ABSTRACT
An assumption often made (implicitly or explicitly) is that the evidence regarding program effectiveness from a randomized trial can directly help inform the effects of a program in potential new locations. This assumption in some ways underlies the thinking behind evidence-based policy; that we can identify “what works” using rigorous designs, and then disseminate or scale up that program or policy to other locations. However, trials are not always directly informative about the effects in a new environment. This chapter provides an overview of threats to the ability to generalize effect estimates from a trial to a new population, with a focus on differences in impacts due to differences between the trial participants and the population of interest on characteristics that also contribute to treatment effect heterogeneity. It discusses design and analysis strategies that can help enhance our ability to take results from trials and inform likely effectiveness of programs and policies in new locations. A key lesson is the need to incorporate available data with subject matter expertise and strong theory, to identify where and when an intervention is likely to be most effective.
