A Comment on "Friendship and female education: Evidence from a field experiment in Bangladeshi primary schools" by Hahn et al

Hahn et al. (2020), published in the Economic Journal, evaluate a treatment randomized across 115 Bangladeshi primary schools that assigned students to study for one week either in groups of friends or with randomly picked peers. They find that the friends treatment improves math performance, but only among low-ability girls. We identify four sets of unresolved discrepancies: inconsistent accounts of implementation responsibilities, irregularities in the sample and group structure, misalignments between reported procedures and observed data, and implausible missing-data and attendance patterns. Taken together, these issues go beyond reporting imperfections and reveal multiple discrepancies between the published record, earlier versions of the study, the replication data, and the authors’ two subsequent replies. These unresolved discrepancies preclude independent verification of material aspects of the study record and raise serious questions about the reliability of the reported findings.