Research Discussion Paper – RDP 34 An Analysis of Econometric Ex-post Forecasting Error


Forecasting is a central issue in any policy making body. Within the Research Department of the Reserve Bank a number of forecasting techniques are used. Over the past few years increasing use has been made of econometric models for this purpose. But development of the techniques of using models in this way and the study of the various properties of this forecasting tool are still in their relative infancy in Australia.

To stimulate debate on these topics we hope to present several papers of which this is the first. Others will probably cover the techniques of using models for forecasting and the predictive value of the forecasts. But the aim of the present paper is limited in that the emphasis is on methods rather than results, although results are given for illustrative purposes.

Specific qualifications should be made about these results. The model to which they relate was an earlier, preliminary version of the Bank's forecasting model and was used for experimental purposes only. The results would not necessarily reflect the properties of the model currently used for forecasting within the Bank, or the properties of the Bank's structural model, RBA.

This paper stresses the need for systematic error analysis to be included as an important integral part of any continuing econometric forecasting project.