Abstract
It is demonstrated that the monetary model of exchange rates is better than the random walk in out-of-sample forecasting if forecasting accuracy is measured by metrics that take into account the magnitude of the forecasting errors and the ability of the model to predict the direction of change. It is suggested that such a metric is the numerical value of the Wald test statistic for the joint coefficient restriction implied by the line of perfect forecast. The results reveal that the monetary model outperforms the random walk in out-of-sample forecasting for four different exchange rates.
Notes
1 The drift factor is effectively the mean percentage change in the exchange rate. It can be estimated by regressing the percentage change in the exchange rate (or the first log difference) on a constant. For the four exchange rates, the estimated values of the drift factor and the t statistics are as follows: JPY/USD (−0.53, −1.30), CAD/USD (−0.13, −0.26), JPY/CAD (−0.26, −0.39) and GBP/USD (0.28, 0.65).