Revisiting the Out-of-Sample Exchange Rate Predictability in the Monetary Model

Abstract

We utilize Monte Carlo simulations to evaluate, in finite samples, the forecasting performance of the monetary model. The data generating process (DGP) is based on the assumptions of Engel and West (2005) about the present-value model for exchange rates, namely that the discount factor is close to unity and the fundamentals have unit-root processes. We evaluate the out-of-sample performance of the monetary model against the random walk model by using the long-run regression test. While the forecasting power of the long-run regression is not strong, the experimental evidence illustrates that the probability of out-of-sample exchange rate predictability at long horizons is generally larger than that at the short horizons. We conclude that the present-value model under Engel and West’s (2005) explanation has a heretofore unrecognized implication of out-of-sample exchange rate predictability at long run horizons.