Targeting in social networks with anonymized information

Abstract

This paper studies optimal targeting when the planner knows the architecture of the network but not the identities of agents occupying different positions in the network. We show that the planner's ability to discriminate among agents depends on the balance between in- and out-neighborhoods in the social network. When influence is reciprocal, the knowledge of the network architecture is sufficient for the planner to implement the first-best actions. When in- and out-neighborhoods are imbalanced, pairs of players have an incentive to jointly misreport their identities. This situation arises when one agent influences all other agents, or when one agent is being influenced by all other agents. It also arises in hierarchical structures with nested neighborhoods where agents at lower tiers of the network are influenced by the same agents as agents at upper tiers who influence them.