Testing for a nonlinear relationship among fundamentals and exchange rates in the ERM

We employ two nonparametric nonlinear testing methodologies, namely a nonparametric nonlinear cointegration approach and a nonlinear Granger causality approach, to test for a nonlinear relationship between macroeconomic fundamentals and exchange rates for two country-pairs, namely the Netherlands-Germany and France-Germany. The results suggest that there is nonlinear cointegration among money, output and exchange rates for Netherlands-Germany, which can be interpreted as evidence of a long-run nonlinear relationship. For France-Germany, we fail to find evidence of nonlinear cointegration, but we find nonlinear Granger causality from French money to the FFr/DM exchange rate. These findings may be interpreted as evidence of a dynamic nonlinear relationship and are consistent with the German dominance hypothesis. On the basis of estimated fractional ARIMA models, we rejected the hypothesis that these nonlinearities are due to bubbles. (C) 2000 Elsevier Science Ltd. All rights reserved.