Family and school social capital, school burnout and academic achievement : a multilevel longitudinal analysis among Finnish pupils

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=rady20 International Journal of Adolescence and Youth ISSN: 0267-3843 (Print) 2164-4527 (Online) Journal homepage: http://www.tandfonline.com/loi/rady20 Family and school social capital, school burnout and academic achievement: a multilevel longitudinal analysis among Finnish pupils P. Lindfors, J. Minkkinen, A. Rimpelä & R. Hotulainen To cite this article: P. Lindfors, J. Minkkinen, A. Rimpelä & R. Hotulainen (2018) Family and school social capital, school burnout and academic achievement: a multilevel longitudinal analysis among Finnish pupils, International Journal of Adolescence and Youth, 23:3, 368-381, DOI: 10.1080/02673843.2017.1389758 To link to this article: https://doi.org/10.1080/02673843.2017.1389758 © 2017 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group Published online: 15 Oct 2017. Submit your article to this journal Article views: 803 View Crossmark data InternatIonal Journal of adolescence and Youth, 2018 Vol. 23, no. 3, 368–381 https://doi.org/10.1080/02673843.2017.1389758 Family and school social capital, school burnout and academic achievement: a multilevel longitudinal analysis among Finnish pupils P.Lindforsa, J.Minkkinena, A.Rimpeläa,b and R.Hotulainenc afaculty of social sciences, health sciences, university of tampere, tampere, finland; bdepartment of adolescent Psychiatry, tampere university hospital, tampere, finland; ccentre for educational assessment, university of helsinki, helsinki, finland ABSTRACT Research on the associations between family and school social capital, school burnout and academic achievement in adolescence is scarce and the results are inconclusive. We examined if family and school social capital at the age of 13 predicts lower school burnout and better academic achievement when graduating at the age of 16. Using data from 4467 Finnish adolescents from 117 schools and 444 classes a three-level multilevel analysis was executed. School social capital, the positive and supportive relationships between students and teachers, predicted lower school burnout and better academic achievement among students. Classmates’ family social capital had also significance for students’ academic achievement. Our results suggest that building school social capital is an important aspect of school health and education policies and practices. Introduction Social capital can be understood as a precondition for healthy social and cognitive development of children and adolescents (e.g. Coleman, 1988, 1990) and it comprises aspects of relationships, networks, norms and trust (Portes, 1988; Schaefer-McDaniel, 2004). For adolescents, the transition from basic education to upper secondary education and training is one of the critical episodes of their life course. During these years, adolescents are becoming increasingly more independent and the demands on scholastic achievements are increased (Danielsen, Samdal, Hetland, & Wold, 2009). Low academic achievement during compulsory education is the strongest predictor of unsuccessful transition to upper secondary education. This, in turn, is a critical predictor of shorter educational and vocational careers overall, of educational dropout, and of lowered well-being in adulthood (e.g. Dufur, Parcel, & Troutman, 2013) The family is a primary context for adolescents’ psycho-social development. The ways in which the parents interact and invest in the relationships with their children may have significant influence on adolescents’ life paths (Dufur et al., 2013; Parcel, Dufur, & Zito, 2010). In terms of learning and achievement, the school may be considered the primary context for educational outcomes, while adolescents’ home environment, socio-economic resources and the content and quality of parent-child interactions affect educational outcomes (Castro et al., 2015; Wilder, 2014). Apart from educational achievements, research ARTICLE HISTORY received 24 august 2017 accepted4 october 2017 KEYWORDS adolescents; social capital; school burnout; academic achievement; multilevel design; longitudinal study © 2017 the author(s). Published by Informa uK limited, trading as taylor & francis Group. this is an open access article distributed under the terms of the creative commons attribution license (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. CONTACT P. lindfors pirjo[email protected] OPEN ACCESS INTERNATIONAL JOURNAL OF ADOLESCENCE AND YOUTH 369 has shown that school context influences students’ development, mental and physical health (Eccles & Roeser, 2011; Gustafsson et al., 2010; Sellstrom & Bremberg, 2006). There is also a strong connection between school performance and indicators for health and well-being during adolescence (Due et al., 2011; Suhrcke & de Paz Nieves, 2011). Thus, there is evidence that home and school contexts have significant impact both on students’ well-being and achievement and that students in better health have higher academic performance. In this study, we investigate whether school and family factors conceptualized as social capital predict better academic achievement and lower school burnout, one