Forecasting: Arima or Kalman Models

Abstract In this article we have compared two of the currently most interesting quantitative models in forecasting applied to the socio-econcmic field, i.e. the ARIMA model and the Kalman filter. The comparison has been based on three fundamental points of view: model adequacy, identification procedure and forecasting function. We have identified two reasons which in our opinion support the ARIMA model: - ARIMA model is more adequate in describing the generation of random events. - ARIMA model owing to the identification procedure, is a source model, not a realization model, which means that on the average, all horizons taken into account, it results in a lower measure of errors in forecasting.