| Summary: | 碩士 === 國立雲林科技大學 === 財務金融系碩士班 === 92 === The exchange rate is always the most important economic variance in the open economic system, as Taiwan is an island economic unit, that only external trade is the way of survivorship. Today facing the freedom and internationalization of global economy, and the increasing development in international trade and multinational corporations, the exchange rate and its risk become more important significantly. Thus, an active exchange rate management is necessary. As well as the forecast of exchange rate, the characteristic of volatility is worth of understanding and discussing.
Generally, the returns of financial assets prices present an asymmetric distribution with high peak and fat tails, and there are three characters on the volatility of prices: (1) Time- varying volatility, it indicates heterogeneous variance. (2) Volatility clustering. (3) Asymmetric volatility. The exchange rate is one kind of financial assets; of course, it should possess the characters that described above. Therefore, this paper applies EGARCH model to test and verify the USD/NTD spot and forward exchange rates in domestic foreign exchange market for researching the volatility of exchange rates between 1994 and 2003. (The term is separated at Oct. 17, 1997 when CBC gave up stabilizing the exchange rates after Asian Financial Crisis.) The empirical results are as follows:
1.The daily returns of USD/NTD exchange rates all present abnormal distribution with asymmetrical bias and high peak, whether the spot or forward rates. In addition, they are not only a stationary time series varying between a stable interval, but also an autocorrelated series with linear dependence, that means the volatility of exchange rates is affected by the past rates varying.
2.By autoregreesive process, the residual series and squared residual series all present autocorrelated, that express heteroskedasticity in the variance of exchange rate returns and hint time-varying volatility. After ARCH LM test, we find that the square residuals are nonlinear dependent, which exhibits higher order correlated. From this we know the returns series varying with time and having ARCH effect of heterogeneous variance.
3.Finally, using EGARCH (1,1) model to test these estimates of parameters in the conditional variance equation. The empirical results are the volatility of USD/NTD exchange rates exiting conditional heteroskedasticity and volatility clustering, but no obvious phenomenon of asymmetric volatility.
|
|---|
