Grade retention and academic outcomes: evidence from Brazil

Repetición escolar y resultados académicos: evidencia de Brasil

Abstrac

Despite the widespread use of grade retention in Latin America, its impact on academic outcomes remains debated. This article examines the effects of retention using a longitudinal student-level dataset (2012-2017) from Rio de Janeiro, Brazil’s second-largest municipal education system. Applying a difference-in-differences approach with student fixed effects, we find that grade retention significantly improves short-term academic performance, with consistent results across robustness checks. Our analysis suggests that this improvement is partly driven by the fear of retention and the additional support provided to students at risk of repeating the school year. However, the impact of retention extends beyond these factors, indicating its role as a direct intervention in enhancing learning outcomes. This study contributes to the retention literature by offering evidence from a developing country, highlighting the potential benefits and complexities of this widely debated policy. While retention incentivizes short-term academic gains, its long-term effects and associated costs warrant further consideration.

Keywords:
grade retention; grade repetition; primary education; academic outcomes

Resumo

Apesar do uso disseminado da repetência escolar na América Latina, seu impacto sobre os resultados acadêmicos segue sendo objeto de debate. Este artigo examina os efeitos da repetência utilizando um conjunto de dados longitudinais em nível de aluno (2012-2017) do município do Rio de Janeiro, a segunda maior rede municipal de ensino do Brasil. Aplicando um modelo de diferenças em diferenças com efeitos fixos para o aluno, encontramos que a repetência melhora significativamente o desempenho acadêmico de curto prazo, com resultados consistentes em diversas verificações de robustez. Nossa análise sugere que essa melhoria se deve, em parte, ao medo da repetência e ao suporte adicional oferecido aos alunos em risco de repetição. No entanto, o impacto da repetência vai além desses fatores, indicando seu papel como intervenção direta na melhoria da aprendizagem. Este estudo contribui para a literatura sobre repetência ao oferecer evidências em um contexto de país em desenvolvimento, destacando os benefícios potenciais e a complexidade dessa política amplamente debatida. Embora a repetência incentive ganhos acadêmicos de curto prazo, seus efeitos de longo prazo e custos associados merecem maior consideração.

Palavras-chave:
repetência escolar; reprovação; ensino fundamental; desempenho acadêmico

Resumen

A pesar del uso generalizado de la repetición de grado en América Latina, su impacto en los resultados académicos sigue siendo debatido. Este artículo analiza los efectos de la repetición utilizando un conjunto de datos longitudinales a nivel de estudiante (2012-2017) del municipio de Río de Janeiro, el segundo sistema educativo municipal más grande de Brasil. Aplicando un enfoque de diferencias en diferencias con efectos fijos por estudiante, encontramos que la repetición de grado mejora significativamente el rendimiento académico a corto plazo, con resultados consistentes en diversas pruebas de robustez. Nuestro análisis sugiere que esta mejora se debe en parte al temor a repetir el año y al apoyo adicional brindado a los estudiantes en riesgo de repetir. Sin embargo, el impacto de la repetición va más allá de estos factores, indicando su papel como intervención directa en la mejora del aprendizaje. Este estudio aporta a la literatura sobre repetición al ofrecer evidencia en un contexto de país en desarrollo, destacando los posibles beneficios y las complejidades de esta política tan debatida. Aunque la repetición fomenta avances académicos a corto plazo, sus efectos a largo plazo y los costos asociados merecen ser considerados cuidadosamente.

Palabras clave:
repetición escolar; reprobación; educación primaria; rendimiento académico

1. INTRODUCTION

Grade retention is widely practiced in Latin America, including Brazil. In 2018, 6% of 15-year-olds in Organisation for Economic Co-operation and Development (OECD) countries reported repeating a grade in primary school, compared to 13.9% in Latin America (OECD, 2020). Despite its additional costs and debated impacts, retention remains popular among teachers.

This paper examines the effects of retention using a longitudinal dataset (2012-2017) from Rio de Janeiro, Brazil’s second-largest municipal education system. We estimate the impact of retention in the 3^rd and 4^th grades on 4^th and 5^th-grade performance using a difference-in-differences (DID) approach with student fixed effects. Our findings reveal positive impacts of retention on academic achievement, with effect sizes of 0.40 and 0.43 standard deviations, consistent across robustness checks.

