You are on page 1of 16

J Youth Adolescence (2012) 41:499–514

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 / Published online: 13 May 2011
Ó Springer Science+Business Media, LLC 2011

Abstract The schools-as-communities perspective pro- misconduct. No differences were found by socio-ethnic
vides a popular explanation for school-disruptive behavior, context. Implications are discussed.
stating that interpersonal bonding at school and feelings of
school belonging prevent misconduct. In this article, we Keywords Schools-as-communities perspective  School
build on this perspective in three ways. First, we test misconduct  Peer attachment  Teacher support  School
whether the preventive influence of school belonging acts belonging  School composition
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 Introduction
perceived teacher support, school belonging, and peer
attachment relate differently to school misconduct. Lastly, A recurrent theme within educational and adolescent
the present study investigates whether the associations of research is the explanation of school misbehavior (e.g.
bonding with teachers, peers and the school with school Freidenfelt Liljeberg et al. 2011). A popular explanation
misconduct differ by socio-ethnic school context. Multi- revolves around the schools-as-communities perspective
level analyses were performed on data from the Flemish (Battistich et al. 1995), a broad line of inquiry advocating
Educational Assessment. The sample consisted of 11,872 that schools should be organized as caring school com-
students (51.4% female) in 85 schools, most of whom were munities. In such caring communities, students feel emo-
natives (88.8%), with immigrants (11.2%) mostly having tionally connected to their peers, teachers, and school,
Turkish or Moroccan backgrounds (both about 30% of which, among other beneficial outcomes, fosters less
immigrants in the sample), and others Southern-European school misconduct in students (Battistich and Hom 1997).
(16%), Eastern-European (8%), North-African (5%), or However, researchers are still unsure whether the pre-
other (17%) backgrounds. Results showed that the stu- ventive effects of belonging act at the individual or the
dents’ individual feelings of bonding with peers, teachers school level. While some scholars state that the beneficial
and school associate with school misconduct, rather than effects of school belonging result from students’ personal
the overall school cohesion. Results further showed that, feelings (Wehlage et al. 1990; Goodenow 1993), others
while higher perceived teacher support and school hold that, in order to combat school deviancy, efforts
belonging related to less school misconduct, higher peer should be made to establish cohesion between actors at the
attachment was associated with higher rates of school 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 indi-
J. Demanet (&)  M. Van Houtte
vidual or the school level. Therefore, the first aim of this
Department of Sociology, Research Team CuDOS, Ghent
University, Korte Meer 3-5, 9000 Ghent, Belgium study is to assess the relative contribution of feelings of
e-mail: jannick.demanet@ugent.be belonging at the school level—that is school cohesion—
M. Van Houtte and at the individual level in preventing school
e-mail: mieke.vanhoutte@ugent.be misconduct.

123
500 J Youth Adolescence (2012) 41:499–514

Researchers from the schools-as-communities perspec- Socioeconomic Status (SES) composition and the ethnic
tive also state that attachment to different actors at school composition (Demanet and Van Houtte 2011). Hence, our
and the school itself impedes school deviancy. However, third research question is whether the relationships
while studies have established that feelings of school between general school belonging, peer attachment and
belonging (Dornbusch et al. 2001) and teacher attachment perceived teacher support are stronger in schools with a
(Freidenfelt Liljeberg et al. 2011) are associated with lower lower SES composition, and schools where immigrants are
rates of deviant behavior, research in this tradition has yet overrepresented. As such, our study tries to find evidence
to investigate whether attachment to peers is related in the for a long-held claim that supporting students in disad-
same way to school misconduct. The schools-as-commu- vantaged schools has even more beneficial effects, because
nities perspective currently draws, in regard to its expec- these students often lack such sources of support outside
tations concerning deviancy, on insights from early social the school context.
control theory (Hirschi 1969), a theoretical approach that
expects all strong social bonds to prevent delinquency. The Schools-As-Communities Perspective
However, this theoretical approach failed to address the
role of deviant peer influences (Erickson et al. 2000). In The schools-as-communities perspective is a popular
fact, research shows that peers can cause each other to viewpoint in educational research. Barber and Olsen (1997)
break school rules, especially in the case of cohesive emphasized three aspects of adolescent socialization:
friendship bonds between deviant peers (Kandel 1978; connection to significant others, regulation of behavior,
Wellman and Frank 2001; Espelage et al. 2003). It is and psychological autonomy. The schools-as-communities
important to consider these different sources of attach- perspective focuses on the first, stating that schools fulfill
ment—school, teachers, and peers—together, to gain their socialization function best when organized as caring
insight in the unique role played by each in preventing or communities (Battistich et al. 1995). Such school com-
promoting school misconduct. The few studies that have munities are defined in diverging ways, but scholars put
incorporated both teachers and peers as sources of support forward common elements (see Battistich et al. 1997,
relate this to achievement (Klem and Connell 2004), p. 137). Broadly stated, communal schools make students
smoking (Karcher and Finn 2005), and other health-risk feel emotionally connected to one another—i.e., they feel
behaviors (McNeely and Falci 2004). These find higher attached (see Libbey 2004, p. 274)—and feel respected and
teacher support to advance achievement and impede helped by their peers and teachers—in other words, they
health-risk behaviors, and higher peer support to yield a perceive themselves as supported (see Libbey 2004,
higher likelihood of smoking and engaging in other health- p. 281). In communal schools, students are made to feel at
risk behavior. In our study, a specific aim is to incorporate home at school. Furthermore, students in caring school
peer attachment, teacher support and general school communities feel that they make important contributions:
belonging and test their relative contribution to school hence, they are given a certain amount of influence in the
deviancy. Contrary to earlier studies on school deviancy, school’s activities and decision-making process. Lastly, in
which deal with student delinquency (e.g. Crosnoe 2002), such schools, there is some sort of common value system
we focus on school misconduct, a minor form of deviancy, (Battistich et al. 1997). Communal schools yield a wide
consisting of rule-breaking behavior such as cheating on array of positive effects in their students, including higher
tests, skipping lessons, and arriving late at school (e.g., school enjoyment, academic achievement, and less school
Stewart 2003; Demanet and Van Houtte 2011), as we can disruption (Battistich et al. 1995; Battistich and Hom
imagine that peer and teacher attachment is more likely to 1997).
influence minor forms of rule-breaking behavior than The preventive effect of school bonding on deviancy has
delinquency at school. Hence, we operationalize feelings of been replicated by many studies. Dornbusch et al. (2001)
school belonging multi-dimensionally, discerning three showed in a longitudinal study that school attachment
aspects—general school belonging, peer attachment and reduces the overall frequency, prevalence, and initiation of
perceived teacher support—to test whether these three deviant involvement, and that this association held across
dimensions relate differently to school misconduct. males and females, different community contexts and
In a third contribution to the schools-as-communities regardless of ethnic groups. Another longitudinal study
perspective, we embed all this within the social context of distinguished between three forms of student engagement
schools. Researchers state that receiving support prevents at school—emotional, behavioral, and cognitive engage-
school deviancy even more effectively in disadvantaged ment—confirming that emotional engagement had strong
schools (Battistich et al. 1995), but evidence concerning preventive effects on the occurrence of delinquency (Hir-
this is mixed. In the current study, we focus on two schfield and Gasper 2011). In an influential study, Finn
indicators of disadvantage in schools, namely the (1989) proposed the participation-identification model,

