Abstract
We rigorously test for ethnic discrimination in high school grading in Sweden. A random sample of the national tests in the Swedish language is graded both non-blind by the student’s own teacher and blind without any identifying information. The increase in the test score due to non-blind grading is significantly higher for students with Swedish background compared to students with foreign background. This discrimination effect is sizeable, and explains the entire difference in test scores between students with Swedish and foreign background.