國內中小企業財務危機預警模型之研究

Abstract

我國中小企業在台灣經濟發展的過程中一直扮演重要的角色，其相對於大企業，營運較有彈性而更能因應市場的變化。但是，近年來全球經濟持續不景氣，使得台灣多數以外銷為主的中小企業經營日益困難。另一方面，銀行對中小企業融資的意願普遍不高(受限於擔保品不足及會計制度較不健全等因素)，也加速中小企業面臨倒閉的威脅。本文研究的目的即是以羅吉斯迴歸探討從量化的財務資料中建構出適用於中小企業的財務危機預警模式，並分析不同模型效果。本研究希望藉此可讓銀行債權人或經營者適時發現企業財務體質惡化的徵兆，以便及早提出解決方案，避免企業發生財務危機。 根據實證結果，本研究得到以下結論： 1.模式一的變數在危機前一年有負債比率、總資產週轉率、總資產報酬率；危機前二年有流動比率、淨值成長率；危機前三年有流動比率、淨值週轉率。模式二的變數在危機前一年有流動性與償債能力、獲利能力；危機發生前二年有償債能力與成長力、獲利能力；危機發生前三年為獲利能力與經營效能。 2.二個模式的分類預測能力均為顯著，具有可行性及預測效果。其中，六個子模式的整體預測正確率達85﹪以上，對危機企業預測正確率最低亦有83.3﹪。 3.整體而言，模式二之預測能力相對優於模式一，表示以主成份分析後，資料構面經過縮減，對危機預警模型的解釋與預測能力具有提高的效果。

The small and medium enterprises(SMEs) in Taiwan play an important role in the process of economic development all the while. Relative to large enterprises, the operation is more elastic in the face of market change. Nevertheless, caused by the global economic downturn recently, it becomes more difficult to handle the business for SMEs depend on export trade mostly. Besides, an inclination that the bank loan to SMEs is commonly low-level through insufficient collaterals and the imperfect accounting system, and it accelerated the bankrupts of SMEs. The purposes of a study were to build a financial distress prediction model for SMEs using Logistic Regression from financial ratios and to analyze various model-effects. We attempt to make creditors or managers find timely signs of the aggravation of the financial constitution by the model, so as to address some solutions in advance to avoid the happening of financial Distress. According to the results of empirical study, the conclusions are as follows: I.Variables of model 1 in year 1 are Liability Ratio、Total Asserts Turnover、Return on Total Asserts; in year 2 are Current Ratio、Equity Growth Ratio；in year 3 are Current Ratio、Return on Equity. Variables of model 2 in year1 are Liquidity and Solvency、Profitability; in year 2 are Solvency and Growth Ratio、Profitability; in year 3 is Profitability and Operation Activity. II.The classifications of both models are significant, and possessed of feasibility and effects. In six sub-models, the prediction ratios for all are above 85%. There are 83.3% at least for distressed enterprises. III.As a whole, model 2 is better than model 1 because of cutting down the data sections, which is using Principle Component Analysis.