Disciplining Expectations and the Forward Guidance Puzzle

Abstract

Forward guidance operates via the expectations formation process of the agents in the economy. In standard quantitative macroeconomic models, the expectations are unobserved state variables and little scrutiny is devoted to analysing the dynamic behaviour of these expectations. We show that the introduction of survey and financial market-based forecasts in the estimation of the model disciplines the expectations formation process in DSGE models. When the model-implied expectations are matched to observed expectations, the additional information of the forecasts restrains the agents’ expectations formation. We argue that the reduced volatility of the agents’ expectations dampens the model reactions to forward guidance shocks and improves the out-of-sample forecast accuracy of the model. Furthermore, we evaluate the case for introducing a discount factor as a reduced form proxy for a variety of microfounded approaches, proposed to mitigate the forward guidance puzzle. Once data on expectations is considered, the empirical support to introduce a discount factor dissipates.

Keywords: Bayesian estimation, DSGE models, expectations formation, forecasting, monetary policy

JEL Classification: C13, C52, E3, E47, E52