Design-Based Estimation and Central Limit Theorems for Local Average Treatment Effects for RCTs

Abstract

There is a growing literature on design-based methods to estimate average treatment effects for randomized controlled trials (RCTs) using the underpinnings of experiments. In this article, we build on these methods to consider design-based regression estimators for the local average treatment effect (LATE) estimand for RCTs with treatment noncompliance. We prove new finite-population central limit theorems for a range of designs, including blocked and clustered RCTs, allowing for baseline covariates to improve precision. We discuss consistent variance estimators based on model residuals and conduct simulations that show the estimators yield confidence interval coverage near nominal levels. We demonstrate the methods using data from a private school voucher RCT in New York City USA.