Teachers’ Beliefs About the Consequences of Grade Retention: Scale Validation and Differences Across Individual and School-Level Factors

Abstract

This study validates and adapts the Grade Retention Survey for Chilean fourth-grade school teachers, focusing on their beliefs about the consequences of grade retention. A questionnaire of 17 items was administered to 4297 fourth-grade elementary school teachers across Chile. We analyze the internal consistency, reliability, and unidimensionality of the Beliefs on Consequences of Retention (BCR) scale. The validated version comprises six items and demonstrates excellent psychometric properties. Our results show that teachers’ beliefs about grade retention vary significantly according to school retention rates, administrative dependence, socioeconomic status, and gender—findings similar to those observed in the empirical literature, reaffirming the validity of our scale. Specifically, we found that teachers in public schools or those serving lower socioeconomic communities tend to have more negative views of grade retention. In contrast, female teachers and those in schools with higher retention rates exhibit more positive perceptions. This scale provides valuable insights into teachers’ beliefs about grade retention, as these beliefs may shape the implementation of educational policies aimed at modifying the use of these mechanism.

1. Introduction

Grade retention, the practice of requiring a student to repeat a grade, is a controversial topic in educational policy and practice. Commonly implemented to force those who do not comply with an educational system’s desired academic benchmarks to reach the desired competencies, it remains a practice whose effectiveness and consequences continue to be debated. Likewise, its application largely depends on the perceptions about grade retention of those who decide how to manage heterogeneity within educational systems and schools.

Within schools, student heterogeneity comes from different abilities, motivations and personal objectives, which can be related to different sociocultural backgrounds. In light of this, education systems worldwide use different practices to manage this heterogeneity within schools and classrooms (Dupriez et al., 2008; Hermann & Kopasz, 2021).

The practices most often used to manage heterogeneity worldwide are tracking, ability grouping, personalized teaching practices and grade retention (Hermann & Kopasz, 2021; Mons, 2007). Tracking and grouping allow students to be classified by ability, motivation or other variables, creating more homogeneous groups. In theory, this improves the chances for low- and high-performance students to receive appropriate teaching for their learning level (Deunk et al., 2018). Individualized teaching practices imply teachers’ dedication of more time and attention to students with lower achievement levels (Demanet & Van Houtte, 2019; Dupriez et al., 2008). In contrast, Grade retention artificially decreases the degree of heterogeneity within classes by holding back students who do not meet a minimum required standard of achievement or maturity (Hermann & Kopasz, 2021).

It is necessary to highlight that the current literature has identified earlier retention, low achievement scores, the number of years being retained, behavioral problems, the degree of parental involvement, low attendance, being male, and socio-economic situation as some of the main factors that increase the probability of repeating a grade (Huang, 2014; Jimerson, 2001; Jimerson et al., 1997; Meriño-Montero, 2021; Valbuena et al., 2021). Similarly, the research on grade retention has supported a variety of perspectives regarding the effects associated with this mechanism, ranging from claims of short-term academic improvement to long-term detriments to emotional and motivational well-being and reduced likelihood of university enrolment. (Díaz et al., 2021; Goos et al., 2021; Valbuena et al., 2021; Van Canegem et al., 2024; Vandecandelaere et al., 2016), as well as no effects or differences from their promoted peers (Pipa et al., 2024). Likewise, academic retention would significantly impact students’ lifelong outcomes and future societies, considering that it is directly associated with a lower probability of entering university (Santos et al., 2023a; Van Canegem et al., 2024).

Considering the wide variety of empirical findings, the importance of understanding grade retention practice, its impact on students, and the factors and perceptions that shape its implementation cannot be understated.

As educational systems globally struggle to balance policy and practice that maintain standards while supporting student growth, educator perspectives on grade retention play a significant role in shaping policy and practice. For this reason, it is important to highlight that empirical research has found that teachers’ beliefs can significantly influence the application of national educational policies and new regulations, shaping how they interpret and implement reforms (Cipriano & Martins, 2021; Fives & Buehl, 2016).

The literature makes it clear that educators’ beliefs about grade retention not only reflect personal experiences and biases, but also respond to the broader educational context (Santos & Monteiro, 2024; Tomchin & Impara, 1992; Walton, 2018). Teachers’ beliefs are particularly important, as they directly influence retention decisions and interact with systemic factors that may determine or increase the probability of grade retention (Feathers, 2020; Kerr, 2007; Walton, 2018; Young et al., 2019). Exploring these beliefs in greater depth can provide valuable insights into the mechanisms that shape grade retention practices within an educational system.

Building on this foundation, our study’s main objective is to adapt and validate a scale to assess 4th-grade teachers’ perceptions of the impact of grade retention on educational trajectories, focusing on measuring their beliefs about the consequences of grade retention. The scale was developed based on items selected from the Grade Retention Survey originally developed by Manley (1988). Additionally, we aim to examine how teachers’ beliefs about grade retention, as measured by the new scale, are influenced by individual and school-level variables (e.g., gender, administrative dependence, socioeconomic status, among others). By doing so, we seek to provide insights into how these beliefs shape and are shaped by the Chilean educational and institutional context, particularly to implement inclusive policies.

Challenging Traditional Perspectives: How a New Public Policy Seeks to Shift Beliefs on Grade Retention in Chile

In Chile, renewed attention to grade retention practices has emerged with the recent implementation of Decree 67 (MINEDUC, 2018). This Decree has brought a heightened examination of existing practices and the potential effects of this policy shift. In 2018, with the calling of the bill that “Approves minimum national standards on evaluation, qualification and promotion” (Decree 67, MINEDUC, 2018), a new conceptual framework for grade retention was established, which began to take effect from 2020 onward and, for the first time, eliminated automatic grade retention. For nearly a century, Chile’s education system maintained a largely unchanged set of rules on retention and promotion, even as student evaluation standards evolved over time (Allende et al., 2024b).

The new Decree 67 does not explicitly prohibit grade retention but specifies that it should be employed only as an exceptional measure by principals and/or director teams. Unlike previous decrees, which mandated automatic criteria for grade retention, the new decree requires a more comprehensive approach. It emphasizes that all relevant information about the students must be considered during the decision-making process, with a primary focus on their well-being. The decree stipulates that the final decision should account for the student’s learning progress and the extent of their learning gaps, as well as the perspectives of teachers, parents, and students.

Schools must develop assessment, grading, and promotion regulations detailing how they will handle situations where students do not achieve the required knowledge or development, focusing on their well-being and comprehensive development (MINEDUC, 2019). This shift highlights a broader commitment by the government to fostering inclusive and supportive learning environments that address both academic and socio-emotional needs (Valenzuela & Allende, 2023).

However, this new regulation contrasts with the empirical evidence about teacher beliefs in Chilean schools, where most teachers blame school failure on the students themselves and their parents (Román, 2013; Treviño et al., 2016; Meyer, 2022). Only a very low percentage of teachers relate failure to how knowledge is delivered by them and the school, or with the quality of teaching in the classroom (Román, 2013). Additionally, there is a prevailing belief or ideology of “demandingness” (García-Huidobro, 2000), in which a “good teacher” is perceived as one who fails a larger number of students, while a “bad teacher” is seen as someone who promotes everyone.

This educational system has historically promoted and positively valued the administration of exclusionary practices (López et al., 2024), such as grade retention. At the same time, many families do not view learning or comprehensive education as the primary goal of schooling. Instead, they value student promotion regardless of whether actual learning is taking place (García-Huidobro, 2000).

The new Decree is introduced in an educational system where grade retention has been made almost invisible by the authorities by failing to provide easy-to-access data about the magnitude of this practice, the factors associated with the likelihood of repeating a year, and the associated effects.

