The relationships between the macroeconomical variables and exchange rates for Turkey and some other countries

In this study, for developed and the developing countries, in determination of the exchange rate which is an important parameter, a prediction depending on the exchange rates was tried to do focusing on which macroeconomic variables are effective, by researcing the relationship between each others. In Turkey, Brasil, Mexico and South Africa, with the study on monthly data between January of 2003 and December of 2018, dependent variable is nominal USD/EUR exchange basket rates. Independent variables are also the budget deficit, current account deficit, the foreign trade deficit, interest rate, oil prices, industrial production index, consumer prices index and, producer prices index. With this study that is considered in a certain time interval, three different models such as VAR Model, Granger Causality Analysis, Multi -Variable Linear Regression Analysis and Artificial Neural Nets have been applied and hence, the best model performance was determined. While using the EViews 11 Econometric Software Program for first two model, The MATLAB R2019a Statistical Package Program was used for the analysis which is done by the Artificial Neural Nets. As conclusion of the study, it was determined which the variables such as the current account deficit, the foreign trade deficit, the consumer price index, the producer price index and the oil price index are significant and effective on the exchange rates. Also, the Granger Causality relationships based upon the one way and two ways were determined by using the current account deficit, the foreign trade deficit, the consumer price index, the producer price index and oil prices index. In terms of the results of the Artificial Neural Nets (ANN) based analysis, it was understood that the ANN approach is more successfully comparing with the classical methods.