In this paper we study the effect of family and school background on students' academic performance, who attended public schools in Medellín. Two approaches are implemented in this work. Firstly, we investigate the effect of covariates on different segments of scores' distribution by quantile regression. Secondly, we are interested in estimating the school effect on average academic performance by estimating a linear regression with a large dummy-variable set. This procedure fits a one-way fixed-effects model. We found that the effect of the individual characteristics on academic achievement is stronger for best students. Additionally, we detect that the school effect is weaker in public schools than the private ones.