Seasonal integration in economic time series

This paper considers seasonality in Australian macroeconomic time series, emphasizing the roles of unit roots and the selection of differencing filters. The consequences of seasonal unit roots and the importance of correct variable transformation are analyzed. For certain variables, in addition to unit roots at the usual zero frequency, it is found that the hypothesis of seasonal unit roots cannot be rejected. In many macroeconomic time series, the commonly used first-differencing filter is insufficient for the removal of seasonal unit roots, and the resultant bias in the critical values of various tests remains if seasonal integration is not considered.