AbstractThis paper examines the effectiveness of forward guidance shocks in the US. We estimate a New Keynesian model with imperfect central bank credibility and heterogeneous expectations using Bayesian methods and survey data from the Survey of Professional Forecasters (SPF). The results provide important takeaways: (1) The estimated credibility of the Fed’s forward guidance announcements is relatively high, but anticipation effects are attenuated. Accordingly, output and inflation do not respond as favorably as in the fully credible counterfactual. (2) The so-called “forward guidance puzzle” arises partly from the unrealistically large responses of macroeconomic variables to forward guidance under perfect credibility and homogeneous fully informed rational expectations, assumptions which are found to be jointly inconsistent with the observed US data. (3) Imperfect credibility provides a plausible explanation for the empirical evidence of forecasting error predictability based on forecasting disagreement found in the SPF data. Thus, we show that accounting for imperfect credibility and forecasting disagreements is important to understand the formation of expectations and the transmission mechanism of forward guidance.