Relations between media effects, social support and self ...
ABSTRACT
This paper uses an ecological perspective to explore the risk factors associated with bullying behaviors among a representative sample of adolescents aged 11 to 14 [pic] [pic] Data derived from the Health Behavior in School Children: WHO Cross-National Survey were used to model the relationship between bullying and media effects, peer and family support systems, self-efficacy, and school environment. Overall, the results of this study suggest that bullying increases among children who watch television frequently, lack teacher support, have themselves been bullied, attend schools with unfavorable environments, have emotional support from their peers, and have teachers and parents who do not place high expectations on their school performance. In addition, we found an inverse relationship between being Asian or African American, feeling left out of school activities and bullying. Our results lend support to the contention that bullying arises out of deficits in social climate, but that social support systems mediate bullying behavior irrespective of the student’s racial/ethnic characteristics, parental income levels or media influences. Because the number of friends and the ability to talk to these friends increases the likelihood of bullying, we suggest that bullying is not simply an individual response to a particular environment but is a peer-group behavior. We conclude that limiting television viewing hours, improving student’s abilities to access family support systems and improving school atmospheres are potentially useful interventions to limit bullying behavior.
INTRODUCTION
Contemporary society is increasingly concerned with the right to be free from victimization because of one’s sexuality, religion, race or physical characteristics/ limitation(s) (Smith, 2004). Bullying, which may occur in up to 30 percent of school-age children in the U.S. (Nansel et al., 2001; Zimmerman, 2005), is an aggressive behavior that targets individuals based on these characteristics. The many distinct forms of harassment that occur in the context of bullying, ranging from verbal threats to bodily injury and sexual assault, are viewed as detrimental to the physical, emotional and psychological well-being of victims as well as perpetrators. Bullying has repeatedly been shown to be related to more aggressive forms of violence and to be associated with negative outcomes in adulthood (Olweus, 1991; Perry et al., 1988; Tritt & Duncan, 1997). Bullying is viewed as a serious public health problem worthy of research and intervention (Zimmerman, 2005).
The recent attention given to understanding bullying behaviors has lead to numerous definitions of bullying. For example, bullying has been defined as “a systematic abuse of power” (Smith and Sharp, 1994, p. 2, emphasis added) and as repetitive, unpleasant behavior that takes place over time (Smith & Brain, 2000). These definitions are broad and have been applied to both adults and children in varied settings. The definition of bullying has recently expanded to include indirect aggression, relational aggression, and social aggression, including withdrawal of friendship, spreading rumors and excluding individuals from social groups (Smith, 2004; Crick, 1999; Underwood, 2002; Espelage & Swearer, 2004). In general, bullying is a form of child/adolescent aggression characterized by three primary and distinguishing features (Limber, 2004; Olweus et al., 1999): 1) behavior with the intent of doing harm to another individual, 2) behavior repeated over time, and 3) behavior that occurs in an interpersonal context involving an imbalance of power.
The Multiple Contexts of Youth Development
Prior work on bullying behaviors in children and adolescents involves a broad array of individual and contextual factors. An ecological framework that focuses on the interplay of these individual characteristics in multi-level contexts of development is particularly useful for understanding the complex dimensions of bullying and for developing sensitive and effective interventions (Limber, 2006; deLara, 2006; Garbarino & deLara, 2002; Bronfenbrenner & Morris, 1998).
The primary focus of the ecological perspective to human development (Bronfenbrenner & Morris, 1998) is on the dynamic interaction between the bully and the victim in the immediate and more distal contexts which include this behavior. The bullying relationship in question is defined by the interaction of the bully and the victim, including the characteristics of each, from the most immediate and primary to the more distal but significant influences (see Figure 1). At the core of the bullying relationship are the individual characteristics of the bully and the victim which, in turn, play out in the various contexts, leading to relationship attributes such as dependency and/or conflict. Emphasis is placed on understanding the bully’s individual characteristics in relation to the multiple social systems of which he or she is an inseparable part.
[Figure 1 here]
The most immediate context of development is the microsystem which includes the direct settings in which individuals develop (see Figure 1). A microsystemic influence focuses on the individual’s most immediate environment, such as support for the child’s behavior in the classroom or the larger school setting. These influences include, for example, parental involvement or interference in the bully/victim interaction and the character of bully/victim relationships in school settings. The mesosystem involves the interaction of two or more microsystems in influencing behavior. The joint contributions of two or more microsystems, such as family and school, can have a powerful impact on the positive development of children and youth, particularly in supporting academic achievement. Likewise, parent-teacher collaboration could prevent or mitigate physical and/or psychological damage from bullying and thus may have a significant impact on the welfare of both individuals. From the perspective of the bully or victim, the exosystem is a more distal context which does not directly include the participants but which may nonetheless significantly impact them. In the case of school bullying, exosystem factors include, among other things, the cumulative effect of school policies that shape institutional contexts and exert an influence on specific behaviors of teachers and/or students. For example, exosystem contextual factors could include specific staff training to reduce or prevent bullying.
