We examine individuals' abilities to identify the highly central people in their social networks, where centrality is defined by diffusion centrality (Banerjee et al., 2013), which characterizes a node's influence in spreading information. We first show that diffusion centrality nests standard centrality measures -- degree, eigenvector and Katz-Bonacich centrality -- as extreme special cases. Next, we show that boundedly rational individuals can, simply by tracking sources of gossip, identify who is central in their social network in the specific sense of having high diffusion centrality. Finally, we examine whether the model's predictions are consistent with data in which we ask people in each of 35 villages whom would be the most effective point from which to initiate a diffusion process. We find that individuals accurately nominate central individuals in the diffusion centrality sense. Additionally, the nominated individuals are more central in the network than "village leaders" as well as those who are most central in a GPS sense. This suggests that individuals can rank others according to their centrality in the networks even without knowing the network, and that eliciting network centrality of others simply by asking individuals may be an inexpensive research and policy tool.
Gossip: Identifying Central Individuals in a Social Network
Abhijit Banerjee, Arun G. Chandrasekhar, Esther Duflo, Matthew O. Jackson