2004年國內企業如博達、皇統、陞技等公司發生財務危機,此外接二連三發生財報疑雲公司頻傳,顯示這些問題企業並未從1998年起國內多家上市櫃公司發生財務危機的事件中獲得教訓。財務報表數字常常透過不同的方法進行窗飾,藉此美化財務報表資料,大幅降低僅考慮財務變數資料之正確性。因此本研究加入公司治理方面的資訊,並以季資料建構模型,期許增進模型本身的預警能力。本研究以台灣上市櫃電子類股為研究對象,分析2003年至2005年間發生財務危機之公司與正常公司以了解危機公司與正常公司間財務比率與公司治理變數的差異性,並希望藉由公開之財務及公司治理資訊建立有效之財務危機預警模式,保障投資人及債權人之權益。 本研究首先使用因素分析將各變數縮減成數個構面,以其做為羅吉斯回歸模型之自變數,並採取逐步回歸之向前進入法,選取具有影響力之因素,分別建構危機發生前三季之預警模型。研究結果顯示:危機前三季模型之正確區別率均達85%以上,顯示模型具有相當良好之區別能力。
In 2004, some of the domestic enterprises occurred financial crisis such as Bo Da , Huang Tong and Sheng Ji etc. Besides, a series of corporations false account affairs avose, these enterprises didn’t obtain a moral lesson from the financial meltdown in 1998. The financial statements window dressing through diverse ways frequently, herewith glamorize the information in the financial statements, and decrease the correctness enormously owing to the merely observation the financial variables. Hence, this study joins information about Corporate Governance and uses the quarterly data to construct Model expected to improve the Warning ability. This study uses listed electronic companies as an object of study to analyze the difference between the crisis and the normal companies, finding that efficient Fincial Distress warning Models by public financial and Corporate Governance information can protect the rights of the investors and creditors. First, the research uses factor analysis approach to reduce variables, in order to shrink dimension and independent variables of the logit regression models. The adoption of Forward LR also constructs the models before the financial crisis arise. The result of the research shows : the percentage of the discrimination ability is more than 85%, which means that the models possess fine discrimination ability.