跳躍訊息與不對稱訊息對股價報酬的衝擊

本研究利用Maheu and McCurdy（2004）提出的GARJI模型來探討跳躍干擾對台灣與美國股價報酬的影響，此模型延伸GARCH模型，並跳躍強度參數化為ARMA（1,1）模式與考量加入不對稱性的設定於跳躍行為，以捕捉GARCH模型無法觀察到的動態跳躍過程。本研究實證結果發現（1）台灣與美國市場皆存在著異常資訊所造成瞬時的跳躍行為與跳躍頻率是隨著時間變動，而跳躍過程所引發的變異佔整體變異比重分別為22.8%與15.3%，因而跳躍過程是不可忽視的重要因素。（2）不論前期是否發生跳躍行為時，壞訊息對市場的衝擊影響比好訊息的影響大。當前期為壞訊息時，則發現跳躍時的回饋係數皆較大於正常時的回饋係數，而當前期為好訊息時，發生跳躍時的回饋係數與正常時的回饋係數相差不大。（3）當發生跳躍時訊息衝擊曲線變得較對稱，即考慮加入跳躍干擾有關之訊息融入目前股價的速度較為迅速，但S&P 500指數在好訊息與壞訊息之間依然存在明顯不對稱情形。（4）在重大金融危機事件的跳躍變異數皆大於所有樣本的平均跳躍變異數，除了阿根廷金融危機對於S&P 500的影響略小於平均跳躍變異數。

關鍵字

GARJI模型；訊息干擾；不對稱效果；資訊影響曲線

並列摘要

This study adopts the GARJI model, proposed by Maheu and McCurdy (2004), to investigate that the jump and asymmetric effects impact on stock returns for Taiwan and US. The model has extended the GARCH model, the conditional jump intensity by an ARMA(1,1) process and the asymmetric effect.The empirical results that, firstly, jump intensity and frequency are time-varying in stock markets and jump-induced volatilities are 22.8% and 15.3% for Taiwan and US, which jump process is greatly importance. Second, The bad news induces volatility greater than the good news whatever the stock markets exit jump behavior. The feedback of the jump will be greater than diffusion in the bed news when early day is the bad news, but the feedback of the diffusion and the jump don’t difference when early day is the good news. Third, the news impact curve became symmetry when the stock markets exit jump behavior; nevertheless, S&P 500 index still exits asymmetry of news. Fourthly, jump volatilities for the serious financial crisis are greater than the mean of sample except financial crisis of Argentina.