Interpreting Null Effects in Education: Insights from a Tutoring Experiment

We learn not only from when interventions work—but also from why they do not. Policymakers embraced high-impact tutoring to address pandemic learning loss, but scaling promising practices is difficult, and not all implementations prove effective. Using random assignment, we find null (math and a low-stakes reading test) and negative (−0.09 SD on state reading scores) impacts from assignment to a virtual tutoring program. Using frameworks for interpreting null results in education experiments, we identify subject-specific heterogeneity and contribute to the literature on dosage challenges for in-school tutoring. We also identify alignment of tutoring content with classroom instruction as a moderator warranting further study. This paper offers strategies for uncovering mechanisms behind unexpected results and designing theory-driven, policy-relevant evaluations.