“…To provide evidence that our use of across-cohort, within school variation is valid and uncontaminated by other unobservables, we conduct a series of balancing tests (following Bifulco et al 2011, Lavy and Schlosser 2011, Billings et al 2012) that estimate the associations between the cohort measures and individual-level exogenous attributes, such as age, health status, nativity status, etc. In Table 3, we regress cohort composition over maternal education, race and ethnicity on ten exogenous attributes of students, omitting the student themselves from 11 Although the focus on same-grade nominations may appear constricting, we note that over 80% of all nominations we capture in the data are for individuals in the same grade.…”