Persuasion on Networks
We analyze persuasion in a model in which each receiver can buy a direct access to the sender's signal or rely on her network connections to get it. For the sender, a higher bias increases the impact per direct receiver, yet diminishes the willingness of agents to receive information. Contrary to naive intuition, the optimal propaganda might target peripheral, rather than centrally-located agents, and is at its maximum levels when the probability that information flows between agents is close to zero or one, but not in-between. The impact of network density depends on this probability as well.
The authors are grateful to Renee Bowen, Ozan Candogan, Martin Cripps, Piotr Dworczak, Wioletta Dziuda, Emir Kamenica, Konrad Mierendorff, Roger Myerson, Elliot Lipnowski, Nikita Roketskiy, and seminar participants at UCL, Berkeley, UCSD, HKBU/NTU/CEIBS/NUS for their helpful comments. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research.