離散倖存模型與Merton模型的實證--以台灣上市公司為例

Abstract

本文以Campbell et al.(2008)所用的離散倖存模型及Merton違約距離模型來探討台灣上市公司資料的情形。Campbell et al.(2008)特別改善了Shumway (2001) 的變數，以加入股東權益市值的資產總額來取代過去的資產帳面價值，不僅如此，也加入了Merton違約距離模型的違約機率作為離散倖存模型的解釋變數，但本文特別加入了Naive Merton模型的違約機率來作為預測違約風險的變數。由於本文使用季資料，不同於Campbell et al.(2008)使用月資料，因此本文有三個發現: (1) Campbell的變數對違約風險預測能力確實比Shumway的變數佳; (2) 在放入其他會計變數及市場變數時，Naive Merton 模型 與Merton 模型所預測的違約機率作為離散倖存模型的解釋變數並沒有解釋能力，但單獨放入離散倖存模型裡，Naive Merton 模型的解釋能力比Merton模型來的好，這與Bharath and Shumway (2008) 的結果是一致的; (3) 離散倖存模型中無論放入Campbell的變數或Shumway 的變數，其模型精確度都比Naive Merton 模型或Merton模型來的佳。本篇論文結果可為金融機構提供在公司信用風險評估上的一個參考。

This paper uses the discrete-time survival model and the Merton DD model to exploit the companies listed on TSE based on Campbell et al.(2008). Campbell et al.(2008) particularly improve the Shumway’s variables, and replace the book value of the assets with the market value of the assets including the market value of the equity. In addition, we use the default probability of Merton DD model and Naive Merton model as the explanatory variable. Because this paper uses quarterly data which is different from the monthly data used by Campbell et al.(2008), there are three findings: (1) Campbell’s variables have the better predictive accuracy than Shumway’s variables.; (2) The default probability of Merton DD model or Naive Merton model as the explanatory variable in the discrete-time survival model has the relatively little forecasting power after conditioning on the other variables. But when the default probability of the both model is only included in the discrete-time survival model, the explanatory ability of the Naive Merton model is better than the Merton DD model. This result is consistent with Bharath and Shumway (2008); (3) No matter what Campbell’s variables or Shumway’s variables is included in the discrete-time survival model, its accuracy is better than the Merton DD model or the Native Merton model.