Behavioral nudges prevent student dropouts in the pandemic

School closures in the context of COVID-19 have been shown to magnify that problem, with at least seven million additional dropouts worldwide in 2020. Despite efforts from governments around the world to mitigate learning gaps by the time in-person classes return, interventions to motivate students to remain in school until then have been overlooked. In particular, behavioral nudges sent to parents' cell phones through text messages had shown promise in preventing student dropouts in developing countries before the pandemic. Having said that, such nudges typically work by leading parents to show up in school to a greater extent and monitor teachers more closely - a mechanism that might not be meaningful in the absence of in-person classes. Methods: We conducted a cluster-randomized control trial with 18,256 high-school students in the State of Goiás, Brazil, randomizing 2/3 of them to receive behavioral nudges through text messages, between June and December 2020. The control group did not receive any messages. Within the treatment group, we additionally randomized students to variations in nudges' content, to study whether behavioral insights linked to framing and social pressure would lead to higher impacts. We estimate the impacts of nudges on dropout risk over the course of the school year, using administrative data from the State Secretariat of Education on whether students took math and Portuguese exams each school quarter. We also estimate heterogeneous impacts of nudges by risk levels (higher for boys, sophomore and junior students, and those below-median first-quarter Portuguese GPA), by whether messages were sent to parents' phones or directly to students, and by whether schools already offered online academic activities prior to the pandemic. Findings: Nudges decreased dropout risk by around 26% over the course of the school year. Effects increased with exposure, and were concentrated in students at the highest risk of dropouts. Nudges only worked when sent directly to students' phones, and in schools that already offered online academic activities prior to the pandemic. Framing content in terms of the upside of graduating high school led to higher impacts than framing it in terms of the downside of dropping out. Alluding to peer motivation to return to in-person classes to leverage social pressure had no additional effects on dropout risk. Interpretation: Results show that behavioral nudges can partially mitigate the dramatic increase in student dropouts during school closures by keeping adolescents motivated to stay in school. The patterns of heterogeneous treatment effects are consistent with complementarities between motivation and academic instruction. All in all, our results showcase that insights from the science of adolescent psychology can be leveraged to shift developmental trajectories at a critical juncture, but also raise caution against indiscriminately applying behavioral insights derived from evaluations of similar interventions in contexts of in-person classes or static decision-making.