USE OF PRELIMINARY VALUES IN FORECASTING INDUSTRIAL-PRODUCTION

The first preliminary values of the monthly index of Finnish industrial production undergo a major revision once a year. The problem discussed in this paper is how to take this into account when forecasting the index using a transfer function model based on quick indicators. The preliminary values may be interpreted as measurements with error and handled in the state space framework as proposed in the literature. However, if one forecasts beyond the period for which preliminary values are available, this does not lead to any significant improvements in prediction accuracy compared to the case in which the preliminary values are treated as final values without any error. This outcome simplifies the forecasting procedure in this specific example, and has to do with improved quality of the preliminary values of the index over time. © 1991.