Beyond the Average Treatment Effect: Causal Inference on Networks and Across Thousands of Strata

Associate Professor of Political Science and Statistics, Jacob Bowers, from the University of Illinois, shows how a testing-based approach to causal inference can be used to complement estimation-based approaches in complex but common situations. This can include when an experiment occurs in thousands of sites, and policy makers' interest may lie in detecting effects in specific sites rather than in estimating an average effect within each site.