Can Factor Models Improve Output Gap Estimates in Real Time? Knut Are Aastveit † and Tørres G. Trovik ‡ * † University of Oslo and Norges Bank ‡ Norges Bank and The World Bank This version: October 7, 2008 Draft - please do not quote Abstract Measurement of the output gap in real time is very unprecise. By using a dynamic factor model we can substantially improve the reliability of output gap estimates in real time through two mechanisms: First, as the factor model extracts a common com- ponent in the data and disregards an idiosyncratic component, data revisions will have less impact to the extent that these revisions are due to new idiosyncratic information or measurement errors. Second, through the factor model we are able to handle a ragged edge in the data and incorporate new information on non-synchronized vari- ables as they become available. Thus, the end-of-sample problem is reduced through the favorable nowcasting properties of the model and can be further reduced by com- bining it with forecast augmentation methods. Our approach reduce the forecasting error for the real time output gap to one quarter of the standard method. Keywords: Output gap, real time analysis, monetary policy, forecasting, factor mod- els JEL Classification: C33, C53, E52, E58 * We have received helpful comments from Hilde C. Bjørnland, Jean Boivin, Domenico Giannone, Ragnar Nymoen and Shaun Vahey as well as conference participants at ISF Nice and CEF Paris and seminar participants at Norges Bank, University of Oslo and Universitat Pompeu Fabra. E-mail: Knut Are Aastveit (corresponding author) at k.a.aastveit@econ.uio.no and Tørres G. Trovik at Torres.Trovik@norges-bank.no