The purpose of this paper is to research the countries' exchange rates. The paper is organized around the empirical modeling method which is devised by Van Dijk, D.; Terӓsvirta, T.; Franses, P.H.(2002) and Van Dijk, D.; Terӓsvirta, T.; González, A.(2005). It consists of estimation, specification tests, and forecast stages.
The data is chosen from 1974M1 to 2008M12. We set the data from 1974M1 to 2003M12 as in-sample period, and from 2003M1 to 2008M12 as out-of-sample period. The macroeconomic variables considered in this paper include nominal exchange rates and real exchange rates.
We get four important empirical results in this paper. First, the STAR model shows that every countries exists “eleventh month effects”. Second, the coefficient of nonlinear in Denmark’s, France’s, Germany’s, Italy’s, Spain’s and UK’s model are statistically significant. This result implies that the government of Japan had been intervening significantly in foreign exchange markets. And the government of other countries had been not intervening in foreign exchange markets. Third, the gamma value of the Panel model is statistically significant but slight. We can conclude that nominal exchange rates and real exchange rates exist relationship in long terms but not in short terms. Fourth, the forecast abilities of two models are almost better than random walk model.