Seasonal unit roots and forecasts of two-digit European industrial production

Monthly industrial production in important sectors of the German, French an d UK economies are shown to exhibit very strong seasonality, such that typi cally 80% or more of the variation in monthly growth can be attributed to s easonal effects. Seasonal unit root test results imply that most of these s eries should be modelled using conventional first differences with the incl usion of monthly dummy variables, rather than as specifications involving o ther levels of differencing. However, when the post-sample forecast accurac y is compared for various models, annual difference specifications often pr oduce the most accurate forecasts at horizons of up to a year. First differ ence models appear to be most accurate at short forecast horizons for serie s where seasonality is particularly marked. The study also examines the imp act of updating coefficient estimates and model respecification within the forecast period. (C) 1999 Elsevier Science B.V. All rights reserved.