These results contribute to the retention literature. Early studies (1980s-1990s) largely found negative effects (Holmes, 1989; Jimerson, 2001), but more recent work highlights methodological flaws, particularly selection bias (Allen et al., 2009; Lorence, 2006). Emerging evidence suggests positive short-term retention effects, especially in early grades (Figlio & Ozek, 2020; Jacob & Lefgren, 2004; Nunes et al., 2018; Schwerdt & West, 2013; Schwerdt et al., 2017), though these benefits typically fade within three years. Additionally, disentangling retention effects from remedial interventions remains a challenge.

Conversely, some studies find null or negative effects (Chen et al., 2010; Roderick & Nagaoka, 2005; Valbuena et al., 2021). Retention may harm psychological well-being, non-cognitive skills (Peixoto et al., 2016), and long-term outcomes, such as high school completion (Jacob & Lefgren, 2009), dropout rates (Eren et al., 2017; Manacorda, 2012), and adult criminal activity (Eren et al., 2018).

The literature offers no consensus, making Brazil a critical case. Brazilian studies largely focus on automatic grade promotion (Machado & Vasconcelos, 2020), with few papers using longitudinal data to analyze retention (Correa et al., 2014; Gomes & Hanushek, 1994; Luz, 2008; Riani et al., 2012). Most report negative effects but rely on designs comparing students based solely on observable characteristics.

This paper is structured as follows: Section 2 outlines Rio de Janeiro’s retention policy, Section 3 discusses the dataset, Section 4 details the empirical strategy, Section 5 presents results, and Section 6 provides robustness checks. Section 7 explores drivers, followed by conclusions in Section 8.

2. INSTITUTIONAL CONTEXT

Rio de Janeiro operates Brazil’s second-largest municipal education system, serving approximately 400,000 students in 2019. Starting in the 3^rd grade, students are evaluated bi-monthly by the Municipal Education Department (MED) through standardized assessments. While the MED develops and oversees the exams, teachers administer them in classrooms and grade the writing sections. These exams, taken by all students in the same grade, determine proficiency levels (Resolução SME nº 1123, de 24 de janeiro de 2011).

Students are categorized into four proficiency levels: Inadequate, Regular, Good, and Very Good. Teachers assess performance bi-monthly, considering factors like quizzes, development, and behavior. Students rated Inadequate in the final two months must attend remedial classes and pass an exam. Those scoring below five on this exam are retained.

While legislation governing Rio’s education system does not mandate specific interventions for retained students, it emphasizes prevention. Schools must offer extra activities during the school year for students with an Inadequate rating or low scores. Additionally, these students must attend extra classes during the July recess and undergo reevaluation in August (Resolução SME nº 1123, de 24 de janeiro de 2011).

3. DATA

The dataset comprises a longitudinal panel of primary education students in Rio de Janeiro (2012-2017). It includes bimonthly academic scores from MED evaluations but lacks information on proficiency levels and attendance. We estimate the impact of 3^rd grade retention on 4^th grade performance and 4^th grade retention on 5^th grade performance. Retention is identified when students remain in the same grade for two consecutive years. The analysis includes 30,825 students for 3^rd grade and 29,013 for 4^th grade.

The treatment group comprises students who repeated the 3^rd or 4^th grade in 2015, divided into three subgroups: (a) those who resumed a regular path afterward, (b) those who repeated the next grade in 2017, and (c) those who repeated two consecutive years. The control group consists of students following a regular progression.

Our complete database initially contained 892,146 students. First, we excluded students with potentially incorrect information, such as different grades in the same year. This exclusion resulted in a loss of 1,893 students, less than 0.2% of the dataset. Additionally, we excluded students who were absent in the dataset for the entire period, as we needed to track their progress throughout their school years. By doing so, we lost 727,691 students, which accounts for 81% of the dataset.

