An Empirical Study On Rmb Exchange Rate Forecasting Based On ARCH Family Models

Since the 1970 s, the progress of world economy integration had gradually been deepened. As exchange rates are bridges between connected currencies in the world, they play a significant role in international economy and trade, their fluctuations would influence the stability of countries' financial system and macroeconomic significantly. In recent years, as China's economy continually grows and China's marketization process gradually deepens since its accessesion to WTO, the RMB's exchange rate fluctuations attract tremendous attention around the world.Firstly, this thesis investigated domestic and overseas research progress and frontier issue up to date concerning volatility and prediction of exchange rate method employed in China and the rest of the world. According to the characteristics of fluctuation clustering of exchange rate, we make a selection of models and determine their forecasting abilities. Based on econometric theory, it's investigated and compared that the principle and analysis method of the autoregressive conditional heteroscedasticity(ARCH) Model, the generalized autoregressive conditional heteroscedasticity(GARCH) Model and asymmetric GARCH Model(including the EGARCH and TGARCH Model). On the basis of review and analysis of evolution of China's exchange rate mechanism, an empirical study of the RMB exchange rate prediction are conducted using 5372 data samples during January 1994 and May 2005 adopted from the Database Pacific Exchange Rate Exchange Rate Service Database Retrieval System(v2.15).Then, ARCH model and ARCH family models are used to fit the logarithmic difference form of RMB/GBP, RMB/EUR, RMB/JPY and RMB/USD exchange rates between January 1994 and May 2015, their fitting method and the main criteria of choose the optimal model are clarified. The empirical study on the efficiency of ARCH model and ARCH family models in describing the RMB exchange rate fluctuations and in exchange rate prediction has carried, and the prediction errors and their statistical indicators are calculated. Considering exchange rate prediction ability of different type of models, the efficiency in RMB exchange rate prediction between ARCH family models and random walk model have been compared, so that the predictability of RMB exchange rate can be identified.Empirical analysis results indicate that there are significant leverage effects shown in RMB/USD exchange rate returns while there is no leverage effect shows in RMB/GBP, RMB/EUR and RMB/JPY exchange rate returns. On the other hand, the expected returns of RMB/USD exchange rate in the future is highly related with the degree of risks the investors are facing with in the foreign exchange market, while in RMB/GBP, RMB/EUR and RMB/JPY exchange rate returns, there is no relationship between expected returns of RMB/USD exchange rate and the degree of risks the investors are facing with. In general, the ARCH and GARCH models are very effective in describing the exchange rate volatility and in exchange rate prediction. And the comparison of exchange rate prediction ability between ARCH family models and the random walk model shows that the ARCH family models perform better. This result suggests that exchange rates are predictable and the conditional heteroscedasticity of model residuals explains fluctuations of exchange rates to a large extent.