Estimation Methods for Cluster Randomized Trials with Noncompliance: A Study of A Biometric Smartcard Payment System in India
share
Many policy evaluations occur in settings with randomized assignment at the cluster level and treatment noncompliance at the unit level. For example, villagers or towns might be assigned to treatment and control, but residents may choose to not comply with their assigned treatment status. For example, in the state of Andhra Pradesh, the state government sought to evaluate the use of biometric smartcards to deliver payments from antipoverty programs. Smartcard payments were randomized at the village level, but residents could choose to comply or not. In some villages, more than 90 treatment, while in other locations fewer than 15 When noncompliance is present, investigators may choose to focus attention on either intention to treat effects or the causal effect among the units that comply. When analysts focus on effects among compliers, the instrumental variables framework can be used to evaluate identify causal effects. We first review extant methods for instrumental variable estimators in clustered designs which depend on assumptions that are often unrealistic in applied settings. In response, we develop a method that allows for possible treatment effect heterogeneity that is correlated with cluster size and uses finite sample variance estimator. We evaluate these methods using a series of simulations and apply them to data from an evaluation of welfare transfers via smartcard payments in India.
READ FULL TEXT
