The paper uses pupil level figures to analyze determinants of school repetition in Rio de Janeiro public schools. This is a worldwide phenomenon that presents significant variation among educational systems. We define school repetition or failure as an educational outcome that prevents the pupil to move forward to the next grade. This decision happens at the end of each school year.

In Brazil, school failure has been declining since the beginning of 1990s, however, the figures are still considered high when compared to other countries in the region – South America. In 2012, Brazil presented the fifth highest rate among all PISA participants. In the case of Rio de Janeiro, the rates have been increasing since 2007 (Crahay; Baye, 2013). It is possible to observe two transitions where pupils are more likely to fail – 3^rd to 4^th grade, eight year old pupils, and 5^th to 6^th grade, eleven-year-old pupils. More recent data from 2014, shows that around one quarter of all pupils enrolled in 3^rd grade in Rio public schools were retained at the end of the school year. For the 6^th grade, around 18% were retained and obligated to attend the same grade in the following year.

The subject of school repetition has been a concern for educational researchers in Brazil for at least 30 years. Nonetheless, very few studies have analyzed, with robust designs, the effects of repetition on pupils´ academic trajectory and, more importantly, the determinants that can help predict the phenomena (Riani; Silva; Soares, 2012). Which variables are associated with school failure? This study will consider pupil and school level variables to analyze the phenomenon using a longitudinal model.

There is one main argument used by teachers and educational researchers in order to justify the need to hold back pupils at the end of a school year. The rationale suggests that these pupils have not learned what was expected for each particular grade. In this way, it would be necessary, and even beneficial, for the pupils to have a second chance to learn and consolidate the subjects. There is an idea that allowing the pupil to progress without knowing the subjects can be even more harmful. 

However, more recent and robust studies analyzing the impact of school failure suggest that this practice is not associated with an increase pupils’ academic achievement. In fact, there is some evidence suggesting that repetition is associated with a decline in academic performance. Not only that, facing school failure increases the chances of a second repetition or school dropout (Hattie, 2009; Manacorda, 2006; 2012).

Many cities in Brazil, including Rio de Janeiro, present high proportion of pupils with multiple school failures. As an example, in 2014, considering all pupils enrolled from 1^st to 9^th grades – pupils from 6 to 14 years old – in Rio public schools, around half had experienced at least one school failure and around 20% two or more school failures. This particular scenario creates an additional problem. It would not be unlikely to observe pupils with very different ages studying in the same classroom. The age difference is described by teacher, head teacher, principals and even parents as a problem to “classroom climate”. To deal with this situation, many educational systems in Brazil have created “special classes” for pupils with some early grade retentions. These classes were intended to provide a better opportunity for those pupils. The concern is that clustering disadvantaged pupils in specific classrooms or schools could create even worst educational opportunities as observed before in the regular classes (Bartholo, 2014). 

Bartholo, T. L. School Segregation in Rio de Janeiro Public Schools: Causes and Consequences. PhD Thesis: Federal University of Rio de Janeiro, 2014. Crahay, M. & Baye, A. (2013) Existem escolas justas e eficazes? Esboço de resposta baseado no PISA 2009. Cadernos de Pesquisa, v.43 n.150 p.858-883. Diggle. P. J. The analysis of Longitudinal Data. Oxford Statistical Science series. Oxford University Press, 2001. Goldstein, H. Multilevel Statistical Models. Kendall’s library of statistics; 3.Arnold London, 2003. Hattie, J. Visible Learning for Teachers. Routledge: New York, 2009. Manacorda, M. Grade Failure, Drop out and Subsequent School Outcomes: Quasi –Experimental Evidence form Uruguayan Administrative Data. In: Centre for the Economics of Education Seminar, 13 Dec 2006, London, UK. ftp://www.cemfi.es/pdf/papers/wshop/Manacorda.pdf ________. The cost of Grade Retention. The Review of Economics and Statistics, v. 94, n.2, May, 2012, p. 596-606. Riani, J. L. R.; Silva, V. C. & Soares, T. M. Reapeating or advancing? An analysis of school failure in public schools of Minas Gerais. Educação e Pesquisa, São Paulo, v. 38, n. 03, p. 623-636, jul-set., 2012. http://www.scielo.br/pdf/ep/v38n3/en_06.pdf Singer, J. D. Willet, J. B. Applied Longitudinal Data Analyysis: Modeling Change and Ocurrence. Oxford University Press, New York, 2003.