Survey Expectations and Learning

In this paper, we evaluate a model that describes real-time inflation data together with the expectations measured by Survey of Professional Forecasters (SPF). We work second-order autoregressive in which agents learn over time intercept and persistence coefficients based on data. To process revisions real data, allow for news noise disturbances. contrast to usual time-varying parameter vector autoregression, use non-linear Kalman filter techniques estimate underlying process. identify systematic changes long-run expected rate are implied model. The forecasts then compared SPF forecasts. As cannot reject hypothesis produced our model, re-estimate using nowcasts as additional observables. This augmented does not change nature magnitude variation but it help reduce uncertainty estimates. Overall, estimated time-variation confirms results perceived present DSGE models learning (Slobodyan Wouters, 2012a, 2012b).