Randomization Inference on Policy Assignments

Randomization inference is quickly becoming a widely used statistical approach in the social, behavioral, and natural sciences. It is an extremely useful tool for the evaluation of public policies and programs, and thus a necessary step for the design of more effective policies that target human development and poverty mitigation. The existing literature proposes a randomization test that is quite attractive because it exactly controls the probability of false positives. One of its limitations is that it assumes perfect knowledge of the policy decision process, e.g., the likelihood that each individual is eligible to receive a policy benefit. This is not always true in practice, so this project seeks to relax this assumption and propose a new randomization test. We plan to illustrate the advantages of our test using simulations as well as real-world data from a social program that provides preschool, health, and other social services to poor children age three to five and their families.