Nonlinear adjustment towards purchasing power parity: the Swiss Franc-German Mark case

We test the hypothesis of nonlinear adjustment towards the purchasing power parity as suggested by Dumas'' (1992) model. We estimate a stable exponential smooth transition regression model (ESTAR) for the Swiss franc/German mark exchange rate over the 1960-1998 period, where the adjustment to the steady state takes place rapidly. The results reveal that, for small deviations, the real exchange rate is best described by a random walk, whereas for larger deviations the real exchange rate is clearly mean-reverting. The same results are found when the sample is reduced to cover only the post Bretton-Woods period. A Monte Carlo simulation shows that our nonlinear models are clearly stable. When the real exchange rate is outside the no-arbitrage band, the estimated deviation half-lives are about 1.5 and 2.5 years for respectively the entire and the restricted sample.