本文創新提出一般化自我迴歸條件密度Moments-in-Mean(ARCD Moments-in-Mean)模型,除了允許高階動差受外在訊息影響而隨時間變化外,並將條件變異數、條件偏態及條件峰態序列納入報酬方程式,評估不同風險型態動差特性對預期報酬的影響。實証結果顯示,股票市場偏態及峰態特性受外在訊息影響而隨時間改變;其次,高階動差風險訊息對價格變動具顯著影響,說明完整描述高階動差特性的重要性;再者,衡量市場上漲下跌傾向的偏態特性,對價格呈現高度解釋效力。此外,波動不對稱性可能來自於偏態特性的欠缺考量,証實分佈假設對於波動行為探討的重要性。
This paper presents a flexible Autoregressive Conditional Density Moments-in-Mean (ARCD Moments-in-Mean) model. Our innovative approach not only allow the higher order moments to be time-varying function of conditioning information, but also extend the traditional ARCH-M model by explicitly modeling the influence of conditional moments (conditional variance, skewness and kurtosis) on the conditional expectation of the data series. The empirical results suggest a preponderance of evidence to support the performance of ARCD Moments-in-Mean model on competing alternatives. Of particular, the time varying skewness, capturing the up/downside risk, is found to exhibit significant effect in explaining the expected return than that of second and fourth moment. In addition, it is also found that the asymmetry of conditional distribution is potentially possible to substitute the asymmetry in the volatility specification demonstrating the importance to specify the proper distribution specification on financial econometric modeling.