變幅波動與GARCH模型之波動預測績效比較—台灣加權股價指數之實證

Volatility Forecasting Performance of Range-based Volatility and GARCH Model-An Empirical Analysis of Taiwan Stock Index

張瑞杰(Jui-Chieh Chang)

指導教授 ： 邱建良 ； 洪瑞成

淡江大學/商學院/財務金融學系碩士在職專班/碩士(2009年)

https://doi.org/10.6846/TKU.2009.00218

摘要

選擇權訂價模型中，波動性的估計是重要探討之議題，學者專家們相繼提出各種波動性之估計模型，欲尋找出何種波動性模型所得之選擇權理論價格與實際市場價格差距最小，本文中以台灣加權股價指數及台灣股價指數選擇權為研究標的，利用變幅估計式（Range Based Estimators）估計其波動性，結合採用不對稱GJR-GARCH模型及MA、AR、EWMA、RW及ARMA等時間序列模型等作為預測模型，並以平均絕對誤差（mean absolute errors， MAE）、均方誤差（mean squared errors，MSE）、平均混合誤差（mean mixed error，MME）等傳統損失函數及優勢預測能力檢定（Superior Predictive Ability Test，SPA）模型，來衡量不同方法模型中何種模型預測績效較能貼近實際市場波動性，希望能藉此找出一適合的模型，可較準確地預測出台灣股價指數波動度，藉以降低台指選擇權之交易風險。結果顯示結合變幅估計式之預測模型對於台灣加權股價指數有較佳的預測效果，而對台灣股價指數選擇權，則以GJR-GARCH模型預測效果較佳。

關鍵字

變幅估計 ； 優勢預測能力檢定

並列摘要

On the field of the option pricing model, the prediction of volatility is one of the important topics among others. For the purpose of searching for a model reflecting the narrowest gap between the theoretical prices of option and the actual market price, the scholars and experts have proposed a variety of models for volatility prediction. This article takes TAIEX and TAIEX Option as the research object, using Range-Based Estimators to estimate the volatility, along with asymmetric GJR-GARCH model and the MA, AR, EWMA, RW and ARMA time series model as the models for the prediction of TAIEX and TAIEX Option. Furthermore, some traditional loss function such as mean absolute errors, MAE, mean squared errors, MSE, and mean mixed errors, MME as well as the Superior Predictive Ability Test, SPA are applied in this article to determine which model with the foregoing methods has better accuracy in predicting the volatility of the actual market. In addition, the goal of this article is in a hope to search out a proper model which can predict the volatility of TAIEX more accurately, and to reduce the transaction risk of TAIEX Option by way of such model. In conclusion, the result indicates that the combination of Range-Based Estimators of the predicting model presents a better effect on prediction for TAIEX, while GJR-GARCH is better for TAIEX Option.

並列關鍵字

GJR-GARCH ； Range-Based Estimators ； Superior Predictive Ability Test

參考文獻

Lin C. T., Hung J. C. and Wang Y. H. (2005), Forecasting the one-period-ahead volatility of the internatilal stock indices：GARCH Model vs. GM(1,1)-GARCH Model, Journal of Grey System, Vol. 8, No. 1, 1-12.

涂惠娟、蔡垂君（2006），「台指選擇權風險值之研究」，交大商管學報，第11卷，第2期，頁57-69。

莊益源（2003），「波動率模型預測能力的比較-以台指選擇權為例」，台灣金融財務季刊，第4輯第2期，頁41-63。

李命志、洪瑞成、劉洪鈞（2007），「厚尾GARCH模型之波動性預測能力比較」，輔仁管理評論，第14卷第2期，頁47-72。

葉銀華、蔡麗如（2000），「不對稱GARCH族模型預測能力之比較研究」，輔仁管理評論，第7卷第1期，頁183-196。