La prévision à l’aide des modèles ARMMI et d’information à priori Forecasting by means of ARIMA models and prior information

ABSTRACT

The method of ARIMA forecasting with benchmarks developed in this paper allows the production of univariate forecasts which take into account the historical information of a series, captured by an ARIMA model (Box and Jenkins, 1970), as well as partial prior information on the future behaviour of the series. The prior information, or benchmarks, stems from the conclusions of a study on the phenomenon to be extrapolated, from forecasts of an annual econometric model or simply from pessimistic, realistic or optimistic scenarios contemplated by the current economic analyst. It may take the form of annual levels to be achieved, of neighbourhoods to be reached for a given time period, of movements to be displayed or more generally of any linear criteria to be satisfied by the forecasted values. By means of this method, the forecaster may then exercize his current economic evaluation and judgement to the fullest extent in deriving the forecasts, since the labouriousness and the "trial and errors" experienced without a systematic method are avoided.