摘要 Fama and French (1988) 指出以股價報酬的平均數復歸(mean-reversion)現象發現股價報酬具有負向自我相關,且其長期的估計結果較短期更為顯著。Fama (1990)也指出股價報酬會隨著時間的增長,其解釋能力會有較佳的表現。傳統文獻上分析股價報酬的長期行為均是以線性模型為主,例如Pesaran and Timmerman (1995, 2000) 利用線性遞迴迴歸模型(linear recursive regression)來預測股價長期報酬的行為。然而McMillan (2001)指出以非線性模型來估計股價行為的表現要比線性模型更適合,因此本研究欲以平滑轉換的非線性迴歸估計方法 (smooth transition regression, STR)來探討股價的長期報酬行為。本研究以亞洲四小龍及日本的股價報酬為研究目標,並考量美國股市對的影響力,以 Bacon and Watts (1971) 首先提出,經Tong (1978)、Teräsvirta and Anderson (1992) 等陸續加以發揚光大的 STR 估計方法來探討股票報酬的非線性關係,並將線性與非線性兩者迴歸估計的樣本內配適度加以比較,且運用Rapach and Wohar (2006) 提出的拔靴法(Bootstrapping)樣本外 (out-sample)預測方法,藉此比較線性與非線性模型之樣本外預測能力。本研究的資料蒐集期間自1988年1月起至2006年8月,而資料來源是從全球金融資料庫中所取得的月資料。
Abstract Fama and French (1988) point out that mean-reverting price component of stock prices shows negative autocorrelation in returns, that is weaker for the daily and weekly holding period but stronger for long-horizon returns. Fama (1990) identifies that the regression increases with the time horizon for the stock return series. Previous studyies working on long-horizon predictability have mostly dealt with the behavior of stock returns in a linear model context. For example, Pesaran and Timmerman (1995, 2000) use linear recursive regression to predict the behavior of long-horizon stock markets. McMillan (2001), however, considers that to predict the stock returns nonlinear models are more fitted than linear models. Therefore, we try to explore more predictive behavior of long-horizon stock returns by utilizing the smooth transition regression (STR), a nonlinear econometric model originated by Bacon and Watts (1971) and exalted later by Tong (1990), Teräsvirta and Anderson (1992). This study uses the data from Japan and the East Asian Tigers’ stock markets, contemplating the influence of American stock markets, and utilizes the STR estimation to examine the nonlinear relationship of stock return series and to make comparison of the fitness in-sample and out-sample forcasting. The data span the period January 1988 to August 2006 using monthly data from The Global Financial Database.