本篇研究主要的目的是分別探討台灣、南韓、新加坡、香港、日本、中國與美國共七國股價指數報酬的非線性動態調整行為。使用本身股價指數報酬之落後期為解釋變數，並且應用自我迴歸平滑轉換模型來描述股價指數報酬非線性調整的部分。採用 Teräsvirta（1992）所提出之自我迴歸平滑轉換模型下，實證結果發現： 首先，使用本身股價指數報酬落後期為解釋變數之下，不同國家，各個股價指數之解釋變數之落後期不全相同。再者，所研究的七國股價指數報酬均呈現非線性調整，並且每一個股價指數報酬都適合使用logistic型態之轉換型態說明非線性調整行為。所有股價指數非線性調整非但存在門檻，且都存在雙門檻，並非單純的單門檻調整。在越過各個不同的門檻值之後，股價報酬序列有著不同的動態調整行為。所有國家中，上海綜合指數報酬之轉換速度最快，新加坡海峽時報指數報酬之轉換速度最慢。最後，門檻差距最大的為上海綜合指數報酬，最小的為台灣加權股價指數報酬。 This paper examines the nonlinear dynamics in stock returns which includes Taiwan、 South Korea、Singapore、Honk Kong、Japan、United States of America and China by using Smooth Transition Autoregressive Model (STAR) and using the lag of stock return as the transition variable. Under the STAR model made by Teräsvirta (1992), we have several results. First, the lags of stocks return are different in each country. Meanwhile, all countries have two thresholds and three regimes. Moreover, all stock returns can be explained by Quadratic Logistic Smooth Transition Autoregressive Model (QLSTAR). By crossing the thresholds fastest and smoothest, the stock returns will have different nonlinear dynamics behaviors. Furthermore, the faster and smoothness regime change is the Shanghai Composite Index; the slowest is the Straits Times Index. Finally, the Shanghai Composite Index has the largest distance between two thresholds; the TSEC weighted index has the smallest distance between the two thresholds.
