Abstract
This paper tests whether teachers unequally evaluate students based on their classroom behavior, rather than their scholastic competence. Evidence is drawn from a unique dataset on students from Brazil, which allows us to contrast teacher- and blindly-assigned scores on achievement tests that are high-stakes and cover the same material. We find that teachers inflate test scores of better-behaved students, and deduct points from worse-behaved ones. We also find that teachers’ decision to promote students is influenced by how they behave in class. Our results (i) document a potentially inefficient way of assessing students’ knowledge, (ii) explain a large part of grading discrimination against boys, and (iii) reveal a causal effect of noncognitive skills on educational outcomes that is unrelated to student proficiency.