Challenges in impact evaluation of development interventions: opportunities and limitations for randomized experiments

In recent years debates on as well as funding of impact evaluations of development interventions have flourished. Unfortunately, controversy regarding the promotion and application of randomized experiments (RE) has led to a sense of polarization in the development policy and evaluation community. As some proponents claim epistemological supremacy of REs (with respect to attribution) the counter reaction among others has been rejection. Needless to say, such extreme positions are counterproductive to reaching a goal that is commonly endorsed: to learn more about what works and why in development. This paper discusses the prospects and limitations of REs from the perspective of three categories of challenges in impact evaluation: delimitation and scope, attribution versus explanation, and implementation challenges. The implicit lesson is twofold. First of all, the question ‘to randomize or not to randomize’ is overrated in the current debate. Limitations in scope, applicability as well as implementation will necessarily restrict the use of REs in development impact evaluation. There is a risk that the current popularity of REs in certain research and policy circles might lead to a backlash as too high expectations of REs may quicken its demise. More importantly, given the nature and scope of the challenges discussed in the paper, more energy should be devoted to developing and testing ‘rigorous’ mixed method approaches within a framework of theory-driven evaluation.