School belonging and school misconduct. The differing role of teacher and peer attachment. more

J Youth Adolescence DOI 10.1007/s10964-011-9674-2 EMPIRICAL RESEARCH School Belonging and School Misconduct: The Differing Role of Teacher and Peer Attachment Jannick Demanet • Mieke Van Houtte Received: 25 February 2011 / Accepted: 3 May 2011 Ó Springer Science+Business Media, LLC 2011 Abstract The schools-as-communities perspective provides a popular explanation for school-disruptive behavior, stating that interpersonal bonding at school and feelings of school belonging prevent misconduct. In this article, we build on this perspective in three ways. First, we test whether the preventive influence of school belonging acts at the individual or school level. Secondly, we test whether a distinction should be made between the different actors with whom students bond at school, by assessing whether perceived teacher support, school belonging, and peer attachment relate differently to school misconduct. Lastly, the present study investigates whether the associations of bonding with teachers, peers and the school with school misconduct differ by socio-ethnic school context. Multilevel analyses were performed on data from the Flemish Educational Assessment. The sample consisted of 11,872 students (51.4% female) in 85 schools, most of whom were natives (88.8%), with immigrants (11.2%) mostly having Turkish or Moroccan backgrounds (both about 30% of immigrants in the sample), and others Southern-European (16%), Eastern-European (8%), North-African (5%), or other (17%) backgrounds. Results showed that the students’ individual feelings of bonding with peers, teachers and school associate with school misconduct, rather than the overall school cohesion. Results further showed that, while higher perceived teacher support and school belonging related to less school misconduct, higher peer attachment was associated with higher rates of school J. Demanet (&) Á M. Van Houtte Department of Sociology, Research Team CuDOS, Ghent University, Korte Meer 3-5, 9000 Ghent, Belgium e-mail: jannick.demanet@ugent.be M. Van Houtte e-mail: mieke.vanhoutte@ugent.be misconduct. No differences were found by socio-ethnic context. Implications are discussed. Keywords Schools-as-communities perspective Á School misconduct Á Peer attachment Á Teacher support Á School belonging Á School composition Introduction A recurrent theme within educational and adolescent research is the explanation of school misbehavior (e.g. Freidenfelt Liljeberg et al. 2011). A popular explanation revolves around the schools-as-communities perspective (Battistich et al. 1995), a broad line of inquiry advocating that schools should be organized as caring school communities. In such caring communities, students feel emotionally connected to their peers, teachers, and school, which, among other beneficial outcomes, fosters less school misconduct in students (Battistich and Hom 1997). However, researchers are still unsure whether the preventive effects of belonging act at the individual or the school level. While some scholars state that the beneficial effects of school belonging result from students’ personal feelings (Wehlage et al. 1990; Goodenow 1993), others hold that, in order to combat school deviancy, efforts should be made to establish cohesion between actors at the school level (Bryk and Driscoll 1988; Battistich et al. 1995). However, few studies have tested specifically whether the effects of school belonging act at the individual or the school level. Therefore, the first aim of this study is to assess the relative contribution of feelings of belonging at the school level—that is school cohesion— and at the individual level in preventing school misconduct. 123 J Youth Adolescence Researchers from the schools-as-communities perspective also state that attachment to different actors at school and the school itself impedes school deviancy. However, while studies have established that feelings of school belonging (Dornbusch et al. 2001) and teacher attachment (Freidenfelt Liljeberg et al. 2011) are associated with lower rates of deviant behavior, research in this tradition has yet to investigate whether attachment to peers is related in the same way to school misconduct. The schools-as-communities perspective currently draws, in regard to its expectations concerning deviancy, on insights from early social control theory (Hirschi 1969), a theoretical approach that expects all strong social bonds to prevent delinquency. However, this theoretical approach failed to address the role of deviant peer influences (Erickson et al. 2000). In fact, research shows that peers can cause each other to break school rules, especially in the case of cohesive friendship bonds between deviant peers (Kandel 1978; Wellman and Frank 2001; Espelage et al. 2003). It is important to consider these different sources of attachment—school, teachers, and peers—together, to gain insight in the unique role played by each in preventing or promoting school misconduct. The few studies that have incorporated both teachers and peers as sources of support relate this to achievement (Klem and Connell 2004), smoking (Karcher and Finn 2005), and other health-risk behaviors (McNeely and Falci 2004). These find higher teacher support to advance achievement and impede health-risk behaviors, and higher peer support to yield a higher likelihood of smoking and engaging in other healthrisk behavior. In our study, a specific aim is to incorporate peer attachment, teacher support and general school belonging and test their relative contribution to school deviancy. Contrary to earlier studies on school deviancy, which deal with student delinquency (e.g. Crosnoe 2002), we focus on school misconduct, a minor form of deviancy, consisting of rule-breaking behavior such as cheating on tests, skipping lessons, and arriving late at school (e.g., Stewart 2003; Demanet and Van Houtte 2011), as we can imagine that peer and teacher attachment is more likely to influence minor forms of rule-breaking behavior than delinquency at school. Hence, we operationalize feelings of school belonging multi-dimensionally, discerning three aspects—general school belonging, peer attachment and perceived teacher support—to test whether these three dimensions relate differently to school misconduct. In a third contribution to the schools-as-communities perspective, we embed all this within the social context of schools. Researchers state that receiving support prevents school deviancy even more effectively in disadvantaged schools (Battistich et al. 1995), but evidence concerning this is mixed. In the current study, we focus on two indicators of disadvantage in schools, namely the Socioeconomic Status (SES) composition and the ethnic composition (Demanet and Van Houtte 2011). Hence, our third research question is whether the relationships between general school belonging, peer attachment and perceived teacher support are stronger in schools with a lower SES composition, and schools where immigrants are overrepresented. As such, our study tries to find evidence for a long-held claim that supporting students in disadvantaged schools has even more beneficial effects, because these students often lack such sources of support outside the school context. The Schools-As-Communities Perspective The schools-as-communities perspective is a popular viewpoint in educational research. Barber and Olsen (1997) emphasized three aspects of adolescent socialization: connection to significant others, regulation of behavior, and psychological autonomy. The schools-as-communities perspective focuses on the first, stating that schools fulfill their socialization function best when organized as caring communities (Battistich et al. 1995). Such school communities are defined in diverging ways, but scholars put forward common elements (see Battistich et al. 1997, p. 137). Broadly stated, communal schools make students feel emotionally connected to one another—i.e., they feel attached (see Libbey 2004, p. 274)—and feel respected and helped by their peers and teachers—in other words, they perceive themselves as supported (see Libbey 2004, p. 281). In communal schools, students are made to feel at home at school. Furthermore, students in caring school communities feel that they make important contributions: hence, they are given a certain amount of influence in the school’s activities and decision-making process. Lastly, in such schools, there is some sort of common value system (Battistich et al. 1997). Communal schools yield a wide array of positive effects in their students, including higher school enjoyment, academic achievement, and less school disruption (Battistich et al. 1995; Battistich and Hom 1997). The preventive effect of school bonding on deviancy has been replicated by many studies. Dornbusch et al. (2001) showed in a longitudinal study that school attachment reduces the overall frequency, prevalence, and initiation of deviant involvement, and that this association held across males and females, different community contexts and regardless of ethnic groups. Another longitudinal study distinguished between three forms of student engagement at school—emotional, behavioral, and