Exchange rate forecasts by the means of econometric models

This paper focuses on the reliable and effective methods of exchange rate forecasting using econometric approach and statistical and economic methods of evaluating the out-of-sample exchange rate predictability. The method tested in this paper is the efficient kitchen sink model, which basically is a regression incorporating multiple predictors. The null hypothesis is that of the equal predictability of the efficient kitchen-sink model and the benchmark random walk model. Using a wide range of statistical and economic tools, the methods are then assessed to find the best performing model. Among empirical models based on uncovered interest parity, purchasing power parity, monetary fundamentals, asymmetric Taylor rule and combination forecasts the efficient kitchen sink method improved by the elastic-net shrinkage method which reduces the effect of the less informative predictors in the regression, is the best statistically and economically, compared to other alternatives