| Summary: | 碩士 === 朝陽科技大學 === 財務金融系碩士班 === 94 === Over the years, many empirical literatures analyzed financial underlying, such as the stock price, future basis and exchange rate. These literatures demonstrated how transaction cost, technical analyst or official interventions induce nonlinear process. Especially in a foreign exchange markets with a large number of investors, regime change might be smooth rather than discrete was due to heterogeneous investment horizons and varying learning speeds, so a smooth transition between regimes seemed more appropriate.
Nonlinear regime switching and structural change are often perceived as competing alternative to linearity, but Lundbergh et al. (2003) indicated nonlinear and structural change were relevant for many time series, and furthermore these features had mainly been analyzed in isolation. Lundbergh et al. (2003) combined these features to develop TV-STAR (time-varying smooth transition autoregressive) model.
In this article, we apply the STAR (smooth transition autoregressive) model and TV-STAR model to analyze real effective exchange rates of Taiwan, China and Japan. The empirical results are as follows:
1.The predictive performance of TV-STAR model in Taiwan is better than STAR model.
2.The predictive performance of TV-STAR model in Japan is worse than STAR model.
3.The predictive performances of STAR and TV-STAR models in China are the same.
Keywords: smooth transition autoregressive, time-varying, real effective exchange rate.
|
|---|