cognitive engagement—confirming that emotional engagement had strong preventive effects on the occurrence of delinquency (Hirschfield and Gasper 2011). In an influential study, Finn (1989) proposed the participation-identification model, 123 J Youth Adolescence which highlights the role of identification with and participation in school in preventing school dropout, and, as dropout is linked to other problem behavior (see Finn 1989, p. 118), also school misconduct. The preventive effect of establishing strong social and emotional connections at school on student deviancy is thus well-established in research. There is some confusion as to whether the effects of belonging should be seen as resulting from students’ personal feelings—an individual-level effect (Wehlage et al. 1990; Goodenow 1993)—or as an effect of the school—a group-level effect (Bryk and Driscoll 1988; Battistich et al. 1995). The first position argues that misconduct is an individual reaction to a lack of belonging, hence, efforts should be directed at making disruptive students feel supported at school. The second view considers communality to be a group level concept, albeit with roots in individual feelings (Battistich et al. 1995). This side states that students have basic needs for belonging that the school should be equipped to fulfill, making misconduct a failure of the school. However, few researchers have tested specifically whether the preventive effect of feelings of school belonging on students’ deviancy is due to students’ personal feelings of belonging or their being part of a school that has a certain level of communality. Some studies have shown that it is important to account for the multilevel nature of belonging at school. One study finds that the quality of teacher-student relationships in a school can counterbalance negative effects of the school climate— seen as school-wide bullying—on the academic achievement of students (Konishi et al. 2010). Another study investigated school-level influences on students’ school belonging, finding that students’ sense of belonging differs more within schools than between schools, and is affected particularly by the students’ self-esteem (Ma 2003). Hence, it could be argued that the sense of belonging is something individual, and, thus, that trying to establish a caring school community does little to impact students. However, no study has investigated the relative effects of school cohesion—school belonging situated at the school level—versus the individual characteristics of peer attachment, teacher support, and general school belonging—situated at the student level—on students’ deviancy. Thus, while the multilevel nature of the sense of belonging has been widely acknowledged, there is still no clear answer to whether the effects of school belonging result from students’ personal feelings or from the school as a whole. The Role of Peer Attachment in Adolescence Although the schools-as-communities perspective holds that all strong social bonds prevent school misconduct, there are reasons to think that peer bonds in adolescence are unlikely to have the same effect. In stating that bonding prevents disruptive behavior, the schools-as-communities perspective echoes the premises of the earliest version of control theory (Hirschi 1969). This theory holds that all individuals are inclined to deviancy, but having strong social bonds to others prevents its manifestation. Hirschi (1969) applied his view to adolescents, asserting that they must have strong bonds with school, parents and peers in order to behave properly. Empirical research based on this theory, however, has shown that this earliest version of control theory has some important flaws (Erickson et al. 2000). Most importantly, it has neglected the characteristics of the actor with whom one should bond. Applied to peer relationships, early control theory is unable to cope with findings that friends often share a comparable level of deviancy (Bendixen et al. 2006). This deviance ‘‘homophily’’ (see Espelage et al. 2003) has been explained by means of two processes: selection and influence (Kandel 1978). Deviant students tend to choose each other as friends, but they also influence each other to commit deviant acts (Bendixen et al. 2006). It is noteworthy that influence occurs especially when friendship bonds with deviant peers are quite cohesive. Deviant peer influence has been explained by differential association theory (Sutherland and Cressey 1978), which states that deviancy is learned from others who support deviancy as justifiable behavior. Therefore, having close bonds with others will increase the chances of being deviant, that is, when these others endorse deviancy. It is likely that antisocial values will prevail in adolescent peer groups. Research indeed shows that peer norms in adolescence tend to favor minor forms of deviancy (Moffitt 1993; Agnew 2003; Allen et al. 2005). Authors argue that while youngsters are biologically mature, they are not allowed to fulfill social roles with complete adult privileges and responsibilities, creating a maturity gap. Adolescents thus have been viewed as endorsing deviancy as a way to show autonomy. As students are socialized into the norms held by the majority of their peers (Allen et al. 2005), and socialization especially occurs in cohesive friendship relationships (Sutherland and Cressey 1978), higher peer attachment in adolescence can be conducive to school misconduct. In the schools-ascommunities perspective, then, it may be essential to consider the possible deviance-yielding effect of cohesive bonds between youngsters. The Context of Communality Effects Researchers suggest that bonding impedes school misconduct most in disadvantaged schools (Battistich et al. 1995; Battistich and Hom 1997). This is not to say that school misconduct is more prevalent there, only that 123 J Youth Adolescence supporting students in such schools makes a larger difference. This occurs because advantaged students are likely to have sources of support outside of school (Battistich et al. 1995). For disadvantaged students, external sources of support are less prevalent (Stanton-Salazar and Dornbush 1995). Therefore, for them, receiving support at school has a major impact. Testing this, Battistich et al. (1995) confirmed that belonging had more positive effects on school and class enjoyment in poorer schools. Battistich and Hom (1997), however, showed that this does not apply to school delinquency, as feelings of communality were more preventive in low and moderate poverty schools than in high poverty ones. At the neighborhood level, scholars have found no evidence for this assertion (Dornbusch et al. 2001). Hence, scholars have provided mixed evidence to the claim that enhancing school belonging should be even more beneficial in poorer schools. Next to poverty, a school’s concentration of different ethnic groups also can create a disadvantaged situation in that school. It has been shown that students achieve less in schools with a high concentration of migrant or ethnic minority students (Felouzis 2003; Rumberger and Palardy 2005). Moreover, teachers in those schools report job dissatisfaction, lower academic expectations, and more difficulty establishing relationships with students (Freeman et al. 1999). Importantly, some research has demonstrated that these disadvantages are entirely due to the poor socioeconomic context of these schools (e.g. Rumberger and Palardy 2005). Ultimately, students in such schools perceive the school as more disadvantaged, and expect less from their academic career (Bankston and Caldas 1996). This is especially the case in Flanders. In present-day Flanders, migrants and their descendants are generally referred to as ‘‘allochthonous’’, a term pertaining to persons residing in Belgium, regardless of their nationality, having at least one parent or grandparent born outside West Europe, and holding a disadvantaged position in society because of their ethnicity (Brans et al. 2004). Their situation resembles that of the black minorities in the US, typified by Ogbu (1978) as caste-like minorities. The second- and third-generation in particular—who were born in Belgium—did not choose to be here. They represent a cultural model characterized by oppositional identity, distrust of society and negative perceptions of the opportunity structure (Hermans 2004). Hence, in Flanders, most students in ethnically concentrated schools do not perceive upward mobility to be likely, which characterizes these schools as disadvantaged. Hence, in line with the claim presented above, it could be argued that supporting students should be even more preventive of school misconduct in schools where ethnic minority students are overrepresented. The Current Study The purpose of this study is to expand upon the schools-ascommunities perspective. Using a multilevel framework, we consider three research questions. To investigate possible mediation or selection effects, we use sequential multilevel analyses. First, as few researchers have investigated whether the effects of school belonging constitute a group-level effect or an individual-level effect, we will test the relative effect of school cohesion (a