The more we know, the less we agree: Higher-order expectations and public announcements

Polarization of opinions after public announcement is widely observed, but often considered to be inconsistent with Bayesian learning. I show that this is not the case in environments where higher-order expectations play a role. I characterize informational structures where public announcement leads to polarization in all higher-order expectations, but not in first-order expectations. To illustrate the economic consequences, I modify two workhorse models of asymmetric information. I show in a Morris-Shin(1998) model that more disclosure of public information can increase the chance of successful currency attacks. I show in a dynamic Grossman-Stiglitz(1980) model that hectic trading around public announcements is consistent with common priors and Bayesian learning.