Survey expectations, learning and inflation dynamics

Abstract

We propose a framework that exploits survey data on inflation expectations to refine the identification of processes that drive inflation in DSGE models. By decomposing fundamental markup shocks into persistent and transitory components, our approach effectively integrates timely survey information about the nature of inflation shocks, enhancing forecasts of inflation and other macroeconomic variables. Models with expectations based on a learning setup can more effectively utilize signals from the combined datasets of realized inflation and survey forecasts compared to their Rational Expectations counterparts. The learning model’s ability to generate time variation in the perceived inflation target, inflation persistence, and sensitivity to various shocks enables it to detect changes in the fundamental processes driving inflation. These features help overcome limitations of survey data and enhance forecast accuracy, particularly during periods when survey forecasts exhibit systematic prediction errors. Specifically, the model with learning successfully identifies the more persistent nature of the recent inflation surge.