Different possibilities could explain this data loss. First, we might be losing students due to administrative inaccuracies. Each student has a unique code in the system, which should not change as they progress along the school years or if they change schools. However, we know from direct communication with the MED that such mistakes happen. Second, students might be leaving the municipal education system for three reasons: i) families could be migrating to another municipality, ii) parents could be transferring their children to private schools, and iii) students could be evading the education system altogether. These losses can be problematic for us if any of these three reasons are related to retention. However, migration and transfers to private schools, presumably, are more associated with socioeconomic fluctuations and changes in parents’ job market. Moreover, primary schooling is mandatory in Brazil, so the evasion rate at this stage is low.¹

In Box 1, we examine our sample compared to all students on the complete dataset in 2014 and 2015. While our sample presents a lower retention rate and higher academic scores, indicating that our sample does not represent all of Rio’s students (referring to the external validity), this is true for both the control and treatment groups, suggesting that our results are internally valid. Additionally, our regression estimations include school and grade, or student fixed effects, dealing with socioeconomic characteristics of the students that may differentially affect promoted or retained students’ performance.

BOX 1
COMPARISON BETWEEN OUR SAMPLE AND THE COMPLETE DATASET

Our dependent variable is the math, writing, and reading score of the evaluation administered bimonthly by the MED. Since the exam changes yearly, we cannot evaluate growth over the years. However, since the exam is the same in all schools, we can compare the results across treatment and control students by standardizing the scores, using the mean and standard deviation according to the month, grade, and year the exam took place.

For our objective, we only need the exam to be comparable across treatment and control groups and not over the years. Each class has retained and promoted students, all taking the same exam. Differences in the test application by teachers are distributed across treatment and control students and controlled by school and grade or student fixed effects in the regression estimations. For the writing exams, which are graded by the teachers, different teachers may grade differently. However, since our analysis is within schools and grades, we can still use the test to compare treatment and control students.

Figure 1 shows how the scores behave over the grades for retained and promoted students. As expected, retained students have lower scores than promoted ones. We can see that all retained students had an increase in their scores after repeating the 3^rd or 4^th grade. The DID method assumes that students’ scores would have followed the same trajectory over time in the absence of retention. Except for reading, the two groups of students have similar trends before the retained grade, so Figure 1 mostly supports this assumption.

FIGURE 1
SCORES OVER THE GRADES BY TREATMENT STATUS

4. METHODS

We use a same-grade comparison approach (Lorence, 2006) to evaluate the impact of 3^rd and 4^th grade retention, employing a DID framework. The data is balanced across years, and all specifications include subject and municipal administration fixed effects, accounting for the three municipal administrations during the study period. Year trends are omitted to maintain a same-grade comparison, avoiding cross-grade contrasts within the same year.

The estimated equations are the following:

$Y_{itdu}={\beta }_0+{\beta }_1{Tr}_i+{\beta }_2{D}_t+{\beta }_3\left({Tr}_i*D_t\right)+{\delta }_{itsu}+\eta +{\gamma }_{it}+{ϵ}_{itsu}$ (1)

In this equation, i represents the student, t denotes the grade, and su refers to the subjects (math, writing, and reading). Y _itsu corresponds to the standardized score for the 4^th or 5^th grade, while Tr _i indicates the treatment (retention in the 3^rd or 4^th grade), and D _t is a period dummy variable (=1 if post). The interaction term Tr _i * D _t captures the Treatment#Post effect. represents subject fixed effects, accounts for municipal administration fixed effects, and includes school fixed effects. Finally, is the error term.

$Y_{itdu}={\beta }_0+{\beta }_1{Tr}_i+{\beta }_3\left({Tr}_i*D_t\right)+{\delta }_{itsu}+\eta +{\gamma }_{it}+{\omega }_t+{ϵ}_{itsu}$ (2)

$Y_{itdu}={\beta }_0+{\beta }_1{Tr}_i+{\beta }_2{D}_t+{\beta }_3\left({Tr}_i*D_t\right)+{\delta }_{itsu}+\eta +{\rho }_i+{ϵ}_{itsu}$ (3)

Equations (2) and (3) are similar to Equation (1), with representing grade fixed effects and denoting student fixed effects.

5. MAIN RESULTS

Box 2 presents the main findings: retaining the 3^rd or 4^th grade significantly improves academic performance in the short term. For 3^rd grade retention, the effect size with student fixed effects is 0.43, slightly higher than the 0.40 observed for 4^th grade retention. Results are consistent across fixed effects specifications. Henceforth, we focus on results using student fixed effects, our preferred model.

