Grade repetition (GR), the practice to retain an underperforming student and do not allow him to progress to the next school grade, has been often called into question for its potentially unfair selection mechanisms. The present paper intends to contribute to the literature investigating how schooling institutions’ react to student’s failure and address the existence of discrimination in teachers’ decisions. By exploiting a large dataset which collect administrative microdata for an entire cohort of Northern-Italian students, we measure the effect of social background and migration on the probability to get retained in 9^th grade. Controlling for a broad set of ability measured at the end of lower secondary school, we perform a three-level mixed-effect logistic regression with random intercepts at the class and school level. The dependent variable is the occurrence of grade repetition, and covariates of main interest are the socio-economic index and migrant background. We include controls also for tracks and class-level means of each first-level variable. We find that students from disadvantaged family background are exposed to much larger probabilities of being retained. After taking under scrutiny alternative concurrent explanations, we discuss how discrimination could emerge in GR’s decisions.