Impact evaluation based on randomized controlled trials (RCTs) offers a powerful tool that has fundamentally reshaped development economics by offering novel solutions to long-standing problems of weak causal identification. Nonetheless, RCTs suffer important and underappreciated pitfalls, some of which are intrinsic to the method when applied to economic problems, others that are the result of methodological boosterism. Among the pitfalls are ethical dilemmas, uncontrollable treatments that result in a 'faux exogeneity,' distortion of the research agenda, and a tendency to estimate interventions' abstract efficacy rather than their effectiveness in practice. We illustrate these points through the literature on smallholder capital access and productivity growth. Ultimately, we argue for a methodological pluralism that recognizes all identification strategies' limitations. Copyright 2010, Oxford University Press.