BOX 2
MAIN RESULTS - ESTIMATED EFFECTS OF GRADE RETENTION ON ACADEMIC ACHIEVEMENT

Box 3 examines heterogeneous effects within the treatment group. Students who maintained a regular school path or repeated the same grade twice experienced positive impacts. However, students who repeated consecutive grades showed negative effects, suggesting that consecutive retentions harm academic outcomes. For students with a regular school path post-retention, benefits persisted, though diminishing over time (Box 4).

BOX 3
HETEROGENEITY IN THE TREATMENT GROUP

BOX 4
RETENTION EFFECTS BY GRADE

6. ROBUSTNESS CHECKS

To address the endogeneity of grade repetition, we re-estimate the main results, excluding the repeated grade. Specifically, we omit the 3^rd grade from the 3^rd grade retention analysis and the 4^th grade from the 4^th grade retention analysis. Box 5 shows consistent results, confirming the robustness of our findings.

BOX 5
REMOVING THE REPEATED YEAR

Additional robustness checks (available upon request) reinforce the results. For instance, analyzing only math scores - given potential biases in reading and writing - yields similar findings. We also investigate the role of socioeconomic factors, including age, race, gender, and mother’s education. Retention is more common among boys, black and brown students, and children of less-educated mothers, consistent with previous studies (Oliveira & Soares, 2012; Valbuena et al., 2021).

7. UNDERSTANDING OUR RESULTS

What explains the improved performance of retained students? To explore this, we first compared the treatment group with a more comparable control group. This control group consisted of students who scored below 4 (on at least one or all subjects) in the 3^rd or 4^th grade in 2014 but were promoted to the next grade in 2015. These low-performing students were likely candidates for retention in 2015. Results in Box 6 (columns 1 and 3) show that retained students outperform promoted low-performing peers in subsequent academic performance.

We also compared two groups of promoted students: those scoring below four and those scoring between four and seven. Higher-performing students face an upper limit to score improvement (e.g., a student scoring eight cannot exceed ten) while lower-performing students have greater room for improvement. Columns 2 and 4 in Box 6 present these findings, revealing that promoted low-performing students outperform promoted higher-performing peers.

BOX 6
ESTIMATION WITH A MORE COMPARABLE CONTROL (COLUMNS 1 AND 3) AND A COMPARISON BETWEEN PROMOTED STUDENTS (COLUMNS 2 AND 4)

The results are positive and significant: Promoted lower-performing students outperform promoted higher-performing students academically. Box 6 highlights that lower-performing students, likely close to being retained, show higher performance. As noted in Section 2, these students often receive remedial classes to avoid retention, which may explain their improvement. Additionally, the threat of retention may drive them to perform better, a phenomenon noted in the literature (Carvalho & Firpo, 2014).

However, these factors do not fully explain the findings, as retained students outperform promoted lower-performing peers. Retention has an impact beyond its threat or remedial activities. Notably, the threat of retention is effective only because retention itself is a real policy intervention.

8. CONCLUDING REMARKS

This paper estimates the impact of 3^rd and 4^th grade retention on 4^th and 5^th grade performance using a DID with fixed effects approach, based on a panel of Rio de Janeiro primary students from 2012 to 2017. Our results show that repeating a grade significantly improves academic achievement, consistent across robustness checks.

We suggest that the fear of retention may partly drive this effect, though retention itself contributes more than just the threat. Our findings indicate that retention serves as a short-term incentive for improved academic performance. However, as other studies have shown that retention has been associated with negative outcomes in the long term and affects students’ psychological well-being, policies should consider leveraging other educational strategies - such as targeted remedial interventions and providing continuous feedback or additional support - before resorting to grade repetition.

Requiring a student to repeat a grade is not the only way to ensure they are prepared for more advanced material, and it might not be the most cost-effective strategy. Effective after-school or summer learning programs have been identified as promising interventions (McCombs et al., 2011; Terzian et al., 2009). Early identification of struggling students and tailored interventions, such as family engagement, could help reduce retention rates and their associated negative effects. Family engagement strategies can include supporting learning both at school and home, shared governance of school decisions, and coordination with community programs (Morris & Nóra, 2024). Retention policies must consider all its possible effects on students and its high costs for the school system, especially considering its prevalence among vulnerable students.

