ABSTRACT
One of the challenges to evaluating a program at scale is that as a program is scaled up and reaches more participants, there are increasing interactions between individuals who receive the program and those who do not, a phenomenon known as spillovers. In this chapter, we discuss how spillovers can make effective programs appear ineffective at scale, and ineffective programs appear effective at scale. We review a field experiment in early childhood education to show how failing to account for spillovers can significantly bias estimates of the effect of the intervention. We discuss evaluation design in the presence of spillovers, offering lessons from the literature on best practices in evaluation design and the tradeoffs of alternative approaches.
