Race classification during childhood and adolescence and its implications for education statistics

Researchers and policy analysts routinely rely on race to characterize social inequalities in official statistics and survey data. An active debate asks whether race classification itself might respond to those very inequalities – as subjects and interviewers internalize social hierarchies in place. This paper provides first-hand evidence on the features of racial classification during childhood and adolescence, using novel data from a representative survey of Brazilian K-12 schools that uniquely collects student race reported both by schools and by students themselves. Here we show that self- and school-reported race disagree for 36.5% of students in the country (Study 1). Strikingly, this is not because adults are more ‘objective’ (e.g., because of racial literacy or socialization), as school-reported race fluctuates with quasi-random changes in school staff composition induced by close local elections (Study 2). Rather, relative to self-reports, schools ‘whiten’ students who read proficiently while ‘darkening’ those falling behind or those affected by sensory and motor disabilities (Study 3). Our results have implications for major issues in the global educational debate – from measuring disproportional representation in special education to evaluating race-based school interventions.