New models for nowcasting


Michael Andersson and Ard den Reijer describe how a central bank can forecast the current state and immediate future of the economy with the use of what are known as nowcasting models. As outcome data is published with a time lag and the current situation is not yet known, such forecast models need to utilise the large number of indicators available in real time.

The article describes two methods for making such forecasts in the short term. The first method estimates many small models and then weighs their forecasts together. The second method weighs together information in several series and then makes a forecast. The new methods also make it possible to understand and interpret the forces that drive economic development and that are reflected by forecasts and forecast revisions.


The article also describes how the Riksbank's nowcasting system has been expanded with models that take account of the different frequencies at which indicator variables are observed and the different delays in their publication. The authors show how well a dynamic factor model, with the help of a more than 100 indicator variables at a monthly frequency, can forecast quarterly percentage changes in GDP. An application shows how GDP forecasts during the fourth quarter of 2008 were gradually revised downwards because the availability of new indicators changed the assessment of how much the global financial crisis affected the Swedish economy.


Michael K. Andersson and Ard H.J. den Reijer
At the time of writing, Michael Andersson worked in the Monetary Policy Department, but he now works at Finansinspektionen. Ard den Reijer works in the Monetary Policy Department.

Last reviewed

Content expert

Contact content expert

Fill in the information

To minimize automated spam, please answer the question in the box below.

4 + 2 ?