Introduction

The importance of the relation between family and school, in the education field, is undeniable, and constitutes, for several decades one of the key aspects in sociological investigation, both in Brazil and abroad (^{ALVES et al., 2013}). At least since the 60s, the strong association between social origin and school performance is known (^{BOURDIEU; PASSERON, 1975}; ^{COLEMAN, 1966}). However, this relation only indicates a strong influence, rather than a cause and effect relationship. In this sense, the association between students’ social origin and school performance does not constitute a deterministic relationship, as high school success rate cases in popular means exist and may be enhanced (^{LAHIRE, 2004}; ^{VIANA, 2000}; ^{ZAGO, 2000}). According to ^{Bourdieu (2004}), both the structure and the volume of family capital account to different forms of academic performance. The social networks in which a family, and consequently, their children, circulate are among the factors that influence each child’s academic performance. That is to say, they are part of the diversified, imbricated and often subtle factors that condition school success or failure, as ^{Souza e Silva (2003)} points out.

Since the 50s and 60s, several ample statistic surveys were conducted in the United States and in Europe (^{COLEMAN, 1966}; ^{PLOWDEN REPORT, 1967}), and defined the relevance of factors associated with social and family origin in children academic performance, identifying their prevalence over typically academic variables. Among the family-related variables identified, parents’ income, occupation and education level, race or skin color, number of children in the family and the child’s position in the siblings’ birth order, gender, among others, should be highlighted (^{NOGUEIRA, 1998}). During the two following decades, due to the hegemony of reproduction theories in field of sociological investigation on the educational system, researchers start to perform more profound analyses and to increase data collection on school inequalities, emphasizing the cultural heritage of families, and their expectations regarding children schooling, based on their class position (^{ALVES et al, 2013}).

According to these studies, family origin and parents’ education level are more relevant to explain academic performance inequalities than the school environment itself. ^{Bourdieu’s (1975}) opinions on schools start with these findings and investigate cultural and school components that articulate to influence different school results. In this sense, the importance of the families’ cultural capital for the definition of school paths. A great part of the studies considered parents’ education level and families’ socioeconomic status as the variables to be used as parameters to investigate the influence of family origin on children’s educational performance. Especially in Brazil, several researches have considered parents’ education level as proxy for socioeconomic status (SES), based on the high association between education level and families’ socioeconomic conditions, identified in the country (^{BARBOSA, 2009}; ^{SOARES, 2004}).

During the 80s, the key role played by families in sociological studies about education is consolidated with the emergence of what was called from that point on, as “sociology of family-school relationships”. Researches on the area, which have multiplied ever since, started to investigate internal dynamics and family socialization processes, seeking to understand the dispositions and strategies developed by parents regarding their children’s school lives, in addition to analyzing how these elements contribute to configure different school outcomes (^{CHARLOT, 1996}; LAHIRE, 1997; ^{NOGUEIRA; ROMANELLI; ZAGO, 2000}). ^{Viana (2005}), among others, emphasizes the need to deepen studies about familial socializing processes in different social contexts, seeking hypotheses that explain school success, measured based on more long-lasting paths, for instance. For the author, such socialization processes produce “dispositional traits {that} potentially trigger elements facilitating - or complicating - academic success, based on how close - or how distant - these traits are in relation to school”. (^{VIANA, 2005}, p.121, emphasis added by author).

The present work is designed to contribute to this long tradition of studies, discussing results of a research investigating the influence of some aspects in family conditions and dynamics on elementary school students’ performances. Given the interest in knowing better what takes place in the family environment, which may affect the schooling process, we have investigated how the family’s belonging to a certain religious group may influence the student’s academic success. Although this topic has been little explored, recent research in Brazil (^{CUNHA; RIOS-NETO; OLIVEIRA, 2014}) and several international researches (^{GLANVILLE; SIKKINK; HERNÁNDEZ, 2008}; ^{LEHRER, 1999}; among others) have highlighted the importance of religion in educational performance studies. This article is designed to contribute to this discussion in the Brazilian case.

Religious affiliation in Brazil: what do researches reveal?

