Bayesian Impact Evaluation with Informative Priors : An Application to a ColombianManagement and Export Improvement Program

Policymakers often test expensive new programs on relatively small samples. Formally incorporatinginformative Bayesian priors into impact evaluation offersthe promise to learn more from these experiments. A Colombian government program which aimed to increaseexporting was trialed experimentally on 200 firms with this goal in mind. Priors were elicited from academics,policymakers, and firms. Contrary to these priors, frequentist estimation can not reject 0 effects in 2019, andfinds some negative impacts in 2020. For binary outcomes like whether firms export, frequentist estimates arerelatively precise, and Bayesian credible posterior intervals update to overlap almost completely with standardconfidence intervals. For outcomes like increasing export variety, where the priors align with the data, the value ofthese priors is seen in posterior intervals that are considerably narrower than frequentist confidence intervals.Finally, for noisy outcomes like export value, posterior intervals show almost no updating from the priors,highlighting how uninformative the data are about such outcomes.