本研究分為二部分，首先採用KMV信用風險模型以選擇權評價模型計 算樣本公司預期違約機率是否明顯不同；接著將無風險利率替換為ROA及資產成長率後以分別檢測在納入不同變數後是否更能有效捕捉台灣上市企業發生違約之風險；接著將預測樣本(2005~2007)之各項財務變數及非財務變數納入Logistic模型，並配對正常公司樣本，以比較兩者違約機率；並利用預測樣本建構出的違約機率模型預測確認樣本(2008)財務危機出現的機率，並觀察預測正確率。 實證結果發現，Logistic與KMV模型在違約前一季均可有效區別正常公司與違約公司，提供預警效果。Logistic模型中，正常樣本公司在考量財務因子或結合財務及非財務因子下而得的違約機率在時間序列上均無明顯變化；而財務危機樣本公司的違約機率則均明顯較正常樣本公司為高，且隨違約時間點接近有明顯提升的趨勢，唯加入非財務因子後違約機率並無明顯改變；但代入確認樣本觀察預測準確度後發現，加入非財務因子之正確率較高。在KMV模型中，則發現越接近違約事件發生時間，違約機率有明顯提升的趨勢，且危機樣本與正常樣本間的差異亦逐漸顯著；若以ROA及資產成長率替代無風險利率參數，可得一考量個別公司狀況下更合理的違約機率，且兩種類型公司的判別度亦有效提升。 This study evaluates two kinds of credit risk models. First one is Moody’s KMV model, and the other is Logistic Model. First, in KMV model, we calculate the average default rate during 260 days before the event time. In advance, we replace the parameter of risk free rate by ROA and Asset Growth Rate to evaluate the effective of these three parameters in our KMV Model. We also collect samples include firms which have ever declared some financial distress firms and normal firms. Then, in Logistic Model, we imply model with only financial variables and model with both financial and non-financial variables to calculate the potential default rate during the sample period. And moreover, we further test if Logistic model can identify the default events in 2008. Our results suggest that both Logistic and KMV models can successfully identify the default firms. In Logistic Model, we find the default rate show a positive trend as the default time being close. On the other hand, although we cannot get a significant different default rate under models with only financial variables and with financial and non-financial variables, model includes non-financial variables can more exactly identify default firms in 2008.KMV model also suggests an increasing default rate on default samples as event time being close, while default rate keeps in consistently low level on normal firms. Besides, after we replace risk free rate by ROA and Asset Growth Rate, we get a higher default rate among total samples because of consideration about specific firm’s risk condition rather than risk-neutral assumption. Finally, by including ROA and Asset Growth Rate in the models, we find much significant difference of default rate among financial distress firms and non-distress firms.
