運用公司信用評等之信用風險模型與破產預測

Abstract

信用評等之用意，乃在於透過授予不同評等，反映公司之間破產機率的差異；然而，相同評等的公司，在破產機率上仍有所不同。故本文提出一個新的混合式信用風險模型，藉由同時涵蓋公司信用評等與公司價值資訊，本模型可以同時捕捉不同評等之間，以及單一評等內的破產機率差異。此外，我們尚導出此模型下破產時間期望值與尾端機率的估計式，以作為模型配適與破產預測之用。最後，我們用實證資料解說我們的模型與市場債券殖利率結構配適良好，同時提出此模型在信用評等研究上的可能應用。

Rating systems imply that firms with distinct ratings have different chances to bankruptcy, but firms within the same rating still defer in bankruptcy probabilities. In this paper, we propose a credit risk model in order to capture these inner- and inter-rating differences. A hybrid model is formulated to incorporate both credit rating and firm individual information in order to reflect these differences. Explicitly computable approximations for the expected bankruptcy time and the tail probabilities of the bankruptcy time are developed for model calibrations and bankruptcy predictions. We further demonstrate our model’s capability in fitting empirical bond-yield structure, and propose an application in rating system evaluation via this procedure.