Empirical Study On Combining Forecasts Method

Since China improved the RMB exchange rate regime in July, 2005, RMB exchange rate fluctuates greater than before. As a result, the risk brought about by fluctuation of exchange rate is more drastic responsively, in addition, because of the greatest weight U.S dollar got in the currency basket, so it is very helpful for the government formulating monetary policy and companies avoiding exchange rate exposure to analyze and forecast the fluctuation of exchange rate correctly. Then combining forecasts method is taken to analyze the movement of RMB exchange rate against U.S dollar in this thesis, and a comparison on forecast performance is made between three different ways of deciding weights.This paper is divided into five chapters to illustrate the methods of RMB exchange rate forecasting.Chapter one is a revision of the theory of exchange rate forecasting. Generally speaking, there are generally three kinds of models for exchange rate forecasting, economic model, time series model and combining forecasts model.Chapter two first reviews the evolvement of RMB exchange rate regime. Then with the help of statistical discrete index, author analyzes the fluctuations of RMB exchange rate since the exchange rate reforming of July, 2005 and its economic factors.Chapter three is about the assessments of exchange rate forecast model.Chapter four is the empirical analysis of the RMB exchange rate against U.S dollar through the statistical trending model. Monthly data of RMB exchange rate against U.S dollar and some other economic variables between Jan 1997 and Dec 2007 has been collected to support the analysis. The econometrical methods in this paper include co-integration, ARMA model, constant weight combining forecasts model, optimal weighted combining forecasts model, deterministic time-varying parameters combining forecasts model. Finally root mean square error and regression-test are performed to compare the forecast performance.Chapter five is about the movement analysis of RMB exchange rate against U.S dollar in 2008, with the help of both forgoing econometrical models and subjective analysis.The essence of combining forecasts method is statistical weights applied on forecast. An important problem of it is how to assign weights, which could either be deterministic, or stochastic. As well as we know, there are various factors having impact on exchange rate, rather, their impact are usually time-varying. Therefore, author prefers to make parameters time-varying in order to grasp the movement of exchange rate as accurately as possible, and apply the model proposed by Diebold and Pauly(1987), the regression-based approach with deterministic time-varying parameters, in forecasting RMB exchange rate against U.S dollar. Those are right where the innovations of this paper lie on. The inadequacies of this thesis are as follows. First, the way used in this article is limited to deterministic time-varying weights method, whereas this way has its own disadvantages which enable the model to do nothing to randomness of exchange rate fluctuation. Second, the combining forecasts model in this paper just takes account of the combination of ARMA model and economic model basing on PPP and IRP, without other combinations of different exchange forecast models; Third, deterministic time-varying weights model is only applied in forecasting RMB exchange rate against U.S dollar, not in other exchange rate, so it would be unwise to generalize the conclusion on forecast performances.