Abstract

This paper uses a randomized experiment to compare to methods of implementing an early primary literacy program in Northern Uganda. Our first treatment group receives the program as implemented by the organization that designed it; the second treatment group received a reduced-cost version of the program that was designed to simulate it might be scaled up. The full version of the program has extremely large impacts on student learning: it improves student recognition of letter names by 1.0 SD- one of the largest impacts ever measured in a randomized trial of an educa-tion program. The reduced-cost version improves this “headline ” outcome measure by 0.4 SD, making it slightly more cost-effective than the full version. However, it has statistically-insignificant effects on overall literacy and large negative effects on higher-level writing. This suggests that cost-effectiveness in improving “headline ” outcome measures can come at the cost of lower performance in other areas.