Political Science 236 Randomization inference

Most of this course will be devoted to the study of treatment effects in the absence of random assignment of subjects to treatments. As we will see, performing causal inference in the absence of random treatment assignment requires that we make fairly strong assumptions. In contrast, when treatment is assigned randomly, treatment effects can be estimated with very mild assumptions and, very importantly, the hypothesis of no treatment effect can be tested without assumptions of any kind. In this section, we will study the basics of randomization inference so that in the remaining of the course we can think of observational studies as departures from this benchmark. Let the term “experimental unit” refer to the opportunity to apply or withhold the treatment. In general, experimental units will be persons who will either receive or not receive the treatment. But units could also be families, classrooms, or, as we have seen in lecture, cups of tea. The randomization model has several distinctive features: