A Class of Unbiased Estimators of the Average Treatment Effect in Randomized Experiments

  title={A Class of Unbiased Estimators of the Average Treatment Effect in Randomized Experiments},
  author={P. Aronow and Joel A. Middleton},
Abstract We derive a class of design-based estimators for the average treatment effect that are unbiased whenever the treatment assignment process is known. We generalize these estimators to include unbiased covariate adjustment using any model for outcomes that the analyst chooses. We then provide expressions and conservative estimators for the variance of the proposed estimators. 

Tables from this paper

Design-Based Ratio Estimators for Clustered, Blocked RCTs
This article develops design-based ratio estimators for clustered, blocked RCTs using the building blocks of experiments. We consider finite population regression estimators, allowing for weights andExpand
A Unified Theory of Regression Adjustment for Design-based Inference
Under the Neyman causal model, it is well-known that OLS with treatment-by-covariate interactions cannot harm asymptotic precision of estimated treatment effects in completely randomized experiments.Expand
Regression‐assisted inference for the average treatment effect in paired experiments
Summary In paired randomized experiments, individuals in a given matched pair may differ on prognostically important covariates despite the best efforts of practitioners. We examine the use ofExpand
Improving Average Treatment Effect Estimates in Small-Scale Randomized Controlled Trials
Researchers often include covariates when they analyze the results of randomized controlled trials (RCTs), valuing the increased precision of the estimates over the potential of inducing small-sampleExpand
Sampling-based randomized designs for causal inference under the potential outcomes framework
It is found that pre-treatment measurements are often unhelpful in improving the precision of average treatment effect estimators under this randomize-first-sample-second design, unless a large number of pre- treatment measurements that are highly associative with the post-treated measurements can be obtained. Expand
Design-Based Ratio Estimators and Central Limit Theorems for Clustered, Blocked RCTs
This article develops design-based ratio estimators for clustered, blocked RCTs using the building blocks of experiments, with an application to a federally funded, school-based RCT testing theExpand
Regression Adjustments for Estimating the Global Treatment Effect in Experiments with Interference
This paper proposes regression adjustment estimators for removing bias due to interference in Bernoulli randomized experiments, and proposes an estimator that allows for flexible machine learning estimators to be used for fitting a nonlinear interference functional form. Expand
Estimating Average Causal Effects Under Interference Between Units
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 definesExpand
Optimal Estimation when Researcher and Social Preferences are Misaligned
Econometric analysis typically focuses on the statistical properties of fixed estimators and ignores researcher choices. In this article, I approach the analysis of experimental data as aExpand
On mitigating the analytical limitations of finely stratified experiments
While attractive from a theoretical perspective, finely stratified experiments such as paired designs suffer from certain analytical limitations not present in block-randomized experiments withExpand


On equivalencies between design-based and regression-based variance estimators for randomized experiments
Abstract This paper demonstrates that the randomization-based “Neyman” and constant-effects estimators for the variance of estimated average treatment effects are equivalent to a variant of the WhiteExpand
Unbiased Estimation of the Average Treatment Effect in Cluster-Randomized Experiments
Many estimators of the average treatment effect, including the difference-in-means, may be biased when clusters of units are allocated to treatment. This bias remains even when the number of unitsExpand
Bias of the regression estimator for experiments using clustered random assignment
This paper shows that regression may be biased for cluster randomized experiments. For one application bias tends to zero when the number of clusters is large but for another, regression is notExpand
Simple, Efficient Estimators of Treatment Effects in Randomized Trials Using Generalized Linear Models to Leverage Baseline Variables
It is proved the maximum likelihood estimate of the coefficient corresponding to the treatment variable is an asymptotically unbiased estimator of the marginal log rate ratio, even when the working model is arbitrarily misspecified. Expand
Bounds on least squares estimates of causal effects in the presence of heterogeneous assignment probabilities
In many contexts, treatment assignment probabilities differ across strata or are correlated with some observable third variables. Regression with covariate adjustment is often used to account forExpand
On regression adjustments in experiments with several treatments
Regression adjustments are often made to experimental data. Since randomization does not justify the models, bias is likely; nor are the usual variance calculations to be trusted. Here, we evaluateExpand
A Class of Robust Sampling Designs for Large-Scale Surveys
Abstract A combined estimation and selection scheme is proposed for use in the context of large-scale enterprise and establishment surveys. The estimator is asymptotically design unbiased and subjectExpand
Is regression adjustment supported by the Neyman model for causal inference
This paper examines both theoretically and empirically whether the common practice of using OLS multivariate regression models to estimate average treatment effects (ATEs) under experimental designsExpand
Survey Design under the Regression Superpopulation Model
Abstract The construction of sample designs and estimators under a linear regression superpopulation model is considered. The anticipated variance, the variance of the predictor computed with respectExpand
SUMMARY THE most general type of linear estimate is defined for a general sampling design. It is demonstrated that an unbiased linear estimate with least variance does not exist uniquely for theExpand