Identifying Central Individuals without Social Network Information

By targeting central individuals in a social network, marketers and policymakers can design seeding interventions that promote information diffusion, product adoption, or behavior change. Social network analysis allows the identification of central individuals, however complete social network information is often unavailable or costly to collect. Two strategies are commonly used in the absence of complete network information: one-hop sampling and social sensing. Yet, these two strategies have not been compared on the same networks, and previous work offers conflicting predictions on which identifies more central individuals (higher-degree seeds). We provide a direct comparison of the two strategies, using original data from a cocoa-farming village in Côte d'Ivoire (502 adults) and public data from 33 villages in India (6,464 households) and 11 classrooms in Spain (487 students). We report two findings. First, social sensing identifies seeds with 9 to 58 percent higher degree than one-hop sampling across all three settings. Second, conditional on degree, social sensing nominations select on demographic dimensions while one-hop selections are fully accounted for by network position. Social sensing thus buys higher centrality at the price of selecting on demographic dimensions beyond network position.