The Use of Preliminary Data in Econometric Forecasting

Abstract

An examination of two types of simple forecasting models using preliminary data compares the merits of optiml versus traditional predictors and indicates the relationship of delay in the availability of data revisions to forecasting accuracy. The Kalman filter approach is used in the first model, based on the optimal use of data containing errors in forecasting. Several suboptimal predictors, which ignore preliminary data, treat it as error-free, or adjust for bias and serial correlation, are then compared. A significant improvement in accuracy is demonstrated with the optimal use of forecasting models. Whether accuracy will improve with more complex models is not yet known. 10 references.