The macrosystem consists of those factors affecting the welfare of the individuals in a most distant and least direct manner (see Figure 1). These factors include broader societal attitudes towards violence and carrying a weapon, whether or not hazing is an accepted part of school sports teams, or whether physically violent bullying is treated as a “boys will be boys” or “girls will be girls” behavior. Another macrosystem variable pertains to the role of the media, which reflects cultural or subcultural values and attitudes.
The chronosystem represents the effect of time on the behavior and on the context in which that behavior takes place. For example, a new child at school may initially engage in bullying and/or victim behaviors. Over time these behaviors may or may not become less prevalent. As another example, societal attitudes towards bullying may change over time. Given that a key element of bullying behavior is repetition of that behavior over time, the chronosystem, and the continuity of the character of the underlying systems (see above), provide a degree of stability or habitualness to bullying behavior. Of course, an important feature of ecologically informed interventions is that the underlying components of the chronosystem may be influenced or changed over time, comprising or enhancing the welfare of individuals and their relationships.
INDIVIDUAL CHARACTERISTICS AND CONTEXTS OF BULLYING BEHAVIOR: A REVIEW OF RESEARCH
Consistent with an ecological perspective, the review of research that follows addresses the key individual characteristics of bullies and victims as well as the significant contexts of bullying, including the school, the family, the peer group and the community as expressed through the media. While attention is given primarily to research on children in the United States, there are instances where international research is helpful, particularly in understanding and gaining perspective on bullying among and between ethnic groups (Elsea & Mukhtar, 2000) and where international perspectives offer insight into theoretical and methodological perspectives to bullying (Veenstra et al., 2005).
Individual characteristics of bullies. Two decades of research has painted a fairly clear picture of the individual-level correlates of adolescent bullying. Individuals who have been bullied in the past are more likely to bully others, have negative attitudes towards school, and engage in unhealthy behaviors such as tobacco and alcohol use (Nansel et al., 2001). Physical bullying is more prevalent among males whereas females are typically involved with verbal or psychological bullying (Olweus, 1991). Borg (1998) presents evidence of a parabolic relationship between bullying and age, emphasizing that the nature of bullying changes in form from more overt/physical behaviors, common among young children, to more covert behaviors as children get older. The relationship between ethnicity and bullying has not been widely studied (Seals & Young, 2003), therefore findings pertaining to the prevalence of bullying among ethnic minority students are inconclusive. For example, some researchers have found no significant differences in bullying behaviors among different racial/ethnic groups (Seals & Young, 2003). On the other hand, Elsea & Mukhtar (2000), who studied bullying in British schools, found that linguistic, religious and cultural differences among diverse ethnic groups are factors that precipitate bullying. Given the small number of studies on the topic, the inconsistent findings, different cultural contexts of international studies, and the failure to control for other factors previously found to be related to bullying in multivariate analyses, bullying among ethnic minority children is a topic that merits further attention (Elsea & Mukhtar, 2000).
Previous studies report a positive association between self-esteem and the ability to cope with stressors (Natvig et al., 2001; Bandura, 1997). In this vein, research on bullying behavior has addressed the relationship between bullying and self-esteem and depression (Smith & Young, 2003; Kaltiala-Heino et al., 1999). However, findings on these individual characteristics in relationship to bullying behavior are not clear (Smith, 2004). Some studies have shown that the enhancement of self-esteem and self-efficacy can be important protective factors for bullying (Rigby & Cox, 1996; O’Moore & Kirkham, 2001), others have suggested the exact opposite, including them instead among the many risk factors for bullying (Pearce & Thompson, 1998), and some have found no relationship at all between self-esteem, self-efficacy and bullying (Seals & Young, 2003). These equivocal outcomes are probably due to varying and inconsistent operational definitions of depression, self-efficacy, self-confidence, and self-esteem. Consequently, the relationship between these psychological moods and bullying is a research topic that merits further investigation.
School context and school climate. Relatively little is known about contextual/environmental factors that may predispose youths to bully others (Limber, 2006; Zimmerman et al. 2005). While the majority of bullying occurs on school grounds, little is known about the effect of school factors on bullying behavior in primary school children (Wolke et al., 2001). Researchers who have studied school-related factors tend to focus on the effect of school size, class size, (Batsche & Knoff, 1994; Olweus, 1997; Whitney & Smith, 1993; Wolke et al., 2001) and competition at school (Olweus, 1997) on subsequent bullying (Natvig et al., 2001). Few studies have focused on the pupils’ perception of different aspects of the school environment as potential risk factors for bullying behavior. One exception is a study by Natvig et al. (2001) who found that school-related stress and school alienation are potential risk factors for bullying behavior among Norwegian adolescents. The current research extends their results by incorporating notions of students’ perceptions of the school environment. Accordingly, we consider the relationship between school stress, as measured by excessive parent/teacher expectations, and stress relief, such as the ability to talk to parents and friends about difficult situations, and bullying in US schools.
