本文研究投機泡沫是否存在於股票價格。其模型誤設及投機泡沫若存在，都是不
可觀察的變數且 state-space 模型中傳遞方程與觀測方程中的干擾項為同期相關。 於是我
們採用 Jazwinski （ 1970 ）的一般化 Kalman filter 來估計。使用 Malkiel（ 1979 ）
和 Pindyck （ 1984 ）的變動風險貼水股價模型來設定誤設變數。 在得到變數估計值後，
我們沿續 Durlauf and Hooker （ 1994 ）及 Chen （ 1995 ）來做正交檢定分析。結果是
CRSP 的資料，變動風險貼水模型具有良好的解釋，且無泡沫存在於股價中。 但 MSCI 資料
中，法國、德國、義大利之股價，無誤設的流量設定式未與訊息集合變數正交。故變動風險
貼水的模型未提供好的解釋，泡沫的存在性也未得到明確的結論。

　　　　　This paper examines whether bubbles or time-varying risk premiums
affect stock prices. A model with speculative bubbles and misspecifications,
factors unobserved in stock prices, is considered. The setting of a time varying
risk premium proposed by Malkiel (1979) and Pindyck (1984) is applied to capture
the possibility of misspecification. The errors in the measurement equation and
transition equation in the state-space model are correlated. Thus, we employ the
generalized Kalman filter developed by Jazwinski (1970) to estimate the
parameters. After we get the estimates, we follow the orthogonality test
discussed in Durlauf and Hooker (1994) and Chen (1995) to analyze the flow and
stock constraints on different information sets. The results show that the time
varying risk premium model provides a good explanation for the CRSP data set and
there is no presence of speculative bubbles. The France, German, and Italy in
MCSI data sets indicate that the time varying premium model does not provide a
suitable explanation. No further conclusion can be drawn for whether or not
speculative bubbles exist in these three countries.