Learning to constrain: Political competition and randomized controlled trials in development

This paper provides a political economic analysis of impact evaluation experiments con- ducted in international development. We argue that in more politically competitive environments, where incumbents face a higher probability of losing power, govern- ments have stronger incentive to run Randomized Controlled Trial (RCT) experiments to constrain successors' margin of policy discretion. Moreover, the effect of compe- tition on the probability to host RCTs is stronger in more polarized societies since the incumbent's cost of losing power is higher. We first propose a formal model and then empirically examine its theoretical predictions using a unique data set on RCTs that we have compiled. Over a panel of Indian states and a cross-national panel, we nd that certain RCTs are more likely to occur in electorally competitive jurisdictions, and that the effect is amplified by political polarization. We demonstrate that politics matter for when, where, and with which partners RCTs in development happen.