Nonlinear Approximation to the Overshooting Model Using Multilayer Neuronal Networks for the USD/COP Exchange Rate

Since the 1970s much work has been done attempting to provide empirical support for some models that offered a linear explanation for the exchange rate dynamic of a country, including that of Dornbusch. So far none have been conclusive and the random walk is considered the best model to which it can be adjusted. De Grauwe has shown that, with the presence of nonlinear relationships and heterogeneous expectations on the part of speculators, the exchange rate can have an apparently random behavior, but with a deterministic explanation. This article presents the Dornbusch model in the nonlinear version proposed by De Grauwe and Dewachter (1993) and an approximation using multilayer neuronal networks applied to the case of the USD/COP exchange rate