Estimating Average Causal Effects Under Interference Between Units

By P. M. Aronow, Cyrus Samii. 2013

251 citations recorded by the source corpus.

This paper presents a randomization-based framework for estimating causal effects under interference between units. The framework integrates three components: (i) an experimental design that defines the probability distribution of treatment assignments, (ii) a mapping that relates experimental treatment assignments to exposures received by units in the experiment, and (iii) estimands that make use of the experiment to answer questions of substantive interest. Using this framework, we develop the case of estimating average unit-level causal effects from a randomized experiment with interference of arbitrary but known form. The resulting estimators are based on inverse probability weighting. We provide randomization-based variance estimators that account for the complex clustering that can occur when interference is present. We also establish consistency and asymptotic normality under local dependence assumptions. We discuss refinements including covariate-adjusted effect estimators and ratio estimation. We illustrate and assess empirical performance with a naturalistic simulation using network data from American high schools.

Subfields: Education, Statistical and Nonlinear Physics, Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Education

In Statistical and Nonlinear Physics

In Statistics and Probability

Citations in the corpus

The Econometrics of Randomized Experiments (2016, background): “Aronow and Samii (2013) study estimation in this general network setting.”
Random graph asymptotics for treatment effect estimation under network interference (2020, methodology): “We seek to estimate the direct, indirect and total effects of the treatment on the outcome, τ̄DIR(π) = 1 n ∑ i Eπ [Yi(wi = 1; W−i)− Yi(wi = 0; W−i)|Y (·)] , τ̄IND(π) = 1 n ∑ i ∑ j 6=i Eπ [Yj(wi = 1; W−i)− Yj(wi = 0; W−i)|Y (·)] , τ̄TOT(π) = d dπ { 1 n ∑ i Eπ [Yi|Y (·)] } , (1) where the expectations”
Panel experiments and dynamic causal effects: A finite population perspective (2020, background): “To derive the asymptotic distribution of causal estimators in such settings, researchers typically require the assignment mechanism to be independent (Chin (2018)) or at least have only limited dependence structure across units (Aronow and Samii (2017)).”