In this paper we propose novel techniques for the empirical analysis of adaptive learning and sticky information in inflation expectations. These methodologies are applied to the distribution of households’ inflation expectations collected by the University of Michigan Survey Research Center. To account for the evolution of the cross-section of inflation forecasts over time and measure the degree of heterogeneity in private agents’ forecasts, we explore time series of percentiles from the empirical distribution. Our results show that heterogeneity is pervasive in the process of inflation expectation formation. We identify three regions of the distribution that correspond to different underlying mechanisms of expectation formation: a static or highly autoregressive region on the left hand side of the median, a nearly rational region around the median and a fraction of forecasts on the right hand side of the median formed in accordance with adaptive learning and sticky information.