indicator of academic well-being, in a longitudinal design. Social capital Social capital as a concept or theoretical framework is widely used throughout social sciences but it is a contested concept, lacking a single, generally accepted definition (Morrow, 1999; Portes, 1988; Waithaka, 2014). The concept of social capital related to children was developed most systematically by Coleman (1988, 1990) and social capital is generally linked to the research on adolescents and children mostly through Coleman’s work (Korkiamäki & Ellonen, 2008; Morrow, 1999). In this study we also follow Coleman’s (1988, 1990) line of thought according to which social capital can be understood as an underlying construct influential in both families and schools. For Coleman, social capital is a characteristic of the social structure and not of each individual within it. Social capital for Coleman is a resource emerging from social structure and ties between persons which can be a significant determinant for adolescents’ well-being and educational achievement (Plagens, 2011). School social capital is considered to manifest in various combinations of relationships between students, teachers and parents and which support academic achievements and has implications for well-being. (Coleman, 1988, 1990; Parcel et al., 2010). Research have found evidence that school social capital is related to better academic achievement (Haghighat, 2005; Tsang, 2010). Research on the associations between school social capital and health outcomes, including academic well-being, is scarce, yet (Virtanen, Ervasti, Oksanen, Kivimäki, Vahtera, 2013). However, in their recent review, Kidger, Araya, Donovan, and Gunnell (2012) found some evidence that students’ sense of connectedness to their schools and their perceptions regarding teacher support has an effect on their emotional health. In this study, we define school social capital as referring to the positive and trustful relationships between students and teachers. Currently multilevel studies on school social capital are scarce and in order to deepen our understanding on the effects of school social capital, a multilevel approach is needed. Family social capital refers to the bonds between parents and children useful in promoting various social outcomes, such as child well-being. These bonds include time and attention which parents spend in interaction with children and in monitoring their activities (Parcel et al., 2010). In Coleman’s theory family social capital encompasses five main components: family structure, quality of parent–child relations, adult’s interest in the child, parents’ monitoring of the child’s activities and obligations of trust and reciprocity and established norms and values in relationships (Ferguson, 2006). In terms of health outcomes, research evidence has shown that social capital measured as a high level of cohesion in family relationships, and parental surveillance and interactions with their children predicted better mental health in adolescents (Virtanen et al., 2013). Ferguson’s (2006) review of family social capital found that intensive social interactions decreased adolescents’ likelihood of dropping out of school. Also high levels of parental monitoring of children’s’ activities are shown to be associated with positive outcomes in academic performance and higher levels of psychological adjustment. (Coleman & Hoffer, 1987; Voydanoff & Donnelly, 1999). The meta-analysis by Wilder (2014) indicated that the relationship between parental involvement and academic achievement was positive and consistent across different grade levels and ethnic groups. However, when parents’ monitoring of schoolwork is perceived as too controlling (such as practices characterized by pressure, intrusiveness), even well-intended monitoring has been found to be associated with lowered motivation and achievement (Gonida & Cortina, 2014). There are several studies, which have indicated that higher social capital, either at school, at home, or both, is related to better academic achievement (Dufur et al., 2013; Parcel et al., 2010). In terms of 370 P. LINDFORS ET AL. health outcomes, Eriksson, Hochwälder, Carlsund, and Sellström (2012)) found that social capital in the family, school and neighborhood predicted lower levels of health complaints and higher levels of well-being among children aged 11–15years. In their review article on social capital and health, Morgan and Haglund (2009) showed that social capital at home, school and neighborhood mattered for adolescent health and health-related outcomes. For the present, multilevel studies of social capital among adolescents are scarce and most studies have focused on the individual-level social capital in various contexts. Our study contributes to this gap of knowledge by using a multilevel approach and exploring the effects of both student, class and school-level social capital on adolescents’ academic well-being and academic achievement. A multilevel approach enables the examining of contextual social capital’s influence above students’ individual effects. We consider class and school-level social capital as potential collective resources for adolescents beyond their individual experiences of oneto-one interactions. School