Over the past three decades there has been a growing interest in modelling and forecasting exchange rate movements. Broadly speaking, two modelling approaches have been pursued. The fundamental approach tries to explain the fluctuations of exchange rates in terms of exogenous macroeconomic variables. Alternatively, the technical approach centres on finding patterns in the movements of the historical data. A number of papers have dealt with the in-sample forecasting performance of empirical exchange rate models. However, systematic studies of their out-of-sample performance are relatively scarce. One major study in this area is by Meese and Rogoff (1983). They found that several important, conventional, structural models based on the monetary/asset theories of exchange rate determination were outperformed, in terms of out-of-sample forecasting accuracy, by a simple random walk. The forecasting performance of the models was poor even though their estimation was based on actual realized values of future explanatory variables. Boothe and Glassman (1987) subsequently confirmed these findings for a number of key exchange rates over the period 1976–84, i.e. the post-Bretton Woods era. When time-varying parameters are incorporated in the models to improve their forecasting performance, both Alexander and Thomas (1987) and Wolff (1987) have shown that these models are still outperformed by the simple random walk forecasting rule.
Weitere Kapitel dieses Buchs durch Wischen aufrufen
- Modelling Exchange Rates Using MARS
Jan G. de Gooijer
- Palgrave Macmillan UK
Neuer Inhalt/© Stellmach, Neuer Inhalt/© Maturus, Pluta Logo/© Pluta, Frankfurt School