傳染效果違約時距之研究：自我迴歸條件時距模型之應用

An Application of the ACD Model to the Process of Default Contagion Duration

周恆志(Heng-Chih Chou)

《運籌與管理學刊》 8卷2期 (2009/12) Pp. 13-23

https://doi.org/10.7047/JLM.200912.0013

摘要

違約事件並非獨立發生，經常會出現違約傳染效果，違約傳染效果的嚴重性會表現在前後違約事件間隔時距的變化上。本文採用自我迴歸條件時距模型捕捉前後違約事件間隔時距的波動過程，並分析違約時距的時間序列特性。我們以台灣上市上櫃公司的違約事件爲實證樣本，並採用Liu and Singh (1992)的移動區塊拔靴複製法來複製樣本資料，樣本期間自1982年10月至2007年12月止。實證結果發現無論是全部樣本，或是電子資訊業與營建業的子樣本資料，違約時距數列皆呈現顯著的條件自我相關特性與群聚效果，ACD模型有助於捕捉違約事件的傳染效果。

關鍵字

違約時距 ； 違約風險 ； 傳染效果 ； 拔靴複製法 ； 自我迴歸條件時距模型

並列摘要

Credit events do not happen independently, but often cause the relating parties to default. That is so-called contagion effect of credit events. The study thus adopts Autoregressive Conditional Duration Models (ACD model) to catch the duration process of credit events and explore the statistical characteristics of the time series data of credit durations. The credit event data, from October 1982 to December 2007, of both listed and OTC companies in Taiwan are chosen as the empirical samples. The block moving bootstrap (BBM) proposed by Liu and Singh (1992) is used to simulate our samples. Our empirical results show that credit durations demonstrate the characteristics of autoregression and also volatility clustering.

並列關鍵字

Default duration ； Default risk ； Contagion effect ； Bootstrapping ； ACD model