burnout and academic well-being Students’ academic well-being is an important indicator of the educational process (Rueger, Malecki, & Demaray, 2010; Tuominen-Soini, Salmela-Aro, & Niemivirta, 2012). However, there is no consensus regarding either the definition or operationalization of academic well-being. In the present study, we focus on school burnout as one essential indicator of academic well-being. Research has shown that both positive and negative emotional school engagement have remarkable interconnections with academic and psychological functioning (Li, Lerner, & Lerner, 2010; Li & Lerner, 2011; Wang & Degol, 2014; Wang & Fredricks, 2014; Wang, Chow, Hofkens, & Salmela-Aro, 2015). School burnout is a concept related to school ill-being. Burnout was originally regarded as a work-related disorder (Maslach, Schaufeli, & Leiter, 2001), but the concept has recently been found to be useful and transferrable to the school setting (Kiuru, Aunola, Nurmi, Leskinen, & Salmela-Aro, 2008; Salmela-Aro, Kiuru, Leskinen, & Nurmi, 2009; Walburg, 2014). School-related burnout can be defined as a combination of exhaustion at schoolwork, cynicism toward the meaning of school, and sense of inadequacy as a student. It can be caused by discrepancies between student’s internal resources, school workload and expectations of school results which can all be considered critical signs of emotional disengagement and which are related to low academic achievement (Kiuru et al., 2008; Wang et al., 2015). Academic achievement Academic achievement builds on interaction between many variables, such as subjects’ characteristics, both cognitive and non-cognitive (personality traits, self-perceptions etc.), classroom practices (teacher-student interaction), and contextual variables (home and community context) (Hattie, 2009). Research has shown that differences between neighborhoods (home and community context) affect school achievement causing betweenschool variations. Meta-analyses of school effects on achievement outcomes have produced varied results (cf. Sellstrom & Bremberg, 2006). School effect (controlled for SES) is around 8% (Scheerens & Bosker, 1997), and the classroom effect begins at 16% and reaches 60% (Alton-Lee, 2003). Hattie (2009) cited various meta-analyses emphasizing within-school effects over between-school effects while first, teacher-student relationship has shown to have a strong effect on academic achievement and second, class composition seems to be an efficient solution to meeting the differing needs of individual students. In Finland, the school-level effect is one of the lowest among the OECD countries, accounting for only around 5% to 8% according to the PISA results (OECD, 2008). However, between-class differences are higher when compared to other Nordic countries (Yang Hansen, Gustafsson, & Rosén, (2014). INTERNATIONAL JOURNAL OF ADOLESCENCE AND YOUTH 371 Aims Considering the centrality of home and school in the lives of adolescents and the importance of social capital for academic well-being and academic achievement, it is important to study both of these outcomes simultaneously. Research has shown that academic well-being and academic performance are intertwined: good academic performance is positively related to schoolwork engagement (Salmela-Aro & Upadaya, 2012) and negatively related to school burnout (Salmela-Aro et al., 2009). So far, longitudinal and multilevel studies on associations between school and family social capital, school burnout and academic achievement are limited. Since research has shown between-school variance in students’ health and education-related outcomes, we are especially interested in whether there is a school or class-level impact of social capital on the outcomes. We address the following research questions: (1) Do school and family social capital, at the beginning of lower secondary school at the age of 13, predict school burnout and academic achievement at the age of 16? (2) Do associations between (a) family and school social capital and school burnout and (b) family and school social capital and academic achievement hold true when earlier burnout, academic achievement and parents’ education are controlled for? Methods The present study was based on three school surveys conducted in 2011 and 2014 and on the register data from the Finnish post-compulsory education application register in 2014 (Table 1). The datasets were linked together. Two baseline surveys conducted in 2011 (health and learning surveys in the seventh grade) and one follow-up in 2014 (health survey in the ninth grade) in the Helsinki metropolitan area of Finland which spans 14 municipalities. All seventh-graders (ages 12–13) were invited to participate in the baseline surveys in 2011 (N=13,012). The recruitment occurred through the educational authorities of the municipalities, each of which gave permission for the study. The Ethical Committee of the National Institute of Health and Welfare approved the protocol. Because the study was part of normal schoolwork, parental consent was not required. However, two of the municipalities obliged parental consent statements, which were collected. An informational letter was delivered to parents in the remaining 12 municipalities. Of the