Brazil’s religion composition underwent great changes in the past decades, evidencing the number of Catholics dropped, as the number of evangelicals and the number of people who declared themselves as not belonging to any religious group increased. During the 40s, Catholics represented 95.3% of the population, whereas between the 40s and the 80s the rate of adhesion to Catholicism dropped 6 percentage points, and 24.6 other percentage points during the following 30 years, reaching 64.7 of the population in 2010. During this period, the increase in the number of those who identify themselves as not religious, as well as in the group of evangelicals was significant. In the 80s, the group of not religious people represented 1.6 of the population, increasing to 8% in 2010, which represents 6.4 percentage points (equal to 13.4 million inhabitants). The number of evangelicals, however, increased from 6.6% to 22.2% of the population, representing the religious group which increased the most during the last inter-census period, with an increase of approximately 16 million people, or 15.6 percentage points (^{NERI, 2005}; ^{JACOB; HEES; WANIEZ, 2013}), as shown in Table 1 below.

Although the number of people who declare themselves as not religious is constantly increasing in Brazil, it may be observed that, during the last decade, its increase rate has decreased, as the number of those who declared themselves as not religious went from 7.3% in 2000 to 8% in 2010. Notwithstanding, this category still occupies third place in the country, considering the statements made for the 2010 census, being located, in relative terms, immediately after the group of Catholics and the group of evangelicals (^{JACOB et al, 2013}). In accordance with these authors, demographic censuses reveal a progressive trend towards religious diversification in Brazil, with the decrease in number of Catholics, a significant increase in the number of those who declare themselves as not religious, and an exceptional increase in the number of evangelicals, especially from the Pentecostal group. According to data from the IBGE Demographic Census recorded in Table 1, in 2010, the number of evangelicals increased to 22.2%, totaling 42,275,636, of which 25,370,484 people identified themselves as Pentecostal evangelicals.

Despite the high number of Pentecostal denominations in the country, the groups Assembleia de Deus (Assembly of God), Congregação Cristã no Brasil (Christian Congregation in Brazil) and Universal do Reino de Deus (Universal Church of the Kingdom of God), together, concentrate 64.9% of the Pentecostalists, or 16 million followers (IBGE, 2012). As ^{Mariano (2004}) stated, the Pentecostal expansion is not recent nor episodic, and has been happening for half a century in a constant manner, which enables Pentecostalism to become the second greater religious group in the country. While the group of Pentecostalists has increased 8.9% per year, historical Protestants have increased 5.2%, so that the first account for two thirds of the evangelicals.

The socioeconomic and demographic profiles of Pentecostal evangelicals and historical evangelicals are quite different. Data from the latest census reveal that most Pentecostals have lower income and education level than the average of the Brazilian population. Most of them earn up to three times the minimum wage and have domestic jobs, usually simple and precarious, at a rate much higher than the national average. In contrast, historical evangelicals have high income and education levels, both much higher than the Brazilian average, and are well distributed in the high school, college and post-graduate educations, and in the income brackets between six and twenty times the minimum wage. Both groups are mostly urban, and the number of women is higher than the number of men (^{PIERUCCI, 2004}).

As to the distribution of followers by skin color, the average of black and dark-skinned among Pentecostal evangelicals is higher than the average of the population, whereas historical evangelicals are characterized by a large number of white people. Pentecostals hold more children and teenagers than adults, whereas historical evangelicals hold more adults and senior citizens than youths. For ^{Mariano (2004}), these differences also reflect their social class distinction. It is important to emphasize, however, that an under-reporting is assumed, especially in the cases of belonging to African religions, such as umbanda and candomblé. Furthermore, information collection at the demographic census does not allow deeper information on religious mobility situations or belonging to two religions.

As ^{Cunha, Rios-Neto and Oliveira (2014}) point out, the history of the Catholic Church in Brazil accompanies the country’s history. From the discovery to the proclamation of the republic, Catholicism was Brazil’s official religion. In 1889, when the republic was proclaimed, Brazil was declared a secular state. In its history, the catholic religion is characterized by including both highly religious people, as well as those who simply attend church to comply with social or passage rites, but do not consider themselves as active Catholics.

^{Coutinho (2011)} points out, about historical Protestantism in Brazil, that, after the ports were opened and the Portuguese royal family moved to Brazil in early 19th century, British merchants brought the Anglican Church, and German merchants brought Lutheranism to Brazil. The construction of protestant churches was allowed, as long as they did not intend to convert Brazilians, serving only as worship place for immigrants.

