Modelling seasonal patterns and long-run trends in U.S. time series

This paper discusses univariate methods used to deal with seasonal data with special emphasis on seasonal unit root and periodic models. These models are examined in the context of 15 US macroeconomic variables. Many of these variables can be described as being periodically integrated, that is, having stochastic trends and seasonal elements that cannot be separated. Periodically varying parameter models are fitted to these variables and evaluated against other seasonal models using forecasting tests.