Linear models in which exchange rates are driven by stochastic `news' have generally proven disappointing. In Discussion Paper No. 370, Research Fellow Paul de Grauwe and Kris Vansanten present a non-linear model of exchange rate determination which incorporates concepts from chaos theory, as applied initially in the natural sciences. Non-linearities in their model are created by a J-curve effect in the trade balance and by modelling both `chartist' and `fundamental' influences in expectations formation. The model is capable of producing unpredictable exchange rate movements, of the sort observed in reality, without resort to stochastic `news'.
One property of chaotic models is the extreme sensitivity of the time series they generate to only small changes in their initial conditions or parameter values. De Grauwe and Vansanten report the results of several such experiments. After only a few periods, the time path of the exchange rate appears to have been generated from a totally different model. This implies that forecasts based on exchange rate models are effectively impossible, because the tiniest errors generate totally different paths for the exchange rate. These findings will be reported in more detail in Bulletin No. 37, in connection with a lunchtime meeting by Paul de Grauwe.

Deterministic Chaos in the Foreign Exchange Market Paul De Grauwe and Kris Vansanten

Discussion Paper No. 370, January 1990 (IM)