This note shows that the unbiasedness and the weak rationality hy potheses are not rejected for inflation forecasts surveyed by the Central Bank. However, a clear pattern of auto-correlation of forecast errors is found. Furthermore, increases (decreases) in inflation are systematically associated with underestimations (overestimations) of inflation in the following month. This suggests that models in which past realizations of inflation have greater weight in the formation of expectations are more accurate than the assumption of rational expectations. Models aimed at explaining how expectations are formed should be able to explain these stylized facts as well as the hysteresis of forecasts.