In terms of magnitude, average grade retention rates in primary education have remained below 5% since 2002, reaching a historic low of under 2% in 2022. In secondary education, retention rates have generally stayed below 10% since 2002, with the exception of 2006 and 2011, when massive demonstrations caused a temporary spike. These rates then continuously declined, reaching approximately 3.5% by 2022. However, a broader perspective emerges when examining the cumulative percentage of students who have repeated a grade during their educational trajectory. By 2022, approximately 6% of students in the first cycle of primary education (1st–4th grade) and 12% in the second cycle (5th–8th grade) had repeated at least one grade, compared to 18% across Latin America (UNESCO, 2021). In secondary education, 21% of students in 2022 had repeated a grade, lower than the 23.2% reported for Chile in the PISA test but significantly higher than the OECD average of 11% (OECD, 2020).

All of this shows us that schools, as well as teachers, could be predisposed to continue with this practice, simply out of habit. As García-Huidobro (2000) mentions, “The functioning of the school cannot be understood without the adhesion of its members and users. Since there is a deep-rooted cultural conviction about the advantages of grade retention, it would not be effective to abolish retention by decree. Before that, it is necessary to wage an important cultural fight to change the prevailing perceptions among us” (p. 6).

2. Literature Review

2.1. Theoretical Approach to Grade Retention and Related Perceptions

Theoretically, grade retention could operate under the logic of negative reinforcement (Galán & Ursúa, 2016; Skinner, 2007) to transform the event of repeating a grade into a stimulus capable of modifying the behavior of all students, encouraging them to increase their effort and, therefore, their academic performance. This is particularly evident among students with lower academic performances, disruptive behaviors, less cultural capital, and fewer social conditions, who tend to have a higher probability of being retained (Goos et al., 2021). Ultimately, the intention is to increase student effort and improve school performance, e.g., by establishing homogeneous class groups. This perspective places the responsibility for grade retention primarily on students, removing responsibility from schools, teachers, and families, reinforcing a positive perception of retention. Overlooking potential negative consequences, such as school disengagement or rejection of school (Valbuena et al., 2021; Kearney, 2008), worsening behavioral issues, and impacts on self-esteem and self-concept (Goos et al., 2021; Van Canegem et al., 2021), all in the pursuit of improving the academic performance of those who struggle the most.

This argument in support of grade retention is particularly accepted in Chile, where this mechanism is not seen as a cost for students but rather as a stimulus to increase their effort and performance (García-Huidobro, 2000; Koppensteiner, 2014; Torres et al., 2015). Behind this argument lies the belief, internalized within Chilean society, that individuals can improve their situation on their own and that achievement is solely under the control of the individual (Peña & Toledo, 2017). This perception is related to Chile’s highly valued meritocratic discourse (PNUD, 2017), often associated with effort and motivation (Fercovic, 2022; PNUD, 2017). Meritocracy represents a social order where merit functions as the distributive criterion (Atria et al., 2019).

In this context, personal effort is an important part of Chilean morality (PNUD, 2017), mainly because it is often seen as a catalyst for demands for greater justice and equality (PNUD, 2017). The meritocratic discourse raises expectations for students and leads them to believe that their personal effort is enough to succeed and that personal effort is the sole explanation for educational success and school failure (Peña & Toledo, 2017). According to this logic, a lack of effort causes students to repeat, and retained students must “try harder” to move on to the next grade.

Another common argument supporting grade retention is to give students a second chance to master the knowledge they did not learn in a given year (OECD, 2020) to help them recover the necessary knowledge and skills to avoid future academic failure (Vandecandelaere, 2015). This argument is based on the belief that a student unable to handle the specific curriculum of their grade will not be able to succeed in subsequent grades (Piaget & Inhelder, 1973). This logic overlooks other reasons that may lead to lower grades and thus to repeating a grade, such as illnesses and other adverse childhood experiences, including material hardship, family and domestic dysfunction, and exposure to violence, among others (Hinojosa et al., 2019). Consequently, the perceptions of the impact of grade retention formed by adhering to this type of argument will be largely positive regarding the outcomes it can achieve. However, they will need to consider potential adverse effects, such as peer rejection and increased behavioral problems (Feathers, 2020).

An additional theoretical argument presented in the literature holds that grade retention could help teachers adjust their teaching methods according to each student’s needs, allowing those with stronger abilities to receive instruction suited to their capacity without affecting (or leaving behind) students with weaker skills. This makes teachers’ work easier by giving them more homogeneous classes (Deunk et al., 2018; Reschly & Christenson, 2013; Steenbergen-Hu et al., 2016). Perceptions based on these arguments would favor grade retention to manage heterogeneity within schools.

Conversely, critics of this practice argue that it is inherently exclusionary, aiming to create homogeneous spaces by disregarding differences in academic performance (Hermann & Kopasz, 2021). They highlight its potential negative effects on psychological aspects (e.g., reduced self-esteem, low self-concept), economic outcomes (e.g., lower income), and health (e.g., stress, depression, anxiety) in the short, medium, and long term (Goos et al., 2021). Additionally, it is unlikely to be an effective public policy, as its negative consequences may outweigh any potential benefits (Valbuena et al., 2021). These perspectives also tend to consider the multitude of personal, family, institutional, economic, and social factors that are intertwined within the educational system, ultimately determining a student’s situation.

As a result, perspectives aligned with these criticisms recognize the many factors that can contribute to social stratification, which practices like grade retention may reinforce. This can make it harder for students with less social and cultural capital to progress and succeed in their educational trajectories.

2.2. Perceived Consequences of Grade Retention

Teachers’ beliefs and decision-making styles serve as a filter, defining what factors they consider relevant to making a decision regarding grade retention (Santos & Monteiro, 2024). Likewise, teachers’ decision-making processes around grade retention predominantly rely on intuitive beliefs (Santos et al., 2023b; Santos & Monteiro, 2024; Vanlommel et al., 2017). This suggests that personal and professional history informs pedagogical decisions by maintaining their beliefs in the effectiveness of grade retention and guiding the factors considered during the decision-making (Santos & Monteiro, 2024). Nevertheless, ethical and emotional conflicts often arise when teachers attempt to balance students’ academic needs with their emotional well-being (Thomas, 2018). Meanwhile, teachers’ beliefs can significantly influence the application of national educational policies and new regulations, shaping how they interpret and implement reforms (Cipriano & Martins, 2021; Fives & Buehl, 2016).

Regarding the beliefs about grade retention, it has been shown that many teachers view retention as an effective practice (Goos & Londers, 2023; Haro, 2015; Walton, 2018), especially when they believe it will help students mature (Santos & Monteiro, 2024; Tomchin & Impara, 1992; Young et al., 2019). Mostly because of the belief that it can provide a necessary pause for learners who are not yet ready to progress academically (Walton, 2018). Beliefs that are rooted in the perception that grade retention can prevent future academic failure, maintain educational standards and motivate students to attend school (Range et al., 2012; Tomchin & Impara, 1992; Walton, 2018; Young et al., 2019). This notion of effectiveness was particularly noted for educators with more teaching experience or graduate degrees than those with only bachelor’s degrees (Feathers, 2020).

It is necessary to consider that, in many cases, personal experiences with grade retention may influence teachers’ beliefs and practices regarding student retention (Santos et al., 2023b). Teachers’ attitudes towards retention are affected by personal and professional experience, which, as already said, plays an essential role in their pedagogical decisions (Santos et al., 2023b). Teachers who have themselves experienced grade retention or have retained students before are more likely to believe in its necessity (Santos et al., 2023b).

