過去已有許多學者針對股價報酬之可預測性提出不同的觀點,本文改以整體經濟風險因素為背景來切入股價指數可預測性的論點,運用向量自我迴歸模型(VAR)、誤差修正模型(ECM)、一般自我迴歸條件異質非對稱變異數模型(EGARCH)、卡爾曼濾嘴模型(Kalman Filter)及馬可夫狀態轉換模型(Markov switch),分別就美國、英國、德國與日本等四個工業大國的股價指數作跨市場不同預測模型的績效比較。實證結果指出,在英德股市預測中,馬可夫狀態轉換模型較能在短期間內預測股價報酬率的脈動。在長期來看,誤差修正模型則有較佳的預測能力。對日本股市而言,於短期預測中的最佳模型亦為馬可夫狀態轉換模型,而在長期預測下的最佳模型為誤差修正模型。而在預測美股報酬走勢方面則產生了明顯不同的績效衡量結果,在MAD法中以聯立式向量自我迴歸模型有較佳的表現,在RMSE法中則認為EGARCH會有較佳的預測結果。
Although the predictability of stock return has been an object of study for a long time, there is little agreement as to the forecast the stock index returns. In this paper we choose macroeconomics risk factors and employ time series models: Vector Autoregressive (VAR) model, Error Correction Model (ECM), Generalized AutoRegression Conditional Heterskedasticity (GARCH) model, Kalman filter Model (KFM) and Markov switch model to forecast the stock index return of US, UK, Germany and Japan. The results show that the forecasting performance of the Markov switch model is better than other models in short run and the ECM is the best model in long run in UK and Germany. For Japan, the Markov switch and the ECM have better forecasting performances in short run and long run respectively. For US, the VAR and EGARCH have better predictabilities in short run and long run.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。