Abstract
<jats:p>This study evaluates a nationwide malaria prevention campaign delivered through social media in India using a district-level cluster-randomized controlled trial. We combine two survey samples(N = 8,253) with administrative health records to assess effects on offline malaria-related behaviors and health outcomes. In the survey data, average treatment effects on protective behaviors, care-seeking intentions, and self-reported malaria incidence were small and insignificant. Administrative data similarly showed no detectable effect on overall recorded incidence. Post-hoc heterogeneity analysis, however, suggests that detectable effects were concentrated among households with higher socioeconomic status (SES) and in urban areas: for higher-SES households, bednet use increased during the campaign and self-reported malaria incidence declined afterward. In administrative records, monthly malaria incidence in urban areas fell by 6.2 cases per million people, a 33% decline relative to the pre-treatment rate. By contrast, estimates for lower-SES households and rural areas, which face higher malaria risk, were small and insignificant. In a second individual-level feed experiment (N = 1,542) that made ad delivery and exposure more comparable across SES groups, the ads increased bednet use and timely treatment-seeking intentions, with no detectable differences by SES. These patterns are consistent with delivery frictions contributing to the campaign's limited average impact, highlighting the importance of better targeting and measurement in social media public-health campaigns.</jats:p>