Beliefs and Fluctuations: A Quantitative Exploration

We develop a methodology that allows DSGE models to capture the empirical content of informational frictions while bypassing the computational challenges that come with conventional formalizations of such frictions. We apply this methodology in the context of an RBC model augmented with a type of belief shocks that helps disconnect expectations of economic activity from fundamentals such as preferences and technologies, and that help capture the notion of "market sentiment", or "confidence". Our baseline specification is calibrated to match some standard moments of the business cycle. We find that belief shocks generate the right co-movement in all key macroeconomic variables; they help matching a number of important business-cycle facts; and they are important contributors to short-run fluctuations. We next estimate the model using a Bayesian maximum likelihood technique and find similar results. We finally allow our belief shocks compete with alternative shocks traditionally introduced in the DSGE literature and find that this does not fundamentally change the overall picture: belief shocks are quantitatively important.