本文探討條件變異的不對稱性、變異結構改變、以及利用前瞻性預測檢定法,在Engle (2002)動態條件相關(dynamic conditional correction, DCC)係數的觀點上,重新檢驗台灣股、匯市與美國股市之間的蔓延效果。利用Inclán and Tiao (1994)建議的疊代累積平方加總運算法(iterated cumulative sums of squares algorithm, ICSS)檢測市場報酬變異的結構性改變,設立虛擬變數,估計一般化誤差分配的EGARCH(exponential generalized autoregressive conditional heteroscedasticity, EGARCH)模型,以及動態條件相關多變量GARCH(generalized autoregressive conditional heteroscedasticity)模型以估算動態條件相關係數,再利用1步預測檢定與N步預測檢定法檢驗蔓延效果。實證結果顯示,台灣加權股票指數、台幣對美元匯率、以及美國紐約綜合股價指數可檢驗出訊息不對稱的槓桿效果,1步與N步預測檢定結果,檢驗出相關係數顯著為“正”與“負”的蔓延效果,以及隱含蔓延效果為一長期現象。
This research investigates asymmetries in conditional variances and structural changes in variance and also retests the contagion effect between the Taiwan, foreign-exchange, and US stock markets through the use of a forward-forecasting testing method developed by Engle (2002) for dynamic conditional correction (DCC). The iterated cumulative sums of squares (ICSS) algorithm developed by Inclán and Tiao (1994) is used to detect the structural breaks in market return, create dummy variables, estimate the conditional generalized error distribution of an EGARCH (exponential generalized autoregressive conditional heteroscedasticity) model, compute the dynamic conditional correction coefficients of the DCC multivariate GARCH (generalized autoregressive conditional heteroscedasticity) model, and employ both a one-step and an N-step forecast test to check for the contagion effect. The results demonstrate the utility of the asymmetric leverage effect test and also indicate that correction coefficients have significant ”positive” and ”negative” contagion effects; moreover, these implicit effects are a long-term phenomenon.