Assessed by a teacher like me: race, gender and subjective evaluations

The underrepresentation of minority teachers and male teachers remains an issue in US elementary education, and there is evidence that racial interactions partly shape student performance. However there is little work on discrimination within the classroom. Do teachers give better grades to children of their own race, ethnicity or gender? A US nationally representative longitudinal dataset that includes both test scores and teacher assessments offers a unique opportunity to answer this question. I look at the effect of being assessed by a same race or same gender teacher conditionally on test scores, child effects and teacher effects. This strategy controls for three confounding effects: (i) children of different races and genders may react differently in the classroom and during examinations (ii) teachers may capture skills that are not captured by test scores and (iii) tough teachers may be matched with specific races or genders. Results indicate that teachers give higher assessments to children of their own race, but not significantly higher assessments to children of their own gender. Also, this effect comes from the differential assessments given to non-hispanic black and hispanic children. White teachers give significantly lower assessments to non-hispanic black children and to hispanic children. Results are robust to various checks on endogenous mobility, measurement error and reverse causality. Moreover children's behavior is not a significant determinant of same race or same gender matching. Finally relative grading does not explain the main results of this paper.