Heterogeneity and learning in inflation expectation formation: an empirical assessment

Relying on Michigan Survey's monthly micro data on inflation expectations we try to determine the main features - in terms of sources and degree of heterogeneity - of inflation expectation formation over different phases of the business cycle. Different learning rules have been applied to the data, in order to test whether agents are learning and whether their expectations are converging towards perfect foresight. Results suggest that behaviour of agents in the right hand side of the distribution is more associated with learning dynamics. Tests for "static" and "dynamic" versions of sticky information are also conducted. Only agents in the middle of the distribution are regularly updating their information sets. Evidence of rational inattention has been found for agents comprised in the upper end of the distribution. We identify three regions of the overall distribution corresponding to different expectation formation processes, which display a heterogeneous response to main macroeconomic indicators : a static or highly autoregressive (LHS) group, a "nearly" rational group (middle), and a group of agents (RHS) behaving in accordance to adaptive learning and sticky information. The latter, generally speaking, are too "pessimistic" as they overreact to macroeconomic fluctuations.