123
J Youth Adolescence (2012) 41:499–514 501

which highlights the role of identification with and par- are unlikely to have the same effect. In stating that bonding
ticipation in school in preventing school dropout, and, as prevents disruptive behavior, the schools-as-communities
dropout is linked to other problem behavior (see Finn 1989, perspective echoes the premises of the earliest version of
p. 118), also school misconduct. The preventive effect of control theory (Hirschi 1969). This theory holds that all
establishing strong social and emotional connections at individuals are inclined to deviancy, but having strong
school on student deviancy is thus well-established in social bonds to others prevents its manifestation. Hirschi
research. (1969) applied his view to adolescents, asserting that they
There is some confusion as to whether the effects of must have strong bonds with school, parents and peers in
belonging should be seen as resulting from students’ per- order to behave properly. Empirical research based on this
sonal feelings—an individual-level effect (Wehlage et al. theory, however, has shown that this earliest version of
1990; Goodenow 1993)—or as an effect of the school—a control theory has some important flaws (Erickson et al.
group-level effect (Bryk and Driscoll 1988; Battistich et al. 2000). Most importantly, it has neglected the characteris-
1995). The first position argues that misconduct is an tics of the actor with whom one should bond. Applied to
individual reaction to a lack of belonging, hence, efforts peer relationships, early control theory is unable to cope
should be directed at making disruptive students feel sup- with findings that friends often share a comparable level of
ported at school. The second view considers communality deviancy (Bendixen et al. 2006). This deviance ‘‘hom-
to be a group level concept, albeit with roots in individual ophily’’ (see Espelage et al. 2003) has been explained by
feelings (Battistich et al. 1995). This side states that stu- means of two processes: selection and influence (Kandel
dents have basic needs for belonging that the school should 1978). Deviant students tend to choose each other as
be equipped to fulfill, making misconduct a failure of the friends, but they also influence each other to commit
school. However, few researchers have tested specifically deviant acts (Bendixen et al. 2006). It is noteworthy that
whether the preventive effect of feelings of school influence occurs especially when friendship bonds with
belonging on students’ deviancy is due to students’ per- deviant peers are quite cohesive. Deviant peer influence has
sonal feelings of belonging or their being part of a school been explained by differential association theory (Suther-
that has a certain level of communality. Some studies have land and Cressey 1978), which states that deviancy is
shown that it is important to account for the multilevel learned from others who support deviancy as justifiable
nature of belonging at school. One study finds that the behavior. Therefore, having close bonds with others will
quality of teacher-student relationships in a school can increase the chances of being deviant, that is, when these
counterbalance negative effects of the school climate— others endorse deviancy.
seen as school-wide bullying—on the academic achieve- It is likely that antisocial values will prevail in ado-
ment of students (Konishi et al. 2010). Another study lescent peer groups. Research indeed shows that peer
investigated school-level influences on students’ school norms in adolescence tend to favor minor forms of devi-
belonging, finding that students’ sense of belonging differs ancy (Moffitt 1993; Agnew 2003; Allen et al. 2005).
more within schools than between schools, and is affected Authors argue that while youngsters are biologically
particularly by the students’ self-esteem (Ma 2003). Hence, mature, they are not allowed to fulfill social roles with
it could be argued that the sense of belonging is something complete adult privileges and responsibilities, creating a
individual, and, thus, that trying to establish a caring school maturity gap. Adolescents thus have been viewed as
community does little to impact students. However, no endorsing deviancy as a way to show autonomy. As stu-
study has investigated the relative effects of school cohe- dents are socialized into the norms held by the majority of
sion—school belonging situated at the school level—ver- their peers (Allen et al. 2005), and socialization especially
sus the individual characteristics of peer attachment, occurs in cohesive friendship relationships (Sutherland
teacher support, and general school belonging—situated at and Cressey 1978), higher peer attachment in adolescence
the student level—on students’ deviancy. Thus, while the can be conducive to school misconduct. In the schools-as-
multilevel nature of the sense of belonging has been widely communities perspective, then, it may be essential to
acknowledged, there is still no clear answer to whether the consider the possible deviance-yielding effect of cohesive
effects of school belonging result from students’ personal bonds between youngsters.
feelings or from the school as a whole.
The Context of Communality Effects
The Role of Peer Attachment in Adolescence
Researchers suggest that bonding impedes school mis-
Although the schools-as-communities perspective holds conduct most in disadvantaged schools (Battistich et al.
that all strong social bonds prevent school misconduct, 1995; Battistich and Hom 1997). This is not to say that
there are reasons to think that peer bonds in adolescence school misconduct is more prevalent there, only that

123
502 J Youth Adolescence (2012) 41:499–514

supporting students in such schools makes a larger dif- The Current Study
ference. This occurs because advantaged students are
likely to have sources of support outside of school (Bat- The purpose of this study is to expand upon the schools-as-
tistich et al. 1995). For disadvantaged students, external communities perspective. Using a multilevel framework,
sources of support are less prevalent (Stanton-Salazar and we consider three research questions. To investigate pos-
Dornbush 1995). Therefore, for them, receiving support at sible mediation or selection effects, we use sequential
school has a major impact. Testing this, Battistich et al. multilevel analyses. First, as few researchers have inves-
(1995) confirmed that belonging had more positive effects tigated whether the effects of school belonging constitute a
on school and class enjoyment in poorer schools. Battis- group-level effect or an individual-level effect, we will test
tich and Hom (1997), however, showed that this does not the relative effect of school cohesion (a school character-
apply to school delinquency, as feelings of communality istic) and students’ feelings of peer attachment, perceived
were more preventive in low and moderate poverty teacher support and general school belonging (individual
schools than in high poverty ones. At the neighborhood characteristics), on school misconduct. In the analyses, we
level, scholars have found no evidence for this assertion will investigate the role of school cohesion in the first
(Dornbusch et al. 2001). Hence, scholars have provided model, to determine whether it has an effect on the
mixed evidence to the claim that enhancing school dependent variable. In a second and third model, we add
belonging should be even more beneficial in poorer the three individual-level measures of belonging, to test
schools. whether the school effect of cohesion is maintained. Our
Next to poverty, a school’s concentration of different second research question is whether peer attachment, per-
ethnic groups also can create a disadvantaged situation in ceived teacher support and general school belonging relate
that school. It has been shown that students achieve less in differently to school misconduct. Although researchers
schools with a high concentration of migrant or ethnic from the schools-as-communities perspective (Battistich
minority students (Felouzis 2003; Rumberger and Palardy et al. 1997) assume that all sources of bonding prevent
2005). Moreover, teachers in those schools report job dis- school misconduct, based on our theoretical exploration,
satisfaction, lower academic expectations, and more diffi- we expect that peer attachment is related in a different
culty establishing relationships with students (Freeman manner to school misconduct than perceived teacher sup-
et al. 1999). Importantly, some research has demonstrated port and general school belonging are. To investigate this,
that these disadvantages are entirely due to the poor socio- we add peer attachment in our second model, followed by
economic context of these schools (e.g. Rumberger and perceived teacher support and general school belonging in
Palardy 2005). Ultimately, students in such schools per- the third model. As a third research question, we consider
ceive the school as more disadvantaged, and expect less the claim made by researchers (Battistich and Hom 1997)
from their academic career (Bankston and Caldas 1996). that sources of support prevent school deviancy more
This is especially the case in Flanders. In present-day effectively in a disadvantaged school setting. In this regard,
Flanders, migrants and their descendants are generally we consider the schools’ SES and ethnic composition as
referred to as ‘‘allochthonous’’, a term pertaining to persons indicators of school disadvantage. In our models, we test
residing in Belgium, regardless of their nationality, having this by adding cross-level interaction effects between the
at least one parent or grandparent born outside West Eur- three bonding measures and SES and ethnic composition in
ope, and holding a disadvantaged position in society our fourth model. This enables us to test whether the
because of their ethnicity (Brans et al. 2004). Their situa- relationships between peer attachment, perceived teacher
tion resembles that of the black minorities in the US, support, general school belonging, and school misconduct
typified by Ogbu (1978) as caste-like minorities. The sec- vary by the SES and the ethnic composition of the schools.
ond- 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, dis- Methods
trust of society and negative perceptions of the opportunity
structure (Hermans 2004). Hence, in Flanders, most stu- Sample
dents in ethnically concentrated schools do not perceive
upward mobility to be likely, which characterizes these The data were part of the FlEA (Flemish Educational
schools as disadvantaged. Hence, in line with the claim Assessment), gathered in the 2004–2005 school year in
presented above, it could be argued that supporting Flemish secondary schools. Our sample consisted of
students should be even more preventive of school mis- 11,872 students across 85 Flemish schools. Flanders is the
conduct in schools where ethnic minority students are northern, Dutch-speaking part of Belgium, and constitutes
overrepresented. a political entity with its own parliament and government.