Teachers might confront ethical and emotional conflicts when they consider retention in an attempt to satisfy academic needs without undermining emotional well-being (Thomas, 2018). Despite this, it has been observed that teachers tend to emphasize academic performance as the primary reason for retaining students (Range et al., 2012). However, even when teachers are inclined to retain students, they remain concerned about the potential long-term harm to children’s emotional and academic development (Feathers, 2020).

The literature also highlights differing attitudes among teachers and educational professionals, such as principals, school counsellors and psychologists. School principals often adopt a more cautious approach, reflecting greater concern for the potential negative outcomes of retention (Range et al., 2012). However, for many, their decision-making is influenced by the pressures of accountability (Colón, 2021), a result that could explain the one found by Feathers (2020), which showed that administrators significantly favor retention as an effective intervention strategy. Psychologists tend to focus on its negative effects on students’ self-esteem and motivation, based mostly on the effectiveness of grade retention literature, warning against the stigmatization and disengagement that can accompany being held back (Haro, 2015). School counsellors support retention more often, acknowledging that in specific cases, particularly in cases of immaturity and poor attendance (Haro, 2015), grade retention might be useful if carefully considered and implemented as a last resort (Kerr, 2007).

This study allows us to deepen our understanding of teachers’ perceptions of grade retention to partially understand how these beliefs could influence the implementation of the aforementioned public policy. By validating a scale that measures beliefs about grade retention and examining the relationship between individual and school variables, we can provide valuable insights into how to align retention practices with more inclusive and supportive educational reforms. To achieve this, it explores the perceptions of grade retention among 4th-grade teachers in Chilean schools following the implementation of Decree 67.

3. Material and Methods

3.1. Data Collection

We adapted and analyzed a questionnaire that gathered perceptions about grade retention from elementary school teachers. Our research team submitted the questionnaire applied to the “Call for Researchers to Propose Questions for Quality and Context of Education Surveys” for the 2023 Chilean school system’s standardized tests (Sistema de Medición de la Calidad de la Educación [SIMCE]). SIMCE is a national test that assesses academic achievement in reading and mathematics. It is applied every year to fourth-grade students and alternately to sixth-, eighth-, and tenth-grade students (Allende et al., 2024a), this includes complementary questionnaires for teachers, principals, families, and students.

The project titled “Perceptions of School Repetition in the Context of Implementing Decree 67” was selected for this purpose, allowing us to include 17 items in the complementary questionnaires for teachers of 4th grade applied in person together with the SIMCE national test during November 2023.

3.2. Participants

The Education Quality Agency administered the questionnaire to all Chilean head fourth-grade teachers as part of the complementary teachers’ survey conducted alongside the SIMCE 2023 national test. To access the public databases, a formal request must be submitted through the official website of SIMCE (See Data Availability Statement). The total number of surveyed teachers who completed the items on perceptions of grade retention was 4297 observations.

According to the State of Chile’s protocols for handling personal data, all datasets used had a blind personal identifier provided by the Education Quality Agency. The identifier allows blind access to the responses of all teachers, safeguarding the confidentiality of the information used. At the same time, this dataset allows cross-referencing such data with data at the school level existing in Chile.

3.3. Instrument

The survey instrument used was an adaptation of The Grade Retention Survey, originally developed by Manley (1988). This instrument was selected primarily because it has been widely used in numerous studies over time since its publication (e.g., Feathers, 2020; Kerr, 2007; Richardson, 2010) but, to the best of our knowledge, has never undergone a formal validation process.

This survey originally consisted of 35 items on teachers’ perceptions and beliefs about grade retention, measured using a six-option Likert scale ranging from “strongly disagree” to “strongly agree”, with a reliability of 0.72 measured by Cronbach’s Alpha (Manley, 1988). For the present study, we selected a series of items from the version of this questionnaire applied by Feathers (2020), which consisted of 27 items, where some questions were rephrased from the original version of the survey. We include 14 items from the original scale. The selected items were focused on the perceived consequences of grade retention and perceptions regarding who should bear the responsibility for applying this type of sanction.

In addition to the selected 14 items from the original questionnaire, 3 items relevant to the Chilean context were added. These questions specifically addressed whether teachers associate their teaching methods with the number of students who repeat a grade and their perceptions of who is responsible for academic failure. These specific questions are relevant to the Chilean reality, considering that the limited research available for Chile has shown that only a very low percentage of teachers relate failure to how knowledge is delivered by them and the school or with the quality of teaching in the classroom (Román, 2013; Meyer, 2022). Likewise, it is important given the meritocratic discourse in Chile, where they are led to believe that their personal effort is the sole explanation for educational success and school failure (Peña & Toledo, 2017). Additionally, teachers were asked whether the availability of complementary support could determine grade retention, as the new Decree 67 explicitly states that students repeating a grade must receive academic support.

The selected items were initially translated, rephrased, and reviewed by the research team. This translation was reviewed by three educational research experts, who validated the initial translation. Additionally, a third review was conducted by the staff of the Education Quality Agency, which administered the final questionnaire. Both revisions provided us with suggestions to improve understanding and application of the questionnaire. This process refined the wording and enhanced the phrasing of some questions when translated into Spanish, ensuring the correct application of the new questionnaire.

The original scale was adjusted to a four-option Likert scale (Strongly Disagree, Disagree, Agree, Strongly Agree). This avoids neutral responses and forces teachers to choose an option concerning the provided alternatives.

Table A3. Perceptions Questionnaire on School Repetition. Fourth-Grade Teachers (Spanish version).

¿Cuán de acuerdo está con las siguientes afirmaciones relacionadas con la repitencia escolar?
Marque con una equis (X) una sola alternativa para cada afirmación.
	Muy en desacuerdo	En desacuerdo	De acuerdo	Muy de acuerdo
Tiene efectos positivos en el aprendizaje y desempeño académico.
Puede llevar al desapego o rechazo hacia la escuela.
No es un costo para los alumnos, sino un estímulo para incrementar su esfuerzo y rendimiento.
Da tiempo a los(as) estudiantes para crecer y madurar.
Incrementa los problemas de comportamiento.
Es una segunda oportunidad para dominar los conocimientos y habilidades necesarias.
Los(as) estudiantes que repiten son rechazados por sus pares.
La forma de enseñar tiene directa relación con la cantidad de estudiantes que reprueban una asignatura.
Es mejor que un(a) estudiante repita ya que no tendrá el suficiente apoyo y/o acompañamiento el siguiente año.
Genera clases homogéneas en términos de habilidades, siendo más fácil enseñar.
La decisión de repetir debería ser tomada solo por el o la docente.
Un(a) estudiante debiera repetir si reprueba una asignatura importante.
Repetir suele ser culpa de los(as) estudiantes y sus familias.
Puede generar efectos negativos en el autoconcepto de los(as) estudiantes.
Estudiantes con muchas inasistencias deberían repetir automáticamente.
La opinión del estudiante debiera ser considerada para tomar la decisión de repetir de curso.
Nadie debiera repetir de curso.

in Appendix A.2 shows all the items included in Spanish, and

Table A4. Perceptions Questionnaire on School Repetition. Fourth-Grade Teachers (English version).

How Much Do You Agree with the Following Statements Related to Grade Repetition?
Mark with an X Only One Alternative for Each Statement.
	Strongly Disagree	Disagree	Agree	Strongly Agre
Has positive effects on learning and academic performance.
It can lead to disaffection or rejection of school.
It is not a cost to students but a stimulus to increase their effort and performance.
It gives students time to grow and mature.
It increases behavioral problems.
It is a second chance to master the necessary knowledge and skills.
Students who repeat are rejected by their peers.
How you teach is directly related to the number of students who fail a subject.
A student should repeat a subject as he/she will not have sufficient support and/or accompaniment the following year.
It generates homogeneous classes in terms of skills, making teaching easier.
The decision to repeat should be made by the teacher alone.
A student should repeat if he/she fails an important subject.
Repetition is often the fault of students and their families.
It can have negative effects on students’ self-concept
Students with many absences should automatically repeat.
The student’s opinion should be considered in deciding to repeat a grade.
No one should repeat a grade.

in Appendix A.2 provides a direct English translation of the items.