school characteristic) and students’ feelings of peer attachment, perceived teacher support and general school belonging (individual characteristics), on school misconduct. In the analyses, we will investigate the role of school cohesion in the first model, to determine whether it has an effect on the dependent variable. In a second and third model, we add the three individual-level measures of belonging, to test whether the school effect of cohesion is maintained. Our second research question is whether peer attachment, perceived teacher support and general school belonging relate differently to school misconduct. Although researchers from the schools-as-communities perspective (Battistich et al. 1997) assume that all sources of bonding prevent school misconduct, based on our theoretical exploration, we expect that peer attachment is related in a different manner to school misconduct than perceived teacher support and general school belonging are. To investigate this, we add peer attachment in our second model, followed by perceived teacher support and general school belonging in the third model. As a third research question, we consider the claim made by researchers (Battistich and Hom 1997) that sources of support prevent school deviancy more effectively in a disadvantaged school setting. In this regard, we consider the schools’ SES and ethnic composition as indicators of school disadvantage. In our models, we test this by adding cross-level interaction effects between the three bonding measures and SES and ethnic composition in our fourth model. This enables us to test whether the relationships between peer attachment, perceived teacher support, general school belonging, and school misconduct vary by the SES and the ethnic composition of the schools. Methods Sample The data were part of the FlEA (Flemish Educational Assessment), gathered in the 2004–2005 school year in Flemish secondary schools. Our sample consisted of 11,872 students across 85 Flemish schools. Flanders is the northern, Dutch-speaking part of Belgium, and constitutes a political entity with its own parliament and government. 123 J Youth Adolescence Since 1988, the Flemish government has gained the jurisdiction to implement and govern its own educational system, which limits the study to the students and schools in this region. Of the respondents in the sample, 51.20% attended the third grade, and 48.8% the fifth grade (corresponding to grades 9 and 11 in the US system). Hence, the majority of students are 15 (34.8%) or 17 (32.6%) years old, with other students being a bit older than most students in their grade, mostly due to being held back a grade because they did not succeed in a previous year (11.3% being 16 years old; 14.3% 18 years old; 4.6% 19 years old, and 1.4% being 20 years old). The sample was equally divided by gender (51.4% girls; see Table 1). The majority of respondents were natives (88.8%). Most immigrants (11.2%) had Turkish or Moroccan backgrounds (both about 30% of immigrants in our sample), some had SouthernEuropean (16%), Eastern-European (8%), North-African (5%), or other (17%) backgrounds. Family occupational background (SES, see below) was unskilled manual labor for 7.9% of students, specialized manual labor for 6.3%, skilled manual labor for 6.3%, routine non-manual employees for 11.2%, farmers and smallholders for 9.3%, lower grade employees and administrators for 23.6%, higher-grade administrators and executives for 17.6%, and professionals and large proprietors for 11.5%. Most respondents attended the general track (46.7%), with 28.5% attending the technical, 22% the vocational, and 2.7% the arts track. Schools were equally divided across school sector (50.6% public, 49.4% private), and had an average school size of 461.55 (SD = 285.27). For the sampling procedure, we used multistage sampling. At first, we selected proportional-to-size postal codes, size being defined by the number of schools within each postal code, information provided by the Educational Department. From the 240 postal codes, we selected a random sample of 48 codes. Because of this strategy, postal codes of large municipalities had a greater chance of selection. This strategy was used to capture the fact that larger municipalities have a greater number of schools and a greater number of migrants: we had to make sure that—in correspondence with their overrepresentation in the Flemish context—schools in larger municipalities had a higher chance to be selected, making sure at the same time that we selected a critical amount of schools with a majority of migrant students. The sampling resulted in the desired overrepresentation of larger municipalities. Consequently, we selected all regular secondary schools in the chosen postal codes that provided a third and fifth grade Table 1 Descriptive statistics for variables: frequencies (%), means (M), standard deviations (SD), Cronbach’s alpha, and N Variables Dependent variable School misconduct School level Cohesion Ethnic composition SES composition School sector Public School size Student level Peer attachment Perceived teacher support General school belonging Gender Female Grade Third Ethnicity Immigrant SES Parental attachment Vocational track Vocational Prior achievement % Mean SD Cronbach’s alpha N 30.04 61.24 16.45 4.80 50.60 461.55 15.81 23.99 37.11 51.40 51.20 11.20 5.20 28.32 22.00 69.42 8.47 2.81 21.70 1.23 0.87 11,561 85 85 85 85 83 285.27 2.76 3.99 6.42 0.74 0.75 0.80 11,554 11,621 11,543 11,843 11,872 11,870 2.10 5.61 0.83 11,137 11,722 11,872 10,713 9.22 123 J Youth Adolescence (corresponding to years 9 and 11 in the US system), yielding a response rate of 31%. This low response rate is due to schools in Flanders being swamped with research requests. Schools choose the research they take part in on a first-come, first-served basis. Analyses in which we compared our sample to the Flemish school population, based on information attained through the Flemish Educational Department, showed that the participating schools did not differ from those that opted out in terms of school sector, size, curriculum, or student composition. Hence, no systematic biases occurred, and the 85 schools in the sample are representative of the Flemish situation (Van Houtte et al. 2005). In the participating schools, we asked all thirdand fifth-grade students present at the time of the visit to fill out the questionnaire. All students participating in the study did so with their parents’ consent. Students filled out the questionnaire in class, supervised by members of the research team and a teacher. A few students were not present, due to absence or field trips. A total of 11,945 students completed the questionnaire, of which 11,872 (response rate: 87%) proved valid. The questionnaires were not anonymous because we needed to couple other data provided by the school with the students’ responses. Ultimately we removed all names, so all analyses were performed on anonymous data. Measures School Misconduct We measured school misconduct using a scale inspired by Stewart (2003, p. 602–604), consisting of 17 items (See ‘‘Appendix’’). Students were asked how often they committed deviant acts, such as ‘‘smoked at school,’’ ‘‘been late for school,’’ ‘‘done drugs at school,’’ and so forth. Students could answer using a 5-point scale, ranging from never (1) to very often (5). Scores were summed to a scale ranging from 17 to 85 (mean = 30.04, SD = 8.47, alpha = 0.87; see Table 1). It has been shown that using self-reported measures is not ideal for measuring deviant acts (Crosnoe 2002), but it nonetheless remains the most common method of gathering such information (e.g. Stewart 2003; Gottfredson et al. 2005). We interpolated missing values by item correlation substitution (Huisman 1999): a missing item was assigned the value of the most highly correlated item. ICS is a simple and relatively easy to implement technique, used to handle missing responses to a scale. Although they tend to overestimate the scale quality, procedures which use the relationships between items—such as ICS—perform best, especially in scales with few response options and a low percentage of missing values (see Huisman 1999). As is common for delinquency measures (Crosnoe 2002; Stewart 2003), the dependent variable was significantly skewed (1.58, SE = 0.023) toward its lower end. Hence, we performed an overdispersed Poisson model with constant exposure on the data to account for possible non-linearity, but the same basic picture emerged as with linear methods. Below we present the results of the more readily interpretable linear technique. Bonding Measures Before we introduce our bonding measures, a word is in order about the use of terminology in this kind of research, which has been lacking in clarity and consistency (for reviews, see Libbey 2004; Barber and Schluterman 2008). School connectedness has been designated with all kinds of different terms, including ‘‘bonding’’, ‘‘attachment’’ and ‘‘connection’’, without clearly specifying what these terms mean. Here, we follow the conceptualizations proposed by Libbey (2004). Our first measure pertains to peers. The 4 items used are: ‘‘I wish I had other friends at school’’, ‘‘My friends accept me as I am’’, ‘‘I