123
J Youth Adolescence (2012) 41:499–514 503

Since 1988, the Flemish government has gained the juris- professionals and large proprietors for 11.5%. Most
diction to implement and govern its own educational sys- respondents attended the general track (46.7%), with
tem, which limits the study to the students and schools in 28.5% attending the technical, 22% the vocational, and
this region. Of the respondents in the sample, 51.20% 2.7% the arts track. Schools were equally divided across
attended the third grade, and 48.8% the fifth grade (cor- school sector (50.6% public, 49.4% private), and had an
responding to grades 9 and 11 in the US system). Hence, average school size of 461.55 (SD = 285.27).
the majority of students are 15 (34.8%) or 17 (32.6%) years For the sampling procedure, we used multistage sam-
old, with other students being a bit older than most students pling. At first, we selected proportional-to-size postal
in their grade, mostly due to being held back a grade codes, size being defined by the number of schools within
because they did not succeed in a previous year (11.3% each postal code, information provided by the Educational
being 16 years old; 14.3% 18 years old; 4.6% 19 years old, Department. From the 240 postal codes, we selected a
and 1.4% being 20 years old). The sample was equally random sample of 48 codes. Because of this strategy, postal
divided by gender (51.4% girls; see Table 1). The majority codes of large municipalities had a greater chance of
of respondents were natives (88.8%). Most immigrants selection. This strategy was used to capture the fact that
(11.2%) had Turkish or Moroccan backgrounds (both about larger municipalities have a greater number of schools and
30% of immigrants in our sample), some had Southern- a greater number of migrants: we had to make sure that—in
European (16%), Eastern-European (8%), North-African correspondence with their overrepresentation in the Flem-
(5%), or other (17%) backgrounds. Family occupational ish context—schools in larger municipalities had a higher
background (SES, see below) was unskilled manual labor chance to be selected, making sure at the same time that we
for 7.9% of students, specialized manual labor for 6.3%, selected a critical amount of schools with a majority of
skilled manual labor for 6.3%, routine non-manual migrant students. The sampling resulted in the desired
employees for 11.2%, farmers and smallholders for 9.3%, overrepresentation of larger municipalities. Consequently,
lower grade employees and administrators for 23.6%, we selected all regular secondary schools in the chosen
higher-grade administrators and executives for 17.6%, and postal codes that provided a third and fifth grade

Table 1 Descriptive statistics


Variables % Mean SD Cronbach’s alpha N
for variables: frequencies (%),
means (M), standard deviations Dependent variable
(SD), Cronbach’s alpha, and N
School misconduct 30.04 8.47 0.87 11,561
School level
Cohesion 61.24 2.81 85
Ethnic composition 16.45 21.70 85
SES composition 4.80 1.23 85
School sector
Public 50.60 85
School size 461.55 285.27 83
Student level
Peer attachment 15.81 2.76 0.74 11,554
Perceived teacher support 23.99 3.99 0.75 11,621
General school belonging 37.11 6.42 0.80 11,543
Gender
Female 51.40 11,843
Grade
Third 51.20 11,872
Ethnicity
Immigrant 11.20 11,870
SES 5.20 2.10 11,137
Parental attachment 28.32 5.61 0.83 11,722
Vocational track
Vocational 22.00 11,872
Prior achievement 69.42 9.22 10,713

123
504 J Youth Adolescence (2012) 41:499–514

(corresponding to years 9 and 11 in the US system), variable was significantly skewed (1.58, SE = 0.023)
yielding a response rate of 31%. This low response rate is toward its lower end. Hence, we performed an overdi-
due to schools in Flanders being swamped with research spersed Poisson model with constant exposure on the data
requests. Schools choose the research they take part in on a to account for possible non-linearity, but the same basic
first-come, first-served basis. Analyses in which we com- picture emerged as with linear methods. Below we present
pared our sample to the Flemish school population, based the results of the more readily interpretable linear
on information attained through the Flemish Educational technique.
Department, showed that the participating schools did not
differ from those that opted out in terms of school sector, Bonding Measures
size, curriculum, or student composition. Hence, no sys-
tematic biases occurred, and the 85 schools in the sample Before we introduce our bonding measures, a word is in
are representative of the Flemish situation (Van Houtte order about the use of terminology in this kind of research,
et al. 2005). In the participating schools, we asked all third- which has been lacking in clarity and consistency (for
and fifth-grade students present at the time of the visit to fill reviews, see Libbey 2004; Barber and Schluterman 2008).
out the questionnaire. All students participating in the study School connectedness has been designated with all kinds of
did so with their parents’ consent. Students filled out the different terms, including ‘‘bonding’’, ‘‘attachment’’ and
questionnaire in class, supervised by members of the ‘‘connection’’, without clearly specifying what these terms
research team and a teacher. A few students were not mean. Here, we follow the conceptualizations proposed by
present, due to absence or field trips. A total of 11,945 Libbey (2004).
students completed the questionnaire, of which 11,872 Our first measure pertains to peers. The 4 items used are:
(response rate: 87%) proved valid. The questionnaires were ‘‘I wish I had other friends at school’’, ‘‘My friends accept
not anonymous because we needed to couple other data me as I am’’, ‘‘I trust my friends at school’’ and ‘‘My
provided by the school with the students’ responses. Ulti- friends at school respect my feelings and ideas’’. As these
mately we removed all names, so all analyses were per- items refer to emotional feelings of connection (see Libbey
formed on anonymous data. 2004, p. 274), we call this peer attachment. Respondents
could answer on a five-point scale, ranging from Does not
Measures describe me at all accurately to Describes me accurately
(1–5). Scores were summed up to a scale, ranging from 4 to
School Misconduct 20. Our respondents had a mean of 15.81 (SD = 2.76). The
alpha for the scale was 0.74.
We measured school misconduct using a scale inspired by The measures concerning bonding to teachers and the
Stewart (2003, p. 602–604), consisting of 17 items (See school context are derived from the Psychological-Sense-
‘‘Appendix’’). Students were asked how often they com- of-School-Membership-scale (Goodenow 1993). This is a
mitted deviant acts, such as ‘‘smoked at school,’’ ‘‘been late scale consisting of 18 items (see Table 2), to which
for school,’’ ‘‘done drugs at school,’’ and so forth. Students respondents could answer on a five-point scale, ranging
could answer using a 5-point scale, ranging from never (1) from totally disagree to totally agree (1–5). On this scale,
to very often (5). Scores were summed to a scale ranging we performed a Principal Component Analysis (PCA;
from 17 to 85 (mean = 30.04, SD = 8.47, alpha = 0.87; n = 11,872), using Varimax rotation. This resulted in four
see Table 1). It has been shown that using self-reported components, one consisting of teacher or adult related
measures is not ideal for measuring deviant acts (Crosnoe assertions, accompanied by three more general support
2002), but it nonetheless remains the most common method related items (see Table 2; items marked by T). The items
of gathering such information (e.g. Stewart 2003; Gott- loaded with a minimum of 0.454 (item 7) and a maximum
fredson et al. 2005). We interpolated missing values by of 0.705 (item 5). A confirmatory PCA (extracting one
item correlation substitution (Huisman 1999): a missing component; n = 11,872) confirmed the high loadings of
item was assigned the value of the most highly correlated these seven items on this one component (ranging from
item. ICS is a simple and relatively easy to implement 0.472 to 0.750). Items were summed to a scale, ranging
technique, used to handle missing responses to a scale. from 7 to 35. As these items pertain to whether students
Although they tend to overestimate the scale quality, pro- feel valued by their teachers (see Libbey 2004, p. 281), and
cedures which use the relationships between items—such the scale is reported by students, we call this perceived
as ICS—perform best, especially in scales with few teacher support. Our respondents had a mean of 23.99
response options and a low percentage of missing values (SD = 3.99). The alpha was 0.75 (n = 11,612).
(see Huisman 1999). As is common for delinquency As the other three components derived from the PCA
measures (Crosnoe 2002; Stewart 2003), the dependent pertained to whether students feel at home at school, and

