You are on page 1of 31

Journal of Interpersonal


An Ecological Systems Approach to Bullying Behaviors Among

Middle School Students in the United States
Chang-Hun Lee
J Interpers Violence 2011 26: 1664 originally published online 3 June 2010
DOI: 10.1177/0886260510370591

The online version of this article can be found at:

Published by:

On behalf of:
American Professional Society on the Abuse of Children

Additional services and information for Journal of Interpersonal Violence can be found at:

Email Alerts:





>> Version of Record - Apr 17, 2011

OnlineFirst Version of Record - Jun 3, 2010

What is This?
Downloaded from by guest on January 31, 2013
Journal of Interpersonal Violence
26(8) 1664­–1693
An Ecological © The Author(s) 2011
Reprints and permission: http://www.
Systems Approach
DOI: 10.1177/0886260510370591
to Bullying Behaviors

Among Middle
School Students in
the United States

Chang-Hun Lee1

The aim of this study is to identify an ecological prediction model of bullying
behaviors. Based on an ecological systems theory, this study identifies signi­
ficant factors influencing bullying behaviors at different levels of middle and
high school. These levels include the microsystem, mesosystem, exo­system,
and macrosystem. More specifically, the ecological factors investi­gated in
this multilevel analysis are individual traits, family experiences, parental
involvement, school climate, and community characteristics. Using data
collected in 2008 from 485 randomly selected students in a school district,
this study identifies a best-fitting structural model of bullying behavior. Findings
suggest that the ecological model accounted for a high portion of variance
in bullying behaviors. All of the ecological systems as well as individual traits
were found to be significant influences on bullying behaviors either directly
or indirectly.

bullying, ecology, ecological system theory, school climate

University of Arkansas at Little Rock

Corresponding Author:
Chang-Hun Lee, PhD, Department of Criminal Justice, University of Arkansas at Little Rock,
2801 S. University Aveue, Little Rock, AR 72204

Downloaded from by guest on January 31, 2013

Lee 1665

Bullying can have substantial consequences for both victims and perpetrators.
Victims of bullying often suffer from psychological harms, suicidal ideation,
and school violence (Kim, Koh, & Leventhal, 2005; Park, Schepp, Jang, &
Koo, 2006). In some countries, for example, school shootings have been rep­
orted as a disastrous response to bullying victimization (Leary, Kowalski,
Smith, & Phillips, 2003; Verlinden, Hersen, & Thomas, 2000). In the United
States, a total of 220 school shootings occurred between 1994 and 1999,
killing 253 (Anderson et al., 2001).
Setting aside damages imposed to victims of bullying, commitment of
bullying behaviors also could bring serious outcomes to perpetrators of bul­
lying. Researchers have found that bullying behavior is a strong and signifi­
cant predictor of delinquency, antisocial behaviors, and violent crime during
puberty and early adulthood (Andershed, Kerr, & Stattin, 2001; Rigby & Cox,
1996; Wong, 2004). Other researchers have suggested that bullying behavior
in schools is easily aggravated into more severe criminal behavior in and out
of school environments (Kinlock, Battjes, & Gordon, 2004).
One nationwide study to determine the prevalence of bullying in the
United States showed that it affected 29.9% of students in middle schools;
13% of the students were bullies, 10.6% victims, and 6.3% both bullies and
victims (Nansel et al., 2001). Studies on the prevalence of bullying and the
seriousness of outcomes of bullying demonstrate the importance of research­
ers adequately understanding bullying in schools and providing appropriate
methods to deter it. An adequate understanding of causal structures of factors
related to bullying may provide useful prevention strategies for intervention
by school administrators.
Despite its significance and importance, bullying behaviors (i.e., bullying
others) have not been adequately investigated until recently. Researchers have
begun to acknowledge the importance of social contexts that could influence
bullying behavior (Espelage & Swearer, 2003; Morrison, 2001). Lee (2009)
investigated bullying behaviors based on an ecological system theory, devel­
oping a multilevel ecological model of bullying behaviors among South
Korean middle school students. Lee (2009) identified a best-fitting structural
model of bullying behaviors and found that his ecological multilevel model
explained a great amount of the variation within bullying behaviors. How­
ever, the study was limited to only Korean sample, indicating that there was
a necessity of such ecological investigation about bullying in Western coun­
tries. In addition, there are several factors that are left uninvestigated or under­
investigated in a series of studies of bullying in Western countries. These

Downloaded from by guest on January 31, 2013

1666 Journal of Interpersonal Violence 26(8)

factors are school climate and teacher’s attitude toward bullying. Although
the importance of such factors has been acknowledged (Espelage & Swearer,
2003; Farrington, 1993), there is limited empirical study focusing on these
This study aims to address some of aforementioned shortcomings found in
bullying literature. Particularly, the current study will apply ecological system
theory (Bronfenbrenner, 1979) as a theoretical framework to identify differ­
ent levels of factors influencing commitment of bullying behavior in schools.
These levels include individual traits, microsystem, mesosystem, exosystem,
and macrosystem levels. Specifically, individual predictors include prior bul­
lying victimization, dominance, impulsivity, attitude toward aggression, and
fun-seeking tendency. Microsystem level includes four different aspects of
children’s lives: interaction with family, teachers, and peers, and life in schools.
Family interaction predictors include parent’s attitude toward bullying, types
of parenting, and experience of domestic abuse. Interactions with teachers
and peers include attitudes toward bullying, intervention, moral authority of
teachers, power dynamics, coercion, and pseudofriendship among peers.
School climate measures contain academic standard, cooperative climate,
moral atmosphere, and policies/regulations regarding bullying. Mesosystem
level includes parental communication with teachers and peers, and macro­
system level includes collectivism/individualism, social disorganization, and
peer group collectivism. Finally, the current research measures three dimen­
sions of bullying: relational, verbal, and physical bullying.

Review of Literature
Definition of Bullying

The most widely used definition of bullying (Olweus, 1993) states, “A person
is bullied when he or she is exposed, repeatedly and over time, to negative
actions on the part of one or more other persons” (p. 8). This definition, although
it is popular, has been criticized by scholars from different cultural and ethnic
contexts, such as Turkey, Japan, China, Korea, and Norway (Naito & Gielen,
2005; Smith, Cowie, Olafsson, & Liefooghe, 2002). The main reason for the
criticism is the lack of cultural and contextual sensitivity to sociocultural dif­
ferences. In these countries, the term bullying is used interchangeably with
social isolation, alienation, collective ostracism, and collective social exclu­
sion (Lee, 2000; Naito & Gielen, 2005). For that reason, this study emphasizes
relational aggression as an equally important dimension as verbal and physi­
cal aggression in bullying behavior and conceptualizes bullying as repeated

Downloaded from by guest on January 31, 2013

Lee 1667

behaviors by one or more individuals that are intended to impose psychologi­

cal and physical harms to, and social isolation of, one or more victims through
physical, verbal, and relational aggression for an extended period of time.

Ecological Model of Bullying

This study borrows Bronfenbrenner’s (1979) ecological system model as a
theoretical framework to understand multilevel etiological factors affecting
bullying behavior in school. As the theory views school environment as “a
set of nested structures” (Bronfenbrenner, 1979, p. 3), the current theoretical
framework allows investigation into simultaneous effects of individual traits
and interpersonal and contextual factors on bullying behaviors. Ecological
system theory contains five components: (a) children, (b) microsystem,
(c) mesosystem, (d) exosystem, and (e) macrosystem. According to Bronfen­
brenner (1979, 1989), children are viewed as the center of their world, and
they interact with their own ecological environments. This indicates that their
behaviors are influenced by not only their own traits but also by the ecologi­
cal contexts with which they are interacting. Microsystem refers to a pattern
of activities, roles, and interpersonal relationships experienced by the chil­
dren. Mesosystem refers to interconnections among two or more microsystems,
and the children actively participate in this setting, such as relationships
between parents, peers, teachers, or neighborhoods. Exosystem refers to the
social setting in which children can be influenced, but they do not necessarily
actively participate. Finally, macrosystem refers to consistencies found at the
level of the culture, which includes belief system, norms, or ideology.
Children as a center of their ecology. According to Bronfenbrenner (1979),
children are a developmental entity that interacts with their ecological sur­
roundings. Consequently, their behaviors are outcomes of interactions as
well as their own traits (Bronfenbrenner, 1979, 1989). An extensive literature
review identified individual factors that were consistently found to be sig­
nificant on bullying behavior across studies. Those factors are age, prior
experience of bullying victimization, dominance, impulsivity, attitude toward
aggression, and fun-seeking tendency (Bentley & Li, 1995; Bosworth,
­Espelage, & Simon, 1999; Boulton, Trueman, & Flemington, 2002;
­Farrington, 1993; Nansel et al., 2001; Pellegrini & Long, 2002). Studies on
these factors consistently suggest that children with higher levels of domi­
nance, impulsivity, positive attitude toward aggression, and fun-seeking ten­
dency are more likely to bully other children.
Microsystem. According to Bronfenbrenner (1979), the most important
concept for microsystem is experience within interpersonal relationships.