recruited seventh graders, 9497 participated in the health survey (response rate of 73%) and 10,917 in the learning-to-learn survey (response rate of 84%) in 2011. Special schools and classes for children with serious learning difficulties, intellectual disabilities or those situated in pediatric hospital wards were excluded from the sample because of the students’ expected difficulty with answering the questions. Five schools from the city of Helsinki did not participate, and two of the schools had computer classes under construction. One school did not receive the individual passwords in time, and two administratively independent schools were not interested in participating. The other non-respondents included those absent from school on the survey day (typically 10–15% of students each day) and those who refused to participate or whose parents’ consent statement was negative or not received. The participating schools did not have information about whether the pupil had refused to Table 1.data in the study. 2011 2014 2014 Baseline surveys: follow-up survey: Post-compulsory education application register datalearning-to-learn assessment health survey health survey autumn term of 7th grade (age 12–13) end of 9th grade, end of compulsory basic education (age 15–16) students whose grades were available in the register and who had answered all three 3 surveys N=5583students who answered both questionnaires N=9079 students who had answered all three questionnaires N=5742 372 P. LINDFORS ET AL. participate or was absent from school. A follow-up health survey was conducted in spring 2014, at the end of lower secondary school (the ninth grade). The follow-up health survey was conducted applying the same procedure as the health survey of 2011. The data was gathered as part of school routine. The participants completed health surveys online in computer classrooms. The learning-to-learn assessment was conducted using paper and pencil. A total of 5742 students participated in the three surveys. Questionnaires with missing information for class specification were excluded. Classes and schools with less than five participants were excluded in order to eliminate possible bias for aggregated class and school-level variables. Participants who had the complete data in the analysis variables and whose school leaving certificates were available in the register were included in the analyses. The final study population consists of 4467 students (50.8% girls), from 444 classrooms and 117 schools. The number of students from each class varied between 5 and 22 students (M=11.62, SD=3.88) and from each school between 5 and 132 students (M=55.02, SD=27.31). Non-response analysis Students in the multilevel analyses (N=4467) were compared to those 9079 students who completed the two questionnaires in the seventh grade, using Chi-square statistics or the Mann-Whitney U-test, as appropriate, to check to what extent the final sample represents the study population. No significant difference was found in gender distribution, family education, or school burnout in the ninth grade. However, the students in the final had slightly lower seventh-grade school burnout than the whole sample of respondents who completed the questionnaires in the seventh grade (U=19,014,091, r=.21, p<.01). Students’ academic achievement was somewhat higher in the final sample both in the seventh and ninth grade than among those who completed the questionnaires in the seventh grade (respectively, U=17,840,861, r=.21, p<.001; U=18,217,715.5, r=.21, p<.001). Moreover, students in the multilevel analyses had slightly higher seventh-grade school social capital and family social capital than the whole sample of students in the seventh grade (respectively, U=17,721,542.5, r=.22, p<.05; U=18,630,680, r=.22, p<.01). In sum, the final research population had somewhat better academic achievement and higher school and family social capital than the original cohort but the differences were small according to the rank correlations. Measures Following Coleman’s line of thought (1988, 1990), we measure both family and school social capital. School social capital School social capital in the seventh grade was measured by nine items (e.g. ‘I feel that my teachers appreciate me’, ‘I feel that our teachers treat students fairly’, ‘I feel that teachers accept me as I am’, ‘I feel that our teachers acknowledge and respect student’s own opinions’, ‘I usually get along well with my teachers’). The scale was 1–7 (1=not true, 7=very true). Item scores were summed to a total score. Higher scores reflected greater school social capital. The reliability of the sum variable was very good (α=.932). (See Table 2). Family social capital In Coleman’s concept, parents’ monitoring of the child’s activities is one of the dimensions of family social capital. We use the scale by Brown, Mounts, Lamborn, and Steinberg (1993) on adolescent’s behavioral control. A five-item scale on how much the adolescents think their parents ‘really know’ about their activities measured family social capital: ‘Who my friends are’, ‘Where I am after school’ ‘Where I go at night’, ‘How much money I spend’ and ‘Where I am most afternoons after school’. Each INTERNATIONAL JOURNAL OF ADOLESCENCE AND YOUTH 