Public beliefs and monetary policy

We study the evolution of public perceptions of monetary policy and show that Bayesian learning provides a structural account of their time variation. Using a medium-scale DSGE model estimated on U.S. data, we first document substantial and persistent time variation in agents’ perceived monetary policy reaction coefficients. We then show that a Bayesian learning framework, in which agents gradually infer policy conduct under incomplete information, provides a compelling explanation for these dynamics and delivers a markedly better empirical fit than rational expectations. Learning about the inflation reaction coefficient plays a central role. Finally, we demonstrate that evolving perceptions generate state-dependent transmission: weaker perceived inflation responsiveness amplifies and prolongs inflationary responses to shocks, while stronger perceived responsiveness stabilizes inflation, with implications for real activity that depend on the type of shock. Monetary policy actions themselves feed back into beliefs, implying that credibility evolves endogenously and acts as an additional propagation channel.