Downloaded from by guest on January 31, 2013

1668 Journal of Interpersonal Violence 26(8)

Through experience, children can be influenced by their counterparts. Thus,

Bronfenbrenner argued that experience is “most powerful in shaping the
course of psychological growth” (p. 22). There are four microsystems around
children: experience in the family, experience with teachers, experience with
peers, and experience within the school environment. Regarding bullying
phenomenon, factors about experience in the family include parental vio­
lence, authoritarian parenting, and parental permissive attitudes toward bul­
lying (Christie-Mizell, 2003; Curtner-Smith et al., 2006; Espelage, Bosworth,
& Simon, 2000; Shields & Cicchetti, 2001). Studies on these factors suggest
that children with higher levels of authoritarian parenting, exposure to domes­
tic violence, and positive parental attitude toward bullying are more likely to
bully other children.
Regarding the experience with teachers, there is no empirical research on
the effect of teacher’s attitude toward bullying or interaction between stu­
dents and teachers in regard to bullying behavior in schools. Naito (1990)
argued that teachers’ lack of moral authority has been associated with an
increase in bullying in the classroom. Studies in South Korea suggested that
teachers’ ineffective intervention in bullying might be significantly associ­
ated with the prevalence of bullying (Park, 2003; Yang, 2005).
Review of literature on peer relationship identified four factors influenc­
ing bullying behavior. They are power dynamic, the level of difference
acceptance, coercion of public self, and pseudofriendship networks. The con­
cept power dynamic captures a tendency that some children view bullying as
a way to obtain higher social status, and they bully the less powerful indi­
viduals or groups to maintain dominance (Twemlow, Fonagy, & Sacco,
2001). Level of difference acceptance refers to how much children and their
peer groups tolerate dissimilarity with other classmates, and this factor has
been found to be an important predictor of bullying behavior (Coloroso, 2003;
Young & Sweeting, 2004). In addition, a study identified a discrepancy
between perceptions of how peers see children and their self-images (or
“public self” and “internal image” in Cairns and Cairns, 1991, p. 256) as an
important predictor of bullying victimization. Lastly, this study develops a
concept, pseudofriendship network, that refers to a relationship that provides
children with protection from bullying victimization, but it also requires
them to bully others to maintain the relationship with deviant peer groups.
Finally, experience within school environment refers to school climate in the
sense that children experience and perceive the school settings and are influ­
enced by them. School climate has been suggested as one of the most impor­
tant social environments that influence children’s behaviors (Gottfredson,
2001; Gottfredson & Gottfredson, 1985).

Downloaded from by guest on January 31, 2013

Lee 1669

Mesosystem. A mesosystem consists of social interconnections between par­

ticipants, such as students, teachers, and peers. Once a child enters a new
environment (e.g., school), which is called a setting transition, the child
become a primary link between family and school, so as to constitute a meso­
system (Bronfenbrenner, 1979). Once parents talk to teachers and peers in
schools, their relationships become supplementary links. In terms of school
environments, parental communications with teachers and peers in schools
are mesosystems, and such systems are expected to significantly influence
children’s bullying behavior in schools. Several studies have verified that lack
of parental involvement in school significantly contributed to bullying behav­
iors (Flouri & Buchanan, 2003; Harris & Petrie, 2003; Sullivan, Cleary, &
Sullivan, 2004).
Exosystem. Bronfenbrenner (1989) suggested that the exosystem “encom­
passes the linkage and processes taking place between two or more settings,
at least one of which does not ordinarily contain the developing person . . .
(e.g., for a child, the relation between the home and the parent’s work place)”
(p. 227). This study utilizes parent’s socioeconomic status (SES) as an exam­
ple of exosystem for students. Studies on the effects of family SES on bully­
ing behaviors have produced contradicting evidence. Wolke and his colleagues
(2001) found there was a significant, but weak, negative association between
SES and bullying in England and Germany. Shin (2000), however, found a
positive association between the variables in Korea.
Macrosystem. A recent comparative study on bullying found that the likeli­
hood of bullying others depends on societal characteristics, such as individu­
alism and collectivism (Nesdale & Naito, 2005) and social disorganization
(Farrington, 1993; Laub & Lauritsen, 1998). Nesdale and Naito (2005) col­
lected data from 158 Japanese and 157 Australian students on behavioral
intentions and found that Japanese participants had a greater likelihood of
bullying others due to the collectivistic nature of their culture. Within such
collectivistic societies, school bullying can be understood as a group-based
problem. In addition, collective judgment and punishment carried out by a
group depends on the level of social disorganization. Naito and Gielen (2005)
argued that bullying as a form of collective punishment occurs more often
when a collective society is poorly integrated and experiences relentless
group competition. Consequently, this study expects to find a positive cor­
relation between levels of social disorganization and bullying behaviors in
more collectivistic peer groups or neighborhoods.
The current study employs multiple factors at different levels of school
contexts to investigate a possible ecological system model explaining influ­
ences of the factors on bullying behaviors. There are two main reasons to

Downloaded from by guest on January 31, 2013

1670 Journal of Interpersonal Violence 26(8)

employ Bronfenbrenner’s (1979) ecological system theory. First, this theory

allows us to include in the model multiple factors at different levels, such as
individual traits, interpersonal relationships, family characteristics, school
climates, and community characteristics. Second, because researchers argue
that bullying is a part of the developmental process of children and a devel­
opmental approach requires understanding of social context (Espelage, Holt,
& Henkel, 2003; Farrington, 1993), this theory allows us to include contex­
tual factors, such as school climate and community characteristics. In sum,
because the ecological system theory views school settings as complex and
multilayered social contexts in which children and others interact and inter­
connect with each other, this theoretical framework allows simultaneous inves­
tigation of individual as well as contextual effects.


The survey data utilized in this study were collected through a multistage
cluster sampling design. In September 2008, a school district in a southern
state of the United States was selected for this study. Within the school dis­
trict, four middle schools were randomly selected. From the third sampling
frame, 22 classes were again randomly selected. All of the students in the
selected classes were asked to participate. To obtain parental consent and
assent from students, the researcher first handed out consent forms with a
brief description and purpose of the current study. Students were asked to take
them to their parents for permission. During the subsequent visit to schools,
signed consent forms were collected, and students who received parental per­
mission were asked to participate in the survey. 600 questionnaires were
handed out to those who provided assent. A total of 485 students responded
to the survey, for a response rate of 80.8%.

