The organization and quality of education are increasingly in the spotlight, both in academic and policy circles across the world. Specifically, some recent policy initiatives rely on measures of teachers’ quality calculated under the assumption that a given teacher has the same impact on any student. We will focus instead on the potential importance of "teacher-student interactions" relying on a novel research design to credibly ascertain how changes in the composition of students affect teacher behavior: Do teachers teach differently across students? Do they direct effort towards more able or more disruptive students? Does this vary with the class they are assigned, e.g. a high vs. a low achieving class? Does it vary with the school they teach at, e.g. one serving high or low socioeconomic status students? Thus, this project will tackle new research that has not yet gained sufficient scientific attention but is potentially of first–order importance for educational policy aiming at providing children with equal educational opportunities and in fostering economic growth. Exploring such questions entails developing novel survey instruments focused on teacher-student relationships and using large administrative data. We will focus on Romania which provides a unique setting to tackle methodological and data-related challenges. In addition, the project has the potential to provide evidence on the long term consequences of educational quality on student outcomes.