Not until late 19th century were American missionaries able to implement, in Rio Grande do Sul, the first Anglican communities for Brazilians. In 1855, Robert Reid, a Scottish missionary, founded congregational churches, which were non-denominational and had no connections with any international churches. Almost at the same time, in 1862, the Presbyterian Church was founded by Ashbel Green Simonton, in Rio de Janeiro. In 1871, the first Baptist group settled in Santa Bárbara d’Oeste (SP), brought by American missionaries and immigrants escaping the American Civil War. In 1888, the Brazilian Presbyterian Church Synod was created, disconnecting it from North American churches. Staring in the 20th century, several divisions also took place in Presbyterian denominations (^{COUTINHO, 2011}).

Classic Pentecostalism started in Brazil with the arrival of two Swedish men, who had become Baptist pastors in the United States, to the state of Pará. They founded the Missão de Fé Apostólica (Apostolic Faith Mission), which later was named Assembleia de Deus (Assembly of God). The Congregação Cristã no Brasil (Christian Congregation in Brazil) was brought from the United States to Brazil in 1910, by the Italian-American missionary Louis Francescon, who worked in Italian communities in south and southeast Brazil. The second wave (also known as second generation Pentecostals) started to organize in the 50s, when two North-American missionaries from the International Church of the Foursquare Gospel arrived to Brazil and created the National Evangelization Crusade, with a mass evangelization strategy centered on divine cure - using, above all, the radio - contributing to the expansion of Pentecostalism in Brazil. They founded the Church of the Foursquare Gospel, which later on originated the denominations: Brazil for Christ, God is Love², House of Blessing, among several others. (^{COUTINHO, 2011}).

Considered the third Pentecostal wave, neopentecostalism started on the second half of the 70s, in the United States, and deserves special attention, as it has been, since then, the wave during which the number of churches increased the most in Brazil. Based on the Prosperity Theology, neopentecostalism is characterized for preaching that success, happiness and prosperity may be reached in this earthly life, in contrast with a more ascetic trend that characterizes traditional evangelicals. Neopentecostals are quite different from historical Protestants and Pentecostals, as they even reject traditional habits and customs that are considered symbols of conversion and belonging to Pentecostalism. Although they are more liberal regarding moral aspects, neopentecostal denominations, as well as all other evangelical denominations, maintain the prohibition of alcohol, tobacco and drug use, and of extramarital and homosexual sex (^{MARIANO, 2004}).

Religious belonging and education

The changes through which the country went during the past decades, regarding the population religious composition, draw attention to the possible social, economic, cultural consequences of religious belonging. Literature on education has shown that social status, gender and color/race are among the most significant variables for explaining students’ performance (^{BARBOSA, 2000}; ^{CARVALHO, 2004}; ^{ALVES; ORTIGÃO; FRANCO, 2007}). In addition, more recently, several researches have also focused on the potential influence of religion - as declared by the families - in the results of high performance schools (^{COSTA, 2008}; ^{COSTA; PRADO; ROSISTOLATO, 2012}; ^{CUNHA; RIOS-NETO; OLIVEIRA, 2014}), a concern that motivated the present study.

According to ^{Regenerus (2000)}, religious socialization, understood from the involvement in activities related to a given religious affiliation, may be a way to provide integration and motivation for school success, in a process that usually operates besides private belief systems and organizational affiliations. In this sense, it could be viewed as a form of social integration whose consequences may be the reinforcement of values leading to an improved educational performance (^{CUNHA; RIOS-NETO; OLIVEIRA, 2014}; ^{LEHRER, 1999}, 2004). ^{Cunha, Rios-Neto and Oliveira (2014}), analyzing data from the longitudinal research conducted by CEDEPLAR/UFMG between 2007 and 2010, involving high school youths from some cities in Minas Gerais, investigated the relation between religion and educational performance, showing the importance of religion for studying education in Brazil, particularly considering changes in the Brazilian population religious composition in the past decades. The study developed confirmed that

the religion is, in fact, related to students’ performance in different manners along the grade range, that is, it is associated to a student who has low grades (for instance, students in the 10% percentile) differently from students who have good grades (for instance, students in the 90% percentile), as demonstrated in historical protestant students’ Portuguese grades. It is also confirmed that Catholics had better performances in Mathematics than other students. In addition, as demonstrated in the nonparametric analysis, the historical protestant group did better in Portuguese and was the only group to have a performance better than that of the Catholics, whose grades were higher than those of Pentecostal and neopentecostal students. A possible explanation for this result is the focus of historical Protestants in reading the Bible and interpreting texts, activities directly related to the abilities required in learning Portuguese (^{CUNHA; RIOS-NETO; OLIVEIRA, 2014}, p. 108-109, emphasis added by authors).