Family context and family climate. Although most bullying behaviors occur in school, there is nonetheless increasing attention to the role of family socialization and sibling relationships in bullying and victimization. With the prominence of catastrophic high school tragedies (e.g. Columbine High School, Virginia Tech), society has awakened to the role of parents in preventing bullying behaviors (Lickel et al., 2003). Researchers studying families report that the parents of bullies typically lack involvement and warmth (Olweus, 1993). Furthermore, there is evidence that bullies tend to have relatively authoritarian parents who use—and therefore model—power assertive techniques of discipline and physical punishment (Bowers et al., 1994; Rodkin & Hodges, 2003). In contrast to the lack of involvement and warmth in families of bullies, there is evidence that the families of victims are sometimes characterized by overly protective mothers who may discourage the development of independence and self confidence in their children as well as fathers who are distant and overly critical or permissive (Olweus, 1993; Duncan, 2004). Not surprisingly, parental maltreatment of children in the family, including physical, emotional and sexual abuse, has been associated with both bullying and victimization behavior in adolescents (Sheilds & Cicchetti, 2001). In addition to parental influence, there is evidence that bullying-victimization behaviors in school are related to being a bully or a victim in sibling relationships (Wolke & Samara, 2004).
Peer group contextual effects. With the exception of bully-victim relationships, most peer relationships are based on voluntary interactions that reflect mutuality, reciprocity and positive companionship (Hartup & Stevens, 1997; Brown, 2003). Bully-victim relationships, however, are antagonistic and based on the exact opposite characteristics. Research on bully-victim peer associations suggests that both bullies and their victims are rejected by their peers, although bullies are the more aggressive partner in the relationship. In the bully-victim alliance, bullies seek out vulnerable peers to be their victims while victims appear to make themselves available targets (Brown, 2003). Bully-victim consequences are further influenced by other peer group characteristics including ethnicity of the peer group and broader sets of peers at the crowd level (Brown, 2003). For example, research at a multiethnic middle school pointed to the relatively smaller number of victims and larger number of bullies among African American peers (Graham & Juvonen, 2002). Furthermore, with the development of larger peer crowds in adolescence, individuals can be stigmatized as a group and, in turn, victimized or bullied by peers with more status and who are more aggressive (Merton, 1996). Although there are distinctive characteristics influencing victim/bully outcomes, Espelage et al. (2003), in a study of middle school early adolescent 6-8th graders, present evidence that peer group membership and contextual effects influence and shape adolescent aggression in both bullying and fighting behaviors.
The community as context for bullying: The role of media exposure. The role of exposure to media violence has long been a source of concern as a potential catalyst for aggressive behavior in children and adolescents. There is considerable evidence, and a longstanding research focus, on the relationship between children’s television watching and child/adolescent violence and aggression. In particular research on violence in films, television programs, music and video games indicates consistently and unmistakably that media violence increases the possibility of aggression and violence, in both immediate/short term and long term contexts (Johnson, et al., 2002; Anderson, et al., 2003; Robertson, et al., 2004). These findings are based on research with large and diverse samples, diverse methods and diverse media venues. While the relationship between media and violence/aggression is clearest in the case of television and film viewing, an increasing number of video game studies are pointing to similar results (Johnson et al., 2002; Anderson et al., 2003).
The amount of influence television watching and computer game playing have on individual development is conditional on specific factors, such as the age of the child and the level of family supervision. However, the role of media exposure, including television watching and video game playing, is not well understood in relation to bullying. Given the substantial research on violence and children’s television watching, the absence of research on adolescent media experiences and bullying is a significant omission. In fact, only two studies have specifically examined the relationship between either television watching or computer game playing and bullying behavior: a cross-sectional sample of youths in Switzerland and a longitudinal study of four year olds whose bullying behavior was assessed and reported by their mothers. The former study explains media usage in terms of bullying behavior (Kuntsche, 2004). The later study, by Zimmerman and colleagues (2005), found significant relationships between parental emotional support, cognitive stimulation, amount of television watching at four years of age and later bullying behavior reported at ages six through eleven. Cognitive stimulation was assessed using information on outings, reading, playing and parental role in teaching a child while emotional support was based on questions related to whether the child ate meals with both parents, parents talked to the child while working and spanking. According to the authors, “maximizing cognitive stimulation and limiting television watching in the early years of development might reduce children's subsequent risk of becoming bullies” (Zimmerman, 2005; p. 388).
