Periodic integration in quarterly UK macroeconomic variables

This paper presents empirical evidence on the seasonal patterns in several UK macroeconomic variables, additional to related evidence reported in Osborn (International Journal of Forecasting (1990), 6, 327–336). The method used is a test procedure for seasonal unit roots that allows parameters to vary over the seasons. This extension of currently applied procedures can select between seasonal and periodic integration. In a small Monte Carlo experiment, this new method is evaluated with respect to two rival procedures. The empirical results for the UK variables indicate that many of these are periodically integrated. The implications of this outcome on modelling and forecasting are discussed. One of the implications is that a periodic error correction model for the univariate series can outperform non-periodic models with respect to forecasting.