3.4. Data Analysis

First, a descriptive inspection of the items was carried out to determine their level of discrimination. This was done to confirm whether the item-scale relationship was above 0.3 for all considered items, the minimum threshold to ensure sufficient discrimination (Robinson et al., 1991, p. 13). The reliability of the scale was assessed by calculating Cronbach’s Alpha with standardized items and McDonald’s Omega, using the polychoric correlation matrix to account for the ordinal nature of the items (Flora, 2020; Hayes & Coutts, 2020). Values greater than 0.8 for both Cronbach’s Alpha and McDonald’s Omega were considered indicative of a robust measurement instrument (Intimayta-Escalante et al., 2025; DeVellis, 2017; Reise et al., 2013; Robinson et al., 1991, p. 13). Additionally, this inspection verified that all items had the same direction; if this condition was not met, the direction of the variable in question was reversed.

Subsequently, the methodology followed is the EFA within CFA framework (E/CFA) proposed by Jöreskog (Brown, 2015; Jöreskog, 1969; Sanders et al., 2015; Sullivan & Davila, 2022). Jöreskog’s E/CFA methodology begins with estimating an Exploratory Factor Analysis (EFA), which in this case, was conducted using the principal factor method. For this analysis, we consider only the 14 items that belong to the original scale developed by Manley (1988).

Then, since all the items considered were categorical, the Polychoric correlation matrix is used in the analysis to obtain robust estimates (Kolenikov & Angeles, 2009). The number of factors selected is determined through the parallel analysis method (Humphreys & Montanelli, 1975). Then, to define the type of rotation to use, the correlation matrix between the items is checked to see whether it showed high correlations (Tabachnick et al., 2019). The matrix displayed moderate and high correlations among many variables (see Appendix A.3,

Table A5. Polychoric Correlation matrix between considered items training set (N = 2148).

	p17_01	p17_02 *	p17_03	p17_04	p17_05 *	p17_06	p17_12	p17_14 *	p17_17 *
p17_01	1
p17_02 *	0.461	1
p17_03	0.690	0.400	1
p17_04	0.687	0.412	0.640	1
p17_05 *	0.372	0.545	0.279	0.422	1
p17_06	0.659	0.375	0.564	0.747	0.386	1
p17_12	0.262	0.101	0.234	0.216	0.006	0.189	1
p17_14 *	0.455	0.564	0.424	0.387	0.420	0.362	0.122	1
p17_17 *	0.450	0.300	0.357	0.439	0.342	0.414	0.231	0.332	1

* These questions were reversed from the original version.

), leading to the decision to use oblimin rotation (Hair et al., 2019).

Next, the variables associated with each factor were selected based on the obtained factor loadings, selecting those with loadings greater than 0.5. This follows the recommendations of Hair et al. (2019), ensuring that each factor explains at least 25% of the variance for each variable included in it, which, according to the authors, practically guarantees the significance of the factor loadings.

Finally, the structure found was tested using a Confirmatory Factor Analysis (CFA) to verify whether the fit indices obtained met the ranges considered acceptable according to the literature: CFI and TLI very close to 1; RMSEA less than 0.06; SRMR close to zero and never exceeding 0.08; likewise, with chi-square to degrees of freedom ratios generally less than 3 (Brown, 2015; F. Chen et al., 2008; Hair et al., 2019). The diagonal weighted least squares (DWLS) method was used to estimate the CFA parameters. This method produces robust standard errors and adjusts for the lack of normality in the Likert scale used.

To assess cross-validation, which aims to evaluate whether the model obtained is robust and generalizes well to new data, thus preventing overfitting (Brown, 2015) and additionally enabling us to refine the final model and confirm its validity. The model was initially trained on a subset consisting of 50% of the sample (training set; N = 2148) and subsequently tested (validated) on the remaining 50% (test set; N = 2149). The model’s structure was first estimated using the EFA on the training set, followed by the final CFA estimation on the test set. Once the final model was validated, it was re-estimated using the entire sample of fourth-grade teachers (model retraining). With more observations, the goal was to achieve greater robustness in the final estimations.

The factor scores were subsequently predicted from the CFA model for the entire sample of fourth-grade teachers. These scores represent teachers’ perceptions of the effects of grade retention on students, which from now on will be referred to as the Beliefs on Consequences of Retention scale (BCR scale), in order to make construct validation clearer, in terms of what the instrument is actually measuring. The final scale was standardized to have a minimum of 1 and a maximum of 4, aligning it similarly with the Likert scale used for all the items (1 “Strongly disagree”; 2 “Disagree”; 3 “Agree”; 4 “Strongly agree”). A higher score on the scale implies that teachers’ perceptions of grade retention are more positive. In other words, it will indicate whether teachers view grade retention mostly as an opportunity for improvement and growth over time. On the other hand, lower scores will indicate whether they perceive it as an experience that could negatively affect emotional well-being, school trajectories, or the connection with the school.

To test measurement invariance, we employed Multigroup Confirmatory Factor Analysis (Cheung & Rensvold, 2002; Hair et al., 2019). This involved testing for configural, metric, and scalar invariance, as “support for scalar invariance is required if any comparisons of relative construct level (e.g., mean scores) are made across groups. That is, scalar invariance allows the relative amounts of latent constructs to be compared between groups” (Hair et al., 2019, p. 739), which was the final comparison conducted in this study.

We evaluated changes in model fit using multiple indicators, including differences in CFI (ΔCFI), RMSEA (ΔRMSEA), SRMR (ΔSRMR), and TLI (ΔTLI). Measurement invariance was assessed based on the following cut-off criteria: ΔCFI ≤ 0.010, ΔRMSEA ≤ 0.015, and ΔTLI ≤ 0.010. For ΔSRMR, the thresholds were ≤0.03 for metric invariance and ≤0.01 for scalar invariance (Intimayta-Escalante et al., 2025; Khademi et al., 2023; Tian et al., 2022; F. F. Chen, 2007; Cheung & Rensvold, 2002).

Finally, after estimating the BCR scale, the predicted index was analyzed in relation to individual and school variables through ANOVA (Analysis of Variance) to determine potential associations between the constructed scale and other variables. The ANOVA method was used to assess whether there were statistically significant differences in the BCR scale across different groups based on individual and school variables.

Among the individual variables, the analysis included gender (a dichotomous variable coded as 1 for female), beliefs about the relationship between grade retention and teaching methods, the presence of support or companionship, and the perception that grade retention is the responsibility of parents and families. As previously described, these variables were measured on a four-option Likert scale (Strongly Disagree, Disagree, Agree, Strongly Agree).

The school-level characteristics considered included administrative dependence (i.e., public, private-subsidized, and private paid), the socioeconomic level scale constructed by the Education Quality Agency based on family characteristics (i.e., Low, Middle-Low, Middle, Middle-High, High), and the grade retention rate in the first cycle of primary education (grades 1 to 4), which was incorporated into the estimations using tertiles of the school grade retention rate.

The estimations were carried out using STATA 17 (StataCorp, 2021) and R version 4.2.1 (R Core Team, 2023). Employing lavaan (Rosseel, 2012) and semTools (Jorgensen et al., 2022) R packages for structural equation modelling.

