Inclusive Networks

When do efficiency and fairness align in societies where individuals care about one another? We study resource allocation in networks in which agents value both their own welfare and the welfare of others through distance-decaying altruism. We introduce a family of networks, which we call inclusive, defined as networks that are both claw-free and paw-free, or equivalently, networks whose connected components are either sparse, a path or cycle, or sufficiently dense, complete or nearly complete. For these networks, two prominent welfare notions, efficiency and fairness, coincide regardless of the altruism profile. First, we show that inclusivity is equivalent to universal efficiency — under linear payoffs, every interior allocation is robustly Pareto efficient (i.e., Pareto efficient regardless of the altruism profile). For concave private utilities, a robustly Pareto efficient allocation exists if and only if the network is inclusive. We further introduce a decentralized notion of fairness: an allocation is charity-free if no agent wishes to alter it by making unilateral transfers. We show that an allocation that is both charity-free and Pareto efficient for all discount factors exists if and only if the network is inclusive.

Link to the paper: https://drive.google.com/file/d/1peesX6YZI4I0kgGG4DUjrcSy7ahGEf75/view

Arseniy Samsonov's homepage: https://sites.google.com/view/asamsonov