While some attention has been paid to the effect of television watching on bullying, no study to date has addressed the influence of playing computer games on this type of child/adolescent aggression. Given a recent study that the average American child aged 2-17 plays seven hours of video games per week (Gentile & Walsh, 2002), the absence of research again represents a significant omission. This research topic is especially salient in light of the recent attention given to a new game by Rockstar Games (publishers of the Grand Theft Auto games) called “Bully,” released in October 2006, where players adopt the persona of a 15 year-old juvenile delinquent who terrorizes his victims with a range of physically and psychologically abusive behaviors.
METHODOLOGICAL LIMITATIONS OF EXTANT RESEARCH ON BULLYING
Beyond the lack of attention to such potentially significant dimensions of bullying behavior as the impact of video gaming, several methodological limitations can be identified in the research on bullying. A primary consideration is confounding incidence with frequency by combining categories of the response variable pertaining to the extent of bullying behaviors (Zimmerman, 2005). This is problematic because bullying is a sustained behavior occurring over time; dichotomizing the measurement of bullying removes the ability to assess frequency and thus introduces measurement error. We argue that equating “sometime” bullies with those who enact sustained bullying behavior tends to convolute results and contradicts most operational definitions of bullying as sustained, repeated behavior that occurs over time.
A second limitation is related to the operationalization of independent variables in studies of bullying behavior. Above we discussed how varying definitions of self-esteem and self-efficacy have resulted in different conclusions. Another example is Zimmerman et al.’s (2005) measure of parental emotional support, which included items related to eating meals, talking to the child while working and spanking. Eating meals and spanking may be inadequate measures of this concept, however. Spanking a child is a measure of physical maltreatment and not an indicator of emotional support. Any possible relationship to emotional support is indirect therefore rendering it a measure of unknown validity. In addition, Zimmerman relied on mothers’ reports of bullying behaviors in their children. This introduction of measurement error was compounded by the fact that the definition of “bullying” as it appears in the social science literature was not revealed to the mothers.
A third limitation pertains to the opportunity for intervening variables to influence the relationship between media and bullying in extant studies. In the oft-cited Zimmerman study on the relationship between bullying and television watching, there are too many potentially confounding influences to adequately support a direct relationship between the two. For example, in addition to gender, race, age and parents’ income and educational levels, which were controlled, other factors must be accounted for in order to assess the ceteris paribus effect of television watching on bullying. Previous research suggests that level of parental involvement, weapon carrying, self-esteem and helplessness, and teacher and student support are all factors that are related to each other, affect bullying and at a minimum should be included in the analysis. Therefore, we believe that evaluating concurrent television watching and video game playing while holding these other factors constant is a better means of assessing the direct relationship between media influences and bullying.
A fourth and final limitation pertains to focusing on the characteristics influencing bullying behavior in isolation, in univariate analyses, based on the relationship of single independent variables, taken one at a time, in relation to bullying behaviors. Such an approach has critical limitations given the complex and broad array of factors associated with bullying behavior, including gender, SES, parenting behaviors, and individual characteristics such as academic performance. Previous research clearly points to the need for multivariate, multi- contextual analyses such as the current investigation (Veenstra et al., 2005; Haynie et al., 2001).
The present research addresses past methodological limitations and, therefore, provides more reliable findings on early adolescent bullying, in the following ways. First, since parametric and categorical models are not optimal for ordinal level data, this research models the frequency of bullying behavior using ordinal logistic regression to assess how the predictor variables affect frequency. In this way, the contributions of these predictor variables to bullying behavior can be assessed while preserving the variation in the data and remaining “faithful” to the original question as it was asked to survey respondents. Second, proportional odds models have been used extensively to model ordinal outcomes. A crucial assumption of this technique is that the effect of the independent variables on each odds ratio is the same (i.e. for this research an example is that the effect of any mediating factor on the odds of never bullying versus “bullying” once or twice is the same as the effect on the odds of “bullying” once a week versus being a chronic and frequent bully), which is clearly restrictive (Long, 1997). In fact, the proportional odds assumption did not hold empirically with these data. This tends to cast doubt on previous research that models this same data to study the correlates of bullying using the proportional odds model (i.e. Nansel, et al., 2001) without making appropriate statistical adjustments. Additionally, we used the students' own perceptions of cognitive stimulation and emotional support which may be more directly related to bullying. And, while the Zimmerman et al. (2005) paper is innovative and informative, we believe that the measures we used in the current study represent an improvement over the measures of cognitive stimulation and emotional support used in their study because we measured students’ perceptions of cognitive stimulation in their immediate environments (e.g. school and home). Finally, by measuring the concurrent effects of media influences on the propensity of adolescents to bully others, the potential for intervening variables to affect the relationship between exposure and outcome is reduced.
