On the Inefficiency of Social Learning

Florian Brandl

(University of Bonn)

«On the Inefficiency of Social Learning»

Abstract: We study whether a social planner can restore efficient learning a finite expected number of incorrect actions in the canonical sequential social learning model. Agents act in order, each choosing one of two actions based on a private signal and the social information they observe. The planner controls the social information disclosed to each agent and offers transfers contingent on the agent’s action. We show that whenever learning is inefficient without intervention, no combination of disclosure and transfers restores efficient learning at finite cost: the planner must offer subsidies approaching the full value of a correct decision to infinitely many agents, so the expected total offered transfer is infinite. In particular, no disclosure policy alone restores efficiency. Offered transfers, however, differ from realized payments: since a transfer is paid only when the agent takes the subsidized action, the planner can achieve efficient learning while keeping the expected total realized payment arbitrarily small.

Face-to-face Seminar 15.2.71  – Room 15.1.39