Endogenous variable. The outcome variable in this study was bullying behav­
ior, measured through a modified version of Olweus’ bully/victim question­
naire (e.g., Lee, 2004; a = .92). The questionnaire measured three dimensions
of bullying behaviors: relational, verbal, and physical bullying behaviors.
The relational bullying behaviors were measured by four items asking, dur­
ing the previous 6 months, whether students had spread rumors, purposely
left a classmate out of things, ignored a classmate, and left a classmate out of

Downloaded from by guest on January 31, 2013

Lee 1671

conversations. The verbal bullying behaviors were measured by six items

asking their behaviors, such as teasing, taunting, threatening, calling names,
making sexual comments, and talking about physical defects. The physical
bullying behaviors were measured by five items asking whether they had
kicked, hit, and punched; cut with sharp objects; deliberately pushed; and
broke bones. The response options for these scales were (1) never, (2) one or
two times, (3) three times, (4) four times, (5) more than four times.
Exogenous variables. The exogenous variables are individual traits, micro­
system, mesosystem, exosystem, and macrosystem (see Lee, 2007, for more
detailed information about the measures). Specifically, individual traits included
dominance, impulsivity, attitude toward aggression, and fun-seeking tendency.
To measure tendency for dominance and impulsivity, a modified version of
the Standardized Personality Evaluation Test (Lee, Byun, & Jin, 1969) was
used. To measure attitude toward aggression, a modified version of the
Beliefs Supportive of Violence Scale from the Houston Community Project
Scale (Dahlberg, Toal, & Behrens, 1998) was used. This modified version
was developed by Bosworth et al. (1999), and the reliability was .71. Fun-
seeking tendency was measured by four agree-disagree items: “I did the
above things [bullying behaviors] to students for fun,” “My friends enjoy it
[bullying behaviors] because they think I am having fun with them,” “It is
fun to watch students go through these things [bullying victimizations],” and
“There is nothing wrong with these things [bullying behaviors] because they
are just for fun.” These items were developed and used in a study of bullying
in South Korea (Lee, 2007; a = .80). The response options for these indi­
vidual traits measures was a five-point Likert scale, with anchors ranging
from strongly disagree to strongly agree.
The microsystem variables included in this study were family interactions,
teacher interactions, peer interactions, and perceived school climate. Family
interactions were indicated by perceptions of parents’ attitude toward bully­
ing, perceptions and experiences of authoritarian parenting, child’s experi­
ences and witness of domestic abuse, and perceived parental views on the
importance of education (Baldry, 2003; Christie-Mizell, 2003; Shields &
Cicchetti, 2001). Teacher interactions were measured by youths’ perception
of teachers’ attitude toward bullying, perception of effectiveness of teachers’
intervention in bullying situations, and perception of teachers’ moral author­
ity (Naito, 1990; Park, 2003; Rodkin & Hodges, 2003). Peer interactions were
measured by four aspects of peer relationship interactions: power dynamic,
level of difference acceptance, coercion of public self, and pseudofriendship
(Cairns & Cairns, 1991; Coloroso, 2003; Graham & Juvonen, 1998; Naito &

Downloaded from by guest on January 31, 2013

1672 Journal of Interpersonal Violence 26(8)

Gielen, 2005). The school climate variable consisted of five subcategories:

perceived school’s academic standards, perceived levels of cooperative school
climate, perception of moral atmosphere, perceived parental involvement in
school boards, and perceived effectiveness of school’s policies, rules, and
programs for antibullying phenomenon. The response option for all micro­
system measures was a five-point Likert scale, with anchors ranging from
strongly disagree to strongly agree.
Mesosystem variables measure interactions between two microsystems.
Two measures used were perceived frequency of parent–teacher communi­
cation regarding knowledge and information about bullying in school and
perceived frequency of parent–peer communication regarding knowledge
and information about bullying in school (Flouri & Buchanan, 2003; Natvig,
Albrektsen, & Qvarnstrom, 2001). The response options for all mesosystem
measures were (1) never, (2) once, (3) a few times, (4) more than a few times,
and (5) very often.
The exosystem variable included in this study was family SES, which
was measured by perceived levels of family SES, parents’ jobs, and monthly
parental income. Macrosystem variables included levels of individualism/
collectivism and levels of social disorganization. The levels of individual­
ism and collectivism were measured by a new version of the INDCOL
(Individualism and Collectivism) measure modified from the original IND­
COL measure developed in 1995 (Singelis, Triandis, Bhawuk, & Gelfand,
1995). To measure youths’ perception of levels of social disorganization,
this study developed three items based on previous research (Sampson &
Groves, 1989; Sampson, Raudenbush, & Earls, 1997): (a) “My neighbors
do not care what my friends do in this area,” (b) “Teenagers in my neigh­
borhood are out of control,” and (c) “It is difficult for kids to make friends
in my neighborhood.” The response option for all macrosystem measures
was a 5-point Likert scale, with anchors ranging from strongly disagree to
strongly agree.

Demographic Characteristics

Table 1 shows the demographic characteristics of the students who partici­

pated in this study. There were slightly more male students (52.8% of the
total) than female students (47.2%). The age ranged between 10 and 15, and
the sample consisted predominantly of White students (88.7% of the total).

Downloaded from by guest on January 31, 2013

Lee 1673

Table 1. Demographic Characteristics of Sample (N = 485)

n % n %

Sex Father’s education

Male 256 52.8 No education    8   1.6
Female 229 47.2 Elementary school    1   0.2
Schools Middle school   19   3.9
Junior high South 114 23.5 High school 153 31.5
Junior high North 142 29.3 2-year college   78 16.1
Middle school South 139 28.7 4-year university   78 16.1
Middle school North   90 18.6 Graduate school or   97 20.0
Age Mother’s education
10    4   0.8 No education    6   1.2
11 161 33.2 Elementary school    1   0.2
12 161 33.2 Middle school   10   2.1
13 116 23.9 High school 143 29.5
14   25   5.2 2-year college   96 19.8
15    2   0.4 4-year university   76 15.7
Graduate school or more 108 22.3
Grade Monthly income
Sixth grade 229 47.2 US$1000 or less   74 15.3
Seventh grade 143 29.5 US$1001-US$1500   46   9.5
Eighth grade 113 23.3 US$1501-US$2000   44   9.1
Race US$2001-US$2500   62 12.8
White 430 88.7 US$2501-US$3000   70 14.4
African American    6   1.2 US$3001-US$3500   49 10.1
Hispanic   19   3.9 More than US$3500   61 12.6
Asian    2    .4 Academic achievement
Others   18   3.7 Upper 10% of class   96 19.8
Perceived SES Upper 20% of class 101 20.8
Upper   35   7.2 Upper 30% of class   79 16.3
Upper middle 144 29.7 Middle of class 147 30.3
Middle 232 47.8 Lower 30% of class    5   1.0
Lower middle   38   7.8 Lower 20% of class    1   0.2
Lower   13   2.7 Lower 10% of class    4   0.8

Note: SES = socioeconomic status.

Interestingly, graduate school or higher level of education was the second

largest category in both father and mother groups following the high school
graduate category. The majority of the students perceived that their family
SES was middle class or higher.

Downloaded from by guest on January 31, 2013

1674 Journal of Interpersonal Violence 26(8)

First-Order Factor Analysis and Reliability Tests for Scales

Factor analysis and reliability tests were performed to develop scales. First,
factor analyses were performed to determine whether one or more constructs
were being indicated for each hypothesized factor. Table 2 presents factors
identified from the analyses along with reliability coefficients (alpha), means,
and standard deviations. The alpha levels for most of the scales were rela­
tively high and fell in an acceptable range; however, several scales had weak
alpha scores. In addition, items that have opposite direction from other items
for a construct were recoded to have the same direction. Overall, the mean
for relational bullying behaviors was 6.6 (range 4-20), the mean for verbal
bullying was 7.9 (range 6-30), and the mean for physical bullying was 6.1
(range 5-25).

