摘要 本研究在探討是否經由財務理論發展的指標與新預測方法的運用,建構一個具有高度預測能力的財務危機模式,以利當企業可能發生危機之前產生早期預警機制。研究樣本以民國八十七年至九十二年間公開上市公司有董監事股權質押之財務資料為研究對象,危機與正常公司樣本以1:2的配對方式進行,選取相同期間、產業相同、規模相近公司為配對基礎。在研究變數方面則擷取有關於(1)營運獲利能力;(2)流動性能力;(3)財務結構;(4)應計基礎;及(5)風險控制;五大結構財務變數進行探討。研究模型採用二元Logistic迴歸模型與支向機模型,比較兩種模型在相同財務比率資料下,檢視是否在建構企業危機預警模式之預測能力與穩定性方面有不同的結果。 實證結果顯示: 1.應計基礎之人為盈餘管理與公司發生財務危機之可能性是呈現正相關。不論是危機公司或正常公司,管理當局普遍有利用裁決性應計項目來做盈餘管理之傾向。 2.董監事股權質押比率與公司發生財務危機之可能性是呈現正相關。且顯示董監事股權質押比率會因為越接近危機發生時點有越明顯之預警訊號。 3.在相同的財務資料情況下,支向機與傳統Logistic所建立之企業財務危機預警模式,在危機發生當季的預測能力皆為最佳。 4.在相同的財務資料情況下,支向機所建立之企業財務危機預警模式其預測能力及穩定性均較傳統Logistic所建構之模式為佳。
ABSTRACT The purposes of this study is to use financial distress signals and new prediction method to develop a advanced distress prediction model。The distress early warning system developing can assist managers to avoid possible losses and wastage of resources, to act as an important investment guide for the investors as well。The study data are collected from the publication of those companies whose stocks are publicly traded on the Taiwan Stock Market between 1998 and 2003。Using 1:2 matched pairs sample method,distress and non-distress companies selection based on the same period time、same industry and the similar size company as matched sample。The study selects several financial ratios as independent variables including (1)Profitability;(2)Liquidity ;(3)Financial Structure;(4)Accruals;(5)Risk Control。Using both the Support Vector Machine and the Logistic Regression model to build the distress prediction models。Finally, we would compare both the prediction accuracy and the stability between the Support Vector Machine and the Logistic Regression Model。 The findings and conclusion of this study is listed below: 1. The empirical results indicating the association between the discretionary accruals and the company financial distress is positively correlated。Earnings manipulation is a common phenomenon whether the distress company or the non-distress company。 2. The findings also indicating the association between companies’ the extent of shares as collateral by the board of directors and financial distress of the company is positively correlated。The extent of the shares as collateral by the board of directors in the distress company’s will have a significant indicator when company is gradually approached the early distress stage。 3. Under the same financial ratio data, the classification accuracy is the best performance in the current season both using Support Vector Machine and the Logistic Regression Method 。 4. However, under the same financial data,both the predictive ability and the stability of the Support Vector Machine is better than The Logistic Regression Model。