本文結合Engle and Lee (1993)的要素模型(component model)與Chanand Maheu (2002)的ARJI模型，亦即ARJI-Trend模型分析台積電、聯電、日月光、友達及中華電等五檔美國存託憑證(ADR)及其標的物間，其資訊反映之傳遞效果以及報酬之波動性的恆常及暫時性因子。再者，本文進一步納入卡爾曼濾嘴(Kalman filter)模型，將現貨成交金額變動率分為預期和非預期變動率，作為衡量資訊(information)和異常資訊(noise)對於現貨之衝擊程度，以期透過觀察標的現貨之成交量變化，來檢驗ADR對於標的現貨市場不同資訊的反映上是否存在差異。實證結果顯示無論是資訊或異常資訊之衝擊皆會先透過標的現貨之成交量進一步反映到ADR報酬，反映出「量先價行」之現象。而在條件變異數的部分，確實存在恆常要素與短暫要素，而且短暫要素的衝擊效果明顯大於恆常要素，顯然在沒有漲跌停限制的ADR次級交易市場中，一旦面臨市場資訊的衝擊時，短期股價報酬之反映較長期來得劇烈。

In this paper, we apply a ARJI-Trend model, which combining component model and ARJI model, proposed by Engle and Lee (1993) and Chan and Maheu (2002), to study the relationship of information transmission between ADR and underlying stocks and estimate the permanent and transitory factors of volatility. Five ADRs are examined in this study, including TSM, UMC, ASX, AUO, and CHT. Furthermore, the Kalman filter model is used to distinguish the trading amount rate of change into expected and unexpected rate of change, for assessing the degree of information and noise shocks. It is capable of examining whether ADR have different responses to the information in underlying stock market by observing the variations of underlying stock's trading volume. The empirical result shows that the both shock of information and noise have response to the return of ADR by underlying stock's volume and exists the phenomenon of "Quantity first price line". It is also found that both permanent and transitory components of the conditional variance really exist and the shock of the temporary component of conditional variance is larger than the permanent component. Because there is no limit up or down in ADR's secondary market, the temporary shock of stock return is larger than the permanent shock.