In our study, given the impossibility to use individual students’ grades in external evaluations aligned with their familial characteristics listed in the Rio de Janeiro Municipal Education Department enrollment form, to which we had access, we chose to investigate this relation from the students’ age-grade gap, information also available in the empiricism provided for the study. Thus, the age-grade gap will be our dependent variable, that is, the variable we want to estimate. Other variables are investigated in the logistic regression analysis that studies the ratio of probability that students be in age-grade gap situation, in their schooling process, considering their belonging to a given religious denomination (as stated by the people responsible for them in the enrollment form). The unit of analysis is the student, who was enrolled from the first to the ninth year of elementary school in 2011 in the Rio de Janeiro municipal school system,

The age-grade gap rate³ evidences the proportion of students in a given school year, stage or learning level, at ages higher than the ones considered appropriate for each school year in relation to the total enrollment. In the Brazilian educational system, the student is considered in age-grade gap or distortion when the difference between their ages and the expected age for the grade is two or more years. The distortion may happen due to late enrollment in the school system, due to leaving or dropping out of school, and, above all, due to being held back. In this situation, students continue their studies, but maintain the gap in the appropriate age considered for each grade, in accordance with what is proposed by education law. In this case, the student will be recorded with the status of age-grade distortion.

Several researchers (^{KLEIN, 2006}; ^{OLIVEIRA; ARAÚJO, 2005}; ^{OLIVEIRA; SCHWARTZMAN, 2002}) agree that one of the greatest and chronic problems in Brazilian education is that of students being held back, which is evident not only in the high fail rates, but also in the age-grade gap rates (^{CORREA; BONAMINO; SOARES, 2014}). As ^{Ribeiro (1991}) already pointed out, being held back once tends to lead to being held back again, contrary to what the Brazilian pedagogical culture suggests, that repeating a grade helps the child to progress in studies. According to the author, the consequences holding back the population of the first series in elementary school have such magnitude that individuals are old in relation to the grade they are still in and tend to abandon school (^{SOARES, 2007}, p. 138).

It is important to emphasize that, from the first researches performed in Brazil, based on SAEB data, it has become evident that delay in school, measured by the age-grade gap, has a negative impact on academic performance (^{FERRÃO et al., 2001}). Finally, we record ^{Alves, Ortigão and Franco’s (2007}) study, which investigates the association between academic failure and the student’s familial and sociodemographic characteristics, using data obtained from SAEB 2001. Results found by the authors

{…} reproduce usual literature findings, which already signaled that several sociodemographic variables increase the risk of failure (work, being male, being black). Economic capital, however, is a failure protection factor. In this work, we demonstrate that economic capital does not protect everyone equally. Specifically, high economic capital increases the risk of failure in students who identify themselves as being black. This result needs to be understood in the context of failure policy dynamics and of allocation of students in schools. In Brazil, approval and failure are, typically, school unit policies, decided in a relatively autonomous way by the schools. The allocation of students in schools is strongly influenced by the students’ economic status, especially by way of the relation between the families’ economic status and their place of residence. Black families in better economic conditions tend to have better living options, which may provide them with access to schools with better teaching conditions. However, this may place their children in the group with higher risk of failing in these schools. (^{ALVES; ORTIGÃO; FRANCO, 2007}, p. 178).

Such considerations and research findings reinforce the relevance of this study’s option to use the school gap indicator a proxy for academic performance, and have influenced the choice of variables to be investigated, based on the empirical data to which we had access, as recorded below.

Methodology

Information used in this study was supplied by the Rio de Janeiro Municipal Education Department. The research used information from the enrollment form or registration chart for students in the Rio de Janeiro municipal school system, who were in elementary school in 2011. The basis on which this study was initially outlined presented a total of 776,023 students enrolled in 2011. Many times the same student had different codes, cases - for instance - in which the student had requested to be transferred to another school, and his/her student code was still recorded in the origin school, although he/she was being transferred to a different school unit. Furthermore, when the student simply dropped out and, after some time, resumed his/her studies at a different school, he/she frequently was assigned a new code, if the recipient school referred to the SGA⁴ (Academic Management System) to search for the student’s code and did not find it. Considering these situations, a meticulous checking and refining work was performed in the database, in which it was possible to detect perfect homonym cases, in which parents’ names were checked to determine they were not duplication cases.

