Structural Models vs Random Walk: The Case of the Lira/$ Exchange Rate

Abstract

After presenting the structural models of exchange-rate determination, the authors show that their out-of-sample predictive performance of the lira/$ exchange rate is inferior to that of the random walk model. Only by moving away from these single-equation, semireduced form models toward suitable economywide macroeconometric models can one hope to beat the random walk. Following this course, the authors show that the Mark V version of their continuous time macroeconometric model of the Italian economy outperforms both the existing structural models and the random-walk process in out-of-sample forecasting tests of the lira/$ exchange rate.