Learning to forecast the exchange rate: two competing approaches

This paper compares two competing approaches to model foreign exchange market participants' behavior: statistical learning and fitness learning. These learning mechanisms are applied to a set of predictors: chartist and fundamentalist rules. We examine which of the learning approaches is best in terms of replicating the exchange rate dynamics within the framework of a standard asset pricing model. We find that both learning methods reveal the fundamental value of the exchange rate in the equilibrium but only fitness learning creates the disconnection phenomenon and only statistical learning replicates volatility clustering. None of the mechanisms is able to produce a unit root process but both of them generate non-normally distributed returns.