After this base refinement work, in which repetitions were discarded, the total of students enrolled in 2011 dropped to 732,507 students. After that, we selected only students who were enrolled in the regular elementary school, that is, the ones enrolled between the first and ninth grade during that year, reducing the total number of cases in the analyzed universe to 557,400 students.

The enrollment form is completed by the parents or the person responsible for the student and, therefore, it allegedly is more reliable than other questionnaires completed by the students, as is the case of the student’s contextual questionnaire for the Prova Brasil (Brazil Test). Although stating their religion in the registration chart does not constitute the parents’, students’ or families’ admission of religion, it tends to indicate a certain level of belonging or identification with a given religious denomination. In addition, as previously pointed out, the fact that the answers are provided by the parents or persons responsible for the students increases the probability of information consistency. We consider that the indication recorded in the registration chart may be considered, at least, a reasonable approximation for recording the belonging to a certain religious group, or an indicator of probable participation in the social network associated with a given religious denomination.

The registration chart structure includes 53 questions⁵ organized in an Academic Control System (SCA) used by schools and by the education departments to control/manage students’ circulation through the system’s schools. In this registration chart, along with the students’ system code, other information is provided, such as the parents’ education level, parents’ occupation, the person with whom the student lives, information on health and/or special education needs, and the respective special treatment code. The form records the school code⁶, the student’s date of birth, the last school year attended by the student at the school, the grade attended by the student and the last academic result obtained by the student at the school (passed, failed or failed due to absences, whether the student has or had any pending classes in the previous year), as well as the student status (active or inactive). The age/grade gap variable was constructed as a dichotomous variable based on the students’ date of birth in combination with the grade attended by the student in 2011, considering the difference of two years or more between the student’s age and the age expected for that grade.

The registration code also indicates if the student attended preschool and literacy classes, the student’s Número de Inscrição Social -NIS (Social Registration Number), as well as the declared skin color/race and the religion specified by the person responsible for the student. The question on religion is an open question, systematized by the Pereira Passos Institute⁷ as 43 different types. Based on this question, we created the variable “religious denomination”, regrouping the more than 40 different types of religion into five categories: catholic, mission or traditional evangelical, other evangelical denominations (Pentecostal, among others), others, and not religious, which were used in the initial data description stage.

We considered the family’s highest education level as a proxy for socioeconomic level, following the trend of other researches, which have repeatedly stated the high correlation between education level and the families’ socioeconomic status, particularly in Brazil (^{BARBOSA, 2009}; ^{SOARES, 2004}). The variable was developed using also information present in the registration chart about the parents’ education level. Once again, it is noticed that, as the form is completed by the people responsible for the students, this information is probably more consistent than the information found in the Prova Brasil database, obtained by the participating students’ answers to the contextual questionnaires.

Similarly to the students’ case, we used the school students’ parents’ education level average as proxy for the school’s SES, recording as “high education level” those who had attended high school or higher (even if not complete) and as “low education level” those who had attended only up to elementary school (complete or not). In the students’ universe, we detected, on average, 35.86% of parents who had high education level. Considering this average found in the universe of students analyzed, we transformed this proxy into a dichotomous variable: schools in which over 35.86% of parents had not attended high school (=0), and schools in which over 35.86% of parents had attended high school or college (=1). Thus, to each student in each school, a variable (school average SES) was added.

Other context and sociodemographic variables (students’ gender and skin color/race originally expressed based on options used in IBGE’s demographic census, and whether the student attended preschool or not) already indicated in the pertinent literature as relevant for the intended analysis (^{ALVES; ORTIGÃO; FRANCO, 2007}) were also transformed into dichotomous variables⁸. Similarly to other studies which have already identified the relevance of having attended preschool as a factor that favors a longer and more successful schooling process (^{KRAMER, 2006}; ^{FELÍCIO; VASCONCELLOS, 2007}; BANCO MUNDIAL, 2001), we also considered, in the analysis, the variable that refers to the family arrangement, whose relevance for school performance has been highlighted in literature (^{SILVA; HASEMBALG, 2002}). The “family arrangement” variable was created based on the coding of answers to questions on with whom the student lives (mother only, father only, neither or both).

Finally, based on the student’s Número de Inscrição Social -NIS (Social Registration Number) found on the database built with information on the Municipal Education Department registration charts, another variable was created, which we consider an indicator of economic vulnerability for students. The NIS record in the student’s chart indicates the fact that the student and his/her family benefit from some type of welfare - such as the Programa Bolsa Família (Family Stipend Program), for instance, which was considered a poverty indicator.

