本文之主要目的在於探討台灣股市是否存在門檻 CAPM （threshold capital asset pricing model, threshold CAPM）之非線 性動態系統風險，以及何種經濟變數造成系統風險呈非線性改 變。實證結果顯示，我國股市確實存在非線性動態系統風險，其 中，無風險利率、生產者物價指數成長率及石油價格變動率是造 成系統風險呈非線性變動的重要經濟變數。此外，相較於線性及 條件 CAPM，其中包含卡爾曼濾嘴（Kalman Filter）及多變量 GARCH 模型（Multivariate-GARCH），門檻 CAPM 較能充分解 釋時間序列股票報酬，且其具有較強之預測能力，隱含門檻 CAPM 模型為解釋台灣股票報酬之較佳資產定價模型。

The purpose of this study is to test whether, as implied by the threshold CAPM, nonlinear dynamic beta exists in Taiwan and which economic variables lead to beta changes. The evidence confirms this implication and shows that risk free interest rate, producer price index growth rate, and oil growth rate are the three most important determinants for beta shifts. Finally, compare to linear and conditional CAPM (Kalman Filter and Multivariate-GARCH), we find that the threshold CAPM can sufficiently explain time-series variation of stock returns and has better predicting power. Therefore, the nonlinear dynamic beta model is the more powerful asset pricing model for the Taiwan stock market.

資本資產定價模型;系統風險;門檻模型;條件 CAPM

Capital Asset Pricing Model; Systematic Risk; Threshold Model; Conditional CAPM