資本資產定價模型 (CAPM) 發展至今已逾三十年，然大多數定態 (static) CAPM的研究均假設貝他係數固定，不隨外在金融環境變動，致使CAPM常無法有效解釋橫斷面之股票報酬率實證資料。有鑑於此，本文假設貝他係數隨股市多頭與空頭市場而有所變動，亦即多空頭雙貝他CAPM應較單貝他CAPM更適於解釋股市報酬率之變化。研究結果發現，各投資組合多頭與空頭貝他係數雖均為正，但明顯不同，故區分多空頭貝他係數應具正確性。再者，投資組合於多頭 (空頭) 市場的報酬率與系統風險呈正 (負) 相闕，但平減風險後的相對報酬率於多頭 (空頭) 市場反隨風險之增高而遞減 (遞增) ，故高風險高 (低) 報酬多存在於多 (空) 頭市場，惟風險與報酬並不完全對稱。此外，多頭與空頭系統風險對投資組合之報酬率具高度解釋力，且迴歸估計之多空頭實際市場風險溢酬亦典預期之多空頭理論市場風險溢酬具顯著相等性，故區分多空頭之雙貝他CAPM不僅適用於描述台灣多空頭股市的報酬率，亦改善了單貝他CAPM的不適用性，更確立變動系統風險的存在及其解釋資產報酬率的重要性。

Capital asset pricing model (CAPM) has been well developed for over thirty years. However, its static assumption comes into question that CAPM explains the cross-sectional stock returns poorly in practice. This study further assumes that the dual (bull and bear) betas, or varying risk, CAPM is empirically sound. Evidence is provided that although both betas are positive, bull betas are significantly different from bear betas, which supports the validity of dual betas assumption. Bull (bear) betas are positively (negatively) related to the portfolios returns, but their relationships with the relative returns, after deflating risk, are inverse, implying that risk and returns are asymmetric. Furthermore, dual-beta market model captures a significant power in explaining the cross-sectional returns of Taiwan bull and bear stock market, and the actual or realized bull and bear market risk premiums are statistically consistent with those the theory expected. Therefore, dual-beta CAPM is more valid in describing Taiwan stock market than single-beta CAPM, which confirms the importance of varying risk in asset pricing model.