經營效率與KMV模型於信用風險管理機制有效性之研究

The Effectiveness of Operating Efficiency and KMV Model on the Credit Risk Evaluation Model

過去許多學者利用財務、非財務變數來預測公司財務危機，但卻沒有研究將經營效率與違約機率同時納入信用風險管理機制模式中，以檢驗是否能提升信用風險評估之準確率。因此，本文以2005~2009年發生危機事件的54家台灣資訊電子業公司及108家正常公司做為研究對象，建立總共162家公司的研究樣本。實證分析分為三個步驟：首先，利用KMV模型計算樣本公司的預期違約機率（expected default frequency, EDF）；其次，以三階段資料包絡分析法（data envelopment analysis, DEA）計算樣本公司之總技術效率、純技術效率與規模效率；最後，將各財務變數、EDF及三階段DEA所計算的效率值納入logit模型中，以建構完整的信用風險管理模型。實證結果發現：第一階段與第三階段DEA所計算之效率值明顯不同，且危機公司與正常公司之財務變數、KMV模型之EDF及三階段DEA之效率值確實存在顯著差異；而在logit迴歸分析方面則發現，於危機事件發生前一年，EDF及三階段DEA之效率變數具有顯著之解釋能力，再者，納入三階段DEA效率變數及EDF後之模型，其預測準確率較純財務變數模型之準確率提高，顯示將三階段DEA效率變數與EDF納入信用風險管理模型中，將可提高模型之準確性。

Many researchers used various financial and non-financial data to predict a firm’s financial crisis, but there was not a study taking the impact of operating efficiency and expected default frequency (EDF) into consideration.Adopting the sample data of Taiwan electronic industry from 2005-2009, we select 54 financially shaky firms and 108 normal ones as match firms in the empirical implementation.The empirical analysis proceeds in three stages.First, we use KMV model to estimate the EDF of sample companies and then we estimate over-efficiency, technical efficiency and scale efficiency by means of three-stage data envelopment analysis (DEA).Finally, we incorporate the financial variables, EDF and operating efficiency variables into logit regression model to establish a credit risk evaluation model.Experimental results indicate that the efficiency measured by one-stage DEA obviously differs from the one measured by three-stage DEA.Furthermore, there exists differences between shaky firms and normal ones when applied to the financial variables, EDF, and three-stage DEA efficiency variables.In the results of logit regression, we find that the EDF and operating variables have significant explanatory power in one year prior to the financial crises of sample firms.Moreover, the predictive results indicates that the models used financial variables with EDF or financial variables with three-stage DEA efficiency variables are more correct than the model used only financial variables. Thus, the incorporation of operating efficiency and EDF into the credit risk evaluation model proves to enhance its overall accuracy.