企業發生財務危機,不只影響企業本身,也讓投資人承受巨大的風險及損失。過去研究採用不同的統計方法及財務變數建立財務預警模型,提高其預測能力。然而,除了財務變數以外,產業結構因素亦與財務危機息息相關。本文旨在檢視產業、市場集中度與產業失敗率三種市場結構變數是否會影響原先建構預警模型的預測能力。研究發現:類神經網絡的整體預測準確率最高,其次羅吉斯迴歸,區別分析最低。再者,發生危機時點的準確率以危機前一年最高,且離危機發生的時間點愈遠,準確率越低。並且三個市場結構變數有助於提升預警模型在危機發生前二年及前三年的預測準確率。最後,依市場集中度、產業失敗率及電子業三者將樣本分群後,高市場集中度群組、高產業失敗率群組、非電子業的預測準確率明顯優於其他群組,並且離危機發生的時間點愈遠,兩者的準確率差距就愈大,因此納入市場結構變數的確有助於提升財務危機預警模型的預測能力。
When financial distress occurs, not only does it affect the company itself, but it also makes its investors bear enormous risk and loss. Previous studies have used different statistical methods and financial variables to construct financial distress prediction models for improving their forecasting ability. However, besides financial variables, industry structures are also related to financial distress. This study examines whether three market structure variables—industry, market concentration rate, and industry failure rate affect the forecasting ability of distress prediction models. The results show that the prediction accuracy of neural network is the highest, then logistic model, and discriminant analysis is the lowest. Secondly, the accuracy of the previous year is the highest. The longer the distress happens, the lower the accuracy is. Next, three market structure variables will improve the previous two and three year’s prediction accuracy. Finally, after classifying by three market structure variables, the prediction accuracy of high market concentration, high industry failure, and non-electronic groups are apparently superior to other groups. Also, the longer the distress occurs, the larger the accuracy difference is. Thus, market structure variables truly improve the forecasting ability of financial distress prediction models.