123
J Youth Adolescence (2012) 41:499–514 505

Table 2 The psychological sense of school membership (PSSM) Square-Within Mean Square)/Between Mean Square, and
scale (Goodenow 1993, p. 84) should be larger than 0.60 for a legitimate aggregation. For
Items sense of belonging, the ICC was 0.91. Schools differed
significantly on their mean level of belonging (F = 10.736;
1. I feel like a real part of this school
p \ 0.001). This means that the sense of belonging is
2. People here notice when I’m good at something T indeed shared between students from the same school, and
3. It is hard for people like me to be accepted here. (reversed) that it is legitimate to speak about cohesion at the school
4. Other students in this school take my opinions seriously level. The mean value on our cohesion measure over all
5. Most teachers at this school are interested in me T schools was 61.24 (SD = 2.81).
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 Ethnicity
a problem
8. People at this school are friendly to me
We assessed ethnicity using multiple questions. The prin-
9. Teachers here are not interested in people like me. (reversed) T
cipal criterion was the birthplace of the maternal grand-
10. I am included in lots of activities at this school mother. If missing (1%), we considered the nationality of
11. I am treated with as much respect as other students T the parents, as most immigrants are second- and third-
12. I feel very different from most other students here. (reversed) generation citizens and have Belgian nationality. As is
13. I can really be myself at this school common, only West European birthplaces and nationalities
14. The teachers here respect me T were considered as native descent (Timmerman et al.
15. People here know I can do good work T 2002). Additional criteria in case of missing data regarding
16. I wish I were in a different school. (reversed) nationality (father: 4%, mother: 3.3%) were the language
17. I feel proud of belonging to this school spoken at home (other than Dutch), religion (Islam), and
18. Other students here like me the way I am the student’s name (Felouzis 2003). This resulted in a
dichotomous variable (coding: 0 = native, 1 = immi-
grant); 11.2% were immigrants.
each in itself does not yield a straightforward interpreta-
tion, we treated the remaining 11 items together in one Socioeconomic Status
scale (see Table 2; items without T). As these items bear
on a broad notion of whether students affiliate to their The SES was measured by the occupation of the father and
schools (see Libbey 2004, p. 278), we call this scale gen- mother (Erikson et al. 1979), or, if unemployed, their last
eral school belonging. A confirmatory PCA (extracting one occupation (1 = Unskilled manual labor; 2 = Specialized
component; n = 11,872) on these eleven items revealed manual labor; 3 = Skilled manual labor; 4 = Routine non-
factor loadings ranging from 0.387 (item 4) to 0.705 (item manual employees; 5 = Farmers and smallholders; 6 =
6). There was one outlier with a loading of 0.298 (item 10), Lower grade employees and administrators; 7 = Higher-
but item analysis did not show a substantial improvement grade administrators and executives; 8 = Professionals and
of alpha when deleting this. Alpha for this scale was 0.80 large proprietors). We used the highest ranked occupation as
(n = 11,543). The scores were summed, yielding a mini- the family SES. The mean was 5.20 (SD = 2.10).
mum of 11 and a maximum of 55. Students had a mean of
37.11 (SD = 6.42). Ethnic Composition

Cohesion As is common in educational research (see Vervoort et al.


2010; Demanet and Van Houtte 2011), we measured ethnic
To assess school cohesion, we aggregated the students’ composition by the proportion of immigrants at school. We
individual beliefs to the school level, by calculating the asked the administrators to estimate this; however, 12
mean sense of belonging (see Goodenow 1993) in each (14.12%) of the 85 administrators did not respond to this
school (e.g. Hofstede et al. 1990). However, first, we had to question. This poses a problem, as multilevel analysis does
investigate whether the sense of belonging was really not permit missing values at the school level and using this
shared between students of the same school. For this, we measure would mean we have to omit 12 schools from our
calculated the ‘‘mean rater reliability’’ (Shrout and Fleiss sample. However, we additionally computed the proportion
1979; Glick 1985), which is based on the intra-class cor- of immigrants at school using the students’ ethnicity at the
relation (ICC) in a one-way ANOVA with sense of individual level (see above). The correlation of 0.88
belonging as dependent variable and schools as factor. (p \ 0.001) between the two measures validates the use of
The ICC is calculated by the formula: (Between Mean this aggregated measure. The 85 schools in our sample

123
506 J Youth Adolescence (2012) 41:499–514

cover the entire range of ethnic composition, from 0% (6 In our data, students ranged from 41 to 100%, with a mean
schools) to 88.20% (1 school). The mean of this measure of 69.42 (SD = 9.22; see Table 1), corresponding grossly
was 16.45 (SD = 21.70). to a ‘C’ in the US high school system. This measure should
be considered carefully. As no standardized (for example,
SES Composition state administered) tests exist in Flanders, it is hard to
compare measures of academic achievement across
The schools’ SES composition was measured by calculat- schools. Furthermore, being self-reported, this measure
ing the mean SES of students (see above) per school (see could contain biases due to memory problems and cover-up
Rumberger and Palardy 2005; Demanet and Van Houtte strategies. As a result, it has a large number of missing
2011). The mean in the sample was 4.80 (SD = 1.23). values (9.8%). Consequently, we entered the measure only
at the very last step.
Parental Attachment
Data Analysis
To measure parental attachment, we used a scale of 7
As we have a clustered sample—students nested within
items, answerable on a 5-point-scale (Brutsaert 2001).
schools—it is necessary to use multilevel analysis (HLM6;
Examples of items used are ‘‘If I want to tell my parents
Raudenbush and Bryk 2002). First, we estimated an
something, they act as if they don’t hear me’’, and ‘‘I don’t
unconditional ‘‘null’’ model to determine school-level
think my parents believe I can do something good’’. Scores
variance in school misconduct. Then, we added the vari-
on the items were summed, resulting in a scale ranging
ables sequentially as explained above. Throughout our
from 7 to 35. Missing answers were imputed by item
analyses, we controlled for several variables at the school
correlation substitution (Huisman 1999). Respondents had
and individual level. At the school level, we considered
a mean of 28.32 (SD = 5.61). The alpha was 0.83.
school sector, SES composition and ethnic composition.
We should point out that there exists a high correlation
Vocational Track between these two latter features (r = -0.784; p \ 0.001,
See ‘‘Appendix’’). Because other studies have established
We also accounted for whether students attended a voca- an effect of ethnic composition next to the effect of SES
tional track. The Flemish school system can be categorized composition (e.g. Eitle and Eitle 2003), we used the vari-
as ‘‘explicit school-level tracking to different school types ables simultaneously, although the results should be
catering to specific student groups’’, using achievement as regarded with caution due to possible multicollinearity. At
a selection criterion (Trautwein et al. 2006, p. 789). There the school level, it would be important to control for school
are four main tracks in which students can enrol: general size (see Stewart 2003). However, two schools provided no
education, technical, vocational, and artistic education, the information on this variable. As multilevel analysis does
latter being a rather marginal track, in terms of number of not permit missing values at the second level, and analysis
students. The different tracks are commonly classified of 83 schools showed that school size exerts no influence
hierarchically, placing vocational tracks at the lower end. on school misconduct (c = 0.000; SE = 0.001; p [ 0.05),
Hence, attending a vocational track in Flanders is rarely a we omitted this variable from the analyses. At the student
positive choice, and vocational students are all too aware of level, then, it is obvious to control for sociodemographic
their low status in society, yielding more anti-school atti- variables such as SES, gender (coding: 1 = girl), grade
tudes and more misconduct (for an extended discussion of (coding: 1 = fifth grade) and ethnicity (coding: 1 =
the Flemish tracking system, see Van Houtte and Stevens immigrant). Furthermore, we control for parental attach-
2008). Moreover, immigrants in Flanders are overrepre- ment (Dornbusch et al. 2001), and attending a vocational
sented in the vocational track. Because of all this, it was track (coding: 1 = vocational track) (see Van Houtte and
necessary to account for this in our analyses. Among Stevens 2008). In the last model, we controlled additionally
students, 22% attended a vocational track (coding: 1 = for prior achievement. This was done only in the last step,
vocational track). because this measure was a rather crude one (see Measures
section) and had a large number of missing values. How-
Prior Achievement ever, as prior achievement has been linked to both school
deviance and school belonging (Stewart 2003), it was
Prior achievement was measured by GPA (Grade Point necessary to control for this. To ensure model stability, all
Average) from the preceding school year. To grade their but the dichotomous variables were grand mean centered.
students, Flemish high-schools use a percentage, hence, Below, we present both the unstandardized (c) and the
grades range from 0 to 100%, 50% being the passing grade. standardized (c*) results.

