THE DECOMPOSITION OF FORECAST IN SEASONAL ARIMA MODELS

This paper presents a procedure to break down the forecast function of a seasonal ARIMA model in terms of its permanent and transitory compo nents. Both depend on the initial values at the forecast origin, but t heir structures are fixed and independent of this origin. The permanen t component is an estimate of the long-run projection of the correspon ding economic variable and the transitory element describes the approa ch towards the permanent one. Within the permanent component a distinc tion is made between the factors that depend on the initial conditions of the system and those that are deterministic. The procedure is comp ared to other methods presented in the literature and illustrated in a n example.