A comparison of predictability of exchange rates between G7 and Asian developing countries: potential impact of government intervention

The issue of predictability of exchange rate is reexamined by comparing the performance of five forecasting models across six G7 and six Asian developing countries. Our result shows that in terms of prediction accuracy, the random walk model performs poorly for three Asian developing countries, namely, Indonesia, Korea and Malaysia, but it fares well for the G7 countries for the total out-sample period. However, once the period, in which the Asian currency crisis prevailed, is purged from the total out-sample period, none of the models examined here outperform the random walk model consistently or have any market timing ability.