Second-Order Confirmatory Factor Analysis (CFA) and Correlations

Second-order CFA. Before further analysis, the second-order CFA was per­
formed for all eight constructs and their 34 indicators (which were developed
from the first-order factor analysis and reliability test) to develop the endog­
enous and exogenous variables. The eight composite measures are bullying
behaviors, individual traits, family interactions, teacher interactions, peer int­
eractions, parental communication, school climate, and community charac­
ter. Table 3 contains the standardized regression weights (factor loadings)
and the squared multiple correlation coefficients (R2). For testing model fit,
two standardized fit indexes were used: the Jöreskog-Sörbom Goodness of
Fit Index (GFI, which should exceed .9 for a good model) and the Adjusted
Goodness of Fit Index (AGFI). Other fit indexes, such as the Bentler Com­
parative Fit Index (CFI; the good model should exceed .95) and the root
mean square error of approximation (RMSEA; the adequate model fit ranges
between .08 and .05, and the good model fit is .05 or less), were used for
additional information.
Fit indices for the initial CFA model showed an unsatisfactory level of
model fit; GFI = .756, AGFI = .710, CFI = .698, and RMSEA = .083. The
chi-square was significant (c2 = 2,167.306, df = 500), meaning that the model
could be improved. The initial CFA model identified two constructs within
individual traits factors: prior bullying victimization and individual tendency.
Prior bullying victimization includes relational, verbal, and physical bullying
victimization; individual tendency includes dominance, impulsivity, attitude
toward aggression, and fun-seeking tendency. With these two constructs
separated, the second CFA was performed. Although the separated model

Downloaded from by guest on January 31, 2013

Lee 1675

Table 2. Reliability Coefficients for Scales Used and General Characteristics (N = 485)

Number of Items/
Scales/Range Directiona Alpha M SD

Dependent variables
Relational bullying (4-20) 4 (+) .84   6.6 3.3
Verbal bullying (6-30) 6 (+) .88   7.9 3.6
Physical bullying (5-25) 5 (+) .75   6.1 2.4
Independent variables
Individual traits
Prior relational bullying 4 (+) .84   6.6 3.3
  victimization (4-20)
Prior verbal bullying 6 (+) .88   7.9 3.6
  victimization (6-30)
Prior physical bullying 5 (+) .75   6.1 2.4
  victimization (5-25)
Dominance (3-15) 3 (+) .53   8.3 3.0
Impulsivity (3-15) 3 (+) .62   8.4 3.5
Attitude toward aggression (4-20) 4 (+) .77 10.0 4.9
Fun-seeking tendency (4-20) 4 (+) .80   6.3 3.4
Microsystem level
Family interaction
Perceptions of parents’ attitude 3 (+) .91 12.8 3.3
  toward bullying: Parents don’t
  like bullying (3-15)
Perceptions of parents’ attitude 3 (+) .81 13.7 2.5
  toward bullying: Parents care
  about bullying (3-15)
Authoritarian parenting (2-10) 2 (–) .45b   6.6 2.3
Experience/witness of 4 (–) .75 18.1 3.2
  domestic abuse (4-20)
Importance of education (2-10) 2 (+) .47b   6.0 2.4
Teacher interaction
Perceptions of teachers’ attitude 3 (+) .97 13.3 3.5
  toward bullying:Teachers don’t
  like bullying (3-15)
Perceptions of teachers’ 3 (+) .64 13.6 2.5
  attitude toward bullying:
 Teachers care about
 bullying (3-15)
Effectiveness of teachers’ 2 (+) .60   7.3 2.5
  intervention (2-10)
Teachers’ moral authority 3 (+) .67 10.7 3.2


Downloaded from by guest on January 31, 2013

1676 Journal of Interpersonal Violence 26(8)

Table 2. (continued)

Number of Items/
Scales/Range Directiona Alpha M SD

Peer interaction
Power dynamic (4-20) 4 (+) .81   7.3 3.9
Level of difference-acceptance 4 (–) .75   9.1 4.3
Coercion of public self (4-20) 4 (+) .57 11.4 3.9
Pseudofriendship (3-15) 3 (+) .66   6.0 3.1
School climate
Academic standard and 3 (+) .78 11.2 3.3
  general image (3-15)
Cooperative school climate 2 (+) .37b   7.2 2.0
Student–teacher cohesion 2 (+) .69   6.6 2.5
Moral atmosphere (2-10) 2 (+) .49b   6.8 2.2
Parents’ involvement with 2 (+) .49b   6.4 2.2
  school boards (2-10)
Effectiveness of policies, rules, 2 (+) .74   7.5 2.5
  programs (2-10)
Mesosystem level
Parents’ communication with 3 (+) .87   5.0 2.7
  teachers (3-15)
Parents’ communication with 3 (+) .87   4.8 2.7
  peers (3-15)
Exosystem level
Family SES (5-35) 5 (+) .59 23.1 5.1
Macrosystem level
Vertical individualism (2-10) 2 (–) .85   7.4 2.6
Horizontal individualism 2 (–) .68   4.3 2.2
Vertical collectivism (2-10) 2 (+) .82   6.5 2.4
Horizontal collectivism (4-20) 4 (+) .82 15.7 3.9
Levels of collectivism of peer 4 (+) .77 15.2 3.8
  group (4-20)
Social disorganization (3-15) 3 (+) .45b   7.9 3.0

Note: SES = socioeconomic status.

a. + indicates that the higher the value, the greater the magnitude of the characteristics;
– indicates that the higher the value, the smaller the magnitude of the characteristics.
b. Scales with weak alpha scores.

Downloaded from by guest on January 31, 2013

Lee 1677

Table 3. Second-Order Confirmatory Factor Analysis

Initial Separated Dropped Final

Loading R2 Loading R2 Loading R2 Loading R2

Dependent variable: Bullying behaviors

Relational   .70 .49   .70 .49 .70 .49 .76 .58
Verbal   .93 .86   .93 .87 .93 .86 .88 .77
Physical   .77 .59   .77 .59 .77 .59 .80 .64
Independent variable: Individual traits
Prior bullying victimization
Relational   .78 .61   .78 .60 .78 .61 .78 .60
Verbal   .91 .83   .95 .90 .95 .89 .95 .90
Physical   .72 .52   .70 .49 .70 .49 .70 .49
Individual tendency
Dominance –.06 .00   .24 .06
Impulsivity   .33 .11   .41 .17 .43 .18
Attitude on   .32 .11   .67 .44 .68 .46 .63 .39
Fun-seeking   .17 .03   .64 .42 .64 .41 .64 .40
Independent variable: Family interactions
Parents’ attitude:   .24 .06   .26 .07
Parents’ attitude:   .27 .07   .30 .09
Authoritarian   .46 .22   .46 .21 .55 .30 .51 .26
Domestic abuse   .67 .45   .65 .42 .61 .37 .64 .41
Importance of –.43 .19 –.42 .18 .50 .25
Independent variable: Teacher interactions
Teachers’   .06 .00   .06 .00
  attitude: Like
Teachers’   .38 .15   .38 .14 .36 .13
  attitude: Care
Teachers’   .62 .39   .62 .38 .62 .39 .61 .37
Moral authority   .75 .56   .75 .57 .77 .59 .82 .67
Independent variable: Peer interactions
Power dynamics   .66 .43   .71 .50 .72 .52 .74 .54
Difference   .67 .45   .63 .39 .62 .39 .57 .33
Coercive   .43 .18   .39 .15 .40 .10
Pseudofriendship   .27 .08   .26 .07


Downloaded from by guest on January 31, 2013

1678 Journal of Interpersonal Violence 26(8)

Table 3. (continued)

Initial Separated Dropped Final

Loading R2 Loading R2 Loading R2 Loading R2

Independent variable: School climate

Academic   .71 .50   .71 .50 .71 .50 .66 .43
Cooperative   .11 .01   .11 .01
Moral   .69 .48   .69 .48 .69 .48 .70 .49
Parental   .63 .40   .63 .40 .64 .41 .66 .43
Policies,   .76 .57   .75 .56 .75 .56 .76 .58
Independent variable: Parental communication
Parent–teacher   .79 .63   .79 .63 .79 .63 .79 .62
Parent–peer   .80 .65   .80 .64 .80 .64 .81 .65
Independent variable: Community character
Vertical –.49 .24 –.49 .24 .49 .24
Horizontal –.81 .66 –.81 .66 .80 .64 .78 .61
Social   .14 .02   .14 .02
Peer-group   .84 .71   .84 .71 .86 .73 .88 .77

Chi-square 2,167.306 1,765.520 903.222 343.096

Degree of 500 492 289 156
Dc2a 401.789 (df = 8), 862.298 (df = 203), 560.126 (df = 133)
GFI .756 .801 .872 .941
AGFI .710 .759 .833 .904
CFI .698 .769 .871 .955
RMSEA .083 .073 .066 .050
Note: GFI = Goodness of Fit Index; AGFI = Adjusted Goodness of Fit Index; CFI =
Comparative Fit Index; RMSEA = root mean square error of approximation. Low R2 scores
are in bold.
a. Dc2 is significant (p < .05) at the degree of freedom of each step.

