Defined as the ratio used when one currency is converted into another, exchange rate mirrors the relative purchasing power of currency A as compared with currency B and it affects the welfare of people of the issuing country when they consume in a foreign state. Moreover, international trade and cross-border movement of capital are both largely influenced by the exchange rate. For an open economy, its international economic involvement is highly correlated with its domestic economic development, and therefore the exchange rate is a factor of crucial importance to the prosperity of an economy. Meanwhile, it is also a symbol of the economic strength and comprehensive economic power of a country. As the exchange rate is such an important variable in global interaction, to forecast it accurately will be helpful in international economic activities. To conclude, exchange rate forecasting is one of the focuses in exchange rate research.Exchange rate forecasting has drawn wide attention in the academic circle and many researchers have found some clues to the answer of the question. This paper first introduced the achievements made in the study of exchange rate forecasting by foreign researchers and domestic scholars (the research of the latter focusing on USD/CNY exchange rate) and found that most Chinese scholars used univariate methods like ARIMA (Autoregressive Integrated Moving Average Model), wavelet transform as well as their combination. But many of them ignored the impact of macroeconomic fundamentals on exchange rate forecasting. The principle of univariate methods lies in that a time series contains information that can be used to forecast its future movement, which is a data mining method. It is objective as it discovers the information deeply hidden in the time series, but it is also weak in the face of theoretic argument. In the real world, economic factors are correlated and interacted with one another and exchange rate changes out of numerous reasons, known or unknown. It is obviously unreliable to ignore the relation between exchange rate and other economic variables and to forecast future exchange rate simply by previous exchange rate.Having noticed the interaction between the exchange rate and macroeconomic fundamentals, this paper established a VAR (Vector Autoregression) model, which is able to display the interaction between variables, of the three variables:exchange rate, money supply and interest rate, which were chosen on the basis of many exchange rate theories, to provide a more accountable tool of USD/CNY exchange rate forecasting. Meanwhile, the VAR model unveiled the interaction between exchange rate, money supply and interest rate as well as the extent of influence of one variable on another, which was neglected in earlier research.This paper used USD/CNY exchange rate (monthly average), Chibor (China Interbank Offered Rate, monthly weighted average) and money supply (M2) during the period between July,2005and Dec,2014. After stability test, it was decided the VAR model would be established with the first order of USD/CNY exchange rate, written as D(FOREX), the interest rate as IR, and the first order of money supply, written as D(M2), all of which were stable series. With a limited number of samples, this paper tried to set up a VAR model within5lag periods. There were4information criteria saying3lag periods was the optimal and thus a VAR(3) model was finally established in the paper. When parameter estimation was done, it was found, according to the adjusted-R2,38%of the fluctuation of D(FOREX) could be accounted for by the regression equation; so did60%of the fluctuation of IR and30%of the fluctuation of D(M2). F-statistics were all significant. The results of AR roots test admitted the stability of the VAR model.After the stability of the model was confirmed, a Granger causality test was done based on the VAR model. The results showed that D(FOREX) and D(M2) Granger caused interest rate, which meant USD/CNY exchange rate changes and money supply growth were accountable for the changes in the interest rate; D(FOREX) and IR Granger caused D(M2), which meant the PBOC might have taken into consideration USD/CNY exchange rate and IR when deciding money supply growth. However, IR and D(M2) did not Granger cause D(FOREX), which meant IR and M2were not significant factors in exchange rate forecasting.Granger causality test confirmed the interaction between the three. Then the IRFs (Impulse Response Functions) were established based on the VAR model to verify the paths of interaction. It portrayed the long-term relation between M2and USD/CNY exchange rate, the lag structure between M2and IR, etc. Variance decomposition was the last step of VAR analysis. The result showed that most of the fluctuation of USD/CNY exchange rate was accounted for by its lags. Interest rate and money supply could explain only a small part of it. That was why D(M2)and IR did not Granger cause D(FOREX).VAR analysis, on one hand, displayed the interaction between interest rate, exchange rate and money supply; on the other hand, found that interest rate and M2could only explain a small part of USD/CNY exchange rate fluctuation during July,2005and Dec,2014. Finally, this paper did an out-of-sample exchange rate forecast based on the VAR model and measured the accountability of the forecast via its stability and co-integration with the real exchange rate series. It was shown that the forecast was acceptable. This paper also set up an ARIMA model and compared the forecasts given by the two models, finding that currently the USD/CNY exchange rate forecast given by VAR model was as accurate as that given by the ARIMA model. With the growth of the volatility of USD/CNY exchange rate, the VAR model, which is able to link the exchange rate to macroeconomic fundamentals, will definitely provide more precise USD/CNY exchange rate forecasts than univariate models. |