Exchange Rate Prediction with MATLAB

This paper constructed regression models on exchange rate prediction and conducted predicting process with different combinations of regressors. Fullwork

Identification was based on KRW-USD exchange rate data from June 2001 to December 2020. Comparisons of regression results suggest that the fitness of the model is significantly improved when the exchange rate in the former period is introduced as a regressor, but the fitted model does not seem to be desirable for prediction due to the lag of time. Whereas the both domestic and foreign inflation rates are significant for exchange rate prediction, the long run exchange rate is less important than short run exchange rate in most of the models. With respect to above findings and the failure of constructing a model with smaller RMSE than the benchmark model, this paper proposed that the undesirable result of exchange rate prediction may be attributed to ignoring short shocks and reference dependence.