Saddlepath learning, MSV learning and consistency of subjective expectations

Abstract

Ellison and Pearlman (2011) show that determinacy implies e-stability under both full and lagged information if (1) subjective expectations are consistent with a structural model and unbiased, and (2) a learning process is given by the saddlepath relationship. This study clarifies that their equivalence result under any information set is a consequence of their assumptions about bounded rationality. The same result also holds for the standard MSV learning approach under consistency or unbiasedness assumption. Without these assumptions, the equivalence result no longer holds for either saddlepath or MSV learning.