RESEARCH QUESTIONS AND HYPOTHESES
We seek to understand adolescent bullying behavior and how the confluence of individual characteristics and multiple contexts are related to bullying. Ecological theory is designed particularly to model the relationships among individual characteristics and multiple contexts (Bronfenbrenner & Morris, 1998; Espelage, et al., 2003; Swearer & Espelage, 2004), and serves as the underlying conceptual foundation for this study. Prior research provides guidance regarding the specific individual characteristics and contexts that act either as risk factors or mitigating factors for bullying.
Specific research questions arising from the ecological framework include:
1) What individual characteristics and psychological states are most associated with bullying behavior?
2) At the microsystem level, what types of teacher behaviors, peer group interactions and family relationships are most associated with bullying behavior?
3) At the mesosystem level, to what extent does the influence of two or more microsystem contexts (e.g. families and schools) reduce or enhance the likelihood of bullying?
4) At the exosystem level, do contexts that only indirectly include the bully (e.g. school climate) influence bullying?
5) At the macrosystem level, is there a relationship between concurrent media influences (e.g. television watching and computer game playing) and bullying?
Prior research on bullying behavior suggests a parabolic relationship between bullying and age. We therefore hypothesized that the relationship between age and bullying among 11-14 year olds would reflect this pattern, with the overall amount of bullying increasing over time. Given the research support for differential gender prevalence for bullying, we hypothesized that males would be more likely to bully than females. In light of both the sparse and inconclusive findings on the relationship between bullying and race or ethnicity, we had no specific expectation for the direction of our findings. Previous research on the psychological correlates of bullying, such as feeling self-confident or helpless, is unclear as well. Therefore, we did not have specific directional expectations for our findings in this regard either.
Prior research suggests that bullies tend to lack social support systems both at home and at school. It was anticipated that individuals who have positive emotional support from parents and/or peers will be less likely to bully. Given that previous research shows that the parents of bullies are uninvolved in their children's lives (Olweus, 1993), we expected that parental involvement in school and bullying would be inversely related. Peer-group relationships and friendships can influence and shape both positive relationship patterns and, as well, adolescent aggression in bullying and fighting behaviors (Espelage et al., 2003). From a purely quantitative perspective, the directional relationship between number of friendships and bullying remains unclear, and would appear to depend on the type and quality of peer influence. On the one hand, an individual with more friends may be more sociable and therefore less likely to bully. However, peer relationships may sanction undesirable behaviors (Nansel et al., 2001), and hence bullying may increase as peer support for those behaviors increases. We expected that self-identified bullies who have extensive peer connections are likely to be affiliated with peer groups that sanction or endorse fighting or bullying behaviors. Moreover, we hypothesized that apathetic teachers, for example those who show little interest in students as people or treat them unfairly, create a classroom climate that would be conducive to bullying.
Prior work has found that the propensity to bully others may be related to an individual's inability to cope with school-related stressors (Natvig et al., 2001). Consequently, we anticipated that individual's who perceive that their parents and teachers have unreasonable expectations of their school performance would be more likely to bully. Moreover, we expected that children whose parents support their school activities, for example by encouraging them to do well at school or by manifesting a willingness to resolve any problems that they may have at school, would be less likely to engage in bullying.
We hypothesized that the general school atmosphere, as measured by students’ perceptions that their school is welcoming, pleasant and a place where they “belong,” would be negatively related to bullying. The broad framework for peer relationships, for example, student perceptions of whether they enjoy being together, whether they feel accepted, or whether the atmosphere is one of students expressing concern and help, would be related to bullying. We anticipated, then, that students who feel positively supported in such a context are less likely to engage in bullying. Finally, given the larger cultural context that reifies violence in television and video games, we presupposed that excessive media usage is related to an increased prevalence in bullying. Since there is evidence that bullying may be a more frequent problem in urban schools, we hypothesized as well that individuals in urban settings may be more prone to bullying behaviors than individuals in smaller communities.
DATA AND METHODS
Participants
Data were drawn from the 1997 – 1998 World Health Organization’s Health Behavior in School-Aged Children Survey (HBSCS). The HBSCS is a nationally representative school-based study intended to (1) monitor health-risk behaviors and attitudes in youth over time to provide background data and to identify targets for health promotion initiatives, and (2) provide researchers with relevant information in order to understand and explain the development of health attitudes and behaviors through early adolescence. Although these data were collected in 1997-1998, they have nonetheless been recently utilized in important research on bullying (e.g. Nansel et al., 2001). In addition, the HBSCS has several advantages. First, it is a representative sample of 15,686 students in grades 6 through 10 in public and private schools in the United States. We focused on the 9,816 students aged 11 – 14. Our sample of 9,816 11-14 year olds included 5,142 non-Hispanic whites (52.3%), 1,575 non-Hispanic blacks (16%), 2470 Hispanic/Latinos (25.2%), 488 non-Hispanic Asians (5%) and 141 non-Hispanic Native Americans (1.5%). Our final regression analysis included 7,946 cases such that 81% of the cases were retained. Since the survey is nationally representative, our conclusions are generalizable to the population of American children aged 11 – 14. In addition, the large sample size has positive implications for power assessments in statistical analyses. This is a marked improvement over other studies that are based on less representative surveys and/or samples using small sample sizes.