123
J Youth Adolescence (2012) 41:499–514 507

Results school misconduct (c* [ 0.10 in model 3). Of all our


school variables, only school sector has such a strong
From the unconditional model (see Table 3), we see that association with the outcome. While prior achievement
7.1% (r2 = 67.107; s0 = 5.156; p \ 0.001) of the vari- does have a significant relation to school misconduct
ance in school misconduct occurs between schools, war- (c* = -0.139; p \ 0.001), including it in model 5 has no
ranting the introduction of school-level determinants. influence on the other effects.

School and Individual-Level Effects of Bonding Differences by Social-Ethnic Context

Our first model shows that our school-measure of cohesion The variance components in the third model show that
has an association with school misconduct (c* = -0.044; there is significant slope variation between schools for peer
p \ 0.01, see Table 4). Adding the individual-level peer attachment (ru2 = 0.058; p \ 0.05), perceived teacher
attachment in the second model decreases this association support (ru2 = 0.035; p \ 0.01) and general school
slightly, but it remains significant (c* = -0.039; p \ 0.05). belonging (ru2 = 0.011; p = 0.053), warranting the testing
However, when perceived teacher support and general of cross-level interactions. In the fourth model, we entered
school belonging are added (model 3), the association ethnic and SES school composition to account for the slope
between the school-measure of cohesion and school mis- variation of, respectively, peer attachment, perceived tea-
conduct vanishes (c* = 0.018; p [ 0.05). We should also cher support and general school belonging. However, these
point out that the explanatory power of our four school- school factors did not explain the slope variation of the
level variables is limited, as our analyses suggest that they bonding measures, hence no evidence was found for cross-
explain 7% of school-level variance. level interactions. To test whether this lack of significant
interaction effects could be due to the high correlation
The Differential Effects of Peer, Teacher, and School between SES and ethnic school composition (r = -0.784;
Bonding p \ 0.001, See ‘‘Appendix’’), we additionally performed
model 4 with SES composition and ethnic composition
Research question 2 is answered in the second and third separately (not shown). These analyses yielded the same
model. While, in the second model, peer attachment in results as the ones shown in Table 4.
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 Discussion
positive association of peer attachment with school mis-
conduct (c* = 0.070; p \ 0.001), although the standard- School deviance is a widely studied topic in adolescent
ized coefficient shows this to be a weak association. research. One of the most influential explanations for the
Perceived teacher support (c* = -0.170; p \ 0.001) and occurrence of school misbehavior is offered by the schools-
general school belonging (c* = -0.096; p \ 0.001) are as-communities perspective (Battistich et al. 1995), which
significantly negatively related to school misconduct. states that preventing school deviancy requires that stu-
Furthermore, at the individual pupil level, gender, grade dents feel part of a caring school community and have
and parental attachment hold strong associations with 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
Table 3 HLM unconditional model characteristics: Variation contend that feelings of belonging have important pre-
between schools in student school misconduct
ventive effects on school misbehavior (e.g. Dornbusch
Characteristic Value et al. 2001), opinions are divided whether efforts should be
directed at transforming the school into a community by
Intercept 30.180***
building cohesion between actors at the school level
Parameter variance
(Battistich et al. 1995), or whether it is more efficient to
Within school 67.107
impact students’ personal feelings of belonging (Goode-
Between schools 5.156
now 1993) in order to prevent school misconduct. While
HLMa reliability estimate 0.862
some studies in the past have indeed pointed to the mul-
Proportion of variance between schools 0.071***
tilevel nature of the sense of belonging to school, no study
a
HLM hierarchical linear modelling has yet investigated the relative importance of school-level
2
v (84, N = 11,872) = 887.970 cohesion versus individual feelings of peer, teacher,
*** p B 0.001 and school attachment in relation to school misconduct.

123
508 J Youth Adolescence (2012) 41:499–514

Table 4 Association between peer attachment, perceived teacher support, and general school belonging and school misconduct. Results of
sequential multilevel analysis
Model 1 Model 2 Model 3 Model 4 Model 5

Intercept c 29.071*** 29.066*** 29.304*** 29.307*** 29.329***


SE 0.552 0.540 0.536 0.543 0.516
School level
Ethnic composition c 0.001 0.000 0.001 -0.004 0.015
c* 0.002 0.001 0.002 -0.010 0.038
SE 0.013 0.014 0.014 0.014 0.013
SES composition c -0.094 -0.161 -0.259 -0.292 0.055
c* -0.014 -0.023 -0.038 -0.042 0.008
SE 0.219 0.222 0.222 0.228 0.194
Public sector c 1.786*** 1.792*** 1.556*** 1.561*** 1.491***
c* 0.105*** 0.106*** 0.092*** 0.092*** 0.088***
SE 0.322 0.330 0.320 0.321 0.269
Cohesion c -0.131** -0.117* 0.055 0.060 0.048
c* -0.044** -0.039* 0.018 0.020 0.016
SE 0.049 0.052 0.062 0.062 0.057
Student level
Female gender c -3.097*** -3.035*** -2.921*** -2.924*** -2.663***
c* -0.183*** -0.179*** -0.172*** -0.173*** -0.157***
SE 0.217 0.230 0.222 0.223 0.216
SES c 0.131** 0.120** 0.145** 0.144** 0.177**
c* 0.033** 0.030** 0.036** 0.036** 0.044**
SE 0.041 0.042 0.044 0.044 0.050
Fifth grade c 1.178*** 1.160*** 1.100*** 1.101*** 0.975***
c* 0.139*** 0.137*** 0.130*** 0.130*** 0.115***
SE 0.106 0.106 0.100 0.100 0.110
Parental attachment c -0.324*** -0.327*** -0.225*** -0.226*** -0.202***
c* -0.215*** -0.216*** -0.149*** -0.149*** -0.134***
SE 0.018 0.018 0.017 0.017 0.018
Immigrant c -0.576 -0.710 -0.577 -0.550 -0.623
c* -0.021 -0.026 -0.021 -0.020 -0.023
SE 0.389 0.404 0.402 0.403 0.404
Vocational track c 1.249** 1.218** 1.056** 1.081** 1.420***
c* 0.060** 0.059** 0.051** 0.052** 0.069***
SE 0.384 0.405 0.374 0.377 0.345
Peer attachment c -0.012 0.216*** 0.210*** 0.201***
c* -0.004 0.070*** 0.069*** 0.065***
SE 0.035 0.040 0.042 0.041
Perceived teacher support c -0.361*** -0.355*** -0.315***
c* -0.170*** -0.167*** -0.149***
SE 0.032 0.034 0.036
General school belonging c -0.126*** -0.124*** -0.117***
c* -0.096*** -0.094*** -0.089***
SE 0.020 0.021 0.021
Prior achievement c -0.128***
c* -0.139***
SE 0.013