4. Results

4.1. Descriptive Analysis

The dataset comprises 4297 observations, representing all head fourth-grade teachers from Chilean schools who completed the items on perceptions of grade retention, included in the teacher survey administered alongside the SIMCE test in 2023 (Figure A2 in Appendix A.6 shows the distribution of all the items measured). Of these respondents, approximately 88% are women, and about 12% are men. This distribution reflects the reality of primary schools in Chile, where, according to MINEDUC data (MINEDUC, 2023), 78% of teachers are women and 22% are men. It is important to note that the non-response rate for the items on perceptions of grade retention was similar for both genders: 9% for women and 7% for men.

Table 1. Training set descriptive statistics, item-total correlation, Cronbach’s Alpha (N = 2148).

Question	Description	Mean	SD	Min	Max	Item-Rest Correlation	Alpha de Cronbach
†† p17_01	Has positive effects on learning and academic performance.	2.87	0.8	1	4	0.66	0.78
†† p17_02 *	It can lead to disaffection or rejection of school.	2.55	0.8	1	4	0.50	0.79
†† p17_03	It is not a cost to students but a stimulus to increase their effort and performance.	2.64	0.8	1	4	0.58	0.78
†† p17_04	It gives students time to grow and mature.	3.19	0.7	1	4	0.64	0.78
†† p17_05 *	It increases behavioral problems.	2.96	0.8	1	4	0.41	0.80
†† p17_06	It is a second chance to master the necessary knowledge and skills.	3.27	0.7	1	4	0.58	0.78
p17_07 *	Students who repeat are rejected by their peers.	3.39	0.7	1	4	0.29	0.81
p17_10	It generates homogeneous classes in terms of skills, making teaching easier.	2.5	0.9	1	4	0.28	0.81
p17_11	The decision to repeat should be made by the teacher alone.	1.86	0.9	1	4	0.26	0.81
†† p17_12	A student should repeat if he/she fails an important subject.	2.01	0.8	1	4	0.31	0.80
†† p17_14 *	It can have negative effects on students’ self-concept.	2.47	0.8	1	4	0.49	0.79
p17_15	Students with many absences should automatically repeat.	2.55	0.9	1	4	0.29	0.81
p17_16 *	The student’s opinion should be considered in deciding to repeat a grade.	2.72	0.8	1	4	0.28	0.81
†† p17_17 *	No one should repeat a grade.	3.21	0.8	1	4	0.48	0.79

Note: * These questions were reversed from the original version. ^†† Indicates the selected items for the EFA.

shows the descriptive statistics for each of the 14 measured items corresponding to the original scale developed by Manley (1988). As already mentioned, three items were excluded from the construction of the scale because these items were not proposed in the original scale and mainly because these items look to inquire about some specific perceptions relevant to the Chilean context. These items were: “How you teach is directly related to the number of students who fail a subject”; “A student should repeat a subject as he/she will not have sufficient support and/or accompaniment the following year”; “Repetition is often the fault of students and their families”.

In general, the average of most items does not approach the maximum of the Likert scale used (Strongly Agree). Similarly, the standard deviations generally range between 0.5 and 1, indicating high heterogeneity within the measured items. Additionally, the reliability of the scale was assessed, yielding a Cronbach’s Alpha of 0.81 for all items and a McDonald’s Omega of 0.82.

These initial results show that no items could be considered excessively desirable in the sense that everyone would want to be positioned at one extreme or the other of the question. Furthermore, the included items could effectively measure a single construct.

To consider a set of highly intercorrelated items, each item should substantially correlate with the collection of remaining items (DeVellis, 2017). This approach allows for an evaluation of item quality, as a high item-rest correlation generally suggests that the item measures the same construct as the others, thereby contributing to the scale’s internal consistency. Similarly, this indicator helps detect problematic items; a low correlation may indicate issues such as poor alignment with the general construct, ambiguity, or poor wording.

Table 1 shows the results of the item-rest correlation. Following the guidelines from the literature, items with an indicator higher than 0.3 will be retained for the analysis (Zijlmans et al., 2018). As a result, the following five items will not be included in the scale construction: Students who repeat are rejected by their peers; It generates homogeneous classes in terms of skills, making teaching easier; The decision to repeat should be made by the teacher alone; Students with many absences should automatically repeat; The student’s opinion should be considered in deciding to repeat a grade.

Similarly, the estimated Cronbach’s Alpha for the retained items was 0.81, while McDonald’s Omega also remained at 0.82. These values fall within the recommended ranges, qualifying the measurement instrument as good in terms of reliability.

4.2. Exploratory Factor Analysis

To determine the internal consistency and obtain an initial structure for the observed data, an Exploratory Factor Analysis (EFA) was conducted with the ten selected items. The parallel analysis results (Figure 1) indicate that one factor should be selected.

This factor alone would explain more than 90% of the common variance of the selected items. Additionally, the Kaiser-Meyer-Olkin (KMO) test reached a value of 0.873, and Bartlett’s test of sphericity was rejected.

Table 2. Factor Loadings EFA (N = 2148).

Variable	Description	Factor 1	Uniqueness
†† p17_01	Has positive effects on learning and performance.	0.819	0.329
†† p17_02 *	It can lead to disaffection or rejection of school.	0.623	0.612
†† p17_03	It is not a cost to students but a stimulus to increase their effort and performance.	0.732	0.465
†† p17_04	It gives students time to grow and mature.	0.817	0.333
†† p17_05 *	It increases behavioral problems.	0.553	0.694
†† p17_06	It is a second chance to master the necessary knowledge and skills.	0.765	0.414
p17_12	A student should repeat if he/she fails an important subject.	0.258	0.933
†† p17_14 *	It can have negative effects on students’ self-concept.	0.596	0.645
†† p17_17 *	No one should repeat a grade	0.541	0.707

Note: * These questions were reversed from the original version. ^†† Indicates the selected items.

shows the factor loadings obtained for each item after conducting the EFA. The item “A student should repeat if he/she fails an important subject” did not exceed the established threshold (0.5) for inclusion in the final factor solution. As a result, after the EFA, eight variables remain, which are grouped into a single factor.

4.3. Confirmatory Factor Analysis

A confirmatory factor analysis (CFA) was conducted to confirm the structure previously determined with the EFA and to define the final structure of the Beliefs on Consequences of Retention scale (BCR scale). To ensure cross-validation, the CFA model was estimated on the test set (N = 2149). Furthermore, obtaining acceptable goodness-of-fit indices allows us to ensure the robustness of the model and avoid overfitting. It also enables us to confirm that the results are generalizable to different populations (Carter, 2016).

First, the CFA model was estimated with all the previously mentioned items (8 items). From this analysis, it was determined that two additional variables needed to be removed due to high modification indices. As recommended in the literature, the analysis and extraction of variables based on the modification indices was carried out sequentially. The removed variables could indicate multicollinearity or that these items were measuring very similar aspects of the defined construct. The items excluded were “It increases behavioral problems” and “It can have negative effects on students’ self-concept”. Both variables showed that their item error was closely related to the item “It can lead to disaffection or rejection of school”.

Table A6. Measurement model results.