Statistical analyses were carried out using STATA, Version 9.0 (StataCorp, College Station, TX). Partial proportional ordinal logistic regression equations[1] were used to model the effects of these variables on the propensity to bully while accounting for 2-stage cluster sampling survey design effects and covariates related to both the frequency of bullying and other independent variables described above.
Outcome Measure
The outcome variable was designed to measure the self-reported frequency of bullying behaviors among respondents in the sample. As distinguished from aggression, bullying is understood in the context of a power differential relationship between a dominant child and a weaker child characterized as ongoing and repeated. The following definition (Olweus, 1993) was introduced to our final sample of 7,946 students aged 11 – 14. The definition clearly emphasized the repeated and ongoing nature of the bullying relationship and distinguished bullying from aggressiveness perpetrated on a victim who cannot readily defend him- or herself:
“We say a student is being BULLIED when another student, or a group of students, say or do nasty and unpleasant things to him or her. It is also bullying when a student is teased repeatedly in a way he or she doesn’t like. But it is NOT BULLYING when two students of about the same strength quarrel of fight. How often have you taken part in bullying other students in school this term? (italics added)”
Although bullying is conceptually different from aggression, there is a strong relationship between aggressiveness, as operationalized by number of fights the child has fought, and bullying among these school-aged children [pic]
Responses were coded 1 if the student had not engaged in bullying behavior (no bullying), 2 if the behavior occurred “infrequently” (i.e., once or twice), 3 if the behavior occurred “sometimes” (i.e., more than once or twice but not weekly), 4 if the bullying was chronic but not frequent (i.e., occurred on a weekly basis), and 5 if the bullying was both chronic and frequent (i.e., several times per week). In the context of the literature that defines bullying in terms of systematic and repeated behavior (Limber, 2004; Olweus et al., 1999), we define bullying behavior to be chronic and frequent (i.e., bullies can be placed into the last category given above). However, we report results for non-bullying (category 1) versus bullying behavior (category 5) with the understanding that if the proportional odds assumption was not violated for that variable then the odds are the same for each contrast (i.e. category 1 versus 5 is the same as category 1 versus 2, 3 or 4). In reporting our results, the term “bullying” refers to chronic and frequent aggressive behavior fitting into category 5. Table 1 reports the frequency distribution of each category.
Main Predictors and Covariates
Explanatory variables used in this study have been associated with bullying or other aggressive behavior in previous literature and inclusion was based solely on existing theoretical and empirical studies. The variables may be grouped into categories representing the microsystem, mesosystem, exosystem and macrosystem (Table 2). Composite measures of our subscales were derived via exploratory factor analytic techniques and items that did not load above 0.30 were excluded from further analysis with one exception (Table 3) (See also Table 2). Despite the fact that “Easy to talk to mother” did not load above .30, it was included in the analysis for theoretical reasons.
[Table 2 here]
[Table 3 here]
Individual-Level Variables
Individual characteristics. We included as individual characteristics self-reports of age, gender, and ethnicity. Dummy variables were included to estimate the differential effect of being female (female = 1) versus being male (male = 0). In addition, dummy variables coded the effect of being African American (African American = 1), Asian (Asian = 1), American Indian (American Indian = 1), or Pacific Islander (Pacific Islander = 1) versus being White (White = 0). Age is a continuous variable ranging from 11 to 14.
Self-confidence, helplessness and feelings of being left out. The individual characteristics of feeling 1) left out; 2) helpless; and 3) self-confident were included as well. Students were asked: “How often do you feel:” (a) “left out of things,” (b) “helpless,” or (c) “confident in yourself.” Response choices included 1 = never, 2 = rarely, 3 = sometimes, 4 = often, and 5 = always.
Additional individual-level variables were parental education, parental income, the presence of both parents in the household, whether the individual was him- or herself a victim of bullying in the past and whether the individual had previously carried a weapon to school.
Microsystem Variables
Emotional support from parents and friends. Separate scales of parent (Cronbach’s alpha = .54) and friend (Cronbach’s alpha = .60) emotional support were created due to the poor reliability of the combined measures. Students were asked to report the level of ease/difficulty they have talking with their parents or friends. Responses included 1 = very easy, 2 = easy, 3 = difficult, 4 = very difficult, 5 = impossible. Possible values for each scale ranged from 2 to 10, with 2 meaning strong social support and 10 meaning a complete lack of social support.
The role of friends. It is assumed that the number of friends that students have proxies the amount of social support they have and therefore “number of friends” was added to the model.
