利用Kalman Filter Approach估國內公司債違約機率

Abstract

以台灣國內公司債為研究對象，探討利用Kalman Filter Approach估公司債違約機率的排序與台灣外部評等公司對其公司評等排序有無顯著性。研究中沿用Duffee(1999)對違約機率所定義的理論，利用無風險狀態變數、風險溢酬、drift項來描述違約機率，並假設違約機率服從獨立平方根隨機過程。Kalman Filter Approach 為一State Space Model，包含了觀察(Measurement Equation)和轉換(Transition Equation)方程式，利用了遞迴(Recursive Process)的特性，將動態系統引入誤差項，藉由系統值和觀察值不斷的加入、更新，使得所估計的推估值和狀態向量之間的推估誤差具有最小變異。此方法為線性的動態系統利用每月的月交易資料去推估公司債的違約機率，但因公司債之交易資料不如股票或公債交易資料多，所以當公司債資料能更齊全應能更及時反映公司營運狀況。

Regard the bond of domestic corporation of Taiwan as the research object to probe into sequencing utilizing Kalman Filter Approach to estimate the probability in breach of company's bond and the arranges in an order dominance to its company by the company of Taiwan outside commenting etc.. Continue to use the theories defined toward probability in breach of Duffee (1999 ) while studying, having the state space parameter of no-risk , risk premium, drift term to describe the probability in breach, and suppose that the probability in breach obeys the independent square root stochastic process. Kalman Filter Approach is a State Space Model, include Measurement Equation and Transition Equation , utilize the characteristic of recursive, introduce error item into dynamic system, by system value and acceding to, upgrading constant of observing etc., make errors of estimating when pushes valuation and state vector quantity estimated make a variation minimum. This method utilizes the monthly trade materials to estimate the probability in breach of company's bond for the linear dynamic system, but because trade materials of company's bond are not so much as stock or government’s bonds, so it can be more completely to act as company's bond materials and should be able to reflect company's operation conditions in time even more .