The Analysis Of Changes And Combined Forecast On RMB Exchange Rate

Recently, the analysis and forecast of exchange rate has been a hot issue intoday’s economic areas. Since RMB exchange rate mechanism reformed in July21,2005, the floating band of exchange rate has been increasing，which bringsmuch more risks than before. And because the U.S. dollar takes an importantposition in a basket of currencies, analyzing and forecasting the exchange rateof RMB against U.S. dollar correctly has a vital significance, thus thegovernment can correctly formulates monetary policies and financialinstitutions can avoid foreign exchange risk. Therefore, the paper will establishthe models on the exchange rate of RMB against U.S. dollar to analyze andpredict. The main contents of the paper are as follows:Firstly, the fluctuations of RMB exchange rate have been analyzed. Itdescribes five kinds of exchange rate decision theories: Theory of InternationalIndebtedness, Theory of Purchasing Power Parity, Theory of Interest Rate Parity,Monetarism Theory and Asset Portfolio Theory. And the paper summarizes thevarious factors in exchange rate changes, such as national income, balance ofpayments, inflation and relative interest rates, and so on.Secondly, based on ARIMA model, GARCH model and Grey Markov Model,we establish combined forecasting models of the non-stationary time series, anddo empirical analysis on the181weekly data from July2010to December2013,by the three empowering means of arithmetic average, minimization ofprediction error square sum and variable weight. According to the requirementof empirical research, the data is divided into two parts: the former154datafrom July2010to June2013is used to estimate parameters of the forecastingmodels, and the remaining27data is then used to test predicting effects ofmodels.Finally, the paper measures accuracies of all forecasting models by usingfour error indicators, and further to explain the models’ pros and cons. Theresults show that the prediction accuracy of the combined forecasting modelbased on the means of variable weight is better than other forecasting models.