Teacher apathy. Teacher apathy includes measures designed to assess the level to which students feel (1) prohibited to express their views in the classroom, (2) the teacher shows little interest in them, (3) they are treated unfairly, and (4) the teacher is unwilling to provide extra help when it is needed. Possible responses included 1 = strongly agree, 2 = agree, 3 = neither agree nor disagree, 4 = disagree, and 5 = strongly disagree. Total scores ranged from 4 through 20 (Cronbach’s alpha = .78). Higher scores indicate increasing levels of teacher apathy.
Mesosystem Variables
Parental support at school. A parental support scale was constructed based on the degree to which the respondent agreed that their parent(s) were (a) ready to help with problems at school, (b) willing to come to school to talk to teachers, and (c) encouraged them to do well in school. Responses ranged from 1 = never, 2 = rarely, 3 = sometimes, 4 = often, and 5 = always. Total scores ranged from 3 to 15 (Cronbach’s alpha = .82); higher values indicate increasing levels of parental support.
School-related stressors. Our analysis considered two measures designed to assess the level of stress students feel from unreasonable expectations of their school performance. In two separate questions, respondents were asked whether they agreed that their teachers and parents expect too much from them in school. Responses ranged from 1 = strongly agree, 2 = agree, 3 = neither agree nor disagree, 4 = disagree, and 5 = strongly disagree. These separate scores were summed to create an index whose values ranged from 2 through 10 (Cronbach’s alpha = .73); higher values indicate increasingly reasonable expectations.
Exosystem Variables
School atmosphere. We created scale items to represent aspects of school climate. Respondents were asked to indicate the extent to which they agree that (a) the rules in school are fair, (b) the school they attend is a nice place, and (c) they belong in school. Possible responses ranged from 1 = strongly agree, 2 = agree, 3 = neither agree nor disagree, 4 = disagree, and 5 = strongly disagree. Possible values range from 3 through 15 (Cronbach’s alpha = .76). Larger values are indicative of a negative school climate and thus students who score highly on this measure are more likely to be associated with bullying.
Broader peer group relationships. We included a scale designed to measure the extent to which respondents believe that (a) students enjoy being together; (b) other students are kind and helpful; and (c) they are accepted by other students. Possible responses included 1 = never, 2 = rarely, 3 = sometimes, 4 = often, and 5 = always. Scale values ranged from 3 through 15 (Cronbach’s alpha = .73). Higher values indicate increasing levels of peer support and acceptance.
Macrosystem Variables
Media effects. Media effects are proxied by two variables: hours of television and hours spent playing computer games. Respondents were asked to report the number of hours spent per week playing computer games. Possible responses ranged from “not at all” through “10 or more.” They were also asked the number of hours spent per day watching television. Responses ranged from “not at all” through “more than 4 hours.”
Urbanicity. Because bullying has been shown to be more of a problem in urban schools (Fleming et al., 2002), we controlled for urban residence by including, in our model, a dummy variable equal to unity if the respondent lived in a big city.
RESULTS
The beta coefficients (log-odd ratios) and associated p-values resulting from the estimated partial proportional ordinal logistic regression model are presented in Table 4. For purposes of brevity, we only show the results for the “Bully” (i.e. category “5” of our dependent variable, n = 385) versus the “Not a bully” (n = 4529) contrast when the proportional odds assumption was violated. The omitted comparisons include 1 (not a bully) vs. 2, 1 vs. 3 and 1 vs. 4. A dash (-) in the table means that the assumption is satisfied and the beta coefficient can be interpreted as applying to all levels of the dependent variable. The first two columns present the results from the proportional odds model. This model assumes that each independent variable has the same influence on the odds of moving from category 1 (no bullying) to category 2 (once or twice) as that factor does on the odds of moving from category 2 to category 3 (sometimes), etc. The second two columns present the adjustment to these odds ratios when the assumption of proportional odds is violated. The test statistics corresponding to the beta coefficients shown in the second two columns of table 4 indicate that significant deviations from the proportional odds model exist. The overall effect of each independent variable on the odds of being a chronic and frequent bully is determined from the combination of beta coefficients in column 1 and adjustments in column 3.
Table 5 presents the results of the logistic regression in terms of odds-ratios; i.e. it transforms the beta coefficients and proportional odds adjustments presented in Table 4 into a more understandable form. An odds ratio (OR) represents the effect of an independent variable on the likelihood (or odds) of being a bully (n = 468) relative to the likelihood (or odds) of not being a bully (n = 4529) In addition, when substantive changes are of interest, we report standard deviation changes and/or changes in predicted probabilities.