Estimator				DWLS
Optimization method				NLMINB
Number of model parameters				24
Number of observations				2149
Model Test User Model:
Test statistic				74.246
Degrees of freedom				9
p-value (Chi-square)				0.000
Model Test Baseline Model:
Test statistic				23,319.463
Degrees of freedom				15
p-value				0.000
User Model versus Baseline Model:
Comparative Fit Index (CFI)				0.997
Tucker-Lewis Index (TLI)				0.995
Root Mean Square Error of Approximation:
RMSEA				0.058
90 Percent confidence interval—lower				0.046
90 Percent confidence interval—upper				0.071
p-value H_0: RMSEA <= 0.050				0.126
p-value H_0: RMSEA >= 0.080				0.002
Standardized Root Mean Square Residual:
SRMR				0.031
Parameter Estimates:
Parameterization				Delta
Standard errors				Standard
Information				Expected
Information saturated (h1) model				Unstructured
Latent Variables:
	Estimate	Std.Err	z-value	p( > |z|)	Std.lv	Std.all
Fm1 = ~
p17_01	1.000				0.814	0.814
p17_02_1	0.616	0.015	40.439	0.000	0.501	0.501
p17_03	0.969	0.017	56.023	0.000	0.788	0.788
p17_04	1.098	0.018	60.977	0.000	0.894	0.894
p17_06	1.029	0.017	61.308	0.000	0.838	0.838
p17_17_1	0.609	0.016	37.142	0.000	0.495	0.495
Thresholds:
	Estimate	Std.Err	z-value	p(>|z|)	Std.lv	Std.all
p17_01|t1	−1.543	0.043	−36.135	0.000	−1.543	−1.543
p17_01|t2	−0.739	0.03	−24.719	0.000	−0.739	−0.739
p17_01|t3	0.799	0.03	26.257	0.000	0.799	0.799
p17_02_1|t1	−1.273	0.037	−34.666	0.000	−1.273	−1.273
p17_02_1|t2	−0.095	0.027	−3.515	0.000	−0.095	−0.095
p17_02_1|t3	1.15	0.035	33.173	0.000	1.15	1.15
p17_03|t1	−1.308	0.037	−34.993	0.000	−1.308	−1.308
p17_03|t2	−0.321	0.028	−11.648	0.000	−0.321	−0.321
p17_03|t3	1.138	0.034	33.013	0.000	1.138	1.138
p17_04|t1	−1.79	0.05	−35.456	0.000	−1.79	−1.79
p17_04|t2	−1.123	0.034	−32.785	0.000	−1.123	−1.123
p17_04|t3	0.399	0.028	14.345	0.000	0.399	0.399
p17_06|t1	−2.026	0.061	−33.265	0.000	−2.026	−2.026
p17_06|t2	−1.336	0.038	−35.225	0.000	−1.336	−1.336
p17_06|t3	0.238	0.027	8.727	0.000	0.238	0.238
p17_17_1|t1	−1.651	0.046	−36.06	0.000	−1.651	−1.651
p17_17_1|t2	−1.154	0.035	−33.237	0.000	−1.154	−1.154
p17_17_1|t3	0.243	0.027	8.899	0.000	0.243	0.243
Variances:
	Estimate	Std.Err	z-value	p(>|z|)	Std.lv	Std.all
.p17_01	0.338				0.338	0.338
.p17_02_1	0.749				0.749	0.749
.p17_03	0.378				0.378	0.378
.p17_04	0.202				0.202	0.202
.p17_06	0.298				0.298	0.298
.p17_17_1	0.755				0.755	0.755
Fm1	0.662	0.015	45.309	0.000	1.000	1.000

in Appendix A.4 presents the final estimated measurement model, for which the confirmatory factor analysis fit indices show: χ²(df = 9) = 74.3, p < 0.00; CFI = 0.997; TLI = 0.995; RMSEA (95% CI) = 0.058 (0.046;0.071); SRMR = 0.031. These results are all within the limits established by the literature.

There were no statistically significant differences between the configural, metric, and scalar models for the key variables examined in this study: gender, administrative dependency, school socioeconomic level, and school terciles based on grade retention rates. The evidence of scalar invariance allowed for the examination of latent scores’ mean differences across groups.

Table 3. Measurement invariance across gender, school administrative dependence, school socioeconomic level, and School Terciles of Rate of Grade Retention.

Model	χ2(df)	CFI	TLI	RMSEA	SRMR	ΔCFI	ΔRMSEA	ΔSRMR	ΔTLI
Gender
Configural invariance	150.8 (18)	0.997	0.995	0.058	0.029
Metric invariance	164.8 (23)	0.997	0.996	0.053	0.030	0.000	−0.005	0.001	0.001
Scalar invariance	173.6 (34)	0.997	0.997	0.043	0.029	0.000	−0.010	−0.001	0.001
School Administrative Dependence
Configural invariance	167.1 (27)	0.997	0.995	0.059	0.031
Metric invariance	183.5 (37)	0.997	0.996	0.051	0.033	0.000	−0.007	0.001	0.001
Scalar invariance	222.5 (59)	0.996	0.997	0.043	0.031	0.000	−0.008	−0.001	0.001
School Socioeconomic Level
Configural invariance	188.6 (45)	0.997	0.995	0.059	0.033
Metric invariance	258.9 (65)	0.996	0.995	0.058	0.039	−0.001	−0.002	0.006	0.000
Scalar invariance	284.8 (109)	0.996	0.997	0.042	0.034	0.000	−0.015	−0.005	0.002
School Terciles of Rate of Grade Retention
Configural invariance	170.0 (27)	0.997	0.995	0.061	0.032
Metric invariance	201.3 (37)	0.996	0.995	0.056	0.034	0.000	−0.005	0.003	0.001
Scalar invariance	211.2 (59)	0.997	0.997	0.042	0.032	0.000	−0.013	−0.002	0.002

Note: CFI = Comparative Fit Index, TLI = Tucker-Lewis Index, RMSEA = Root Mean Square Error of Approximation, SRMR = Standardized Root Mean Square Residual, ΔCFI = difference in CFI, ΔRMSEA = difference in RMSEA, ΔSRMR = difference in SRMR, ΔTLI = difference in TLI.

presents the fit indices for the estimated invariance models.

Finally, the construct that we will call Beliefs on Consequences of Retention scale (BCR scale) comprises six items.

Table 4. Items for the BCR scale.

How Much Do You Agree with the Following Statements Related to Grade Repetition?
Mark with an X Only One Alternative for Each Statement.
	Strongly Disagree	Disagree	Agree	Strongly Agre
Has positive effects on learning and academic performance.
It can lead to detachment or rejection of school *.
It is not a cost to students but a stimulus to increase their effort and performance.
It gives students time to grow and mature.
It is a second chance to master the necessary knowledge and skills.
No one should repeat a grade *.

Note: Items with * were reversed from the original version to be included in the final scale.

presents the six selected items alongside the original question posed to participants. These six items represent all the measured variables, as they account for a significant proportion of the shared variance among the initially considered items.

Table 5. Beliefs on Consequences of Retention Scale by Individual Variables.

Variable	Values	Mean	SD	N
Total	Total	2.79	0.6	4297
Gender	Men	2.70	0.6	526
	Women	2.80	0.6	3721
Believes about grade retention
How you teach is directly related to the number of students who fail a subject.	Strongly disagree	2.93	0.6	1226
	Disagree	2.77	0.5	1747
	Agree	2.70	0.5	982
	Strongly agree	2.62	0.8	342
A student should repeat a subject as he/she will not have sufficient support and/or accompaniment the following year.	Strongly disagree	2.73	0.7	1075
	Disagree	2.69	0.5	1945
	Agree	2.87	0.5	1010
	Strongly agree	3.36	0.5	267
Repetition is often the fault of students and their families.	Strongly disagree	2.64	0.7	887
	Disagree	2.75	0.5	2135
	Agree	2.92	0.5	1065
	Strongly agree	3.10	0.6	210

and

Table 6. Beliefs on Consequences of Retention Scale by School Variables.

Variable	Mean	SD	N
Total	2.79	0.59	4297
School Administrative Dependence
Public	2.74	0.62	1389
Private Subsidized	2.83	0.58	2206
Private	2.76	0.57	702
School Socioeconomic Level
Low	2.66	0.64	191
Middle-Low	2.78	0.61	1102
Middle	2.80	0.6	1410
Middle-High	2.84	0.56	836
High	2.76	0.57	758
Terciles of the Grade Retention Rate
Tercil 1	2.71	0.61	1437
Tercil 2	2.81	0.59	1428
Tercil 3	2.84	0.57	1432

present the average results of the BCR scale, differentiating by individual characteristics and at the school level, respectively.