123
J Youth Adolescence (2012) 41:499–514 509

Table 4 continued
Model 1 Model 2 Model 3 Model 4 Model 5

Interaction terms
Peer attachment*ethnic composition c -0.003 -0.002
SE 0.003 0.003
Peer attachment* SES composition c -0.029 -0.007
SE 0.044 0.045
Perceived teacher support*ethnic composition c 0.003 0.004
SE 0.003 0.003
Perceived teacher support*SES composition c 0.012 -0.003
SE 0.044 0.045
General school belonging*ethnic composition c 0.000 -0.001
SE 0.002 0.002
General school belonging* SES composition c -0.005 -0.006
SE 0.030 0.027
Variance components
Intercept U0 10.310 8.596 8.996*** 9.201*** 6.538***
Female gender U1 1.452 1.840 1.707*** 1.696*** 1.451***
SES U2 0.022 0.034 0.041 0.041 0.059
Fifth grade U3 0.339 0.349 0.297* 0.300* 0.428**
Parental attachment U4 0.011 0.010 0.009* 0.009* 0.011*
Immigrant U5 3.819 4.441 4.711** 4.834** 3.782**
Vocational track U6 3.474 4.014 3.228*** 3.224*** 2.006***
Peer attachment U8 0.036 0.058* 0.060* 0.046
Perceived teacher support U7 0.035** 0.036** 0.040**
General school belonging U9 0.011$ 0.012* 0.009
Prior achievement U10 0.008**
Model deviance 74,593.94*** 73,420.39*** 72,395.97*** 72,441.45 66,225.11***
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 more in disadvantaged school contexts (Battistich et al.
schools-as-communities perspective should distinguish 1995). However, especially in the case of deviancy, this
between the different actors bonded with at school. claim has received mixed empirical support. In the current
Regarding school deviancy, the perspective currently is study, we considered the SES composition and the ethnic
based largely on early control theory (Hirschi 1969), a composition of schools as indicators of school disadvan-
theoretical approach that has neglected the characteristics tage (Demanet and Van Houtte 2011), to investigate
of the actor involved in the bonding. Differential associa- whether feelings of support were indeed more strongly
tion theory (Sutherland and Cressey 1978), on the contrary, related to less school misconduct in schools with a lower
states that closer peer relationships might generate more mean SES and a higher proportion of minority students.
deviancy if antisocial socialization occurs in the friendship To assess the first research question, namely what the
group. This is likely in adolescence, as deviancy is relative importance is of school-level cohesion versus
accepted by a certain number of adolescents as a valid form individual feelings of belonging, we tested whether the
of behavior (Allen et al. 2005). Led by this theoretical association of school cohesion is maintained when con-
exploration, we expected peer attachment to relate differ- trolling for three aspects of students’ sense of belonging—
ently to school misconduct than general school belonging peer attachment, perceived teacher support, and general
and perceived teacher support. In a third contribution to the school belonging. We found that, in association to school
schools-as-communities perspective, we investigated a misconduct, individual bonding seems more important than
long-held claim by researchers that providing students with overall school cohesion, as no relation of school cohesion
supportive relationships should impact those students even is seen once the three aspects of individual belonging are

123
510 J Youth Adolescence (2012) 41:499–514

taken into account. This finding can be explained in two social and teacher support. Although their measure of
ways. First, it could be due to selection. A selection effect social support does not specifically refer to peers, they
is found when it is clear that school-level effects are conclude that conventional bonds to teachers can counter-
actually due to features of the students enrolled in the balance unconventional peer bonds. Baker (1998, p 35)
school, and as such are not an effect over and above these gives a more process-oriented explanation for our results,
individual student effects. In this case, schools that are stating that when students fail to form meaningful con-
cohesive are simply made up of more students with a high nections to their school and its teachers, they turn to other
sense of teacher support and general school belonging, youngsters to fulfill their need for belonging, subsequently
which is why these students report less school misconduct. forming sub-communities within the school. It is note-
Or second, it could be due to a mediation effect: in this worthy that this occurs especially with disruptive students,
case, cohesion at the school level yields a higher perceived because teachers react more harshly to them (Good and
support received from teachers in students and a higher Brophy 1994), and they are likely to be rejected by their
sense of overall school belonging, which is responsible for prosocial peers (Salmivalli et al. 1997). As such, small,
their lower school misconduct. As we have no longitudinal deviant peer groups are formed in which students incite
design, however, we can not differentiate between the each other to continue to break the rules (Bendixen et al.
selection or mediation explanations. Whichever is the case, 2006).
our results suggest that intervention efforts designed to We should, however, point to a possible alternative
improve a school’s cohesiveness will not affect school explanation for our findings. As this is not a longitudinal
deviancy directly (cf. Nelson 1996). In the mediation case, study, we cannot ascertain the direction of the effects. In
at most these efforts might create more sense of belonging fact, longitudinal research shows that effects are likely to
in individual students, making them less prone to school be bidirectional (Hirschfield and Gasper 2011). Hence, it
misconduct. Therefore, creating a cohesive school may not has been shown that students who are deviant at school also
be a sufficient condition to combat school misconduct. Our are more likely to feel less at home in the school context
results suggest that, to counter school deviancy, interven- and perceive disconnection from their teachers (Karcher
tion efforts should focus directly on individual feelings of 2002). It is also likely that deviant students tend to be
school belonging. However, we should point out that, as popular among their friends (see also Demanet 2008). As in
this is not a longitudinal study, we can not be sure about the adolescence, peer norms tend to favor minor deviant acts,
causal direction of these relationships. students who display school misconduct are likely to be
Regarding the second research question, our results popular (Allen et al. 2005). Rather than being determinants
show that, while higher school belonging and perceived of school misconduct, bonding relationships can be out-
teacher support are related to less school misconduct, comes of school-deviant behavior.
greater peer attachment was associated to more school In response to the third research question, we found no
misconduct. It is noteworthy that the association between evidence for the claim that supportive relationships should
peer attachment and school misconduct only showed up be related even more strongly to school misconduct in
when we controlled for general school belonging and disadvantaged schools: none of the cross-level interaction
perceived teacher support, meaning that both characteris- effects were significant. This concurs with research relating
tics buffer the relation of peer attachment with school this to neighborhood disadvantage (Dornbusch et al. 2001).
misconduct. This occurs because students who feel bonded A possible explanation for the lack of support found for
to their school and teachers generally perceive a high this claim is that having supportive relationships is not
degree of peer attachment as well. Students who are bon- more beneficial for all students in disadvantaged schools.
ded to all three sources are less deviant than others, As described above, the rationale for interaction effects is
because the preventive effect of teacher support and school that disadvantaged students will benefit disproportionally
belonging overshadows the deviance-generating effect of from school communality because they have less social
their high peer attachment. However, students lacking capital outside of school. Hence, this does not apply to
teacher support and school belonging miss this suppression advantaged students in disadvantaged schools. It is possi-
effect, so that the deviance-yielding effect of peer attach- ble, then, that supportive relationships are more beneficial
ment is free to occur. Hence, for students feeling attached for disadvantaged students, whatever the school context
to their peers, yet lacking teacher and school bonding, more they enroll in, than for advantaged students.
misconduct can be expected. This concurs with earlier We should note the limitations of this study and propose
research. In a recent longitudinal study, Freidenfelt Lilje- some remedies to be addressed in future research. First, as
berg et al. (2011) point to the role of teacher attachment in we mentioned earlier, our design was cross-sectional, so we
preventing delinquency. Further evidence is provided by cannot make any causal claims. Research has established
McNeely and Falci (2004) who distinguished between that the association between bonding and misconduct is