was significantly improved (Dc2/Ddf = 401.786/8), the fit indexes suggested

the model still could be improved (GFI = .801, AGFI = .759, CFI = .769, and
RMSEA = .073). The separated model contained several items that had low

Downloaded from by guest on January 31, 2013

Lee 1679

factor loadings. These items were dominance (R2 = .06), parents’ attitude (don’t
like bullying, R2 = .07), parents’ attitude (care about bullying, R2 = .09),
teachers’ attitude (don’t like bullying, R2 = .00), pseudofriendship (R2 = .07),
cooperative climate (R2 = .01), and social disorganization (R2 = .02). These
items were excluded from the subsequent CFA model, and the dropped CFA
model significantly improved (Dc2/Ddf = 862.298/203). However, these model
fit indexes were not satisfactory (GFI = .872, AGFI = .833, CFI = .871, and
RMSEA = .066). Thus, these items with low factor loadings, that is, impul­
sivity (R2 = .18), importance of education (R2 = .25), teachers’ attitude (care
about bullying; R2 = .13), coercive conformity (R2 = .10), and vertical col­
lectivism (R2 = .24), were excluded in the final CFA model. The final model
improved significantly (Dc2/Ddf = 560.126/133). The chi-square was signifi­
cant (c2 = 343.096, df = 156), and all fit indexes indicated a good fit support­
ing the presence of distinct constructs (GFI = .941, AGFI = .904, CFI = .955,
and RMSEA = .050). The remainder of the discussion uses the measurement
model developed from the final CFA model for Structural Equation Model­
ing (SEM).
Correlations. Table 4 presents bivariate correlations among the nine con­
structs developed from the second-order CFA. The results indicate that fac­
tors are moderately, sometimes weakly, correlated to each other. Particularly,
individual tendency and peer interaction were significantly and positively
related to bullying behaviors, suggesting that students who have higher levels
of tendency toward aggression and fun-seeking tendency and who have higher
levels of power dynamics and intolerance are more likely to bully other stu­
dents. In addition, family experiences and teacher interactions were negatively
related to bullying behaviors. This indicates that children with low levels of
authoritarian parenting and domestic abuse and with high levels of positive
attitude toward teacher’s bullying intervention and their moral authority are
less likely to bully others. In terms of social context, school climate and mac­
rosystem were negatively related to bullying behaviors, suggesting that chil­
dren who perceived their schools more positively and their community more
collectively were less likely to bully others. These findings are consistent with
the previous research.

Proposed ecological model (MH ). The model fit summary for the hypothesized
model is presented in Table 5, in the row headed “MH hypothesized model.”
Since the chi-square is 903.357 at degree of freedom 212, the null hypothesis
that the model fits the data should be rejected. All fit indices are out of the

Downloaded from by guest on January 31, 2013

1680 Journal of Interpersonal Violence 26(8)

Table 4. Bivariate Correlations

DV V1 V2 V3 V4 V5 V6 V7 V8

Dependent 1
  variable: Bullying
V1: Prior bullying   .310 1
V2: Individual   .510   .239 1
V3: Microsystem– -.169 -.323 -.340 1
V4: Microsystem- -.365 -.205 -.465   .255 1
V5: Microsystem– .450   .130 .558 -.322 -.522 1
V6: Microsystem– -.162 -.113 -.295   .134   .237 -.169 1
  school climate
V7: Mesosystem   .288   .197   .246 -.097 -.137   .212 -.080 1
V8: Macrosystem -.268 -.119 -.294   .123   .159 -.164 -.204 .603 1
Note: All correlations are significant at either p < .01 or p < .05.

Table 5. Model Fit Summary for All Models

MH hypothesized model 930.357 212 .855 .812 .830 .084
MR revised model 567.611 207 Yesa .904 .872 .915 .060
MF final model 397.731 196 Yesb .932 .905 .951 .047
MM measurement model 343.096 156 .941 .904 .955 .050
Note: GFI = Goodness of Fit Index; AGFI = Adjusted Goodness of Fit Index; CFI =
Comparative Fit Index; RMSEA = root mean square error of approximation.
a. Dc2 = 362.746/Ddf = 5.
b. Dc2 = 164.880/Ddf = 11 (the critical chi-square value with 2 df is 5.991, p < .05.

acceptable range, GFI = .855, AGFI = .812, CFI = .830, and RMSEA = .084.
Table 6 and Figure 1 summarize the results of the hypothesized model. Nine
paths were nonsignificant (see Table 6). Overall, the model explained 41 %
of variation in bullying behaviors (R2 = .41).
The nomological validity of the hypothesized model can be tested by
comparing its chi-square with that of the measurement model (see Tables
3 and 5). The chi-square difference (930.357-343.096 = 587.261) at the

Downloaded from by guest on January 31, 2013

Lee 1681

Table 6. Standardized Path Coefficients and R2 for All Models


Bullying behaviors (R2) (.41) (.50) (.56)

Macrosystem -.20*** -.29*** -.19*
Exosystem: Family SES .00
Mesosystem .13** .09
Microsystem: Family .38** .23** .27**
Microsystem: Teacher .25
Microsystem: Peers 2.68
Microsystem: School climate .13*** .32*** -.34***
Individual traits: Prior bullying victimization .27*** .20*** .20***
Individual traits: Tendency -1.68 .69*** .82***
Microsystem: School climate (R2) (.13) (.56) (.49)
Macrosystem .69*** .68***
Mesosystem .00 .13**
Microsystem: Family .00
Microsystem: Teacher .03
Microsystem: Peers -.34* -.15** -.20***
Microsystem: Teacher interactions (R2) (.58) (.54) (.56)
Macrosystem -.14*
Individual traits: Tendency -.76*** -.74*** -.78***
Microsystem: Peer interactions (R2) (1.03) (.97) (1.07)
Macrosystem .26***
Individual traits: Tendency 1.01*** .98*** 1.09***
Individual trait:Victimization (R2) (.20) (.20) (.22)
Exosystem: Family SES -.09
Mesosystem .19*** .20***
Microsystem: Family -.45*** -.38*** -.42***
Microsystem: Teacher -.21
Microsystem: Peers -.20
Individual trait: Tendency (R2) (.31) (.43) (.46)
Macrosystem -.25*** -.31***
Exosystem: Family SES .25*** .25***
Mesosystem .20*** .23***
Microsystem: Family -.56*** -.55*** -.53***
Note: SES = socioeconomic status.
*p < .05. **p < .01. ***p < .001.

degree of freedom difference (212-156 = 56) is significant at p < .05. This

result suggests that the hypothesized model is not valid in accounting for
the observed relationships between the latent constructs (Anderson &
Gerbing, 1988).

Downloaded from by guest on January 31, 2013

1682 Journal of Interpersonal Violence 26(8)


Family SES

Interactions 0.13

.20 –0.45 0
Individual Traits:
Victimization 0.27
.58 .41
Teacher 0.25 Bullying
.31 –0.76
0.03 2.68
Individual Traits: –1.68
–0.2 1.03 .13
1.01 Peer School
Interactions –0.34 Climate

Figure 1. A proposed initial ecological prediction model of bullying and the results
of SEM
SEM = structural equation modeling.