trust my friends at school’’ and ‘‘My friends at school respect my feelings and ideas’’. As these items refer to emotional feelings of connection (see Libbey 2004, p. 274), we call this peer attachment. Respondents could answer on a five-point scale, ranging from Does not describe me at all accurately to Describes me accurately (1–5). Scores were summed up to a scale, ranging from 4 to 20. Our respondents had a mean of 15.81 (SD = 2.76). The alpha for the scale was 0.74. The measures concerning bonding to teachers and the school context are derived from the Psychological-Senseof-School-Membership-scale (Goodenow 1993). This is a scale consisting of 18 items (see Table 2), to which respondents could answer on a five-point scale, ranging from totally disagree to totally agree (1–5). On this scale, we performed a Principal Component Analysis (PCA; n = 11,872), using Varimax rotation. This resulted in four components, one consisting of teacher or adult related assertions, accompanied by three more general support related items (see Table 2; items marked by T). The items loaded with a minimum of 0.454 (item 7) and a maximum of 0.705 (item 5). A confirmatory PCA (extracting one component; n = 11,872) confirmed the high loadings of these seven items on this one component (ranging from 0.472 to 0.750). Items were summed to a scale, ranging from 7 to 35. As these items pertain to whether students feel valued by their teachers (see Libbey 2004, p. 281), and the scale is reported by students, we call this perceived teacher support. Our respondents had a mean of 23.99 (SD = 3.99). The alpha was 0.75 (n = 11,612). As the other three components derived from the PCA pertained to whether students feel at home at school, and 123 J Youth Adolescence Table 2 The psychological sense of school membership (PSSM) scale (Goodenow 1993, p. 84) Items 1. I feel like a real part of this school 2. People here notice when I’m good at something 3. It is hard for people like me to be accepted here. (reversed) 4. Other students in this school take my opinions seriously 5. Most teachers at this school are interested in me 6. Sometimes I feel as if I don’t belong here. (reversed) 7. There’s at least one adult in this school I can talk to if I have T a problem 8. People at this school are friendly to me 9. Teachers here are not interested in people like me. (reversed) T 10. I am included in lots of activities at this school 11. I am treated with as much respect as other students 12. I feel very different from most other students here. (reversed) 13. I can really be myself at this school 14. The teachers here respect me 15. People here know I can do good work 16. I wish I were in a different school. (reversed) 17. I feel proud of belonging to this school 18. Other students here like me the way I am T T T T T Square-Within Mean Square)/Between Mean Square, and should be larger than 0.60 for a legitimate aggregation. For sense of belonging, the ICC was 0.91. Schools differed significantly on their mean level of belonging (F = 10.736; p \ 0.001). This means that the sense of belonging is indeed shared between students from the same school, and that it is legitimate to speak about cohesion at the school level. The mean value on our cohesion measure over all schools was 61.24 (SD = 2.81). Ethnicity We assessed ethnicity using multiple questions. The principal criterion was the birthplace of the maternal grandmother. If missing (1%), we considered the nationality of the parents, as most immigrants are second- and thirdgeneration citizens and have Belgian nationality. As is common, only West European birthplaces and nationalities were considered as native descent (Timmerman et al. 2002). Additional criteria in case of missing data regarding nationality (father: 4%, mother: 3.3%) were the language spoken at home (other than Dutch), religion (Islam), and the student’s name (Felouzis 2003). This resulted in a dichotomous variable (coding: 0 = native, 1 = immigrant); 11.2% were immigrants. Socioeconomic Status The SES was measured by the occupation of the father and mother (Erikson et al. 1979), or, if unemployed, their last occupation (1 = Unskilled manual labor; 2 = Specialized manual labor; 3 = Skilled manual labor; 4 = Routine nonmanual employees; 5 = Farmers and smallholders; 6 = Lower grade employees and administrators; 7 = Highergrade administrators and executives; 8 = Professionals and large proprietors). We used the highest ranked occupation as the family SES. The mean was 5.20 (SD = 2.10). Ethnic Composition As is common in educational research (see Vervoort et al. 2010; Demanet and Van Houtte 2011), we measured ethnic composition by the proportion of immigrants at school. We asked the administrators to estimate this; however, 12 (14.12%) of the 85 administrators did not respond to this question. This poses a problem, as multilevel analysis does not permit missing values at the school level and using this measure would mean we have to omit 12 schools from our sample. However, we additionally computed the proportion of immigrants at school using the students’ ethnicity at the individual level (see above). The correlation of 0.88 (p \ 0.001) between the two measures validates the use of this aggregated measure. The 85 schools in our sample each in itself does not yield a straightforward interpretation, we treated the remaining 11 items together in one scale (see Table 2; items without T). As these items bear on a broad notion of whether students affiliate to their schools (see Libbey 2004, p. 278), we call this scale general school belonging. A confirmatory PCA (extracting one component; n = 11,872) on these eleven items revealed factor loadings ranging from 0.387 (item 4) to 0.705 (item 6). There was one outlier with a loading of 0.298 (item 10), but item analysis did not show a substantial improvement of alpha when deleting this. Alpha for this scale was 0.80 (n = 11,543). The scores were summed, yielding a minimum of 11 and a maximum of 55. Students had a mean of 37.11 (SD = 6.42). Cohesion To assess school cohesion, we aggregated the students’ individual beliefs to the school level, by calculating the mean sense of belonging (see Goodenow 1993) in each school (e.g. Hofstede et al. 1990). However, first, we had to investigate whether the sense of belonging was really shared between students of the same school. For this, we calculated the ‘‘mean rater reliability’’ (Shrout and Fleiss 1979; Glick 1985), which is based on the intra-class correlation (ICC) in a one-way ANOVA with sense of belonging as dependent variable and schools as factor. The ICC is calculated by the formula: (Between Mean 123 J Youth Adolescence cover the entire range of ethnic composition, from 0% (6 schools) to 88.20% (1 school). The mean of this measure was 16.45 (SD = 21.70). SES Composition The schools’ SES composition was measured by calculating the mean SES of students (see above) per school (see Rumberger and Palardy 2005; Demanet and Van Houtte 2011). The mean in the sample was 4.80 (SD = 1.23). Parental Attachment To measure parental attachment, we used a scale of 7 items, answerable on a 5-point-scale (Brutsaert 2001). Examples of items used are ‘‘If I want to tell my parents something, they act as if they don’t hear me’’, and ‘‘I don’t think my parents believe I can do something good’’. Scores on the items were summed, resulting in a scale ranging from 7 to 35. Missing answers were imputed by item correlation substitution (Huisman 1999). Respondents had a mean of 28.32 (SD = 5.61). The alpha was 0.83. Vocational Track We also accounted for whether students attended a vocational track. The Flemish school system can be categorized as ‘‘explicit school-level tracking to different school types catering to specific student groups’’, using achievement as a selection criterion (Trautwein et al. 2006, p. 789). There are four main tracks in which students can enrol: general education, technical, vocational, and artistic education, the latter being a rather marginal track, in terms of number of students. The different tracks are commonly classified hierarchically, placing vocational tracks at the lower end. Hence, attending a vocational track in Flanders is rarely a positive choice, and vocational students are all too aware of their low status in society, yielding more anti-school attitudes and more misconduct (for an extended discussion of the Flemish tracking system, see Van Houtte and Stevens 2008). Moreover, immigrants in Flanders are overrepresented in the vocational track. Because of all this, it was necessary to account for this in our analyses. Among students, 22% attended a vocational track (coding: 1 = vocational track). Prior Achievement Prior achievement was measured by GPA (Grade Point Average) from the preceding school year. To grade their students, Flemish high-schools use a percentage, hence, grades range from 0 to 100%, 50% being the passing grade. In our data, students ranged from 41 to 100%, with a mean of 69.42 (SD = 9.22; see Table 1), corresponding grossly to a ‘C’ in the US high school system. This measure should be considered carefully. As no standardized (for example, state administered) tests exist in Flanders, it is hard to compare measures of academic