373 subscale comprised three options, which were assessed using a 3-point scale (2=a parent knows well, 1=knows quite well, 0=does not know at all). The same five items were assessed for mother and father separately. A total score was composed of ten items (five items for each parent), with the total score ranging from 0 to 20. Higher scores reflected greater family social capital. The reliability of the sum variable was good (α=.892). School burnout The first dependent variable, school burnout in the ninth grade, was assessed by using the School Burnout Inventory (SBI) developed by Salmela-Aro and colleagues (for validity and reliability, see Salmela-Aro et al., 2009). The inventory consists of three subscales: exhaustion at school (e.g. ‘I feel overwhelmed by my schoolwork’), cynicism toward the meaning of school (e.g. ‘I’m continually wondering whether my schoolwork has any meaning’), and sense of inadequacy as a student (e.g. ‘I often have feelings of inadequacy in my schoolwork’). Each subscale comprised three items, which were assessed using a 6-point Likert-type scale ranging from 1 (Completely disagree) to 6 (Completely agree). A composite score was constructed using nine items, with the total score ranging from 9 to 45 and showing good reliability of the sum variable (α=.886). School burnout in the seventh grade was composed in the same way (α=.880) and used as a controlling variable. The variable of school burnout in the ninth grade was available on 5219 students. Academic achievement The second dependent variable, academic achievement in the school-leaving certificate, was examined using a mean of the sum of the grades in three subjects: mother tongue, mathematics, and foreign language (starting in grades 1–3). The grade scale is 4–10 (4=fail, 10=excellent). The grades were obtained from the Finnish post-compulsory education application register system at the end of students’ upper secondary education 2014 and when they were not available, from the ninth grade survey. The register contained the grades of those 5742 students who had answered all three questionnaires as follows: mother tongue (N=5583), mathematics (N=5583), and foreign language (N=5559). The national register of the joint application system entails only the grades of those students who applied into general upper secondary or vocational upper secondary education, but lacks the grades of those students who applied into specialized vocational education and training institutions, to the 10th grade in lower secondary education, to pre-vocational education, or did not apply. Academic achievement reported by students in the seventh grade was used as a controlling variable. Previous research has detected excellent validity of the self-reported grades in Finland (Kupiainen, Vainikainen, Marjanen, & Hautamäki, 2014). Seventh grade academic achievement was compared to academic achievement in the ninth grade using a mean of the sum of the grades in three subjects (Table 2). Table 2.descriptive statistics (N=4467). aacademic achievement in the ninth grade. higher scores are indicative of higher levels of variables. Mean SD Range school burnout (7th grade) 21.506 7.858 9–45 school burnout (9th grade) 24.714 7.556 9–45 academic achievement (7th grade) 8.325 .856 4–10 academic achievement (9th grade)a8.133 1.058 4–10 school social capital 46.420 9.865 9–63 family social capital 15.232 4.167 0–20 374 P. LINDFORS ET AL. Family education Family education was used as a background variable. Family education was asked separately for mothers and fathers and the highest education level for each parent was included in the analysis. The question was ‘What kind of education do your parents have?’ The options were: basic education only, vocational upper secondary education or vocational college, matriculation examination certificate and vocational college, university degree, no mother/father. University degree was encoded as 1, other options as 0. No mother and father was coded as a missing value. We used all three surveys to form the variable of family education starting from the ninth grade health survey and complementing missing values from other surveys when possible. A total of 38.6 percent of students had at least one parent with university-level education, reflecting a high education level in the population of the Helsinki Metropolitan area. The classand school-level variables The classand school-level variables were aggregated variables taking the classand school-level means of the student-level scores. Data analysis Multilevel modelling was applied, as the method allowed us to examine student-, classand schoollevel effects simultaneously by way of splitting the variance of the observed variables into the variance components for each level (Heck & Thomas, 2009). Linear three-level analyses were conducted using the Mplus statistical package (version 8; Muthen & Muthen, 1998–2012). Maximum likelihood estimation with non-normality robust standard errors was applied (MLR estimator; Muthen & Muthen, 1998–2012). Two-tailed significance testing at the criterion level of p=.05 was used. Intra-class correlations (ICC) of each predictor were calculated by using class and school as clustering variables in order to ascertain the percentage of