- 學位論文
金融與總體經濟變數對股票報酬之影響－Linear與STARX模型之比較分析
The Effect of Financial and Macroeconomic Variables on Taiwan Stock Market Return – Comparison of the Linear and the STARX Models
摘要
近年來有許多實證研究指出股票市場報酬可由不同的金融和總體經濟變數來預測，但實證方法大多以線性模型為主，因此本文參考一般的STR和STAR 模型，此模型的特色是能夠平滑轉換在行為結構間，由於之前的門檻模型其轉換過程相當突然，而STAR 模型可以克服此項缺點，其所持的理由如下，只有當所有的證券交易人員同時行動時才會出現突然轉換的結果，而市場的證券交易人員在交易時都會有些微的時間差異，所以平滑轉換模型是較適合的。其次，STAR 模型允許不同型式的轉換函數以反映不同型式的市場行為，例如Logistic函數允許不同的行為基於報酬是否為正或負的；Exponential函數則是允許不同的行為發生在較大或較小的報酬而卻可忽略符號為正或負，另外，不同的產業對金融和總體經濟變數的反應是否相同也是值得研究的議題，故本研究目的在檢驗不同類股的股票指數報酬率與金融和總體變數的非線性關係，以及非線性的方法是否可以用來改善報酬的預測力。 從本研究的實證結果可以觀察到以下結論，在不同類股的報酬下其顯著的外生變數和落後期數是不相同的，表示由於產業的差異造成對金融和總體變數產生不同的反映，最後我們比較線性模型和STARX模型樣本外的預測力，其結果除了造紙類股外，以STARX模型所做的樣本外預測結果比線性模型來的好，顯示以非線性的方法似乎能夠改善線性模型樣本外的預測能力。
並列摘要
Recent empirical evidence suggests that stock market returns are predictable form a variety of financial and macroeconomic variables. But most of empirical models are linear. This paper considers a version of the general class of STR and STAR models that allows for smooth transition between regimes of behavior. This model is favored over that simple threshold models which imposes an abrupt switch in parameter values, because only if all traders act simultaneously will this be the observed outcome, for a market of many traders acting at slightly different types of market behavior depending on the nature of the transition function. In particular the logistic function allows differing behavior depending on whether returns are positive or negative, while the exponential function allows differing behavior to occur for large and small returns regardless of sign. Besides, the relationship between different industries and financial and macroeconomic variables is worth to note. Hence this paper has tested for evidence of a nonlinear relationship between stock market returns and financial and macroeconomic variables, and whether this nonlinearity can be exploited to improve forecasts of returns. From empirical evidence of this research, we have some conclusions. In different industry stock indexes which the statistically significant exogenous variables and lag length are different. The result represents the financial and macroeconomic variables to generate different reactions due to different industries character. Last we compares the out of sample predictability both linear and STARX models. The out of sample predictability of STARX model outperform the linear model except construction industry. The results indicating an improved marginally superior out-of -sample forecast results.