本文是依 Tsay (1989)所提之門檻自我迴歸模型及Glosten, Jaganathan 與Runkle (1993)所提GJRGARCH 模型之想法，提出一個雙門檻-GARCH 模型探討上海綜合股價指數報酬與香港恆生指數報酬波動之關係，且以香港恆生指數報酬波動的正負值作為門檻，其研究資料期間是採用2000 年1 月4 日到2006 年6 月30 日的上海綜合股價指數與香港恆生指數的資料。而由實證結果顯示AR(8)-雙門檻-GRACH(1，1)模型對探討香港恆生指數報酬波動對於上海股票市場報酬的影響是合適的，且反應出上海股票市場具有不對稱的效果。而由實證結果也顯示香港恆生指數報酬波動將影響上海股票市場報酬，且反應出香港恆生指數報酬波動為負值的情況下將增加上海股票市場報酬波動的變異風險，但GARCH 與GJR-GARCH 模型無法反應此信息，此也反應出雙門檻-GARCH 模型是比傳統之GARCH 與GJRGARCH模型較具有解釋能力。 This paper uses the idea of threshold auto-regression model ( Tsay, 1989) and the idea of GJR-GARCH model (Glosten, Jaganathan and Runkleafter, 1993),the researcher propose a double-threshold-GARCH model to study the relationships of the Shanghai synthesis index returns and Hong Kong hang seng index returns volatility, using the positive and negative values of the volatility of Hong Kong hang seng index returns as the threshold. The study data period is from January 1999 to December 2005. Empirical result shows that the effects of Hong Kong hang seng index returns volatility and Shanghai stock market return can construct on an AR(8)-double threshold-GARCH(1,1) model. This model is also response the asymmetrical effects of the Shanghai stock market returns. Empirical analyses also indicate that the Hong Kong hang seng index returns volatility will negatively affect the stock market returns. As the positive and negative of Hong Kong hang seng index will negatively affect the variation risk of stock return volatility, but the models of GARCH and GJR-GARCH does not respond this information as above. The explanatory ability of proposed model is better than the traditional model of GARCH and GJR-GARCH.