the total variance at each level (see Heck & Thomas, 2009). If the variance component at the student-level variable was statistically significant at the cluster levels, a corresponding cluster variable (an aggregated variable) was included into the multilevel analyses (Table 3). According to this procedure, all student-level variables in the analyses were also analyzed at the class and school levels, taking the mean of the student score within each cluster. In model 1, variables of school and family social capital were included into the model at the student, class and school level. Analyses were executed separately for school burnout (Model 1a) and academic achievement (Model 1b). In order to analyze the power of contextual effects, standardized student-level coefficients were compared to the corresponding classand school-level estimates using z-scores Table 3.Intra-class correlations (Icc) and Variance estimates at the student, class and school levels (standard errors in parentheses). aacademic achievement in the ninth grade. *p<.05; **p<.01; ***p<.001. Variables ICC Class ICC School Within-group variance, student level (SE) Between-class variance (SE) Between-school variance (SE) school burnout (7th grade) .048 .013 48.321 (1.357)*** 2.487 (.540)*** .688 (.321)* school burnout (9th grade) .029 .016 54.723 (1.335)*** 1.645 (.472)*** . 925 (.325)** academic achievement (7th grade) .109 .050 5.830 (.203)*** .762 (.111)*** .343 (.095)*** academic achievement (9th grade)a .092 .086 8.629 (.267)*** .968 (.138)*** .901 (.220)*** school social capital .072 .058 86.591 (2.702)*** 7.216 (1.201)*** 5.789 (1.562)*** family social capital .028 .025 16.761 (.453)*** .488 (.152)** .438 (.116)*** family education .038 .056 .213 (.003)*** .009 (.002)*** .013 (.003)*** INTERNATIONAL JOURNAL OF ADOLESCENCE AND YOUTH 375 (Muthen and Muthen, 1998–2012; see: Paternoster, Brame, Mazerolle, & Piquero, 1998). We utilized group-mean centering by class for the student-level predictors, group-mean centering by school for the class-level predictors and grand-mean centering for the school-level predictors as is recommended for research questions where the effects of the student-level predictors and the corresponding higher level predictors are compared (Enders & Tofighi, 2007). In Model 2, two dependent variables, school burnout and academic achievement, were simultaneously added into analysis in order to control their association. Also, the controlling variables of school burnout and academic achievement in the seventh grade and parents’ education were added into Model 2. Results Students’ school burnout and academic achievement varied significantly across classes and schools in the ninth grade (Table 3). ICC of school burnout was .029 at the class level and .016 at the school level, indicating that 2.9 percent of the variation occurred across classes and 1.6 percent across schools. 9.2 percent of the variation of academic achievement occurred across classes and 8.6 percent across schools. Significant variance was also found across classes and schools in the variables of school social capital, family social capital, school burnout and academic achievement in the seventh grade, and parents’ education. The student-, classand school-level effects of school and family social capital were examined in Model 1a and 1b (Table 4). The dependent variable was school burnout in the ninth grade in Model 1a and the coefficient of determination (R2) was 4.7% at the student level, 20.5% at the class level, and 25.1% at the school level. In Model 1b, the dependent variable was academic achievement in the school leaving certificate and R2 was 3.9% at the student level, 21.0% at the class level, and 45.5% at the school level. School social capital in the seventh grade significantly predicted lower level of school burnout in the ninth grade (Model 1a) and better academic achievement in the school leaving certificate (Model Table 4.Multilevel linear Modeling for school Burnout (Model 1a) and academic achievement (Model 1b) (Nstudent level=4467, Nclass level=444, Nschool level=117). aclass mean. bschool mean. *p<.05; **p<.01; ***p<.001. Parameters Model 1a Model 1b B SE β SE B SE β SE Intercept 24.647*** .158 22.589*** 3.389 24.423*** .093 25.618*** 2.901 Regression coefficients student level, 7th grade school social capital −.147*** .013 −.176*** .015 .053*** .006 .162*** .018 family social capital −.155*** .032 −.081*** .017 .052*** .013 .072*** .018 class level, 7th grade school social capitala−.116* .046 −.306** .115 .058** .018 .233** .066 family social capitala−.235* .117 −.227* .112 .212*** .049 .306*** .064 school level, 7th grade school social capitalb−.131* .061 −.404* .186 .145*** .036 .521*** .125 family social capitalb−.174 .170 −.167 .164 .226* .104 .256* .105 Variance components student level residual variances 51.618*** 1.372 .953*** .007 8.140*** .254 .961*** .007 class level residual variances 1.464*** .395 .975*** .092 .686*** .129 .790*** .052 school level residual variances .891* .354 .749*** .156 .495** .153 .545*** .118 R2 R2 (student level) .047 .039 R2 (class level) .205 .210 R2 (school level) .251 .455 Model fit information loglikelihood h0 Value −15224.061 −11643.379 deviance (BIc) 30,532.166  23,371.183 