Speaker: Carl Heese, Assistant Professor of Economics at the University of Hong Kong

Time: 10:30-12:00 a.m., November 16, 2023, GMT+8

Venue: 302, School of Economics, PKU

Abstract: 

I analyze a model of Bayesian Persuasion with a continuum of receivers and with binary actions. The sender’s payoff depends on the distribution of the share s of the receivers choosing his preferred action. The sender can correlate the receivers’ signals in any way.

I characterize all obedient signal structures in terms of a single majorization constraint, and this yields a geometric approach to calculating the equilibrium distribution of s.

When the sender’s payoff is (Schur)-convex, it is optimal to maximally correlate the receiver’s signals given the constraint. When the sender’s payoff is (Schur-)concave, independent signals are optimal. I provide constructions of an optimal signal in other cases, e.g., when the sender’s payoff is cut-off in s. Finally, I consider applications to voting, sales, and team production, and provide a series of results showing that persuasion becomes harder if a sender faces more disperse senders (in terms of prior beliefs, abilities, etc.).

Source: National School of Development, PKU