Forecasting The RMB Real Effective Exchange Rate Using Mixed Frequency Data

As a kind of Trade-weighted exchange rate index, real effective exchange rategets rid of the price factors and it can reflect the real purchasing power. Also, it is aneconomic indicators that can embody the competitiveness of the trade commodity inthe international market. The change of RMB real effective exchange rate will affectmany parts of domestic economy, such as the foreign trade, employment structure,economic growth and so on. With the economic globalization, the real effectiveexchange rate will play a more important role in the international economy. Thepurchase of the study is to forecast the RMB real effective exchange rate using mixedfrequency data, and improve the accuracy of prediction in order to help the policymaker to make a more appropriate policy. It has a great significance to the economyof our country.In this article, we choose the RMB real effective exchange rate as the lowfrequency data and the USDCNY, EURCNY, JPYCNY, KERCNY weekly data as thehigh frequency data. Then we use the mixed frequency model to forecast the realeffective exchange rate, and compare with the tranditional model such as ADL andAR model. Specifically, we use the MIDAS model and MF-VAR model to forecastthe REER and compare the accuracy with the benchmark model. Through comparingwith the benchmark model we find that the MIDAS model can help to improve theaccuracy of prediction. What’s more, if we import the autoregression term to theMIDAS model, we will find that the accuracy will have a great improvement. And theMF-VAR model is invalid in the short-term prediction, and plays a good performancein the long-term prediction. Moreover, we compare the MF-VAR model with MIDASmodel and MIDAS-AR model, we find that in the short-term prediction, MF-VARmodel perform worse than MIDAS model, and in the long-term prediction, MF-VAR model is better than MIDAS model. But comparing with the MIDAS-AR model, nomatter the short-term or the long-term prediction, MIDAS-AR model will play abetter performance and in the medium-term MF-VAR model do better. Meanwhile,we forecast the next month to next three months’ REER, and we find that the REERwill have a large rise,2%more or less. And at the second and third month in thefuture, REER will still keep appreciated, but the range will decrease.This paper is divided into four chapters. First chapter introduces the significanceand the background knowledge of the study and summarize the representative point ofwestern and Chinese scholars. Second chapter, we introduce the mixed frequencymodel, such as MIDAS and MF-VAR model, and introduce the method of estimation.In the third chapter, we forecast the REER with MIDAS model and compare theaccuracy with benchmark model in order to show its superiority in forecasting. Andbuilding the MIDAS model to do real-time nowcastng and short-term forecasting.Chapter4analyzes the effectiveness and the prediction accuracy by MF-VAR model.Finally, we give the summary of the result.