Table 2 provided next shows how students in the regular elementary school in the Rio de Janeiro municipal system were distributed, as per their families’ statements regarding religious affiliation or denomination on their enrollment records in 2011. Although the group of those families who declare themselves as catholic is predominant, the group of evangelical denominations (traditional, mission or others) is about one fourth of the students.

As the intended analysis depended on information about religion indicated in the registration chart, we decided to consider, as our investigation universe, only students whose registration charts provided information on their religious denomination, which limited the number of studied cases to 453,021 students, distributed as per table 3 next.

Table 3 shows that, from a total of 138,781 evangelicals, only 15.4% declared to be historical or mission evangelicals, whereas other evangelical denominations add up to 117,497 students. Among the evangelical denominations that do not correspond to historical Protestantism, 114,494 are Pentecostal evangelicals - which corresponds to 97.5% of this total, whereas only 3,003 (2.5%) belong to undetermined evangelical denominations. The prevalence of Pentecostal evangelicals in the group of students whose persons responsible for them indicated they are evangelicals led us to the decision to recode the religious denomination variable into four categories only: catholic, evangelical, other religions, and not religious. Therefore, the category “evangelical” - which, as we saw, may be essentially characterized as Pentecostal or neopentecostal - ended up totaling 138,781 students, or 30.64% of the students with declared religious affiliations. The group indicated in the category “others” included spiritualists (0.7% of the total), African origin religions, such as umbanda and candomblé (0.3%), and those declared affiliation to other religions, without specifying which (0.2%).

We know that many inferences may be made based on the correlations among the variables. Although the correlation may be a very useful research tool, it does not tell us anything about the variables’ predictive capabilities (^{FIELD, 2009}). However, through regression analysis, it is possible to predict one of the variables as a function of the other. Regression analysis enables predicting values of the dependent variable based on one or more independent variables. In this study, the logistic regression model was used, which is a multiple regression, with a dichotomous category output variable. One of the principles of logistic regression is that the dependent variable behaves in a binary or dichotomous manner (^{FIELD, 2009}, p. 221). This is a chance model, that is, in which the probability ratio that students whose families declared they belong to a certain religious denomination are in age/grade gap or not - considering INEP’s definition, that is, students with a two-year gap or more - in comparison with students whose parents declare they belong to other religious belonging, keeping other variables constant.

The following is a descriptive analysis of the information collected on students and their families, focusing on the correlations that, according to the literature provided, may have impact on the schooling process, and the success or failure possibilities measured by means of the age/grade gap. After that, we will present the variables considered in the logistic regression analysis, which will study the academic gap change ratio, considering the belonging to a given religious denomination.

Age-grADE gap and religious belonging¹⁰

We sought to investigate possible relations between belonging to a given religious group, and the age/grade gap in the schooling process. We consider that belonging to a given religious denomination could have impact on the chance of academic success, measured by the age-grade gap, and we have initially attempted to identify relative distributions in each case. Table 4 below shows the frequency and percentage of those displaying a greater or smaller gap, based on their belonging to each religious denomination.

The largest percentage of students who do not have an academic gap is found in the evangelical groups, with 82.8% of the cases. As this group presented the largest percentage of students without an academic gap, we chose to consider them as the category or reference group for analysis trough logistic regression, to be provided below.

We have also investigated the correlations between the “gap” variable and the “family arrangement” variable - the only categorical variable considered in the analysis other than religion - due to its potential to interact with a given religious group or not. As explained before, the information regarding the family arrangement was based on answers on with whom the student lived, father, mother, both, or neither. Among the students enrolled in the regular elementary school at the Rio de Janeiro municipal system in 2011, 47.7% lived with both parents, 35.1% lived only with their mothers, 3.9% lived only with their fathers, and 13.3% lived with neither. Several studies in the field of sociology of education indicated the relevance of family arrangement for academic success or failure (^{SILVA; HASEMBALG, 2002}), especially among students from families with more precarious economic conditions. We have also investigated the relation between the “family arrangement” variable and academic gaps, which resulted in table 5 below.

Corroborating other results in several education researches, the highest percentage of students without academic gap (84.3% of the cases) is found in the group of those who live with both parents - biparental families. The following step was studying the relation between belonging to a given religious group and each type of family arrangement, as provided in table 6 next.

