Modelling foreign exchange rates

The theoretical models developed under assumption of purchasing pover parity fail to explain large fluctuations that appear in the currency time series since the beginning of the 1980s. The object of the present paper is to demonstrate, that unrestricted Engle-Granger cointegration model with error correction term is a valid framework for analyzing long-run exchange rate equilibrium as well as short term dynamics. In contrast to previous works, the genetic algorithm input selection is used to identify possible exogenity of variables entering in the model and a simulation based experiment splits short-run dynamics into several consecutive periods to improve model's final empirical validity.

Keywords: cointegration, error correction, genetic algorithms, exchange rates