Error correction modelling and dynamic specifications as a conduit to outperforming the random walk in exchange rate forecasting

The proposition that dynamic exchange rate models can outperform the random walk in out-of-sample forecasting, in the sense that they produce lower mean square errors, is examined and disputed. By using several dynamic versions of three macroeconomic exchange rate models, it is demonstrated that dynamic specifications outperform the corresponding static models but improvement in the forecasting power may not be sufficient for the dynamic models to perform better than the random walk. The results are explained by suggesting that any dynamic specification or transformation of the static model leads to the introduction of a lagged dependent variable, which in effect is a random walk component. The analysis leads to the conclusion that it is implausible to aim at beating the random walk by augmenting a static model with a random walk component.