Empirical Investigation of Real Exchange Rate Misalignments in Morocco: A Machine Learning Approach

Machine learning techniques such as Random Forests and LASSO have become increasingly popular for forecasting time series dynamics. These methods have demonstrated superior predictive accuracy, particularly when dealing with a large number of explanatory variables. The primary goal of this study is to evaluate the effectiveness of these techniques in forecasting Real Exchange Rate misalignments, which refer to the disparity between the Real Effective Exchange Rate and the equilibrium or steady-state Real Exchange Rate. Our dataset consists of around one hundred variables encompassing economic, financial, social, and global environmental indicators. The application of these methods indicates that the predictive performance is satisfactory and aligns with findings from the International Monetary Fund, particularly those generated by the LASSO algorithm. Projections suggest an overvaluation in the second quarter and throughout 2021, with additional forecasts for 2022 indicating a 0.22% overvaluation of the REER.