A Model for Real-time Data Assessment and Forecasting

In this paper we propose a model that describes the regularities of the revision process for real-time macroeconomic data. This model takes the typical publication process of statistical agencies into account: after an initial release, revisions are published in the following periods, as additional information comes in and measurement errors are corrected. Beyond the descriptive purpose of assessing the revision structure for a particular variable, the model is used to derive confidence intervals that allow quantifying data uncertainty, and to predict future releases. An application to real-time GDP growth rates demonstrates the usefulness of the model.


Kai Carstensen
David Liedo