123
J Youth Adolescence (2012) 41:499–514 511

bidirectional (e.g. Karcher 2002; Hirschfield and Gasper do propose that future studies test whether the conclusions
2011), and lacking longitudinal data, we can not ascertain in this study also apply to students in the vocational track
which direction is dominant. Furthermore, previous longi- and to immigrants.
tudinal research showed that exposure to deviant peers was In summary, our study’s main finding is that, although
not equally predictive of later problem behavior at all ages associations appear to be small, bonding with different
(Fleming et al. 2010). Hence, we call for additional lon- actors at school can have mixed connections to deviant
gitudinal research to investigate the intertwined roles of behavior. This contradicts early control theory, on which
school cohesion and individual feelings of bonding, and the schools-as-communities perspective, with reference to
link this to developmental processes. Second, our reliance school deviancy, is now largely based. Following the
upon a single informant—namely the student—for all the addition of differential association theory to the early
individual-level variables could create problems of shared version of control theory, we argue that it is vital to
method variance: because all variables are reported by the account for the characteristics of the actors with whom the
same individual, correlations could be estimated larger student bonds, and especially whether they endorse devi-
than they actually are. Moreover, self-reports are not ancy or not. It seems important to incorporate this in the
always accurate representations of actual levels of both schools-as-communities perspective. Furthermore, our
deviance (Crosnoe 2002) and bonding (Stevens et al. results support earlier findings that receiving teacher
2002). Future studies on this topic should combine reports support yields multiple positive outcomes for students,
by students, teachers, and parents to avoid these problems. including less involvement in health-risk behavior (West
Third, in this study, we did not distinguish between et al. 2004), and higher engagement and achievement
attachment to deviant or non-deviant peers because we (Klem and Connell 2004). Hence, we contribute to the
assumed that all students get socialized in the broadly held evidence base showing that schools should be organized
peer norms that school misconduct is acceptable behavior as caring communities. Adults at school play a large role
(see also Allen et al. 2005), whether their close friends in this, as they are responsible for making all students,
were deviant or not. However, our conclusion led us to even the disruptive and poor achieving ones, feel at home
hypothesize that students with less perceived teacher sup- at school.
port and general school bonding form sub-communities
within the school, in which we expected deviancy to be Acknowledgments The authors are grateful to Simon Boone, the
Editor, and three anonymous reviewers for their helpful feedback on
endorsed. Hence, we arrive at the same conclusion as many this manuscript.
former studies (e.g. Bendixen et al. 2006), namely that
especially attachment to peers who themselves perform
deviant acts promotes school misconduct. Future research Appendix: The School Misconduct Scale (inspired
should distinguish between attachment to prosocial and by Stewart 2003, p. 602–604)
antisocial peers in order to compare their influence on
school misconduct to teacher support and general school How often have you:
belonging. Fourth, as this study used a quantitative
approach, we only could gain general information from our 1. been late for school
respondents, and had less insight into the perceptions of the 2. skipped lessons
actors in question. Future qualitative research should 3. skipped school all day
attempt to understand on a more micro level how sociali- 4. cheated on tests
zation takes place in close interactions, and how the buf- 5. copied someone’s homework
fering effect of receiving teacher support is experienced by 6. not made your homework
7. fought at school
students themselves. Lastly, we should note a small
8. stolen at school
selection bias with regard to our analytical sample. Stu-
9. committed vandalism at school
dents from the vocational track and immigrants were
10. smoked at school
overrepresented among the respondents excluded from the
11. drunk alcohol during school hours
analyses due to missing responses. However, because of
12. done drugs during school hours
the small number of missing cases, this is not likely to
13. talked back at teachers
affect our overall results. Moreover, we tried to account for
14. broke the school rules
this by controlling for the socio-demographic variables in
15. had to do impositions
question. Indeed, further analyses—in which we imputed
16. been sent to detention
all missing values by means of mean imputation—con-
17. been suspended for one or more days
firmed the conclusions of this study. Although it is not
likely that this small selection bias impacts our results, we See Tables 5 and 6.

123
512 J Youth Adolescence (2012) 41:499–514

Table 5 Pearson product moment correlations between individual-level variables


School SES Parental Peer Perceived General school
misconduct attachment attachment teacher belonging
support

SES 0.009
Parental attachment -0.225*** -0.005
Peer attachment -0.093*** 0.085*** 0.231***
Perceived teacher support -0.281*** 0.047*** 0.312*** 0.294***
General school belonging -0.241*** 0.085*** 0.337*** 0.476*** 0.617***
GPA -0.237*** 0.079*** 0.141*** 0.075*** 0.150*** 0.141***
*** p B 0.001

Table 6 Pearson product moment correlations between school-level de l’immigration en Belgique [Research and police: a case study
variables of immigration in Belgium]. Gent: Academia Press.
Brutsaert, H. (2001). Co-educatie. Studiekansen en Kwaliteit van het
Ethnic composition SES composition Schoolleven [Coeducation. Study Chances and Quality of School
Life]. Leuven/Apeldoorn: Garant.
SES composition -0.784*** Bryk, R., & Driscoll, D. (1988). The high school as community:
Cohesion -0.141 0.297** contextual influences and consequences for students and teach-
ers. Madison: Center for Educational Research, University of
** p B 0.01, *** p B 0.001 Wisconsin.
Crosnoe, R. (2002). High school curriculum track, and adolescent
association with delinquent friends. Journal of Adolescent
References Research, 17, 143–167.
Demanet, J. (2008). Populair of verstoten? Een netwerkanalytisch
Agnew, R. (2003). An integrated theory of the adolescent peak in onderzoek naar de sociale kenmerken van pesters in het Vlaamse
offending. Youth and Society, 34, 263–299. secundaire onderwijs [Popular or isolated? A network analytical
Allen, J. P., Porter, M. R., McFarland, F. C., Marsh, P., & McElhaney, study for the social characteristics of bullies in Flemish
K. B. (2005). The two faces of adolescents’ success with peers: secondary schools]. Tijdschrift voor sociologie, 29, 397–423.
Adolescent popularity, social adaptation, and deviant behavior. Demanet, J., & Van Houtte, M. (2011). Social-ethnic school
Child Development, 76, 747–760. composition and school misconduct: Does sense of futility
Baker, J. A. (1998). Are we missing the forest for the trees? clarify the picture? Sociological Spectrum, 31, 224–256.
Considering the social context of school violence. Journal of Dornbusch, S. M., Erickson, K. G., Laird, J., & Wong, C. A. (2001).
School Psychology, 36, 29–44. The relation of family and school attachment to adolescent
Bankston, C., & Caldas, S. J. (1996). Majority African American deviance in diverse groups and communities. Journal of
schools and social injustice: The influence of de facto segrega- Adolescent Research, 16, 396–422.
tion on academic achievement. Social Forces, 75, 535–555. Eitle, D., & Eitle, T. M. (2003). Segregation and school violence.
Barber, B. K., & Olsen, J. A. (1997). Socialization in context: Social Forces, 82, 589–616.
Connection, regulation, and autonomy in the family, school, and Erickson, K. G., Crosnoe, R., & Dornbusch, S. M. (2000). A social
neighborhood, and with peers. Journal of Adolescent Research, process model of adolescent deviance: Combining social control
12, 287–315. and differential association perspectives. Journal of Youth and
Barber, B. K., & Schluterman, J. M. (2008). Connectedness in the Adolescence, 29, 395–425.
lives of children and adolescents: A call for greater conceptual Erikson, R., Goldthorpe, J. H., & Portocarero, L. (1979). Intergener-
clarity. Journal of Adolescent Health, 43, 209–216. ational class mobility in 3 Western European societies—England,
Battistich, V., & Hom, A. (1997). The relationship between students’ France and Sweden. British Journal of Sociology, 30, 415–441.
sense of their school as a community and their involvement in Espelage, D. L., Holt, M. K., & Henkel, R. R. (2003). Examination of
problem behaviors. American Journal of Public Health, 87, peer-group contextual effects on aggression during early ado-
1997–2001. lescence. Child Development, 74, 205–220.
Battistich, V., Solomon, D., Kim, D. I., Watson, M., & Schaps, E. Felouzis, G. (2003). La ségrégation ethnique au collége et ses
(1995). Schools as communities, poverty levels of student conséquences [Ethnic segregation at secondary school and its
populations, and students attitudes, motives, and performance— consequences]. Revue Française de Sociologie, 44, 413–447.
A multilevel analysis. American Educational Research Journal, Finn, J. D. (1989). Withdrawing from school. Review of Educational
32, 627–658. Research, 59, 117–142.
Battistich, V., Solomon, D., Watson, M., & Schaps, E. (1997). Caring Fleming, C. B., Catalano, R. F., Haggerty, K. P., & Abbott, R. D.
school communities. Educational Psychologist, 32, 137–151. (2010). Relationships between level and change in family,
Bendixen, M., Endresen, I. M., & Olweus, D. (2006). Joining and school, and peer factors during two periods of adolescence and
leaving gangs: Selection and facilitation effects on self-reported problem behavior at age 19. Journal of Youth and Adolescence,
antisocial behaviour in early adolescence. European Journal of 39, 670–682.
Criminology, 3, 85–114. Freeman, D. J., Brookhart, S. M., & Loadman, W. E. (1999). Realities
Brans, M., Jacobs, D., Martineillo, M., Rea, A., Swyngedouw, M., of teaching in racially/ethnically diverse schools—Feedback
Adam, I. et al. (2004). Recherche et Politiques publiques, le cas from entry-level teachers. Urban Education, 34, 89–114.