Revised model (MR ). Instead of dropping all the nonsignificant paths, the
hypothesized model was revised based on modification indexes (MI) pro­
vided by AMOS as well as theoretical considerations. First, several nonsig­
nificant paths were dropped in the revised model (see Table 6). Second, the
nonsignificant path between individual tendency and bullying behaviors was
retained in the revised model because previous studies found a significant
relationship between them (Bentley & Li, 1995; Rigby, 2004; Shin, 2000; see
also correlation results). Six new paths were added to the revised model
based on the MI scores and theoretical considerations. The paths were between
macrosystem and school climate (MI = 128.842), macrosystem and individ­
ual tendency (MI = 28.412), SES and individual tendency (MI = 16.006),

Downloaded from by guest on January 31, 2013

Lee 1683

SES and victimization (MI = 8.011), mesosystem and individual tendency

(MI = 16.878), and between mesosystem and victimization (MI = 10.327).
Model fit indexes are presented in Table 5, in the row headed “MR revised
model.” Although the revised model improved significantly (Dc2 = 362.746/
Ddf = 5), this model did not fit data well since the chi-square value was sig­
nificant (c2 = 567.611/df = 207). The several model fit indexes were still out
of the acceptable range, GFI = .904, AGFI = .872, CFI = .915, and RMSEA =
.060. Overall, the revised model explained 50% of the variance in bullying
behaviors (R2 = .50).
Final model (MF). Two nonsignificant paths (between mesosystem and bul­
lying, and between SES and victimization) were dropped in the final model.
In addition, three paths were added to the final model. First, a path between
macrosystem and peer interaction was added based on the MI (5.351). Sec­
ond, a path between macrosystem and teacher interaction was added. Accord­
ing to the ecological system theory, macrosystem may have significant and
distal influence on all aspects of children’s lives. Thus, it is hypothesized that
the macrosystem significantly influences microsystem, such as teacher inter­
action and peer interaction. Finally, a path between mesosystem and school
climate was added. Because previous studies found that parental involve­
ment in school activities would influence the development of positive school
climate, such as high academic standard, moral atmosphere, or strict policies
and regulations, it is hypothesized that mesosystem would influence school
climate (Espelage & Swearer, 2003; Hoshino, 2001; Naito, 1990).
Model fit indexes are presented in Table 5, in the row headed “MF final
model.” The chi-square value was significant (c2 = 397.731/df = 196), sug­
gesting that the null hypothesis that the model fits the data should be rejected.
However, the model fit indexes indicated that the model well accounts for the
relationships between the latent constructs, GFI = .932, AGFI = .905, CFI =
.951, and RMSEA = .047. The test of the nomological validity of the final
model (c2: 397.731-343.096 = 54.635, df: 196-156 = 40) was nonsignificant
at p <.05. This indicates that the final model is valid in accounting for the
observed relationships between the latent constructs. Consequently, this study
retains this model as the final and best-fitting model. Overall, the model
improved significantly (Dc2 = 164.880/Ddf = 11) and explained 56 % of vari­
ance in bullying behaviors (R2 = .56). Figure 2 represents the final ecological
prediction model of bullying behaviors.
As hypothesized, the path model shows that the macrosystem influences
many different aspects of children’s lives. Particularly, the macrosystem posi­
tively influences school climate and peer interactions, and negatively influences

Downloaded from by guest on January 31, 2013

1684 Journal of Interpersonal Violence 26(8)


0.68 0.26 –0.14

1.07 –0.31 .56
School –0.2 Peer Teacher
Climate Interactions Interactions

–0.34 1.09 –0.78

.22 .46
0.13 Individual Traits: Individual Traits:
Victimization 0.2 0.82 Tendency

0.2 –0.42 0.27 –0.53

Mesosystem Family SES

Figure 2. Final model: Ecological prediction model of bullying

SES = socioeconomic status.

individual tendency and teacher interactions. Interestingly, the macrosystem

negatively influences bullying behaviors, meaning that the collective com­
munity and the collective peer group decrease bullying behaviors among
children. In terms of exosystems, family SES influences only individual ten­
dency in a positive direction, indicating that the higher the SES, the higher
levels in attitude toward aggression as well as fun-seeking tendency. School
climate, which includes other types of exosystems, is negatively related to
bullying behaviors. This finding indicates that children who perceived their
school more positively were less likely to bully others. As hypothesized, the
mesosystem influences school climate; however, this study found that the
mesosystem was positively related to individual tendency. Regarding micro­
systems, family interactions influence individual tendency and victimization.
Interestingly, peer interactions do not directly influence bullying behaviors,
but they influence school climate. Another interesting finding is that teacher

Downloaded from by guest on January 31, 2013

Lee 1685

interactions do not influence bullying behaviors. These unexpected findings

are inconsistent with the prior findings regarding bullying phenomenon. Among
these ecological factors, individual tendency was the most important influ­
ence (b = .82) on bullying behaviors. Other important factors for bullying
behaviors were school climate (b = .34), family interactions (b = .27), indi­
vidual victimization (b = .20), and macrosystem (b = –.19).

Discussion and Conclusion

Summary of Findings

Bullying has been understood as an independent incident within complex

social contexts. Recently, scholars have recognized and emphasized the
importance of holistic and ecological approaches to bullying (Espelage &
Swearer, 2003; Morrison, 2001), and empirical studies have been conducted
to understand the complex networks of influences among ecological factors
on bullying behaviors in South Korea (e.g., Lee, 2007). This study aimed to
add empirical evidence to support the ecological approach to bullying behav­
iors and tested an ecological system theory applied to bullying phenomenon
in the United States.
First, this study found that individual traits, particularly individual tendency
for aggression and fun seeking, have the most important influence on bully­
ing behaviors. This finding confirmed the long-lasting effects of psychological
and psychosocial traits on bullying behaviors (Bentley & Li, 1995; Bosworth
et al., 1999; Lee, 2000). For example, Lee (2010) found that the most signifi­
cant factors for different types of bullying behaviors in South Korea were
prior bullying victimization and fun-seeking tendency.
Second, this study found an alternative path between individual tendency
and bullying behaviors. As Figure 2 presents, individual tendency has a direct
and strong influence on bullying behaviors, but it also influences peer inter­
actions, which influences school climate. School climate, in turn, influences
bullying behaviors. This finding is unique in the sense that the alternative
path shows individual tendency has not only the direct influence but also
significant indirect effects on bullying behaviors through social settings,
such as peer relationship and school environments. The finding suggests that
individual tendency positively influences peer group tendency for bullying
behaviors, and such peer group interactions negatively influence school cli­
mate. School climate, then, reduces bullying behaviors among children.
Third, it appears teacher interactions have little influence on bullying
behaviors. In contrast to findings from an ecological study of bullying in

Downloaded from by guest on January 31, 2013

1686 Journal of Interpersonal Violence 26(8)

South Korea (Lee, 2007), this study found a very weak and nonsignificant
relationship between teacher interactions and bullying behaviors. To
indentify alternative paths between teacher interactions and bullying behav­
iors, several paths were added from teacher interactions to peer interactions,
victimization, or school climate. These attempts, however, significantly
diminished the fit indexes of the model in SEM, indicating that those alterna­
tive paths were not helping in the explanation of the relationships involving
teacher interactions. Further study on this issue is strongly recommended.
Fourth, this study found that positive experiences in the family signifi­
cantly and positively influence bullying behaviors, indicating that children
with less authoritarian parenting and less domestic abuse are more likely to
bully others at school. This finding is inconsistent with previous studies
(Baldry, 2003; Christie-Mizell, 2003; Shields & Cicchetti, 2001). A possible
reason for this inconsistent finding is that positive family experiences pro­
duce high levels of self-esteem, which is positively related to bullying others
(Christie-Mizell, 2003; Maday & Szalay, 1976). However, this study also
found that family interactions have significant negative influence on both
individual victimization and tendency for bullying at school. The results sug­
gest that children with less authoritarian parenting and less domestic violence
exposure are less likely to get victimized and are less likely to have higher
levels of tendency for aggression and fun seeking. In other words, children
with high levels of negative experience in the family are more likely to have
high levels of attitude toward aggregation and fun-seeking tendency for bul­
lying behaviors. Thus, it seems that experience in the family influences bul­
lying behavior indirectly through individual victimization of bullying and
individual tendency for aggression and fun seeking. In addition, it seems more
plausible to say that the latter indirect influences (b = –.53 and b = –.42) are
stronger than the direct family interaction influence (b = .27) on bullying
Fifth, this study hypothesized that mesosystem, which is parental com­
munication with teachers and peers, would be negatively related to bullying
behaviors. This study found no significant relationship between them.
Instead, this study found that parental communication with teachers and
peers significantly influenced individual traits and school climate. The
finding on the significant influence on school climate is consistent with the
arguments from previous studies (Hoshino, 2001; Naito, 1990; Tai, 2001).
However, the finding on the significant influences between individual traits
and mesosystem was unexpected. The possible reason for the positive rela­
tionships stems from the nature of cross-sectional data. As there is no defi­
nite time interval between these exogenous factors, it may be more plausible

