|
In this article we model monthly data on the
Japanese nominal exchange rate in relation to the US dollar by means
of fractionally integrated statistical models. For this purpose, we
use both parametric and semiparametric techniques proposed by P.M.
Robinson in a number of papers. The results indicate that the order
of integration of the series is higher than 1 and thus the standard
approach of taking first differences to get series which are
integrated of order 0 (which is required, for example, in the
context of cointegration) may lead to spurious results, the series
still having a component of long memory behaviour. |