Comment on "Telemonitoring and Homeschooling during School Closures: A Randomized Experiment in Rural Bangladesh" by Hassan et al.
Hassan et al. (2024) describe a randomized controlled trial conducted in Bangladesh during COVID-19. This comment identifies data irregularities from the original paper’s replication package and notes its connection to other published papers. Critically, Hassan et al. (2024) and Wang et al. (2024) have 132 individuals matched on CHILD_ID, VILLAGE_ID, and FAMILY_ID in both studies, suggesting contamination of the educational interventions of both studies. Although the endline tests were identical and occurred within weeks of each other, some individual scores and item responses vary considerably, while others have identical answers across all questions. Importantly, the sample that is both control in Hassan et al. (2024) and treated in Wang et al. (2024) appear to systematically score higher in the data of Wang et al. (2024). A statistical comparison of endline data across studies also reveals concerning discrepancies where anomalies in response patterns and total questions answered correctly are consistent with potential data integrity issues. Furthermore, we find that being assigned to treatment in Guo et al. (2024) has large, statistically significant effects in the opposite direction when using educational outcomes in Hassan et al. (2024). We further raise concerns about the connections to other studies, timeline inconsistencies and inexplicable changes of respondent demographics and reported grades.