Endogenous Persistence in an estimated DSGE Model Under Imperfect Information


  • Corresponding author: Paul Levine, School of Economics, University of Surrey, Guildford, Surrey, GU2 7XH, UK. Email: p.levine@surrey.ac.uk.

  • We acknowledge financial support for this research from the ESRC, project no. RES-000-23-1126 and from the EU Framework Programme 7 project MONFISPOL. Earlier versions of this article were presented at the Conference ‘Robust Monetary Rules for the Open Economy’ at the University of Surrey, September 20-21, 2007; the CEF 2008 Annual Conference in Paris, June 26–28, 2008; the annual CDMA Conference, 2009, University of St Andrews, 2–4 September; a MONFISPOL workshop, March 11–12, 2010; the 2010 Royal Economic Society Conference, March 29–31; and seminars during 2010 and 2011 at Sheffield University, the Bank of England, Goethe University and Birkbeck College. Comments from participants at all these events are gratefully acknowledged, particularly those from the discussants at the Surrey Conference (Michel Juillard) and the MONFISPOL workshop (Eleni Iliopulos). We also acknowledge comments from two anonymous referees and the editor of this Journal. The usual disclaimer applies.


A framework for estimating Dynamic Stochastic General Equilibrium (DSGE) models by Bayesian methods and validation under very general information assumptions is applied to a New Keynesian model. The standard asssumption that private agents have perfect information observing all state variables including shocks, whereas the econometrician uses only observable data, is compared with both agents having the same imperfect information (II) set. We also generalise rational expectations to a behavioural composite model that allows some households and firms to form expectations adaptively. We find significant empirical support for II as an endogenous persistence mechanism, but this is dominated by that from habit and adaptive learning.