臺灣企業財務危機之預測與衡量

Abstract

信用風險的議題日受重視，其中又以企業違約機率的衡量最受矚目。本研究以KMV模型為理論基礎，嘗試結合標準普爾與中華信評對臺灣企業所發佈的評等及歷史違約機率，透過調整與轉換，發展一違約機率預測模型，裨以符合新巴塞爾協定的部分要求。 本研究另一關切的重點在於，選擇權模型法與信用評分法何者在企業財務危機的預測上有較佳的準確性。因此，本研究以發展的預測模型，及由近期文獻中選取的本土信用評分法模型，透過不同方法檢驗何種財務危機指標具有較高的預測效力。 實證結果發現： 1.就本研究所發展的預測模型而言，在「股價報酬率標準差」、「違約點」兩變數方面，以「考慮股價報酬一階相關之標準差」、「總負債」為較佳之變數設定，對於企業財務危機有較優越的預測能力。 2.比較選擇權模型法與信用評分法兩者的相對效力，在群內分析法、Logit迴歸法方面，兩種財務危機指標均有相當程度的預測能力；而在檢定力曲線方面，結果則支持本研究模型優於本土信用評分法模型。 3.儘管檢定力曲線顯示本研究模型預警能力較佳，但其他二檢驗法仍可驗證本土信用評分法模型具有一定警訊效果，因此，在評估企業信用風險上，兩種方法均值得交互參考以為運用。

Since the late 1990s, the amount of credit risk taken by banks has increased so that the ability to predict company’s default probability has become a critical issue of risk management. This thesis aims to develop a financial distress prediction model, which is based on the KMV model and utilizes historical default probability data from the S&P and Taiwan Ratings Corporation. We hope this model will meet the requirements under the New Basel Capital Accord. Moreover, this thesis focuses on comparing the forecast accuracy of option pricing model and credit scoring model. We use the KMV-S&P model developed in our research as a representative of option pricing model, and choose a Z-Score model for Taiwan’s companies to represent the credit scoring model. Through intra-cohort analysis, logit regression method and power curve, we investigate the relative informativeness regarding financial distress of the models. The findings of our research are as follows: 1.As for the KMV-S&P model, first-order correlation stock price volatility and total debt are the best agents of equity volatility and default point, respectively. 2.We compare the KMV-S&P model and local Z-Score model in their predicting corporate financial distress with the intra-cohort analysis and logit regression methods; the results indicate that both models have significant predictive ability. However, the alternative method, power curve, concludes that the KMV-S&P model outperforms the local Z-Score model. 3.Although the power curve shows that KMV-S&P model is superior in predicting financial distress of Taiwan’s companies, the intra-cohort analysis and logit regression methods indicate that both models have equal predictive abilities. Therefore, we suggest that both models provide incremental information in measuring credit risk.