The objective of this work is to evaluate if retaining pupils of the third year of fundamental education in 2009 brought benefits to their learning in comparison to pupils that advanced to the fourth year. We seek thereby to contribute to the debate about the impact of school failure in pupils' learning. In Brazil, due to the absence of longitudinal data that would allow a more accurate comparison between those pupils who failed and those that advanced, considering their performance before and after the school failure, the majority of studies focus primarily on the impact of the policies of cycles on the performance of the pupil. Using data from the Literacy Assessment Program (and external assessment conducted every year in public schools of the State of Minas Gerais), it was possible to set up a longitudinal database to identify pupils that had repeated and those who had not. To try to explain the performance of students who had failed and those who had not, we tried to use as a guiding line the approach based on the Educational Production Function, having as a methodological basis the hierarchical models. The results indicate that both categories of students - those who had failed and those who had not - displayed very important growth in proficiency average rates. However, the hierarchical models constructed demonstrated that, between two pupils with the same proficiency in 2008, one of them having failed and the other not, the pupil that did not fail tended to present the better proficiency level in 2009. Another important finding refers to the fact that the result of the pupil is very much influenced by the general result of the school.
School failure; Performance; Hierarchical models
