Abstract: Schools shape skills that matter beyond academic achievement. Teachers are central to this process, yet whether these effects arise from fixed teacher attributes or from malleable instructional practices remains unclear. We investigate this question through a randomized intervention in Spanish secondary schools. The program equips treated teachers with tools to observe students' socio-emotional skills and provide structured feedback during routine instruction, thus generating exogenous variation in instructional practices. We combine external observations of student behavior with independent measures of teaching practices, survey measures of socio-emotional skills, and administrative records of grades and absences. The intervention improves students' socio-emotional behaviors in an externally assessed group activity and increases academic achievement, with particularly pronounced academic gains among students with lower baseline performance. These improvements translate into a substantial reduction in the probability of failing at least one subject. In contrast, we find no average improvement in socio-emotional behaviors observed during regular classroom instruction, although there is suggestive evidence of larger behavioral gains among students with weaker baseline socio-emotional skills and those from disadvantaged backgrounds. We also find evidence of changes in classroom interactions and instructional practices, which may contribute to the observed academic improvements. Together, these findings indicate that malleable teaching practices can improve academic outcomes and the expression of socio-emotional competencies, while highlighting that behavioral responses depend on the context in which they are measured, as well as students' initial characteristics.