Although almost half (47.7%) of the students lived with both parents, relevant variations may be observed in accordance with the religion declared by those responsible for the students. Table 6 shows a slightly higher concentration of biparental families among those who declare themselves as catholic or evangelical (47.5% and 52.1% respectively). Such percentages seem to avow the hypothesis that belonging to the evangelical religions would be positively associated with biparental family arrangements. Previously mentioned studies (^{MARIANO, 2004}), however, enable us to assume that this religious belonging would also be associated with the valorization of individual efforts as a strategy for conquering improvements in a social mobility perspective, especially among the Pentecostal and neopentecostal evangelicals. In this sense, the association between this effort ethos and biparental arrangements may provide an environment that favors successful schooling and, therefore, with a gap equaling zero. Therefore, belonging to a given religious denomination (evangelical, in particular) could increase the probability of family conditions to be more favorable to successful schooling, constituting sort of a virtuous circle.

In order to estimate the chances that the student has an academic gap, given some control categories, a logistic regression model was used. The variables described in Chart 1, provided next, were used in it:

Source: Prepared by the authors based on information from the enrollment database during the year 2011 from the Rio de Janeiro Municipal Education Department.

CHART 1 - Description of variables used in logistic regression analysis 

As a regression principle, the analysis of each of the covariations is done in relation to the dependent variable, whereas all other control variables are maintained constant. As pointed out in Chart 1, for this analysis we used the two situations in which a greater no-gap student percentage was found: the evangelical group and the biparental family arrangement. In case of dichotomous variables, the reference considered is always equal to one. We also considered: Exp (B) - 1 = chance, that is, the influence of each variable on the probability that the student’ schooling process has a gap. In regression, the reliability level of 95% and the significance level lower than 0.05 were imposed.

The results provided in Table 7 indicate the statistical significance of all variables studied. Students whose parents attended high school or college have a 53.8% reduction in the change to have an academic gap, if compared to students whose parents only attended up to elementary school. This result corroborates the vast literature in the sociology of education field, which states that, the higher the parents’ education level, the lower the probability of the child having an academic gap. Furthermore, as pointed out by ^{Ferrão et al (2001}), the negative impact of the age-grade gap in academic performance may have different magnitudes as per the school attended. In this study, it was possible to determine that the student who attends a school in which over 35.86% of the parents have high education level (at least high school or college) reduces the chance of gap at 18.8%, indicating the possibility of the peer effect, identified in several researches (^{ALVES; SOARES, 2007}), which most likely interact with different school processes often associated to school management. Nevertheless, regarding the specific interest of the present study, findings from other research conducted on the same education system, albeit with a more qualitative nature, indicate the probable association between these indicators and the belonging to a given religious denomination, not only by the families but also by other school agents (^{COSTA, 2008}; ^{RAMOS, 2014}) in an interaction that seems to reinforce also interschool stratification mechanisms. In this sense, the very magnitude of the school average SES would recommend further deeper studies about this analysis level.

The data provided indicate that living only with the father increases in 71.5% the chance that the student has an academic gap, if compared to living with both father and mother (biparental family arrangements), thus highlighting the previously pointed out effects of family arrangement on schooling. Once more the possible interaction between this variable and the values of the respective family religious affiliation (^{MARIANO, 2004}), which would tend to enhance a sort of “virtuous circle” favoring good academic performance, should be emphasized.

In accordance with numerous sociological studies - since the ^{Coleman Report in 1966} - about the influence of socioeconomic origin in school results, having a Número de Inscrição Social -NIS (Social Registration Number) increases in 9.4% the chance that a student carries an age/grade gap regarding the student who does not have this indication of economic vulnerability. Furthermore, in the same direction, being white reduces the chance of gap in 30% in relation to non-white students. Although the female presence in school is 48.3% and male presence is 51.7%, the fact of being a boy increases the student’s age/grade gap probability in 59.6%. Results as to gender and skin color/race reiterate numerous studies in the field of sociology of education which have analyzed the perception and differentiated evaluation of teachers in accordance with the students’ gender, as they tend to fail boys more frequently than girls (^{CARVALHO, 2004}) and that indicates that the skin color/race variable predicts a higher fail risk - and consequent age-grade gap - (^{ALVES; ORTIGÃO; FRANCO, 2007}). Also confirming results from other researches previously mentioned on the positive association between attending preschool and school results during elementary school (^{KRAMER, 2006}), the fact that a student attended preschool reduces the chance of academic gap in 43.6%, in relation to students who did not attend preschool.