As mentioned, once the final model was validated, it was re-estimated using the entire sample of fourth-grade teachers (model retraining). For this final model, the confirmatory factor analysis fit indices were: χ²(df = 9) = 146,8, p < 0.00; CFI = 0.997; TLI = 0.995; RMSEA (95% CI) = 0.06 (0.051;0.068); SRMR = 0.029. The factor scores were subsequently predicted from this CFA model. The distribution of the estimated index can be seen in Figure A1 in Appendix A.5.

4.4. ANOVA of the Estimated BCR Scale

Regarding the variables related to the schools, results show statistically significant differences in the BCR scale between teachers in schools with higher retention rates during the first cycle of primary education and those in schools with very low retention rates (F = 10.63 (4, 4292), p = 0.000). In fact, post-hoc analysis using Tukey’s Honest Significant Difference test—Tukey’s HSD—(

Table A7. Tukey’s HSD for Tertiles of the Grade Retention Rate.

Tukey HSD pairwise comparisons
studentized range critical value(0.05, 3, 4294) = 3.3156696
uses harmonic mean sample size = 1432.324
			Mean
grp vs. grp	Group means		dif	HSD-test
1 vs. 2	2.7128	2.8086	0.0958	6.1511 *
1 vs. 3	2.7128	2.8389	0.1261	8.0998 *
2 vs. 3	2.8086	2.8389	0.0303	1.9487

Note: * p < 0.05. 1 = Tercil 1, 2 = Tercil 2, 3 = Tercil 3.

in Appendix A.7) shows that teachers in schools with no repeating students during the first cycle of primary education (1st to 4th grade) had statistically significant lower positive perceptions compared to teachers in schools where there are students who repeat during the first cycle of primary education.

Conversely, the ANOVA results show significant differences based on the administrative dependence of the school to the BCR scale (F = 10.62 (2, 4294), p = 0.000). Post-hoc analysis using Tukey’s HSD showed that compared to private subsidized schools, teachers at public schools had lower BCR scale scores, indicating that teachers in public schools tend to have more negative beliefs regarding grade retention.

The results also show significant differences based on socioeconomic group (F = 4.24 (4, 4292), p = 0.000). Differences observed through post-hoc analysis reveal that teachers in schools with lower socioeconomic levels tend to have more negative perceptions of grade retention (

Table A8. Tukey’s HSD for School Socioeconomic Level.

Tukey HSD pairwise comparisons
studentized range critical value(0.05, 5, 4292) = 3.8593347
uses harmonic mean sample size = 533.749
			Mean
grp vs. grp	Group means		dif	HSD-test
1 vs. 2	2.6618	2.7786	0.1168	4.5681 *
1 vs. 3	2.6618	2.7976	0.1359	5.3146 *
1 vs. 4	2.6618	2.8353	0.1735	6.7873 *
1 vs. 5	2.6618	2.7558	0.094	3.6778
2 vs. 3	2.7786	2.7976	0.0191	0.7465
2 vs. 4	2.7786	2.8353	0.0567	2.2192
2 vs. 5	2.7786	2.7558	0.0228	0.8902
3 vs. 4	2.7976	2.8353	0.0377	1.4727
3 vs. 5	2.7976	2.7558	0.0419	1.6368
4 vs. 5	2.8353	2.7558	0.0795	3.1094

Note: * p < 0.05. 1 = Low, 2 = Middle-Low, 3 = Middle, 4 = Middle-High, 5 = High.

in Appendix A.7).

Regarding individual-level results, significant differences were observed based on teacher gender (t(4245) = −3.42, p = 0.0006.), with female teachers having a higher BCR scale, indicating that they hold more positive perceptions of grade retention than male teachers.

The ANOVA results show significant differences among the responses for the three beliefs relevant to the national context included in the questionnaire (

Table A2. ANOVA results from the BCR scale compared with individual variables.

Variable	SS	df	F	p-Value
Woman	4.09	1	11.69	0.001
Believes about grade retention
How you teach is directly related to the number of students who fail a subject.	43.51	3	42.65	0.000
A student should repeat a subject as he/she will not have sufficient support and/or accompaniment the following year	115.44	3	119.00	0.000
Repetition is often the fault of students and their families	59.56	3	59.03	0.000

in Appendix A.1). Post-hoc results (

Table A10. Tukey’s HSD for “How you teach is directly related to the number of students who fail a subject”.

Tukey HSD pairwise comparisons
studentized range critical value (0.05, 4, 4293) = 3.6347711
uses harmonic mean sample size = 750.416
			Mean
grp vs. grp	Group means		dif	HSD-test
1 vs. 2	2.9296	2.7705	0.1591	7.4733 *
1 vs. 3	2.9296	2.6965	0.2331	10.9486 *
1 vs. 4	2.9296	2.6154	0.3142	14.7584 *
2 vs. 3	2.7705	2.6965	0.074	3.4753
2 vs. 4	2.7705	2.6154	0.1551	7.2851 *
3 vs. 4	2.6965	2.6154	0.0811	3.8099 *

Note: * p < 0.05. 1 = Strongly Disagree, 2 = Disagree, 3 = Agree, 4 = Strongly Agree.

in Appendix A.7) show that teachers who disagree more with the statement “How you teach is directly related to the number of students who fail a subject” tend to have higher BCR scores. Similarly, teachers who agree that “Repetition is often the fault of students and their families” also have higher BCR scales (

Table A11. Tukey’s HSD for “Repetition is often the fault of students and their families”.

Tukey HSD pairwise comparisons
studentized range critical value (0.05, 4, 4293) = 3.6347711
uses harmonic mean sample size = 548.197
			Mean
grp vs. grp	Group means		dif	HSD-test
1 vs. 2	2.6447	2.7497	0.105	4.2374 *
1 vs. 3	2.6447	2.9173	0.2727	11.0073 *
1 vs. 4	2.6447	3.0996	0.455	18.3666 *
2 vs. 3	2.7497	2.9173	0.1677	6.7699 *
2 vs. 4	2.7497	3.0996	0.35	14.1292 *
3 vs. 4	2.9173	3.0996	0.1823	7.3593 *

Note: * p < 0.05. 1 = Strongly Disagree, 2 = Disagree, 3 = Agree, 4 = Strongly Agree.

in Appendix A.7).

These results are complemented by the result obtained for the statement, “A student should repeat a subject as he/she will not have sufficient support and/or accompaniment the following year”, where teachers who agree more with this statement tend to have higher BCR scales (

Table A12. Tukey’s HSD for “A student should repeat a subject as he/she will not have sufficient support and/or accompaniment the following year”.

Tukey HSD pairwise comparisons
studentized range critical value (0.05, 4, 4293) = 3.6347711
uses harmonic mean sample size = 647.271
			Mean
grp vs. grp	Group means		dif	HSD-test
1 vs. 2	2.7343	2.6919	0.0425	1.9001
1 vs. 3	2.7343	2.8736	0.1393	6.2326 *
1 vs. 4	2.7343	3.3587	0.6244	27.9357 *
2 vs. 3	2.6919	2.8736	0.1818	8.1327 *
2 vs. 4	2.6919	3.3587	0.6669	29.8358 *
3 vs. 4	2.8736	3.3587	0.4851	21.7031 *

Note: * p < 0.05. 1 = Strongly Disagree, 2 = Disagree, 3 = Agree, 4 = Strongly Agree.

in Appendix A.7).

