Performance of Periodic Error Correction Models in Forecasting Consumption Data

Periodic time series models have become an appealing tool for the analysis of time series showing distinct seasonal patterns. Since these models condition the data{generating mechanism of a given time series on the season they are able to cope with periodic generalisations of common economic models introducing seasonal preferences, seasonal technologies etc. The paper examines for some macroeconomic time series the forecasting performance of periodic models in comparison to some non{periodic (standard) speciications. A periodic generali-sation of the consumption income relation is derived. Diierent speciications of a periodic Error{Correction Model (ECM) for the consumption series are used for estimation and forecasting. The analysed time series are consumption and income series from UK, Sweden, Germany and Japan. The author thanks the participants of the sessions "Forecasting and Seasonality" for helpful comments.