Finally, as to the topic of the present study, the analysis performed indicated that belonging to a family that declared itself as catholic increases the chance of age/grade gap in 22.2% in relation to students whose family declared itself evangelical, maintaining all other variables as constant. However, belonging to other religious denominations increases the chance of academic gap in 16.6% in relation to students whose families declare to be evangelical. Students whose parents declare to be not religious, on the other hand, increase their chance of age/grade gap in 11.5%, in relation to students coming from families who declare themselves evangelical, maintaining all other variables as constant. These results are consistent with those found by ^{Cunha, Rios-Neto and Oliveira (2014}) in a study involving high school students in some cities in the state of Minas Gerais. In regression analysis, it is possible to compare one of the variables in relation to the other. In the case in question, the logistic regression model enables the comparison between self declared family religious affiliation, in relation to groups whose parents declare themselves evangelical. As previously mentioned, logistic relation does not tell us anything about the variables’ potential to predict (^{FIELD, 2009}). In addition, it is not possible to infer among groups of religious self declared affiliation and those from other religious affiliations among themselves, as ^{Mood (2010}) already pointed out.

These results, albeit fruit of the study of data from a single school system in a single year (2011), indicate the relevance of also considering religious belonging as a potentially intervening variable in academic success construction processes. The analysis brings the probable articulation between religion and other variables - such as family arrangement, skin color/race, gender and school average socioeconomic status - which may favor the development of “virtuous circles” of academic result construction, which deserve to be more densely analyzed in the field of schooling process sociology, and family-school relation analysis. The investigation performed seems to indicate the relevance of considering religious belonging of families and students and their occasional interactions with other individual and school variables (management characteristics, religious belonging in school agents, internal student evaluation criteria and processes, family-school relations, etc.) as study hypotheses that deserve to be developed in an articulate manner in future researches in the educational field.

Final considerations

The study presented indicates that different religious belonging in the families of elementary school students in the Rio de Janeiro municipal school system, among other family, school and social factors are positively associated with schooling paths, confirming previous inferences found in literature - still scarce in the educational field - about the topic. As previously mentioned, given the impossibility to use individual students’ grades in external evaluations aligned with their family characteristics listed in the Rio de Janeiro Municipal Education Department enrollment form, to which we had access, we opted to investigate such relation based on the students’ age-grade gap, which was accessible information in the available data¹¹. Several researches confirm the relevance of using this information as proxy for academic performance (^{FERRÃO et al., 2001}; ^{OLIVEIRA; ARAÚJO, 2005}; ^{OLIVEIRA; SCHWARTZMAN, 2002}; ^{ALVES; ORTIGÃO; FRANCO, 2007}; ^{CORREA; BONAMINO; SOARES, 2014}). These researches, among others, have also guided the selection of other variables used in the current study, notwithstanding the restrictions of data to which we had access.

This study, it should be noted, was conducted in a single - although large, as it includes half a million students - public education system, based on data collected with the families at the time of enrollment in their units, in the year 2011. Starting with the total number of students enrolled in the Rio de Janeiro public system, we reached 557,400 students in regular elementary school, among which only 453,021 students’ parents specified the religion to which they belonged in the registration chart. It seems pertinent replicating this study in other school systems and years, considering, especially, urban realities in other states, with different religious belonging distribution, based on the demographic census performed by IBGE in 2010. It would also be interesting to have more information on the type of religious involvement of families and students, in accordance with their age group and family housing, so as to evaluate better other associations between sociodemographic variables with the issue of belonging to social networks connected to different religious beliefs.

In addition to these limitations, it is important to highlight that, although education literature (^{LEE, 2001}; ^{FERRÃO; BELTRÃO; SANTOS, 2002}) recommends the use of multilevel models, given the hierarchic nature of educational data, as the relevant dimension observed of the effect of school average SES on the results also indicated, it was not possible to gather complementary information about schools that we considered necessary for the present study. We record, therefore, an important limitation we expect to be able to overcome in future research. In this sense, the exploratory nature of this investigation must be registered, as we expect it mobilizes new efforts in the same direction, enlarging the knowledge buildup in the field of sociology of education, about the students’ and their families’ religious belonging.

Everything indicates that considering religious belonging in sociological analyses on the family-school relation may contribute to the increase and renewal of knowledge on the field, favoring a denser understanding of the schooling processes and a more comprehensive look by school agents on the families of the students they teach. The present study was designed to shed a little light on the issue, bringing results expected to be useful to the reflection on the schooling of children and youths in the most economically vulnerable populations.