5. Discussion

This study aimed to validate the selected items from the SIMCE questionnaires on teachers’ perceptions of grade retention, included under the project “Perceptions of School Repetition in the Context of Implementing Decree 67”. For this purpose, 14 items were initially considered, and it was possible to determine the existence of a single unidimensional construct composed of six items, which adequately represent the Beliefs on Consequences of Retention (BCR scale) for elementary school teachers in Chile.

The reliability, discrimination measures, and validation of the internal consistency of the items considered confirm that the estimated construct has strong properties and can serve as a reliable scale for measuring the beliefs on the consequences of grade retention. Additionally, the unidimensional structure was tested through a CFA, which demonstrated a good fit for the items comprising this construct.

The results obtained regarding the BCR scale showed that contextual factors such as school retention rates, administrative dependence, and socioeconomic status significantly shape teachers’ beliefs about grade retention. Our results regarding contextual factors strengthen the validity of the BCR scale, confirming the results obtained in previous research.

In particular, teachers working in schools with higher grade retention rates or serving lower socioeconomic communities tend to hold more negative beliefs about grade retention. This may reflect teachers’ greater reliance on grade retention as a measure to manage heterogeneity within different classrooms, a result consistent with findings documented in the literature (Cipriano & Martins, 2021; Santos et al., 2023b). In contrast, teachers in schools with low retention rates may have less supportive perceptions of retention, possibly due to a stronger focus on alternative educational strategies to manage student heterogeneity. As mentioned by Protheroe (2007) some of these alternatives could include enhancing teacher effectiveness by supporting them in diversifying their instructional approaches to meet the needs of lower-performing students and extending learning opportunities through supplementary programs, such as weekend classes, after-school sessions, and summer schools.

Conversely, we show that teachers in lower socioeconomic schools tend to have more negative perceptions of grade retention, suggesting that they may be more aware of the broader challenges faced by students from disadvantaged backgrounds. This result is also consistent with findings in the international literature (Santos et al., 2023b; Walton, 2018), which highlights how socioeconomic factors could influence teachers’ views about grade retention, as they are more likely to recognize the role that sociodemographic conditions play.

Regarding administrative dependence, our results show that teachers in public schools hold more negative beliefs about grade retention. This disparity in viewpoints between various types of schools demonstrates the influence of institutional context on teachers’ attitudes towards grade retention, a result that has also been supported in the literature (Santos & Monteiro, 2024; Walton, 2018). Specifically, it is likely that teachers in public schools may view grade retention more critically, possibly reflecting differences in student populations, resources, or institutional approaches to managing student heterogeneity.

The BCR scale showed significant gender differences in teacher beliefs, with female teachers showing more positive beliefs about the effectiveness of grade retention than their male counterparts, a finding that has also been reported in the international literature (Santos & Monteiro, 2024). This result suggests that male and female primary educators may adopt different approaches when dealing with students with performance issues. These differences in perception could, in turn, influence their classroom practices and decision-making regarding how to support struggling students.

Concerning the items that sought to explore the Chilean reality, we found that the teachers’ beliefs about responsibility for student outcomes play a pivotal role in shaping their retention views. Our results could indicate that when teachers believe they don’t have any relation with poor student outcomes, they reveal much more positive beliefs toward grade retention. Thus, teachers who do not directly associate student retention with their actions and attitudes but rather with factors beyond their control hold more positive perceptions of grade retention. This could be because they consider grade retention as something beyond their control or something they cannot directly influence, thus reinforcing the belief that grade retention is an effective tool for managing student performance.

These findings align with the meritocratic discourse that prevails in the Chilean educational system, where individual effort plays a key role in students’ success or failure (Peña & Toledo, 2017). This perspective not only places full responsibility on students and their families for academic failure but also removes accountability from schools and teachers. In line with the notion of negative reinforcement, teachers operating under this logic may overlook the potentially negative consequences of grade retention, such as emotional distress, disengagement from school, and damage to self-esteem (Valbuena et al., 2021; Kearney, 2008; Goos et al., 2021; Van Canegem et al., 2021).

On the other hand, belief systems applied to education are the ideas that teachers, students, or institutions hold about teaching, learning, and knowledge (Wolf & Brown, 2023). These systems influence how teachers teach and interact with students, how students learn, and how the educational environment within the classroom is structured (Muijs & Reynolds, 2002; Sabarwal et al., 2022). Furthermore, Santos et al. (2023b) highlighted that teachers’ beliefs about grade retention are deeply intertwined with their broader psycho-pedagogical perspectives. These beliefs are embedded within a complex system encompassing views on learning, intelligence, assessment, and educational fairness.

Teacher beliefs, as defined by Kagan (1992), are a form of personal knowledge generally described as teachers’ implicit assumptions about students, learning, classrooms, and the subject matter to be taught, which are instrumental in determining the quality of interactions among teachers in a given school.

These beliefs—both implicit and explicit—that teachers hold about their students’ abilities and potential can affect how they interact with them, shaping classroom dynamics and influencing student outcomes (Muijs & Reynolds, 2002). Thus, teacher beliefs become a fundamental part of the design and implementation of public policies, as they directly influence how the educational environment is structured (Fives & Buehl, 2016; Sabarwal et al., 2022).

Teachers’ perceptions are crucial when attempting to modify or change specific practices, like grade retention, because perceptions can lead to resistance to change. Since teachers are typically responsible for understanding and implementing the requirements of educational reforms, the impact of a progressive policy like Decree 67 could be diminished or even nullified if teachers’ perceptions do not align with the inclusive and progressive normative imposed by Decree 67.

The implementation of Decree 67, which aligns with global educational trends that seek to move away from punitive approaches to student failure, needs to consider these belief systems that exist in a more profound way, and more specifically, teachers’ beliefs about implementing the law. The decree should also emphasize the importance of robust support mechanisms and political consensus to ensure the successful integration of inclusive educational practices. Moreover, the challenges associated with its implementation underscore the need for sustained advocacy, comprehensive support systems, and cross-sectoral collaboration from the government to achieve its objectives. Additionally, it should prioritize the development of intervention strategies that focus on personalized support to guide teachers in making informed decisions, reducing reliance on intuition or personal beliefs.

By recognizing that grade retention should be an exception rather than the norm, Chile is aligning its policies with contemporary views on education that prioritize equity and the holistic development of students, taking their well-being into account rather than focusing solely on educational outcomes. To do this, it is necessary that all those involved in implementing this regulation align their beliefs, something that has not yet been achieved in Chile, as our results showed.

Finally, this study has produced a concise, validated instrument for assessing teachers’ beliefs about grade retention and its consequences. The instrument is a valuable tool for monitoring public policies aimed at reducing grade retention. It can provide insights into how teachers’ beliefs may influence the implementation and effectiveness of such policies and help assist in designing programs or interventions within schools to improve the implementation of these public policies.

The main drawback of this research is the potential presence of selection bias, as schools with high or low grade retention rates may also have distinct student populations, implement different programs, or be located in different geographical areas. These differences could lead to unobserved factors affecting the beliefs of teachers about grade retention. To partially address this issue, E/CFA with fixed effects at the school level could control for unobserved, time-invariant characteristics. However, the limited number of observations per school—since the survey was only completed by 4th-grade head teachers—does not allow us to carry out this estimation. Furthermore, we do not have a geographical grouping (e.g., neighborhood) sufficiently small to enable a fixed-effects model that adequately controls for unobserved variables at the geographical level. It is also important to note that this approach would not eliminate selection bias arising from non-random differences between groups.

We recommend that future research validates this instrument in secondary schools, as well as in other countries, to enhance its applicability and ensure its reliability across diverse educational contexts. Such efforts would ultimately contribute to more informed and inclusive approaches in education.