Cointegration and the monetary exchange rate model revisited

To test for cointegration in a multiple of countries a number of procedures are available: a panel vector error correction framework, a panel-data version of the Engle and Granger (1987) two-step procedure and the Johansen (1991) trace statistic. We apply these three methods on two four-country datasets consisting of the exchange rate data and monetary fundamentals relative to both the United Kingdom and the United States to test the empirical validity of the monetary exchange rate model. Of the three aforementioned methods only the panel vector error correction approach provides evidence for the validity of both the cointegration restriction as well as the long-run parameter restrictions of the monetary model, irrespective of the numeraire country.