George Tawadros
In this book, Moosa and Burns ‘demystify’ one of the oldest puzzles in International Finance, that being that conventional exchange rate models cannot outperform the naïve random walk model in a simple out-of-sample forecasting exercise. In this comprehensive analysis, Moosa and Burns re-examine this ‘myth’, and show that trying to outperform the random walk should not be the yardstick used to judge the performance of conventional exchange rate models. Instead, they show that the performance of these exchange rate models should be assessed using other criteria, such as correctly predicting the direction of change and the level of profitability (or more generally, utility), the so-called ‘adjusted’ root mean squared error, the use of the Fair and Schiller (1990), and comparing the actual forecast to the perfect forecast. Moosa and Burns then analyse every one of the six explanations usually offered as the reason why the exchange rate models have failed, these being: 1). The stochastic movements in the underlying parameters; 2). Model misspecification; 3). The effect of nonlinearities; 4). Simultaneous equation bias; 5). Sampling errors, and 6). Modelling expectations. They conclude by showing that the random walk cannot be beaten if the forecasting performance of exchange rate models is judged by conventional measures of forecasting accuracy, but can be beaten if the forecasting performance is measured by the accuracy of direction and profitability. Given the thorough analysis undertaken by Moosa and Burns, this book will be the final authoritative word on conventional exchange rate models beating the random walk model in a forecasting exercise. George B. Tawadros Royal Melbourne Institute of Technology