Downloaded from by guest on January 31, 2013

Lee 1687

to say that parental communication has increased to respond to increased

bullying behaviors.
Finally, this study found that the macrosystem significantly influences
bullying behaviors both directly and indirectly. The results indicated that the
collective community and the collective peer group significantly and posi­
tively influence school climate and peer interactions, suggesting that chil­
dren, who perceived their community and peer group more collectively, were
more likely to perceive school climate positively and less likely to have
­bullying-prone peer attitudes. In addition, the results also indicated that the
collective community and the collective peer group significantly and nega­
tively influenced teacher interactions and individual tendency for bullying.
This finding suggests that children who perceive their community and peer
group more collectively were less likely to develop a positive attitude toward
teachers and were less likely to have higher levels of the tendency for aggres­
sion and fun seeking. Overall, collective group and community characteristics
promote positive school climate, reduce the individual tendency for aggres­
sion and fun-seeking behavior, and prevent bullying behaviors.

Limitations and Conclusion

There are several limitations contained in the current study. First, although
a longitudinal design could be the best option for investigating developmen­
tal approach to bullying, the current research employed a cross-sectional
design for three reasons. First, some ecological variables, for example, per­
ceived community characteristics and school climate, could be more easily
and efficiently compared with cross-sectional data rather than longitudinal
data. As there are not many changes over time in such contextual factors,
variation in such factors is best achieved by sampling different contexts
rather than the same context over time. Second, studies have found that the
anonymous questionnaire is a more valid and reliable medium for bullying
study (Ahmad & Smith, 1990; Chan, Myron, & Crawshaw, 2005). Because
longitudinal research design usually requires identifiers to link data col­
lected at different times, it makes any personal information necessary. The
collection of personal information may possibly threaten the validity of data
collected in a longitudinal study. Finally, resources and time constraints
limited options to a cross-sectional design. Future study, however, should
consider a longitudinal design to collect data on bullying behaviors, and
such data will best serve the developmental approach of ecological system
theory to bullying behaviors.

Downloaded from by guest on January 31, 2013

1688 Journal of Interpersonal Violence 26(8)

Second, the current study employed only variables relevant to risk factors
for commitment of bullying. Although several factors at different ecological
systems measured preventive effects on bullying behaviors, for example,
parents’ and teacher’s attitude toward bullying at microsystem level, parental
communication with teachers and peers at mesosystem level, emphasis of
variable selection was given to risk factors rather than resiliency factors. Thus,
generalization and interpretation of the findings from the current study should
be made carefully.
Overall, the current study attempted to identify an ecological prediction
model of bullying behaviors based on an ecological systems theory. Despite
of aforementioned limitations, this study found that all of the ecological sys­
tems significantly influenced bullying behaviors in either direct or indirect
ways. Especially, this study found that individual traits were the most signifi­
cant factors facilitating commitment of bullying and that the ecological app­
roach to bullying could be a useful tool for understanding children’s bullying
in schools.

Declaration of Conflicting Interests

The author declared that they had no conflicts of interests with respect to their author­
ship or the publication of this article.

The author disclosed that they received the following support for their research and/
or authorship of this article: College of Professional Study Summer Research Grant,
University of Arkansas at Little Rock, that promotes research in the field of criminal
justice and criminology.

Ahmad, Y., & Smith, P. K. (1990). Behavioural measures: Bullying in schools. Child
Psychology and Psychiatry, 12, 26-27.
Andershed, H., Kerr, M., & Stattin, H. (2001). Bullying in school and violence on the
streets: Are the same people involved? Journal of Scandinavian Studies in Crimi-
nology and Crime Prevention, 2(1), 31.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice:
A review and recommended two-step approach. Psychological Bulletin, 103,
Anderson, M., Kaufman, J., Simon, T. R., Barrios, L., Paulozzi, L., Ryan, G., et al.
(2001). School-associated violent deaths in the United States, 1994-1999. Journal
of the American Medical Association, 286, 2695-2702.

Downloaded from by guest on January 31, 2013

Lee 1689

Baldry, A. C. (2003). Bullying in schools and exposure to domestic violence. Child

Abuse & Neglect, 27, 713-732.
Bentley, K. M., & Li, A. K. F. (1995). Bully and victim problems in elementary
schools and students’ beliefs about aggression. Canadian Journal of School Psy-
chology, 11, 153-165.
Bosworth, K., Espelage, D. L., & Simon, T. R. (1999). Factors associated with
bullying behavior in middle school students. Journal of Early Adolescence, 19,
Boulton, M. J., Trueman, M., & Flemington, I. (2002). Associations between sec­
ondary school pupils’ definitions of bullying, attitudes towards bullying, and ten­
dencies to engage in bullying: Age and sex differences. Educational Studies, 28,
Bronfenbrenner, U. (1979). The ecology of human development: Experiments by
nature and design. Cambridge, MA: Harvard University Press.
Bronfenbrenner, U. (1989). Ecological systems theory. In R. Vasta (Ed.), Annals of
child development (Vol. 6, pp. 197-249). Greenwich, CT: JAI Press.
Cairns, R. B., & Cairns, B. D. (1991). Social cognition and social networks: A devel­
opmental perspective. In D. J. Pepler & K. H. Rubin (Eds.), The development
and treatment of childhood aggression (pp. 249-278). Hillsdale, NJ: Lawrence
Chan, J. H., Myron, R., & Crawshaw, M. (2005). The efficacy of non-anoymous mea­
sures of bullying. School Psychology International, 26, 443-458.
Christie-Mizell, C. A. (2003). Bullying: The consequences of interparental discord
and child’s self-concept. Family Process, 42, 237-251.
Coloroso, B. (2003). The bully, the bullied, and the bystander. New York: Harper-
Curtner-Smith, M. E., Culp, A. M., Scheib, C., Owen, K., Tilley, A., Murphy, M.,
et al. (2006). Mothers’ parenting and young economically disadvantaged chil­
dren’s relational and overt bullying. Journal of Child and Family Studies, 15,
Dahlberg, L. L., Toal, S. B., & Behrens, C. B. (1998). Measuring violence-related
attitudes, beliefs, and behaviors among youths: A compendium of assessment
tools. Atlanta, GA: Centers for Disease Control and Prevention, National Center
for Injury Prevention and Control.
Espelage, D. L., Bosworth, K., & Simon, T. R. (2000). Examining the social context
of bullying behaviors in early adolescence. Journal of Counseling and Develop-
ment, 78, 326-333.
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.

