• C14;
  • E17;
  • E37
  • differential interpretation;
  • survey expectations;
  • Bayesian learning;
  • disagreement

In this paper, I estimate a simple Bayesian learning model to expectations data from the Survey of Professional Forecasters. I reformulate the model in terms of forecast revisions, which allows one to abstract from differences in priors and to focus the analysis on the relationship between news and revisions. The empirical analysis shows that there is significant heterogeneity in the interpretation of news among forecasters, in particular at longer horizons, while it decreases closer to the forecast target date. The results also indicate a positive relationship between prior sentiment and interpretation of the signal, in the sense that relatively optimistic (pessimistic) forecasters are likely to believe that the signal under (over) estimates the future realization and assign it a low (high) weight in the forecast revision.