自台灣股市發展至今,歷經了暴漲暴跌與早期投資人理財金融知識不足,呈現市場不效率,本研究欲以兩階段迴歸方法(two-stage regression)來驗證Sharpe(1964)與Lintner(1965)發展而成的「平均數—變異數」資本資產訂價模型,以及Hogan and Warren(1974)研究出的「平均數—半變異數」資本資產訂價模型。以觀察台灣投資人對於風險與報酬之間的關係之了解,與台灣股市對於兩種模型的適用性。 本研究的實證有以下結論: (1) 在傳統的「平均數-變異數」資本資產訂價模型中,截距項大致與理論值不同,即 與研究期間的無風險利率大多呈顯著不同,而在「平均數-半變異數」資本資產訂價模型中,在不同區間, 也大多顯著與理論值不同。 (2) 此外,在「平均數-變異數」資本資產訂價模型中,模型的解釋能力偏低,僅達16.16%~18.68%,顯示除了系統風險以外,尚有其它變數可用以解釋臺灣股市的報酬率;而以半變異數為衡量系統風險的方法時,其模型的解釋能力不增反減,因此可以推論,在本研究期間兩種模型均不適合於臺灣股市。 Since the stock market was built in 1960’s, Taiwan investors have experienced ups and downs in this market. With the empirical researches done by people of former times, we know that Taiwan market is not efficient. There are many individual reasons for market efficiency in Taiwan, such as price limits, exchange cost, and the level of risk recognition for investors etc. This paper would like to observe if investors are more knowledgeable about the relationship between risk and return than those twenty year ago. Through the two-stage regression of E-V CAPM (Mean- Variance CAPM) and E-S CAPM (Mean- Semivarance CAPM), we anticipate seeing the conformity of these two models. The empirical results are as follows: (1) In both models, the intercepts of the second-pass regressions significantly differ from the theoretical value; that is,the intercepts is divergent from our risk-free rate. (2) Moreover, the determination coefficients of the two models are no more than 18%, which means there must be other variables that can explain security prices in Taiwan market.