Michael Leung (University of Southern California): "Causal Inference Under Approximate Neighborhood Interference"
Abstract: This paper studies causal inference in randomized experiments with network interference. The most common approach to modeling interference assumes that treatments assigned to alters only affect the ego's response through a low-dimensional exposure mapping. We instead study models satisfying a substantially weaker approximate neighborhood interference assumption that captures the intuition that treatments assigned to units far from the ego should have a small (but potentially nonzero) impact on the ego's response. We show that this assumption is satisfied in well-known models of social interactions, in contrast to the exposure mapping approach. When the data consists of a single large network, we prove that standard inverse probability weighted estimators can consistently estimate treatment and spillover effects and are asymptotically normal. Finally, we propose a new variance estimator.
Cindy Pingry
(847) 467-7263
Email