123
J Youth Adolescence (2012) 41:499–514 513

Freidenfelt Liljeberg, J., Eklund, J. M., Väfors Fritz, M., & af Ogbu, J. U. (1978). Minority education and caste. The American
Klinteberg, B. (2011). Poor school bonding and delinquency system in cross-cultural perspective. New York: Academic
over time: Bidirectional effects and sex differences. Journal of Press.
Adolescence, 34, 1–9. Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear
Glick, W. H. (1985). Conceptualizing and measuring organizational models: Applications and data analysis methods. London: Sage
and psychological climate: pitfalls in multilevel research. Publications.
Academy of Management Review, 10, 601–616. Rumberger, R. W., & Palardy, G. J. (2005). Does segregation still
Good, T. L., & Brophy, J. F. (1994). Looking in classrooms. New matter? The impact of student composition on academic
York: Harper Collins. achievement in high school. Teachers College Record, 107,
Goodenow, C. (1993). The psychological sense of school membership 1999–2045.
among adolescents—Scale development and educational corre- Salmivalli, C., Huttunen, A., & Lagerspetz, K. M. J. (1997). Peer
lates. Psychology in the Schools, 30, 79–90. networks and bullying in schools. Scandinavian Journal of
Gottfredson, G. D., Gottfredson, D. C., Payne, A. A., & Gottfredson, Psychology, 38, 305–312.
N. C. (2005). School climate predictors of school disorder: Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: uses in
Results from a national study of delinquency prevention in assessing rater reliability. Psychological Bulletin, 86, 420–428.
schools. Journal of Research in Crime and Delinquency, 42, Stanton-Salazar, R. D., & Dornbush, S. M. (1995). Social capital and
412–444. the reproduction of inequality: Information networks among
Hermans, P. (2004). Applying Ogbu’s theory of minority academic Mexican-origin high school students. Sociology of Education,
achievement to the situation of Moroccans in the low countries. 68, 116–135.
Intercultural Education, 15, 431–439. Stevens, V., De Bourdeaudhuij, I., & Van Oost, P. (2002). Relation-
Hirschfield, P. J., & Gasper, J. (2011). The relationship between ship of the family environment to children’s involvement in
school engagement and delinquency in late childhood and early bully/victim problems at school. Journal of Youth and Adoles-
adolescence. Journal of Youth and Adolescence, 40, 3–22. cence, 31, 419–428.
Hirschi, T. (1969). Causes of delinquency. Berkeley: University of Stewart, E. A. (2003). School social bonds, school climate, and school
California Press. misbehavior: A multilevel analysis. Justice Quarterly, 20,
Hofstede, G., Neuijen, B., Ohayv, D. D., & Sanders, G. (1990). 575–604.
Measuring organizational cultures—A qualitative and quantita- Sutherland, E., & Cressey, D. (1978). Criminology. Philadelphia:
tive study across 20 cases. Administrative Science Quarterly, 35, Lippincot.
286–316. Timmerman, C., Hermans, P., & Hoornaert, J. (2002). Allochtone
Huisman, M. (1999). Imputation of missing item responses: some jongeren in het onderwijs: een multidisciplinair perspectief
simple techniques. In M. Huisman (Ed.), Item nonresponse: [Migrant youngsters in education: a multidisciplinary perspec-
Occurence, causes, and imputation of missing answers to test tive]. Leuven: Garant.
items (pp. 91–119). Leiden: DSWO Press, Leiden University. Trautwein, U., Ludtke, O., Marsh, H. W., Koller, O., & Baumert, J.
Kandel, D. B. (1978). Homophily, selection, and socialization in (2006). Tracking, grading, and student motivation: Using group
adolescent friendships. American Journal of Sociology, 84, composition and status to predict self-concept and interest in
427–436. ninth-grade mathematics. Journal of Educational Psychology,
Karcher, M. J. (2002). The cycle of violence and disconnection 98, 788–806.
among rural middle school students: teacher disconnection as a Van Houtte, M., & Stevens, P. A. J. (2008). Sense of futility: the
consequence of violence. Journal of School Violence, 1, 35–51. missing link between track position and self-reported school
Karcher, M. J., & Finn, L. (2005). How connectedness contributes to misconduct. Youth and Society, 40, 245–264.
experimental smoking among rural youth: Developmental and Van Houtte, M., Stevens, P. A. J., Sels, A., Soens, K., & Van
ecological analyses. The Journal of Primary Prevention, 26, Rossem, R. (2005). De invloed van structurele en composi-
25–36. torische schoolkenmerken op prestaties en welbevinden van
Klem, A. M., & Connell, J. P. (2004). Relationships matter: Linking leerlingen in het secundair onderwijs: een verklaring via
teacher support to student engagement and achievement. Journal cultuur. [The influence of structural and compositional school
of School Health, 74, 262–273. features on achievement and well-being of students in
Konishi, C., Hymel, S., Zumbo, B. D., & Li, Z. (2010). Do school secondary education: An explanation through culture]. First
bullying and student-teacher relationships matter for academic research report (not published). Ghent: Universiteit Gent,
achievement? A multilevel analysis. Canadian Journal of School vakgroep sociologie, onderzoeksgroep jeugd, educatie en
Psychology, 25, 19–39. geslacht.
Libbey, H. P. (2004). Measuring student relationships to school: Vervoort, M. H. M., Scholte, R. H. J., & Overbeek, G. (2010).
Attachment, bonding, connectedness, and engagement. Journal Bullying and victimization among adolescents: The role of
of School Health, 74, 274–283. ethnicity and ethnic composition of school class. Journal of
Ma, X. (2003). Sense of belonging to school: Can schools make a Youth and Adolescence, 39, 1–11.
difference? Journal of Educational Research, 96, 340–349. Wehlage, G., Rutter, R., Smith, G., Lesko, N., & Fernandez, R.
McNeely, C., & Falci, C. (2004). School connectedness and the (1990). Reducing the risk: Schools as communities of support.
transition into and out of health-risk behaviour among adoles- Philadelphia: Falmer.
cents: A comparison of social belonging and teacher support. Wellman, B., & Frank, K. (2001). Network capital in a multilevel
Journal of School Health, 74, 284–292. world: Getting support from personal communities. In N. Lin, K.
Moffitt, T. E. (1993). Adolescence-limited and life-course-persistent Cook, & R. S. Burt (Eds.), Social capital: Theory and research
antisocial-behavior—A developmental taxonomy. Psychological (pp. 233–274). New York: Walter de Gruyter, Inc.
Review, 100, 674–701. West, P., Sweeting, H., & Leyland, A. (2004). School effects
Nelson, J. R. (1996). Designing schools to meet the needs of students on pupils’ health behaviours: Evidence in support of the
who exhibit disruptive behavior. Journal of Emotional and health promoting school. Research Papers in Education, 19,
Behavioral Disorders, 4, 147–161. 261–291.

123
514 J Youth Adolescence (2012) 41:499–514

Author Biographies Mieke Van Houtte received her Ph.D. in Sociology in 2002 from
Ghent University. Currently, she is lecturer at the Department of
Jannick Demanet received his master in sociology in 2007. Since Sociology, Ghent University, and she is head of the CuDOS research
then, he is assistant at the Department of Sociology, Ghent University, team. Her research interests include diverse topics within the
and he is a member of the CuDOS research team. He is currently sociology of education, covering school effects research, education
preparing his Ph.D.-project, covering structural, compositional and and gender, tracking research, and topics related to antisocial school
cultural school determinants of antischool and antisocial attitudes and behavior and school misconduct.
behavior. His research interests further include bullying/victimization,
and the nature and strength of students’ relations with peers, teachers,
parents, and the school.

123

You might also like