Finding and Correcting for Missing RCTs: A Meta-Analysis of Low-Cost Parent-Interventions Around the World

Abstract

We conduct a meta-analysis of 82 randomized controlled trials across more than 20 countries to estimate the effects of interventions that provide instruction or information to parents about their child's education. These interventions are highly scalable in that they are low cost and delivered remotely via text messages, phone calls, and apps. We build on Andrews and Kasy's 2019 model of publication bias to incorporate sample sizes from studies that were attempted but never written, which we collect from study registries and grant reports. This allows us to identify the share of studies "in the file drawer" as well as to characterize the full distribution of effects for written and unwritten studies. Remote parenting interventions produce small, but almost uniformly positive, effects on academic outcomes that survive publication bias corrections. Bias-corrected mean effects are 0.05 SD for test scores, 0.08 SD for grades, 0.05 SD for attendance, and 0.01 SD for enrollment. Roughly 80% of studies with statistically insignificant results are written up, but the share is lower in domains with fewer available studies. We also characterize how much additional, unobserved file-drawer bias would negate our findings. We conclude by using the model's Bayesian framework to inform where and how an additional RCTs would be most helpful for decision making.