信用風險變遷的動量特性研究

Abstract

馬可夫過程是早期信用風險評價模型的重要假設之一。隨著評價技術的進步，此假設被考慮為過於簡化和失實。在評價信用衍生性商品的過程中，信用評等是必需的資料，但實務上可觀察的信用評等往往是落後指標，導致評價結果偏誤。另外，信用評等的發佈頻率(如每年一次)可能無法滿足即時評等的需要，如銀行授信時需估計企業未來的償還能力，卻只能獲得一年前的公司評等。本研究根據個別公司的特質（如負債比率、產業特色、公司治理狀況等），捕捉違約風險的動量效果(momentum effect)，證明排除馬可夫假設將使違約機率或移轉機率的估計更為準確且貼近現實狀況；接而進一步把違約風險和景氣循環的關係分離，說明動量效果和景氣循環的關係。非馬可夫過程的成立將說明公司未來的信用評等不僅和現在的信用狀況有關，更是和過去的信用狀況息息相關。

The credit migration matrices, which are widely used in credit risk pricing, are typically assumed to be generated by a simple Markov process. But the recent literature and empirical evidence suggest that non-Markovian behavior, or momentum effect, of rating process is mounting. Moreover, techniques that price credit risk derivatives need historical data of credit ratings. However, credit rating is always lagged and thus makes biased valuations. In addition, the frequency of credit rating announcement by rating agencies may not satisfy the purposes of industrial practice. For example, banks would prefer “point in time” credit information to “through the cycle” credit ratings. In this study, we capture the trend of corporate default probability based on the characteristics of the firms and show that the momentum effect on credit transitions is economically meaningful. Additionally, we find that trends of default probability are independent of cyclicality and are related to firm’s quality. The non-Markov property implies that the future distribution of a firm’s ratings depends not only on its current rating but also on its past rating history.