2),股價標準差、每股盈餘及董監質押比率對財務危機的發生有顯著正向影響;在財務危機發生前三年(T-3),總資產週轉率對財務危機的發生有顯著負向影響,假說1-9獲得支持,董監持股率與董監質押比率對財務危機的發生有顯著正向影響。在截斷點為0.5下,本研究的財務危機預警模型分類正確率方面:財務危機發生前三年之分析樣本的整體預測正確率為89.4%、88.3%、88.3%,驗證樣本的整體預測正確率為80.1%、83.7%、79.8%。在最適截斷點(cutoff point)下,型Ⅰ誤差(亦即,預測為財務健全公司,實際上為財務危機公司的機率)與型Ⅱ誤差(亦即,預測為財務危機公司,實際上為財務健全公司的機率)之和最低,本研究發現在財務危機發生前三年分析樣本之截斷點分別在0.61、0.55、0.44;驗證樣本之截斷點在0.96、0.68、0.79,型Ⅰ誤差與型Ⅱ誤差合最小。
This study chooses listed companies which have occurred financial crises to be research samples, and chooses listed companies in the same industry which have similar total assets to be matched companies in 1996 to 2007. This study take the ratio of 1:4 to reduce the shortcomings over sampling of the financial crisis companies, the final samples in this study: 190 companies in financial crisis , 760 companies in financial healthy . The electronic industry and construction industry have the highest ratio of occurring financial crisis, the financial crisis ratio are 34.2% and 16.3%, respectively. This study follows 5 models which included Altman (1968), Zmijewski (1984), Shumway (2001), Hsinan Hsu et al .(2007) and Ruey-Ching Hwang et al (2007), and select 23 variables to proceed independent samples T-test, and find that there exist 14 variables have significant difference between financial crisis and healthy companies. This study then use the 14 variables to proceed Logit regression analysis. The empirical results show that both the standard deviation of stock prices and pledge ratio of supervisors and directors have significantly positive effects on financial crisis, while the ratio of net working capital to total assets, the natural log of net market values and cash flow ratio have significantly negative impacts on financial crisis in the previous year before financial crisis(T-1). The standard deviation of stock prices, earnings per share and pledge ratio of supervisors and directors have significantly positive effects on financial crisis in the two years ago before financial crisis (T-2). The total asset turnover has significantly negative impacts on financial crisis, the ownership of supervisors and directors and pledge ratio of supervisors and directors have significantly positive effects on financial crisis in the three years ago before financial crisis(T-3). Under the cutoff point of 0.5, this study of the financial crisis early-warning model of the correct classification rate: the first three years of financial crisis to analyze samples of the overall prediction accuracy rate of 89.4%, 88.3%, 88.3%, to verify the correct sample rate of the overall forecast for 80.1%, 83.7%, 79.8%. Optimum cut-off point, the typeⅠerror and typeⅡerror and the lowest, the study found that the financial crisis in the first three years of cut-off point to analyze samples 0.61,0.55,0.44, respectively; validation samples in the cut-off point 0.96,0.68,0.79, typeⅠerror and All the smallest typeⅡerror.