Testing the purchasing power parity in pooled systems of error correction models

In this paper we test the purchasing power parity for the post Bretton Woods period for 18 main industrial countries. As base currencies we use alternatively the Deutsche mark, the Japanese yen, and the US dollar. We employ error correction models for single countries and on the level of pooled equations allowing efficient inference on domestic and foreign price elasticities of nominal exchange rates. Likelihood ratio tests are applied to infer on linear restrictions implied by the economic relationship. Critical values for these tests are estimated by means of the wild bootstrap that copes with heterogeneous error distributions and contemporaneous correlation within a pooled system of single equations. Furthermore, the tests are performed recursively in order to address the issue of time dependence of our results. We find that the purchasing power parity provides an accurate description of exchange rate dynamics if the Deutsche mark or the Japanese yen are used as base currencies. Specified towards the US dollar we reject the economic model. It turns out that this overall conclusion is not invariant with respect to the investigated sample period.