achievement across schools. Furthermore, being self-reported, this measure could contain biases due to memory problems and cover-up strategies. As a result, it has a large number of missing values (9.8%). Consequently, we entered the measure only at the very last step. Data Analysis As we have a clustered sample—students nested within schools—it is necessary to use multilevel analysis (HLM6; Raudenbush and Bryk 2002). First, we estimated an unconditional ‘‘null’’ model to determine school-level variance in school misconduct. Then, we added the variables sequentially as explained above. Throughout our analyses, we controlled for several variables at the school and individual level. At the school level, we considered school sector, SES composition and ethnic composition. We should point out that there exists a high correlation between these two latter features (r = -0.784; p \ 0.001, See ‘‘Appendix’’). Because other studies have established an effect of ethnic composition next to the effect of SES composition (e.g. Eitle and Eitle 2003), we used the variables simultaneously, although the results should be regarded with caution due to possible multicollinearity. At the school level, it would be important to control for school size (see Stewart 2003). However, two schools provided no information on this variable. As multilevel analysis does not permit missing values at the second level, and analysis of 83 schools showed that school size exerts no influence on school misconduct (c = 0.000; SE = 0.001; p [ 0.05), we omitted this variable from the analyses. At the student level, then, it is obvious to control for sociodemographic variables such as SES, gender (coding: 1 = girl), grade (coding: 1 = fifth grade) and ethnicity (coding: 1 = immigrant). Furthermore, we control for parental attachment (Dornbusch et al. 2001), and attending a vocational track (coding: 1 = vocational track) (see Van Houtte and Stevens 2008). In the last model, we controlled additionally for prior achievement. This was done only in the last step, because this measure was a rather crude one (see Measures section) and had a large number of missing values. However, as prior achievement has been linked to both school deviance and school belonging (Stewart 2003), it was necessary to control for this. To ensure model stability, all but the dichotomous variables were grand mean centered. Below, we present both the unstandardized (c) and the standardized (c*) results. 123 J Youth Adolescence Results From the unconditional model (see Table 3), we see that 7.1% (r2 = 67.107; s0 = 5.156; p \ 0.001) of the variance in school misconduct occurs between schools, warranting the introduction of school-level determinants. School and Individual-Level Effects of Bonding Our first model shows that our school-measure of cohesion has an association with school misconduct (c* = -0.044; p \ 0.01, see Table 4). Adding the individual-level peer attachment in the second model decreases this association slightly, but it remains significant (c* = -0.039; p \ 0.05). However, when perceived teacher support and general school belonging are added (model 3), the association between the school-measure of cohesion and school misconduct vanishes (c* = 0.018; p [ 0.05). We should also point out that the explanatory power of our four schoollevel variables is limited, as our analyses suggest that they explain 7% of school-level variance. The Differential Effects of Peer, Teacher, and School Bonding Research question 2 is answered in the second and third model. While, in the second model, peer attachment in itself has no association with school misconduct (c* = -0.004; p [ 0.05), entering perceived teacher support and general school belonging in the third model reveals a positive association of peer attachment with school misconduct (c* = 0.070; p \ 0.001), although the standardized coefficient shows this to be a weak association. Perceived teacher support (c* = -0.170; p \ 0.001) and general school belonging (c* = -0.096; p \ 0.001) are significantly negatively related to school misconduct. Furthermore, at the individual pupil level, gender, grade and parental attachment hold strong associations with school misconduct (c* [ 0.10 in model 3). Of all our school variables, only school sector has such a strong association with the outcome. While prior achievement does have a significant relation to school misconduct (c* = -0.139; p \ 0.001), including it in model 5 has no influence on the other effects. Differences by Social-Ethnic Context The variance components in the third model show that there is significant slope variation between schools for peer attachment (ru2 = 0.058; p \ 0.05), perceived teacher support (ru2 = 0.035; p \ 0.01) and general school belonging (ru2 = 0.011; p = 0.053), warranting the testing of cross-level interactions. In the fourth model, we entered ethnic and SES school composition to account for the slope variation of, respectively, peer attachment, perceived teacher support and general school belonging. However, these school factors did not explain the slope variation of the bonding measures, hence no evidence was found for crosslevel interactions. To test whether this lack of significant interaction effects could be due to the high correlation between SES and ethnic school composition (r = -0.784; p \ 0.001, See ‘‘Appendix’’), we additionally performed model 4 with SES composition and ethnic composition separately (not shown). These analyses yielded the same results as the ones shown in Table 4. Discussion School deviance is a widely studied topic in adolescent research. One of the most influential explanations for the occurrence of school misbehavior is offered by the schoolsas-communities perspective (Battistich et al. 1995), which states that preventing school deviancy requires that students feel part of a caring school community and have meaningful and supportive relationships to actors at school (Battistich and Hom 1997). In the present study, we built on this perspective in three ways. While most researchers contend that feelings of belonging have important preventive effects on school misbehavior (e.g. Dornbusch et al. 2001), opinions are divided whether efforts should be directed at transforming the school into a community by building cohesion between actors at the school level (Battistich et al. 1995), or whether it is more efficient to impact students’ personal feelings of belonging (Goodenow 1993) in order to prevent school misconduct. While some studies in the past have indeed pointed to the multilevel nature of the sense of belonging to school, no study has yet investigated the relative importance of school-level cohesion versus individual feelings of peer, teacher, and school attachment in relation to school misconduct. Table 3 HLM unconditional model characteristics: between schools in student school misconduct Characteristic Intercept Parameter variance Within school Between schools HLMa reliability estimate Proportion of variance between schools a Variation Value 30.180*** 67.107 5.156 0.862 0.071*** HLM hierarchical linear modelling 2 v (84, N = 11,872) = 887.970 *** p B 0.001 123 J Youth Adolescence Table 4 Association between peer attachment, perceived teacher support, and general school belonging and school misconduct. Results of sequential multilevel analysis Model 1 Intercept School level Ethnic composition c c* SE SES composition c c* SE Public sector c c* SE Cohesion c c* SE Student level Female gender c c* SE SES c c* Fifth grade SE c c* SE Parental attachment c c* SE Immigrant c c* SE Vocational track c c* SE Peer attachment c c* SE Perceived teacher support c c* SE General school belonging c c* SE Prior achievement c c* SE -3.097*** -0.183*** 0.217 0.131** 0.033** 0.041 1.178*** 0.139*** 0.106 -0.324*** -0.215*** 0.018 -0.576 -0.021 0.389 1.249** 0.060** 0.384 -3.035*** -0.179*** 0.230 0.120** 0.030** 0.042 1.160*** 0.137*** 0.106 -0.327*** -0.216*** 0.018 -0.710 -0.026 0.404 1.218** 0.059** 0.405 -0.012 -0.004 0.035 -2.921*** -0.172*** 0.222 0.145** 0.036** 0.044 1.100*** 0.130*** 0.100 -0.225*** -0.149*** 0.017 -0.577 -0.021 0.402 1.056** 0.051** 0.374 0.216*** 0.070*** 0.040 -0.361*** -0.170*** 0.032 -0.126*** -0.096*** 0.020 -2.924*** -0.173*** 0.223 0.144** 0.036** 0.044 1.101*** 0.130*** 0.100 -0.226*** -0.149*** 0.017 -0.550 -0.020 0.403 1.081** 0.052** 0.377 0.210*** 0.069*** 0.042 -0.355*** -0.167*** 0.034 -0.124*** -0.094*** 0.021 -2.663*** -0.157*** 0.216 0.177** 0.044** 0.050 0.975*** 0.115*** 0.110 -0.202*** -0.134*** 0.018 -0.623 -0.023 0.404 1.420*** 0.069*** 0.345 0.201*** 0.065*** 0.041 -0.315*** -0.149*** 0.036 -0.117*** -0.089*** 0.021 -0.128*** -0.139*** 0.013 0.001 0.002 0.013 -0.094 -0.014 0.219 1.786*** 0.105*** 0.322 -0.131** -0.044** 0.049 0.000 0.001 0.014 -0.161 -0.023 0.222 1.792*** 0.106*** 0.330 -0.117* -0.039* 0.052 0.001 0.002 0.014 -0.259 -0.038 0.222 1.556*** 0.092*** 0.320 0.055 0.018 0.062 -0.004 -0.010 0.014 -0.292 -0.042 0.228 1.561*** 0.092*** 0.321 0.060 0.020 0.062 0.015 0.038 0.013 0.055 0.008 0.194 1.491*** 0.088*** 0.269 0.048 0.016 0.057 c SE 29.071*** 0.552 Model 2 29.066*** 0.540 Model 3 29.304*** 0.536 Model 4 29.307*** 0.543 Model 5 29.329*** 0.516 123 J Youth Adolescence Table 4 