| Summary: | 碩士 === 逢甲大學 === 統計與精算所 === 96 === Stock markets have become the most important investment target. It’s more important to understand the relationship of stock returns between G7 and Taiwan and research about it.
On the purpose of unit root test, we descry every stock index has unit root and all stock returns are all stationary time series. We find three phenomenons in the describe statistics: 1. Average stock return is the highest in America and it’s the lowest in Japan. 2. It has lowest risk on the stock market risk in America and it’s highest in Taiwan. 3. Stock returns are all think tail distribution in G7 and Taiwan clearly.
By the study all stock returns have Autoregressive Conditional Heteroscedasticity, and according to Sign Bias test we can know that every stock market has different volatility to face with different shocks (news).
Using stock returns data to fit two different nonlinear time series model in the eight countries and calculate the News Impact Curve (NIC), The result is the same with above study that the average stock risk is lowest in America, highest in Taiwan.
This section we study Granger Causality test and Impact Response Function by Vector Autoregressive (VAR) model and finding G7’s stock returns ahead of its in Taiwan apparently. The ahead effect is largest in America. Each country’s stock return has biggest effect by it’s pridian return. And America stock return has great effect to Taiwan stock return. Finally we use Time-Varying Correction model to research the correlation coefficient’s long-term trend and stock market risk between G7 and Taiwan. From Asia crisis the correlation coefficient between G7 and Taiwan have big shaking and trend to lowest point until 319 affair and rise now. By this study it has similar trend in the same area in the world. A suggestion for investors disperse risk by investing stock market in different area in the world is a good method.
Key Word: Unit Root test, Sign Bias test, News Impact Curve, Granger Causality test
Vector Autoregression model, Impact Response Function, Time-Varying Correlation model
|
|---|