Downloaded from by guest on January 31, 2013

1690 Journal of Interpersonal Violence 26(8)

Espelage, D. L., & Swearer, S. M. (2003). Research on school bullying and victimiza­
tion: What have we learned and where do we go from here? School Psychology
Review, 32, 365-383.
Farrington, D. P. (1993). Understanding and preventing bullying. Crime and Justice,
17, 381-458.
Flouri, E., & Buchanan, A. (2003). The role of mother involvement and father
involvement in adolescent bullying behavior. Journal of Interpersonal Violence,
18, 634-644.
Gottfredson, D. C. (2001). Schools and delinquency. Cambridge, UK: Cambridge
University Press.
Gottfredson, G. D., & Gottfredson, D. C. (1985). Victimization in schools. New York:
Plenum Press.
Graham, S., & Juvonen, J. (1998). A social cognitive perspective of peer aggression
and victimization. Annals of Child Development, 13, 23-70.
Harris, S., & Petrie, G. F. (2003). Bullying: The bullies, the victims, the bystanders.
Lanham, MD: Scarecrow Press.
Hoshino, K. (2001). Higaishya to kagaishyatono kankei [Relations between bully and
victim]. In Y. Morita (Ed.), Ijime no Kokusai Hikaku Kenkyu (pp. 73-92). Tokyo:
Kaneko Shyobou.
Kim, Y., Koh, Y., & Leventhal, B. (2005). School bullying and suicidal risk in Korean
middle school students. Pediatrics, 115, 357-363.
Kinlock, T. W., Battjes, R. J., & Gordon, M. S. (2004). Factors associated with crimi­
nal severity among adolescents entering substance abuse treatment. Journal of
Drug Issues, 34, 293-319.
Laub, J. H., & Lauritsen, J. L. (1998). The interdependence of school violence with neigh­
borhood and family conditions. In D. S. Elliott, B. A. Hamburg, & K. R. Williams
(Eds.), Violence in American schools: A new perspective (pp. 127-158). Cambridge,
UK: Cambridge University Press.
Leary, M. R., Kowalski, R. M., Smith, L., & Phillips, S. (2003). Teasing, rejection, and
violence: Case studies of the school shootings. Aggresive Behavior, 29, 202-214.
Lee, C. (2000). 왕따의 가해/피해아의 성격 특성과 사회성 측정 지위의 차이
분석 [Analysis of differences in personality traits and sociometric status of bully/
victim]. South Korea: 한국교원대학교 [Korean National University of Education].
Lee, C.-H. (2007). An ecological prediction model of bullying behaviors among South
Korean middle school students. Unpublished doctoral dissertation, Michigan State
University, East Lansing.
Lee, C.-H. (2009). Bullying and ecology: An ecological prediction model of bullying
among South Korean students. Saarbrucken, Germany: VDM.
Lee, C.-H. (2010). Personal and interpersonal correlates of bullying behaviors among
Korean middle school students. Journal of Interpersonal Violence, 25, 152-176.

Downloaded from by guest on January 31, 2013

Lee 1691

Lee, D. (2004). 청소년의 문화성향과 또래동조성 및 집단따돌림의 관계 [Rela­

tionship between bullying and youth culture and peer conformity]. Seoul, South
Korea: Seokang University.
Lee, S., Byun, C., & Jin, E. (1969). 표준화 성격 진단 검사 실시 요강 [Standard­
ized Personality Evaluation Test]. Seoul, South Korea: 중앙적성연구소 [Chung-
Ang Aptitude].
Maday, B. C., & Szalay, L. B. (1976). Psychological correlates of family socialization
in the United States and Korea. In T. Williams (Ed.), Psychological Anthropology.
The Hague, Netherlands: Mouton.
Morrison, B. (2001). Bullying, violence and alienation. Acton, Australia: Centre
for Restorative Justice, Research School of Social Science, Australian National
Naito, T. (1990). Moral education in Japanese public schools. Moral Edcuation
Forum, 15(2), 27-36.
Naito, T., & Gielen, U. P. (2005). Bullying and Ijime in Japanese Schools. In
F. L. Denmark, H. H. Krauss, R. W. Wesner, E. Midlarsky, & U. P. Gielen
(Eds.), Violence in schools: Cross-national and cross-cultural perspectives
(pp. 169-190). New York: Springer.
Nansel, T., Overpeck, M., Pilla, R. S., Ruan, W. J., Simons-Morton, B., & Scheidt, P.
(2001). Bullying behaviors among U.S. youth: Prevalence and association with
psychosocial adjustment. Journal of the American Medical Association, 285,
Natvig, G. K., Albrektsen, G., & Qvarnstrom, U. (2001). School-related stress experi­
ence as a risk factor for bullying behavior. Journal of Youth and Adolescence, 30,
Nesdale, D., & Naito, M. (2005). Individualism—Collectivism and the attitudes to
school bullying of Japanese and Autralian students. Journal of Cross-Cultural
Psychology, 36, 537-556.
Olweus, D. (1993). Bullying at school: What we know and what we can do. Cam­
bridge, MA: Blackwell.
Park, H. S., Schepp, K. G., Jang, E. H., & Koo, H. Y. (2006). Predictors of suicidal
ideation among high school students by gender in South Korea. Journal of School
Health, 76, 181-188.
Park, J. (2003). 청소년 집단따돌림의 원인과 대처방안 [The causes and the settle­
ments of teenagers mobbing]. 인문사회과학연구, 10, 355-369.
Pellegrini, A. D., & Long, J. D. (2002). A longitudinal study of bullying, dominance,
and victimization during the transition from primary school through secondary
school. British Journal of Developmental Psychology, 20, 259-280.
Rigby, K. (2004). Addressing bullying in schools: Theoretical perspectives and their
implications. School Psychology International, 25, 287-300.

Downloaded from by guest on January 31, 2013

1692 Journal of Interpersonal Violence 26(8)

Rigby, K., & Cox, I. (1996). The contribution of bullying at school and low self-
esteem to acts of delinquency among Australian teenagers. Personality and Indi-
vidual Differences, 21, 609-612.
Rodkin, P. C., & Hodges, E. V. E. (2003). Bullies and victims in the peer ecology:
Four questions for psychologists and school professionals. School Psychology
Review, 32, 384-400.
Sampson, R. J., & Groves, W. B. (1989). Community structure and crime: Testing
social-disorganization theory. American Journal of Sociology, 94, 774-802.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent
crime: A multilevel study of collective efficacy. Science, 277, 918-924.
Shields, A., & Cicchetti, D. (2001). Parental maltreatment and emotion dysregula­
tion as risk factors for bullying and victimization in middle childhood. Journal of
Clinical Child Psychology, 30, 349-363.
Shin, Y. (2000). 중학생의 집단따돌림(왕따)에 관한 비교연구: 사회환경을
중심으로 [A comparison study in bullying of middle school students: Focus on
social environment]. Seoul, South Korea: 중앙대학교 [Joongang University].
Singelis, T. M., Triandis, H. C., Bhawuk, D. D., & Gelfand, M. (1995). Horizontal
and vertial dimensions of individualism and collectivism: A theoretical and mea­
surement refinement. Cross-Cultural Research, 29, 240-275.
Smith, P. K., Cowie, H., Olafsson, R. F., & Liefooghe, A. P. D. (2002). Definitions
of bullying: A comparison of terms used, and age and gender differences, in a
fourteen-country international comparison. Child Development, 73, 1119-1133.
Sullivan, K., Cleary, M., & Sullivan, G. (2004). Bullying in secondary schools: What
it looks like and how to manage it. Thousand Oaks, CA: Sage.
Tai, M. (2001). Ijime taiou to sonokouka [Treatments for bullying and their effects].
In Y. Morita (Ed.), Ijime no kokosai hikaku kenkyu (pp. 123-144). Tokyo: Kaneko
Twemlow, S. W., Fonagy, P., & Sacco, F. C. (2001). An innovative psychodynami­
cally influenced approach to reduce school violence. Journal of the American
Academy of Child and Adolescen Psychaitry, 40, 377-379.
Verlinden, S., Hersen, M., & Thomas, J. (2000). Risk factors in school shootings.
Clinical Psychology Review, 20(1), 3-56.
Wolke, D., Woods, S., Stanford, K., & Schulz, H. (2001). Bullying and victimization
of primary school children in England and Germany: Prevalence and school fac­
tors. British Journal of Psychology, 92, 673-696.
Wong, D. S. W. (2004). School bullying and tackling strategies in Hong Kong. Inter-
national Journal of Offender Therapy and Comparative Criminology, 48, 537-553.
Yang, J. (2005). 중학생의 왕따 발생 요인에 대한 연구 [A study on causal factors
of bullying among middle school students]. 한국교육 [Journal of Korean Educa­
tion], 32(2), 31-52.

Downloaded from by guest on January 31, 2013

Lee 1693

Young, R., & Sweeting, H. (2004). Adolescent bullying, relationships, psychological

well-being, and gender-atypical behavior: A gender diagnosticity approach. Sex
Roles, 50(7/8), 525-534.

Chang-Hun Lee, PhD, is an assistant professor at the department of criminal justice,
University of Arkansas, Little Rock. He earned his PhD from the School of Criminal
Justice at Michigan State University. His primary research interests include crimino­
logical theory and integration of theories, juvenile delinquency, and administration of
justice, especially policing.

Downloaded from by guest on January 31, 2013