ARMA-GARCH風險值模型預測績效實證

Empirical Study of Forecasting Performance on ARMA-GARCH VaR Model

洪儒瑤(Ju-Yao Hung) ； 古永嘉(Yeong-Jia Goo) ； 康健廷(Chien-Ting Kang)

《中華技術學院學報》 34期 (2006/06) Pp. 13-35

https://doi.org/10.7095/JCIT.200606.0013

摘要

1996年巴賽爾銀行監理委員會(BCBS)提出巴賽爾協議(Basel Accord)的修正案，將市場風險因素納入考量，以估算銀行機構應計提的資本準備，同時允許銀行擇取自行開發VaR內部模式或直接採用Basle提供的標準模式。我國財政部於1998年5月修正「銀行自有資本與風險性資產之範圍計算方法及未達標準之限制盈餘分配辦法」，將市場風險納入資本適足性之衡量，並於該年7月明訂銀行計算市場風險之方法有二：標準法與自有模型法（即VaR模型）。本文以ARMA(m,n)+GARCH(p,q)風險值模型來預測銀行的VaR值，並與最常被使用的三種模型－歷史模擬法、蒙地卡羅法與變異數－共變異數法作一比對，用以比較模型的預測效果。經過實證比較，發現ARMA-GARCH模型的整體表現比其他三個模型好。

關鍵字

風險值 ； 歷史模擬法 ； 蒙地卡羅法 ； 變異數－共變異數法

並列摘要

Risk management has caught sight gradually since the Basel Accord on BCBS claimed banking institutes to build a systematic risk management procedures to manage quality and procedures of making loans in 1988. Later, the Group of Thirty published report of Value at Risk in 1993 meanwhile J.P. Morgan issued four-fifteen report in 1994 had put the VaR into indicator of risk management. In this study we exploit an ARMA(m,n)+GARCH(p,q) model to predict the VaR of six commercial banks in Taiwan. To firmly decide the effectiveness we also evaluate the predicting performances of the model and compare with the three mostly used model Historical Simulation, Monte Carlo Simulation, Variance/Covariance Approach. Via the empirical results, we find the overall performances of ARMA(m,n)+GARCH(p,q) model are better than the other three models.

並列關鍵字

VaR ； Value at Risk ； Historical Simulation ； Monte Carlo Simulation ； Variance/Covariance Approach ； ARMA-GARCH