How useful are measured expectations in estimation and simulation of a conventional small New Keynesian macro model?

This paper examines how measured expectations (survey data) affect the basic properties of a conventional small New Keynesian macro model. In particular, how do survey expectations change the role of persistence of inflation and output (i.e. coefficients of the corresponding lagged terms)? Survey data are modeled in several different ways, to facilitate an analysis of different relationships with rational expectations. The model is estimated by means of a Bayesian estimator from quarterly euro area data using both aggregated and micro level data from the ECB Survey of Professional Forecasters for 1999Q1–2012Q3. The broad finding is that the use of measured expectations produces more economically meaningful results than does the standard use of model-consistent rational expectations. In particular, survey expectations also reduce the relative weights of the lagged dependent variables in the Phillips curve and the IS curve. All this shows up in impulse responses that turn out to be quite different suggesting that measured expectations are not only proxies of rational expectations. By contrast, measured expectations are related to rational expectations with a way that may well reflect different adjustment and learning processes.