The Local Approach to Causal Inference under Network Interference

TL;DR

This paper introduces a nonparametric framework for causal inference in networked settings, leveraging local agent configurations to estimate treatment effects amidst network interference.

Contribution

It develops a local, configuration-based approach for causal inference under network interference, including finite-sample bounds and a valid hypothesis test.

Findings

Finite-sample bounds on mean-squared error for the estimator

Asymptotically valid test for policy irrelevance

Demonstration on networked causal inference scenarios

Abstract

We propose a new nonparametric modeling framework for causal inference when outcomes depend on how agents are linked in a social or economic network. Such network interference describes a large literature on treatment spillovers, social interactions, social learning, information diffusion, disease and financial contagion, social capital formation, and more. Our approach works by first characterizing how an agent is linked in the network using the configuration of other agents and connections nearby as measured by path distance. The impact of a policy or treatment assignment is then learned by pooling outcome data across similarly configured agents. We demonstrate the approach by deriving finite-sample bounds on the mean-squared error of a k-nearest-neighbor estimator for the average treatment response as well as proposing an asymptotically valid test for the hypothesis of policy irrelevance.