AN EMPIRICAL-STUDY OF SEASONAL UNIT ROOTS IN FORECASTING

We assess the usefulness of pre-testing for seasonal roots, based on t he HEGY approach, for out-of-sample forecasting. It is shown that if t here are shifts in the deterministic seasonal components then the impo sition of unit roots can partially robustify sequences of rolling fore casts, yielding improved forecast accuracy. The analysis is illustrate d with two empirical examples where more accurate forecasts are obtain ed by imposing more roots than is warranted by HEGY. The issue of asse ssing forecast accuracy when predictions of any one of a number of lin ear transformations may be of interest is also addressed. (C) 1997 Els evier Science B.V.