continued Model 1 Interaction terms Peer attachment*ethnic composition Peer attachment* SES composition Perceived teacher support*ethnic composition Perceived teacher support*SES composition General school belonging*ethnic composition General school belonging* SES composition Variance components Intercept Female gender SES Fifth grade Parental attachment Immigrant Vocational track Peer attachment Perceived teacher support General school belonging Prior achievement Model deviance U0 U1 U2 U3 U4 U5 U6 U8 U7 U9 U10 74,593.94*** 73,420.39*** 72,395.97*** 72,441.45 10.310 1.452 0.022 0.339 0.011 3.819 3.474 8.596 1.840 0.034 0.349 0.010 4.441 4.014 0.036 8.996*** 1.707*** 0.041 0.297* 0.009* 4.711** 3.228*** 0.058* 0.035** 0.011$ 9.201*** 1.696*** 0.041 0.300* 0.009* 4.834** 3.224*** 0.060* 0.036** 0.012* 6.538*** 1.451*** 0.059 0.428** 0.011* 3.782** 2.006*** 0.046 0.040** 0.009 0.008** 66,225.11*** c SE c SE c SE c SE c SE c SE -0.003 0.003 -0.029 0.044 0.003 0.003 0.012 0.044 0.000 0.002 -0.005 0.030 -0.002 0.003 -0.007 0.045 0.004 0.003 -0.003 0.045 -0.001 0.002 -0.006 0.027 Model 2 Model 3 Model 4 Model 5 Unstandardized (c) and standardized (c*) gamma’s are presented, with standard errors (SE), variance components U, and Model Deviance, with significance level of the Chi-squared test comparing it to the deviance of the previous model; Model 1 is compared to the null-model * p B 0.05, ** p B 0.01, *** p B 0.001; $ p = 0.053 Moreover, in this article, we investigated whether the schools-as-communities perspective should distinguish between the different actors bonded with at school. Regarding school deviancy, the perspective currently is based largely on early control theory (Hirschi 1969), a theoretical approach that has neglected the characteristics of the actor involved in the bonding. Differential association theory (Sutherland and Cressey 1978), on the contrary, states that closer peer relationships might generate more deviancy if antisocial socialization occurs in the friendship group. This is likely in adolescence, as deviancy is accepted by a certain number of adolescents as a valid form of behavior (Allen et al. 2005). Led by this theoretical exploration, we expected peer attachment to relate differently to school misconduct than general school belonging and perceived teacher support. In a third contribution to the schools-as-communities perspective, we investigated a long-held claim by researchers that providing students with supportive relationships should impact those students even more in disadvantaged school contexts (Battistich et al. 1995). However, especially in the case of deviancy, this claim has received mixed empirical support. In the current study, we considered the SES composition and the ethnic composition of schools as indicators of school disadvantage (Demanet and Van Houtte 2011), to investigate whether feelings of support were indeed more strongly related to less school misconduct in schools with a lower mean SES and a higher proportion of minority students. To assess the first research question, namely what the relative importance is of school-level cohesion versus individual feelings of belonging, we tested whether the association of school cohesion is maintained when controlling for three aspects of students’ sense of belonging— peer attachment, perceived teacher support, and general school belonging. We found that, in association to school misconduct, individual bonding seems more important than overall school cohesion, as no relation of school cohesion is seen once the three aspects of individual belonging are 123 J Youth Adolescence taken into account. This finding can be explained in two ways. First, it could be due to selection. A selection effect is found when it is clear that school-level effects are actually due to features of the students enrolled in the school, and as such are not an effect over and above these individual student effects. In this case, schools that are cohesive are simply made up of more students with a high sense of teacher support and general school belonging, which is why these students report less school misconduct. Or second, it could be due to a mediation effect: in this case, cohesion at the school level yields a higher perceived support received from teachers in students and a higher sense of overall school belonging, which is responsible for their lower school misconduct. As we have no longitudinal design, however, we can not differentiate between the selection or mediation explanations. Whichever is the case, our results suggest that intervention efforts designed to improve a school’s cohesiveness will not affect school deviancy directly (cf. Nelson 1996). In the mediation case, at most these efforts might create more sense of belonging in individual students, making them less prone to school misconduct. Therefore, creating a cohesive school may not be a sufficient condition to combat school misconduct. Our results suggest that, to counter school deviancy, intervention efforts should focus directly on individual feelings of school belonging. However, we should point out that, as this is not a longitudinal study, we can not be sure about the causal direction of these relationships. Regarding the second research question, our results show that, while higher school belonging and perceived teacher support are related to less school misconduct, greater peer attachment was associated to more school misconduct. It is noteworthy that the association between peer attachment and school misconduct only showed up when we controlled for general school belonging and perceived teacher support, meaning that both characteristics buffer the relation of peer attachment with school misconduct. This occurs because students who feel bonded to their school and teachers generally perceive a high degree of peer attachment as well. Students who are bonded to all three sources are less deviant than others, because the preventive effect of teacher support and school belonging overshadows the deviance-generating effect of their high peer attachment. However, students lacking teacher support and school belonging miss this suppression effect, so that the deviance-yielding effect of peer attachment is free to occur. Hence, for students feeling attached to their peers, yet lacking teacher and school bonding, more misconduct can be expected. This concurs with earlier research. In a recent longitudinal study, Freidenfelt Liljeberg et al. (2011) point to the role of teacher attachment in preventing delinquency. Further evidence is provided by McNeely and Falci (2004) who distinguished between social and teacher support. Although their measure of social support does not specifically refer to peers, they conclude that conventional bonds to teachers can counterbalance unconventional peer bonds. Baker (1998, p 35) gives a more process-oriented explanation for our results, stating that when students fail to form meaningful connections to their school and its teachers, they turn to other youngsters to fulfill their need for belonging, subsequently forming sub-communities within the school. It is noteworthy that this occurs especially with disruptive students, because teachers react more harshly to them (Good and Brophy 1994), and they are likely to be rejected by their prosocial peers (Salmivalli et al. 1997). As such, small, deviant peer groups are formed in which students incite each other to continue to break the rules (Bendixen et al. 2006). We should, however, point to a possible alternative explanation for our findings. As this is not a longitudinal study, we cannot ascertain the direction of the effects. In fact, longitudinal research shows that effects are likely to be bidirectional (Hirschfield and Gasper 2011). Hence, it has been shown that students who are deviant at school also are more likely to feel less at home in the school context and perceive disconnection from their teachers (Karcher 2002). It is also likely that deviant students tend to be popular among their friends (see also Demanet 2008). As in adolescence, peer norms tend to favor minor deviant acts, students who display school misconduct are likely to be popular (Allen et al. 2005). Rather than being determinants of school misconduct, bonding relationships can be outcomes of school-deviant behavior. In response to the third research question, we found no evidence for the claim that supportive relationships should be related even more strongly to school misconduct in disadvantaged schools: none of the cross-level interaction effects were significant. This concurs with research relating this to neighborhood disadvantage (Dornbusch et al. 2001). A possible explanation for the lack of support found for this claim is that having supportive relationships is not more beneficial for all students in disadvantaged schools. As described above, the rationale for interaction effects is that disadvantaged students will benefit disproportionally from school communality because they have less social capital outside of school. Hence, this does not apply to advantaged students in disadvantaged schools. It is possible, then, that supportive relationships are more beneficial for disadvantaged students, whatever the school context they