The Predictable Ability of Stock Index Return be Evaluated by Time Series Model

• Main Authors: Huang Chun Yi, 黃駿逸
• Other Authors: Jiann-Liang Chiu
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/05398927470339917697

碩士 === 淡江大學 === 財務金融學系 === 89 === Abstract： Many empirical studies and researches believe that he overall performance of a country’s economy is strongly related to the performance of its stock market. Some empirical studies, however, show that this relationship does not necessarily hold. In other word, sometimes the stock index is viewed as random walk process. Hence, in this research we want to know that whether the stock index is predictable or not? In this research we applied five time series models to investigate on forecasting of the stock index return in American、British、German、Japanese capital markets. The analysis techniques are distinguished in two parts, which are known as the simultaneous equation model and nonlinear unique equation models. Models of the first parts compose of Vector Autoregressive Models (VAR) and Error Correction Models (ECM). The second part of the models are EGARCH、Kalman Filter Model (KFM) and Markov Switch Models (MKS) which are used to account for time-varying structure parameters and conditional variances. The forecasting performance is measure and compared with MAD and RMSE. Hence, in this research we will compare these five models and determine which of these models make the most accurate prediction. The conclusions are shown as the following: 1) In the short-term, the MKS model capture a better rate of return on the German stock price prediction. Conversely, in the long-term market performance, the ECM model possesses a better result. 2) In the British market performance, one can see that linear equation models perform better in forecasting in the first two months. Nevertheless, nonlinear unique equations tend to carry out an excellent prediction in a long-term. The overall market performance is best forecasted by using the KFM model. 3) In the short-term Japanese market performance, the better predictable model almost have the same character about having asymmetry variance assumption. Then ECM best model for the overall performance prediction. Lastly, in the US market forecast, it is noteworthy that different forecasting models generate different effects. However, EGARCH is the best-chosen model use for market performance prediction.