本文採用雙變量GARCH模型估計台灣加權股價指數正負基差的非對稱性，就是考慮正基差與負基差對現貨與期貨報酬的變異數與共變異數影響。為了考慮負面消息對報酬率變異數與共變異數的影響，本文在雙變量GARCH模型中也加入了GJR效果與延展了DCC方法，使用ADCC方法來估計共變異數。此外，本篇再將基差效果加入ADCC中，計算出最佳的避險模型來提供投資人使用。本文比較了 (1) 傳統線性迴歸法(OLS)、(2) 指數加權移動平均法（EWMA）、(3) 雙變量GARCH模型、(4) 雙變量GARCH-DCC模型、(5) 基差不對稱雙變量GARCH-DCC模型、(6) 基差不對稱雙變量GARCH-GJR-DCC模型、(7) 雙變量GARCH-ADCC模型、(8) 基差不對稱雙變量GARCH-ADCC模型、(9) 基差不對稱雙變量GARCH-GJR-ADCC模型。研究發現現貨與期貨的基差效果是不對稱的，負基差比正基差對變異數影響更甚，但正基差對共變異數的影響卻比負基差來的大。在將GJR效果納入考量後，發現負面消息對報酬率的波動是具有影響的。估計結果顯示基差不對稱雙變量GARCH-GJR-ADCC模型為避險績效最佳的模型，這也提供了投資人在台灣加權股價指數基差變動時最佳的避險策略。 This paper use Bivariate GARCH model to estimate the asymmetric basis effect in Taiwan stock index futures. That is, we also considering the positive and negative basis effect on time-varying variance-covariance of spot and future return. In addition, the bad news effect on time-varying variance-covariance of spot and futures, we add GJR effect in the Bivariate GARCH model and extend the DCC model to ADCC model to estimate the covariance. Besides, we add basis effect in ADCC model to evaluate and choose the best model for investors. We compare (1) the OLS model, (2) EWMA model, (3) BGARCH model, (4) BGARCH-DCC model, (5) asymmetric BGARCH-DCC model, (6) asymmetric BGARCH-GJR-DCC model, (7) BGARCH -ADCC model, (8) asymmetric BGARCH-ADCC model, and (9) asymmetric BGARCH-GJR-ADCC model. The empirical results find that spot and future have asymmetric basis effect, and the negative basis has greater impact than the positive basis. After consider GJR effect, we found the bad news have effect on the volatility of return. It is found that asymmetric BGARCH-GJR-ADCC model is the best hedging model. These provide the best hedging policy in Taiwan stock futures market.