enroll in, than for advantaged students. We should note the limitations of this study and propose some remedies to be addressed in future research. First, as we mentioned earlier, our design was cross-sectional, so we cannot make any causal claims. Research has established that the association between bonding and misconduct is 123 J Youth Adolescence bidirectional (e.g. Karcher 2002; Hirschfield and Gasper 2011), and lacking longitudinal data, we can not ascertain which direction is dominant. Furthermore, previous longitudinal research showed that exposure to deviant peers was not equally predictive of later problem behavior at all ages (Fleming et al. 2010). Hence, we call for additional longitudinal research to investigate the intertwined roles of school cohesion and individual feelings of bonding, and link this to developmental processes. Second, our reliance upon a single informant—namely the student—for all the individual-level variables could create problems of shared method variance: because all variables are reported by the same individual, correlations could be estimated larger than they actually are. Moreover, self-reports are not always accurate representations of actual levels of both deviance (Crosnoe 2002) and bonding (Stevens et al. 2002). Future studies on this topic should combine reports by students, teachers, and parents to avoid these problems. Third, in this study, we did not distinguish between attachment to deviant or non-deviant peers because we assumed that all students get socialized in the broadly held peer norms that school misconduct is acceptable behavior (see also Allen et al. 2005), whether their close friends were deviant or not. However, our conclusion led us to hypothesize that students with less perceived teacher support and general school bonding form sub-communities within the school, in which we expected deviancy to be endorsed. Hence, we arrive at the same conclusion as many former studies (e.g. Bendixen et al. 2006), namely that especially attachment to peers who themselves perform deviant acts promotes school misconduct. Future research should distinguish between attachment to prosocial and antisocial peers in order to compare their influence on school misconduct to teacher support and general school belonging. Fourth, as this study used a quantitative approach, we only could gain general information from our respondents, and had less insight into the perceptions of the actors in question. Future qualitative research should attempt to understand on a more micro level how socialization takes place in close interactions, and how the buffering effect of receiving teacher support is experienced by students themselves. Lastly, we should note a small selection bias with regard to our analytical sample. Students from the vocational track and immigrants were overrepresented among the respondents excluded from the analyses due to missing responses. However, because of the small number of missing cases, this is not likely to affect our overall results. Moreover, we tried to account for this by controlling for the socio-demographic variables in question. Indeed, further analyses—in which we imputed all missing values by means of mean imputation—confirmed the conclusions of this study. Although it is not likely that this small selection bias impacts our results, we do propose that future studies test whether the conclusions in this study also apply to students in the vocational track and to immigrants. In summary, our study’s main finding is that, although associations appear to be small, bonding with different actors at school can have mixed connections to deviant behavior. This contradicts early control theory, on which the schools-as-communities perspective, with reference to school deviancy, is now largely based. Following the addition of differential association theory to the early version of control theory, we argue that it is vital to account for the characteristics of the actors with whom the student bonds, and especially whether they endorse deviancy or not. It seems important to incorporate this in the schools-as-communities perspective. Furthermore, our results support earlier findings that receiving teacher support yields multiple positive outcomes for students, including less involvement in health-risk behavior (West et al. 2004), and higher engagement and achievement (Klem and Connell 2004). Hence, we contribute to the evidence base showing that schools should be organized as caring communities. Adults at school play a large role in this, as they are responsible for making all students, even the disruptive and poor achieving ones, feel at home at school. Acknowledgments The authors are grateful to Simon Boone, the Editor, and three anonymous reviewers for their helpful feedback on this manuscript. Appendix: The School Misconduct Scale (inspired by Stewart 2003, p. 602–604) How often have you: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. been late for school skipped lessons skipped school all day cheated on tests copied someone’s homework not made your homework fought at school stolen at school committed vandalism at school smoked at school drunk alcohol during school hours done drugs during school hours talked back at teachers broke the school rules had to do impositions been sent to detention been suspended for one or more days See Tables 5 and 6. 123 J Youth Adolescence Table 5 Pearson product moment correlations between individual-level variables School misconduct SES Parental attachment Peer attachment Perceived teacher support General school belonging GPA *** p B 0.001 0.009 -0.225*** -0.093*** -0.281*** -0.241*** -0.237*** -0.005 0.085*** 0.047*** 0.085*** 0.079*** 0.231*** 0.312*** 0.337*** 0.141*** 0.294*** 0.476*** 0.075*** 0.617*** 0.150*** 0.141*** SES Parental attachment Peer attachment Perceived teacher support General school belonging Table 6 Pearson product moment correlations between school-level variables Ethnic composition SES composition Cohesion -0.784*** -0.141 0.297** SES composition ** p B 0.01, *** p B 0.001 References Agnew, R. (2003). An integrated theory of the adolescent peak in offending. Youth and Society, 34, 263–299. Allen, J. P., Porter, M. R., McFarland, F. C., Marsh, P., & McElhaney, K. B. (2005). The two faces of adolescents’ success with peers: Adolescent popularity, social adaptation, and deviant behavior. Child Development, 76, 747–760. Baker, J. A. (1998). Are we missing the forest for the trees? Considering the social context of school violence. Journal of School Psychology, 36, 29–44. Bankston, C., & Caldas, S. J. (1996). Majority African American schools and social injustice: The influence of de facto segregation on academic achievement. Social Forces, 75, 535–555. Barber, B. K., & Olsen, J. A. (1997). Socialization in context: Connection, regulation, and autonomy in the family, school, and neighborhood, and with peers. Journal of Adolescent Research, 12, 287–315. Barber, B. K., & Schluterman, J. M. (2008). Connectedness in the lives of children and adolescents: A call for greater conceptual clarity. Journal of Adolescent Health, 43, 209–216. Battistich, V., & Hom, A. (1997). The relationship between students’ sense of their school as a community and their involvement in problem behaviors. American Journal of Public Health, 87, 1997–2001. Battistich, V., Solomon, D., Kim, D. I., Watson, M., & Schaps, E. (1995). Schools as communities, poverty levels of student populations, and students attitudes, motives, and performance— A multilevel analysis. American Educational Research Journal, 32, 627–658. Battistich, V., Solomon, D., Watson, M., & Schaps, E. (1997). Caring school communities. Educational Psychologist, 32, 137–151. Bendixen, M., Endresen, I. M., & Olweus, D. (2006). Joining and leaving gangs: Selection and facilitation effects on self-reported antisocial behaviour in early adolescence. European Journal of Criminology, 3, 85–114. Brans, M., Jacobs, D., Martineillo, M., Rea, A., Swyngedouw, M., Adam, I. et al. (2004). Recherche et Politiques publiques, le cas de l’immigration en Belgique [Research and police: a case study of immigration in Belgium]. Gent: Academia Press. Brutsaert, H. (2001). Co-educatie. Studiekansen en Kwaliteit van het Schoolleven [Coeducation. Study Chances and Quality of School Life]. Leuven/Apeldoorn: Garant. Bryk, R., & Driscoll, D. (1988). The high school as community: contextual influences and consequences for students and teachers. Madison: Center for Educational Research, University of Wisconsin. Crosnoe, R. (2002). High school curriculum track, and adolescent association with delinquent friends. Journal of Adolescent Research, 17, 143–167. Demanet, J. (2008). Populair of verstoten? Een netwerkanalytisch onderzoek naar de sociale kenmerken van pesters in het Vlaamse secundaire onderwijs [Popular or isolated? A network analytical study for the social characteristics of bullies in Flemish secondary schools]. Tijdschrift voor sociologie, 29, 397–423. Demanet, J., & Van Houtte, M. (2011). Social-ethnic school composition and school misconduct: Does sense of futility clarify the picture? Sociological Spectrum, 31, 224–256. Dornbusch, S. M., Erickson, K. G., Laird, J., & Wong, C. A. (2001). The