FOUNDATIONS OF DESIGN-BASED INFERENCE UNDER INTERFERENCE
PETER ARONOW – YALE UNIVERSITY
ABSTRACT
I present a design-based framework for the analysis of the effects of randomized interventions on (potentially connected) finite populations. I consider inference on causal parameters that are defined with respect to the probability distribution induced by the intervention. I discuss conditions under which some of these causal parameters can be interpreted as the causal effects of an exposure. I consider the finite- and large-n behavior of Horvitz-Thompson-type estimators for these causal parameters. Applications to probability-based surveys and social network experiments are discussed.