This paper carries out the task of evaluating inflation forecasts from the Livingston Survey and the Survey of Professional Forecasters, using the Real-Time Data Set for Macroeconomists as a source of real-time data. We examine the magnitude and patterns of revisions to the inflation rate based on the output price index. We then run tests on the forecasts from the surveys to see how good they are. We find that there are several episodes in which forecasters made persistent forecast errors, but the episodes are so short that by the time they can be identified, they have nearly disappeared. Thus, improving on the survey forecasts seems to be very difficult in real time, and the attempt to do so leads to increased forecast errors.