Robust tests for unit roots in the foreign exchange market

Empirical studies that have tested for unit roots in foreign exchange rate data have assumed independent and identically distributed errors in order to apply Dickey-Fuller test procedures. This assumption ignores the temporal dependence of the error sequence present in many simple efficient market tests, as discussed in Hansen and Hodrick (1980), as well as Krasker (1980). In this paper we apply unit root test statistics that allow for heterogeneously distributed and mildly dependent error processes, and thus provide superior tests of the unit root hypothesis in the case of foreign exchange rates.