Abstract:
|
We study the methodological issues related to the design and analysis of two-stage randomized experiments, which are useful when the treatment of one unit affects the outcome of another unit. We define various causal quantities of interest, propose consistent estimators, and derive their statistical properties. We also present the power analysis and compare these analytical results with those of standard randomized experiments. Our methodologies are illustrated with the randomized evaluation of the India's national health insurance program, which was conducted using a two-stage randomized design.
|