The Benefits of Condescension in Social Learning

Department d'Economie et Sciences de la Décision  
Intervenant : Itai Arieli
Salle : T-009

Abstract :

We consider the canonical social learning model with misspecified private signals. We assume that each agent can identify his own private signal but he suffers from misspecification with respect to the signals of other agents in the population. We show that within a class of tail-regular information structures, asymptotic learning holds if and only if fast learning holds. Moreover, we provide a necessary and sufficient condition over the parametrization of the tail-regular signals so that asymptotic learning holds. We show that our result also holds in the case where each agent only observes his last predecessor.

(Joint with Y. Babichenko S. Muller, F. Pourbabaee, and O. Tamuz)