In an environment with dispersed information, how much can agents learn from past endogenous aggregate outcomes such as prices or output? We show that, in a rational expectations equilibrium, two possible regimes can arise endogenously: a perfect revealing regime and a confounding regime. The...