|題名: ||資本資產定價模型檢定 : 門檻模型之應用|
|其他題名: ||The test of CAPM : the application of threshold model|
|作者: ||吳佩珊;Wu, Pei-shan|
|關鍵詞: ||資本資產定價模型;三因子模型;門檻;超額報酬;CAPM;Threshold Model;beta;excess return|
|上傳時間: ||2010-01-11 00:43:45 (UTC+8)|
|摘要: ||Sharpe (1964)、Lintner (1965)所提出之資本資產定價模型(Capital Assets Pricing Model, 以下簡稱CAPM )在過去相關的研究文獻中，多以線性型態作為探討風險與報酬之依據，然而，仍有實証指出對 CAPM的完美線性架構存在與否，有檢測之必要性。故本文提出在考量市場實際可能為非線性的狀態之下，以美國股票市場為實證對象，選取1926年7月至2004年的月資料，嘗試以Fama and French (1992,1993) 所提出三因子模型之概念，納入淨值市價比(Book to Market Value Ratio, BM)及公司規模(Total Market Value Equity, ME)這兩個思考角度，輔以Bai and Perron (1998, 2003)所建議之門檻方法 (Threshold Model)探討超額報酬與系統風險之間的表現；藉由可能造成關係改變而產生之門檻值，進而探討在各區段中，是否呼應Sharpe- Lintner CAPM之理論概念，及CAPM實際適用之可能性。|
2、研究發現淨值市價比及公司規模群組，於熊市市場符合Sharpe- Lintner CAPM理論之假設，亦即支持截距項α為0的條件；牛市市場則不然。而各門檻變數下之群組，超額報酬受系統風險的影響程度，依其群組大小而產生不同程度顯著的影響，隱含著β是波動的，並非為固定之值，且與單一線性迴歸結果呈現近乎相同之變動趨勢；此外，在牛市市場的影響程度亦遠較熊市市場深。
3、實證結果發現CAPM門檻模型，雖於門檻前後可能有一致或不一致於Sharpe- Lintner CAPM的情形，指出理論假設可能在某些區域中被接受，亦可能被拒絕於另一些區域之中，然卻提供不同的角度證明Sharpe- Lintner CAPM理論於效率市場仍為有用的定價方法。
According to the Capital Assets Pricing Model (CAPM) proposed by Sharpe (1964) and Lintner (1965), it is the fact that most of previous studies support the linear relationship between return and risk. However, recently there are increasing researches pointing out the portfolio returns are determined by multi- factor, rather than the single one, beta, and argue that it is necessary to further test the perfect conditions that the traditional CAPM has. Therefore, it dries us to understand the possibility of the practical application of CAPM in the economic environment.
Motivated by the three-factor model asserted by Fama and French (1992,1993), we have a try to propose another empirical approach for testing the Sharpe- Lintner CAPM in betas allowing for the threshold model followed directly from the work of Bai and Perron (1998,2003) in order to discuss the relationship between expected excess return and beta. We use the mimicking portfolios composed by NYSE, AMEX, NASDAQ as the samples and sort these samples on the BM (Book to Market Value Ratio) and ME (Total Market value Equity) basis 5 regimes per set in order to capture the variation in the pricing model and the changes in betas under non-linear consideration. By adopting the monthly data in BM and ME portfolios from July, 1926 to Dec.2004,we try to find out whether there is a threshold effect and the behavior of pricing model are responded to the Sharpe- Lintner CAPM while the threshold exists at each regime.
The empirical findings suggest that-
Firstly, there is a strong evidence of threshold effect which separates each regime for all the BM-sorted and ME-sorted portfolios based on the excess return of market portfolio as the threshold variable and at least exists one threshold at each regime. Besides, it appears that most of the expected excess return falling on the level with rate 4%~ 5%.
Secondly, the results for regimes sorting by BM and ME all support the assumption of Sharpe-Lintner CAPM in bearish market, i.e., the intercepts are different significantly from zero; but otherwise in bullish market.
Moreover, the expected excess returns explained by market system risk for each regime are different. It implies that the beta coefficients in dynamic analysis exist instability feature and the traditional assumption of a stable relationship between the return and beta can be questioned. In particular, the trend for the volatility of betas appears the same as the single-equation does. In addition, the effect with regard to betas in bullish market is significantly deeper than the bearish market no matter the regimes formed by BM and ME.
Finally, we found the results of linear specifications can only permit the data to be either consistent or inconsistent with the Sharpe- Lintner CAPM. In other words, under the threshold model some regimes can accept Sharpe- Lintner CAPM, but some are not. However, it still provides different version to observe the validity of the Sharpe- Lintner CAPM in the efficient market.