ACKNOWLEDGMENTS

Roberta Costa acknowledges financial support from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) through a doctoral scholarship (Grant No. 131075/2016-0), active from 2018 to 2021. The authors are also grateful to the Municipal Department of Education of Rio de Janeiro for providing the data and to Letícia Côrtes for the information shared. Special thanks go to Danielle Carusi, Fabio Waltenberg, Martin Carnoy, and Elaine Pazello, whose comments and contributions significantly enriched the work.

REFERENCES

• Allen, C. S., Chen, Q., Willson, V. L., & Hughes, J. N. (2009). Quality of research design moderates effects of grade retention on achievement: A meta-analytic, multilevel analysis. Educational Evaluation and Policy Analysis, 31(4), 480-499.
• Carvalho, S., & Firpo, S. (2014). O regime de ciclos de aprendizagem e a heterogeneidade de seus efeitos sobre a proficiência dos alunos. Economia Aplicada, 18(2), 199-214.
• Chen, X., Liu, C., Zhang, L., Shi, Y., & Rozelle, S. (2010). Does taking one step back get you two steps forward? Grade retention and school performance in poor areas in rural China. International Journal of Educational Development, 30(6), 544-559.
• Correa, E. V., Bonamino, A., & Soares, T. M. (2014). Evidências do efeito da repetência nos primeiros anos escolares. Estudos em Avaliação Educacional, 25(59), 242-269.
• Eren, O., Depew, B., & Barnes, S. (2017). Test-based promotion policies, dropping out, and juvenile crime. Journal of Public Economics, 153, 9-31.
• Eren, O., Lovenheim, M. F., & Mocan, N. H. (2018). The effect of grade retention on adult crime: Evidence from a test-based promotion policy (NBER Working Paper 25384). National Bureau of Economic Research.
• Figlio, D., & Ozek, U. (2020). An extra year to learn English? Early grade retention and the human capital development of English learners. Journal of Public Economics, 186, 104-184.
• Gomes, J. B. Neto , & Hanushek, E. A. (1994). Causes and consequences of grade repetition: Evidence from Brazil. Economic Development and Cultural Change, 43(1), 117-148.
• Holmes, C. T. (1989). Grade level retention effects: A meta-analysis of research studies. In L. A. Shepard, & M. L. Smith (Eds.), Flunking grades: Research and policies on retentions Education Policy Perspectives (pp. 16-33). The Farmer Press.
• Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira. (2020, outubro). Taxas de Transição https://www.gov.br/inep/pt-br/acesso-a-informacao/dados-abertos/indicadores-educacionais/taxas-de-transicao
  » https://www.gov.br/inep/pt-br/acesso-a-informacao/dados-abertos/indicadores-educacionais/taxas-de-transicao
• Jacob, B. A., & Lefgren, L. (2004). Remedial education and student achievement: A regression discontinuity analysis. Review of Economics and Statistics, 86(1), 226-244.
• Jacob, B. A., & Lefgren, L. (2009). The effect of grade retention on high school completion. American Economic Journal: Applied Economics, 1(3), 33-58.
• Jimerson, S. R. (2001). Meta-analysis of grade retention research: Implications for practice in the 21st century. School Psychology Review, 30(3), 420-437.
• Lorence, J. (2006). Retention and academic achievement research revisited from a United States perspective. International Education Journal, 7(5), 731-777.
• Luz, L. S. (2008). O impacto da repetência na proficiência escolar: uma análise longitudinal do desempenho de repetentes em 2002-2003 [Dissertação de Mestrado]. Universidade Federal de Minas Gerais.
• Machado, M. R. C., & Vasconcelos, A. M. N. (2020). Políticas de não-repetência na educação básica: uma revisão da literatura brasileira. Perspectivas em Diálogo: Revista de educação e sociedade, 7(15), 164-184.
• Manacorda, M. (2012). The cost of grade retention. The Review of Economics and Statistics, 94(2), 596606.
• McCombs, J. S., Augustine, C. H., & Schwartz, H. L. (2011).Making summer count: How summer programs can boost children’s learning Rand Corporation.
• Morris, E. M., & Nóra, L. (2024). Six Global Lessons on How Family, School, and Community Engagement Can Transform Education Brookings Institution.
• Nunes, L. C., Balcão Reis, A., & Seabra, C. (2018). Is retention beneficial to low-achieving students? Evidence from Portugal. Applied Economics, 50(40), 4306-4317.
• Organização para a Cooperação e Desenvolvimento Econômico. (2020). Pisa 2018 results (Volume V): Effective policies, successful schools OECD Publishing.
• Oliveira, L. F. B., & Soares, S. S. D. (2012). Determinantes da repetência escolar no Brasil: uma análise de painel dos censos escolares entre 2007 e 2010 (Texto para Discussão 1706). Instituto de Pesquisa Econômica Aplicada.
• Peixoto, F., Monteiro, V., Mata, L., Sanches, C., Pipa, J., & Almeida, L. S. (2016). “To be or not to be retained... that’s the question!” retention, self-esteem, self-concept, achievement goals, and grades. Frontiers in Psychology, 7, 1550.
• Resolução SME nº 1123, de 24 de janeiro de 2011 (2011). Estabelece diretrizes para a avaliação escolar na Rede Pública do Sistema Municipal de Ensino da Cidade do Rio de Janeiro e dá outras providências.
• Riani, J. L. R., Silva, V. C., & Soares, T. M. (2012). Repetir ou progredir? Uma análise da repetência nas escolas públicas de Minas Gerais. Educação e Pesquisa, 38(3), 623-636.
• Roderick, M., & Nagaoka, J. (2005). Retention under Chicago’s high-stakes testing program: Helpful, harmful, or harmless? Educational Evaluation and Policy Analysis, 27(4), 309-340.
• Schwerdt, G., & West, M. R. (2013, April). The effects of test-based retention on student outcomes over time: Regression discontinuity evidence from Florida (CESIFO WORKING PAPER NO. 4203). Center for Economic Studies & Ifo Institute.
• Schwerdt, G., West, M. R., & Winters, M. A. (2017). The effects of test-based retention on student outcomes over time: Regression discontinuity evidence from Florida. Journal of Public Economics, 152, 154-169.
• Terzian, M., Moore, K. A., & Hamilton, K. (2009, July). Effective and promising summer learning programs and approaches for economically-disadvantaged children and youth Wallace Foundation, Child Trends.
• Valbuena, J., Mediavilla, M., Choi, Á., & Gil, M. (2021). Effects of grade retention policies: A literature review of empirical studies applying causal inference. Journal of Economic Surveys, 35(2), 408-451.

