企業的經營狀況最終會反應在財務報表上，但投資人卻需要等到財務報表揭露之後才能得知公司是否出現危機，過去許多學者利用非財務、股價及整體經濟因素等資訊來幫助提早預測企業財務危機的發生，但卻忽略了經營效率對於企業發生財務危機的影響，因此本研究嘗試從過去文獻常使用的財務變數中再加入相對效率值來建構財務危機預警模型，期望能增加預警模型的預測能力，且進一步證明財務危機與經營效率的關聯性。
本研究針對台灣市值比率最大的資訊電子業建立財務危機預警模型，選取民國九十一年初到九十四年底之間被列為全額交割股或下市、櫃的56家公司為財務危機公司，以1：2的配對方式建立總計168家公司的研究樣本。利用過去文獻常用的財務變數以及由資料包絡分析法計算而來的效率值進行預警模型的建構，首先利用財務變數進行基準模型的建構，接著逐一加入三種效率值：總技術效率、純技術效率、規模效率進行改進預測模型。
本研究使用Logit模式以及類神經網路比較四種模型的預測能力，實證結果顯示：在預測正常公司時使用財務變數搭配規模效率的模型以及在預測危機公司時使用財務變數搭配總技術效率或規模效率的兩種模型，相較於單純使用財務變數的預測模型都有較高的預測正確率。而此現象也證明了財務危機的發生確實與公司經營績效的低落有關聯性，且此跡象在發生危機的三年前就已經開始顯現。此外危機公司在發生危機前三年開始每年的固
定資產均有持續減少的現象，因此應該被視為一個重要的區別變數。而本研究採用Logit模型所建構的預警模型之預測正確率相較於以類神經網路所建構的預警模型來的高。
The operating status of a firm will be presented at its financial reports after all. However, it is too late that the investors knew a firm's financial crisis because they generally knew it after the firm disclosed its financial reports. Many researchers used some information such as non-financial data, market price, macro-economic factors, to predict a firm's financial crisis. However, the effect of operating efficiency is also important factor in predicting a firm's financial crisis. In this research, the authors tried to structure the financial predictive model by using some efficient factors and to examine relationship between financial crisis and operating efficiency.
The authors collected 56 crisis firms in Taiwan electronic industry and match with 112 normal companies during 2002 to 2005.The model uses the financial variable and operating efficiency to predict the business crisis. First, predictive model only using financial variable to establish pattern model after including over-efficiency、technical efficiency、scale efficiency in turn.
The authors compared the predictive results from four models by logit model and artificial network approach. It indicates that the approach used both financial variables and scale efficiency to predict a firm's crisis is more correct than the approach used only financial variables. It is also proved that there is a correlation between financial crisis and unhealthy operating performance. Moreover, it is found that the volumes of the fixed assets at a firm with crisis were declined gradually since 3 years ago. This factor should be recognized as an important discriminative factor. In this research, the model structured by logit model is better than the one structured by artificial neural network approach.