relation of family and school attachment to adolescent deviance in diverse groups and communities. Journal of Adolescent Research, 16, 396–422. Eitle, D., & Eitle, T. M. (2003). Segregation and school violence. Social Forces, 82, 589–616. Erickson, K. G., Crosnoe, R., & Dornbusch, S. M. (2000). A social process model of adolescent deviance: Combining social control and differential association perspectives. Journal of Youth and Adolescence, 29, 395–425. Erikson, R., Goldthorpe, J. H., & Portocarero, L. (1979). Intergenerational class mobility in 3 Western European societies—England, France and Sweden. British Journal of Sociology, 30, 415–441. Espelage, D. L., Holt, M. K., & Henkel, R. R. (2003). Examination of peer-group contextual effects on aggression during early adolescence. Child Development, 74, 205–220. ´ ´ ´ Felouzis, G. (2003). La segregation ethnique au college et ses ´ consequences [Ethnic segregation at secondary school and its consequences]. Revue Francaise de Sociologie, 44, 413–447. ¸ Finn, J. D. (1989). Withdrawing from school. Review of Educational Research, 59, 117–142. Fleming, C. B., Catalano, R. F., Haggerty, K. P., & Abbott, R. D. (2010). Relationships between level and change in family, school, and peer factors during two periods of adolescence and problem behavior at age 19. Journal of Youth and Adolescence, 39, 670–682. Freeman, D. J., Brookhart, S. M., & Loadman, W. E. (1999). Realities of teaching in racially/ethnically diverse schools—Feedback from entry-level teachers. Urban Education, 34, 89–114. 123 J Youth Adolescence ¨ Freidenfelt Liljeberg, J., Eklund, J. M., Vafors Fritz, M., & af Klinteberg, B. (2011). Poor school bonding and delinquency over time: Bidirectional effects and sex differences. Journal of Adolescence, 34, 1–9. Glick, W. H. (1985). Conceptualizing and measuring organizational and psychological climate: pitfalls in multilevel research. Academy of Management Review, 10, 601–616. Good, T. L., & Brophy, J. F. (1994). Looking in classrooms. New York: Harper Collins. Goodenow, C. (1993). The psychological sense of school membership among adolescents—Scale development and educational correlates. Psychology in the Schools, 30, 79–90. Gottfredson, G. D., Gottfredson, D. C., Payne, A. A., & Gottfredson, N. C. (2005). School climate predictors of school disorder: Results from a national study of delinquency prevention in schools. Journal of Research in Crime and Delinquency, 42, 412–444. Hermans, P. (2004). Applying Ogbu’s theory of minority academic achievement to the situation of Moroccans in the low countries. Intercultural Education, 15, 431–439. Hirschfield, P. J., & Gasper, J. (2011). The relationship between school engagement and delinquency in late childhood and early adolescence. Journal of Youth and Adolescence, 40, 3–22. Hirschi, T. (1969). Causes of delinquency. Berkeley: University of California Press. Hofstede, G., Neuijen, B., Ohayv, D. D., & Sanders, G. (1990). Measuring organizational cultures—A qualitative and quantitative study across 20 cases. Administrative Science Quarterly, 35, 286–316. Huisman, M. (1999). Imputation of missing item responses: some simple techniques. In M. Huisman (Ed.), Item nonresponse: Occurence, causes, and imputation of missing answers to test items (pp. 91–119). Leiden: DSWO Press, Leiden University. Kandel, D. B. (1978). Homophily, selection, and socialization in adolescent friendships. American Journal of Sociology, 84, 427–436. Karcher, M. J. (2002). The cycle of violence and disconnection among rural middle school students: teacher disconnection as a consequence of violence. Journal of School Violence, 1, 35–51. Karcher, M. J., & Finn, L. (2005). How connectedness contributes to experimental smoking among rural youth: Developmental and ecological analyses. The Journal of Primary Prevention, 26, 25–36. Klem, A. M., & Connell, J. P. (2004). Relationships matter: Linking teacher support to student engagement and achievement. Journal of School Health, 74, 262–273. Konishi, C., Hymel, S., Zumbo, B. D., & Li, Z. (2010). Do school bullying and student-teacher relationships matter for academic achievement? A multilevel analysis. Canadian Journal of School Psychology, 25, 19–39. Libbey, H. P. (2004). Measuring student relationships to school: Attachment, bonding, connectedness, and engagement. Journal of School Health, 74, 274–283. Ma, X. (2003). Sense of belonging to school: Can schools make a difference? Journal of Educational Research, 96, 340–349. McNeely, C., & Falci, C. (2004). School connectedness and the transition into and out of health-risk behaviour among adolescents: A comparison of social belonging and teacher support. Journal of School Health, 74, 284–292. Moffitt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial-behavior—A developmental taxonomy. Psychological Review, 100, 674–701. Nelson, J. R. (1996). Designing schools to meet the needs of students who exhibit disruptive behavior. Journal of Emotional and Behavioral Disorders, 4, 147–161. Ogbu, J. U. (1978). Minority education and caste. The American system in cross-cultural perspective. New York: Academic Press. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods. London: Sage Publications. Rumberger, R. W., & Palardy, G. J. (2005). Does segregation still matter? The impact of student composition on academic achievement in high school. Teachers College Record, 107, 1999–2045. Salmivalli, C., Huttunen, A., & Lagerspetz, K. M. J. (1997). Peer networks and bullying in schools. Scandinavian Journal of Psychology, 38, 305–312. Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: uses in assessing rater reliability. Psychological Bulletin, 86, 420–428. Stanton-Salazar, R. D., & Dornbush, S. M. (1995). Social capital and the reproduction of inequality: Information networks among Mexican-origin high school students. Sociology of Education, 68, 116–135. Stevens, V., De Bourdeaudhuij, I., & Van Oost, P. (2002). Relationship of the family environment to children’s involvement in bully/victim problems at school. Journal of Youth and Adolescence, 31, 419–428. Stewart, E. A. (2003). School social bonds, school climate, and school misbehavior: A multilevel analysis. Justice Quarterly, 20, 575–604. Sutherland, E., & Cressey, D. (1978). Criminology. Philadelphia: Lippincot. Timmerman, C., Hermans, P., & Hoornaert, J. (2002). Allochtone jongeren in het onderwijs: een multidisciplinair perspectief [Migrant youngsters in education: a multidisciplinary perspective]. Leuven: Garant. Trautwein, U., Ludtke, O., Marsh, H. W., Koller, O., & Baumert, J. (2006). Tracking, grading, and student motivation: Using group composition and status to predict self-concept and interest in ninth-grade mathematics. Journal of Educational Psychology, 98, 788–806. Van Houtte, M., & Stevens, P. A. J. (2008). Sense of futility: the missing link between track position and self-reported school misconduct. Youth and Society, 40, 245–264. Van Houtte, M., Stevens, P. A. J., Sels, A., Soens, K., & Van Rossem, R. (2005). De invloed van structurele en compositorische schoolkenmerken op prestaties en welbevinden van leerlingen in het secundair onderwijs: een verklaring via cultuur. [The influence of structural and compositional school features on achievement and well-being of students in secondary education: An explanation through culture]. First research report (not published). Ghent: Universiteit Gent, vakgroep sociologie, onderzoeksgroep jeugd, educatie en geslacht. Vervoort, M. H. M., Scholte, R. H. J., & Overbeek, G. (2010). Bullying and victimization among adolescents: The role of ethnicity and ethnic composition of school class. Journal of Youth and Adolescence, 39, 1–11. Wehlage, G., Rutter, R., Smith, G., Lesko, N., & Fernandez, R. (1990). Reducing the risk: Schools as communities of support. Philadelphia: Falmer. Wellman, B., & Frank, K. (2001). Network capital in a multilevel world: Getting support from personal communities. In N. Lin, K. Cook, & R. S. Burt (Eds.), Social capital: Theory and research (pp. 233–274). New York: Walter de Gruyter, Inc. West, P., Sweeting, H., & Leyland, A. (2004). School effects on pupils’ health behaviours: Evidence in support of the health promoting school. Research Papers in Education, 19, 261–291. 123 J Youth Adolescence Author Biographies Jannick Demanet received his master in sociology in 2007. Since then, he is assistant at the Department of Sociology, Ghent University, and he is a member of the CuDOS research team. He is currently preparing his Ph.D.-project, covering structural, compositional and cultural school determinants of antischool and antisocial attitudes and behavior. His research interests further include bullying/victimization, and the nature and strength of students’ relations with peers, teachers, parents, and the school. Mieke Van Houtte received her Ph.D. in Sociology in 2002 from Ghent University. Currently, she is lecturer at the Department of Sociology, Ghent University, and she is head of the CuDOS research team. Her research interests include diverse topics within the sociology of education, covering school effects research, education and gender, tracking research, and topics related to antisocial school behavior and school misconduct. 123