• 1
  Between 2014 and 2015, the evasion rate was 2.1 for the 3rd grade and 2.0 for the 4th grade (Instituto Nacional de Estudos e Pesquisas Educacionais Anísio Teixeira [INEP], 2020).

• DATA AVAILABILITY
  The full dataset supporting the findings of this study is available upon request to the Municipal Department of Education of the city of Rio de Janeiro. The data are not publicly available for two reasons: (i) they contain sensitive information that could compromise the privacy of students in the municipal school system; and (ii) the agreement established with the Department at the time of data acquisition restricts sharing with third parties.
• 3
  [Original version]
• Reviewers:
  Marco Antonio Catussi Paschoalotto (Universidade de São Paulo, Ribeirão Preto / SP - Brazil) https://orcid.org/0000-0003-2276-8531
• Reviewers:
  Bruno Sendra de Assis (Universidade de São Paulo, São Paulo / SP - Brazil) https://orcid.org/0000-0001-9925-3080
• Reviewers:
  One reviewer did not authorize the disclosure of their identity.
• Peer review report:
  The peer review report is available at this link https://periodicos.fgv.br/rap/article/view/94495/88065

Edited by

• Editor-in-chief:
  Alketa Peci (Fundação Getulio Vargas, Rio de Janeiro / RJ - Brazil) https://orcid.org/0000-0002-0488-1744

• Associate editor:
  Gabriela Spanghero Lotta (Fundação Getulio Vargas, São Paulo / SP - Brazil) https://orcid.org/0000-0003-2801-1628

Data availability

The full dataset supporting the findings of this study is available upon request to the Municipal Department of Education of the city of Rio de Janeiro. The data are not publicly available for two reasons: (i) they contain sensitive information that could compromise the privacy of students in the municipal school system; and (ii) the agreement established with the Department at the time of data acquisition restricts sharing with third parties.

Publication Dates

• Publication in this collection
  05 Dec 2025
• Date of issue
  2025

History

• Received
  25 July 2024
• Accepted
  08 June 2025