Networks Webinar - Yiqing Xing

Abstract:

We adopt a screening perspective to study network interventions when each agent privately observes their network information. Under this framework, an intervention policy consists of a menu of allocation-transfer bundles for the agents to choose from based on their out- and in-degrees (susceptibility and influence). We apply this framework to two problems: a monopoly’s problem to maximize the expected profits, and a benevolent planner’s problem to maximize the expected social welfare (subject to a budget constraint). In both problems, the optimal intervention policy features different treatments of out- and in-degrees: it directly screens susceptibility and indirectly infers influence from susceptibility (via a friendship paradox effect). Moreover, the social planner’s problem can be fully characterized by a combination of two benchmarks: the first-best setup without information asymmetry, and the monopoly’s problem. Finally, while linear interventions can achieve a decent outcome in some networks (e.g., with uniform degree distributions), the gain from screening is particularly large in networks with high-degree dispersions (e.g., with power-law degree distributions).