[Table 4 here]
[Table 5 here]
Individual Characteristics of Bullies
The first entry in Table 5 shows that the odds of being a bully are 0.80 times as high for females as for males, i.e. females are 20% less likely to be bullies although this result is not significant at conventional levels (Z = -1.89, p = .059). The odds of bullying increase significantly with age. Being a year older increases the odds of bullying by 6%, holding other variables constant (Z = 2.24, p = .025). The odds of bullying are increased by 24% when age is allowed to vary over its range (i.e., from its minimum value of 11 to its maximum value of 14). African American (Z = 4.21, p = .000) and Asian (Z = -3.46, p = .001) adolescents are significantly less likely to engage in bullying behaviors than White adolescents. The odds of bullying among African Americans and Asians are 25% and 30% less than they are for Whites, respectively. None of the other race/ethnicity variables were statistically significant. Our results further indicate that as parental income increases so does bullying (Z = 3.27, p = 0.001). The variables measuring mother and father education and living with both parents, however, were statistically insignificant.
With respect to the psychological correlates of bullying, we found a significant and negative relationship between feeling “left out” and the frequency of bullying behaviors among school aged children (Z = -4.53, p = 0.00). Additionally, although not statistically significant, we found that the more students feel “helpless,” the more likely they will bully (Z = 1.69, p = .092). Lacking self-confidence was not statistically significant and hence we are unable to conclude that it is a risk factor for bullying.
Individuals who were previously bullied at school are significantly more likely, in turn, to be bullies. The odds of bullying are 34% greater among individuals who reported being bullied at school during the current school term (Z = 12.82, p = 0.000). Finally, the odds of bullying are 46% higher among students who reported carrying a weapon to school (Z = 8.50, p = 0.00).
Microsystem Factors
Emotional support from parents and friends. Our measure of parental emotional support was statistically significant and positively related to bullying (Z = 4.88, p = 0.000). Students who have greater difficulty discussing their problems with parents are more likely to engage in bullying behaviors. For example, the odds of bullying are 7% higher among individuals who lack emotional support from their parents. Additionally, our measure of friend emotional isolation strongly predicted bullying. Students who are less isolated from their friends are more likely to engage in bullying (Z = -6.56, p = 0.000).
Peer group contextual effects. The role of one’s peers in facilitating bullying behaviors extends as well to the number of friends one has: each additional friend increases the odds of bullying by 12% (Z = 5.43, p = 0.000).
Teacher apathy. Teachers who are supportive, take an active interest in students, and treat them fairly, create an environment where bullying is less likely. For example, a two standard deviation decrease in our measure of teacher apathy decreases the odds of bullying by almost 24% (Z = 4.39, p = 0.000).
Mesosystem Factors
Parental support at school. Our measure of parental support at school was statistically insignificant; therefore we cannot conclude that inadequate parental involvement in school is a risk factor for bullying.
School-related stressors. Overly permissive parental and teacher relationships with students are likely to foster an environment where bullying results. More specifically, the odds of bullying increase by almost 5% among students whose parents and teachers hold low expectations of their school performance (Z = 1.97, p = 0.048), contrary to our prediction.
Exosystem Factors
School atmosphere. Our measure of school climate was statistically significant as well. Each standard deviation increase in our measure of school atmosphere increases the odds of being a bully by 44% (Z = 3.38, p = 0.001). Recall from above that our measure of school atmosphere consisted of respondents’ perception that the rules in school are fair, the school they attend is a nice place, and that they belong in school. Our results therefore indicate that transitioning from a students’ perception of a pleasant school atmosphere characterized as fair, welcoming and pleasant to one characterized as unpleasant, unfair and unwelcoming increases the predicted probability of being a bully by 6%, holding other variables constant. Our measure of broader peer group relationships, however, was not statistically significant.
Macrosystem Factors
The community as context for bullying: The role of media exposure. While our results indicate that urbanicity is not related to bullying, our measure of television watching has a very large and statistically significant effect on the probability of being a bully (Z = 7.84, p = 0.000). Each standard deviation increase in hours of television watched per day increases the odds of being a bully by 21%, holding other variables constant. The coefficient on our variable corresponding to hours of computer game playing per week was statistically insignificant. For comparison purposes, each standard deviation increase in hours of computer game playing per week increased the odds of bullying by only 1.5%.
In Figure 2, we let the number of television hours per week vary and plotted the effect of television watching on the probability of bullying while the other variables were held constant at their means. At zero hours per day of television viewing the probability of being a bully is approximately 1.5%; as television viewing hours increase the probability of bullying nearly doubles to 2.8% for an adolescent who watches four hours of television per day (just under the sample average of 4.1 hours). Moving to the sample maximum of six hours per day, the probability of bullying nearly doubles again, to exceed 5%.
[Figure 2 here]
The Set of Patterned Interrelationships Among Subsystems: Interaction Effects
In keeping with an ecological approach, we were interested in the multiple contextual effects of our variables and their interactions. To investigate these effects, we defined combinations of characteristics corresponding to ideal types in the population based on [pic]combinations of characteristics. These characteristics included the self-identified bully’s gender and racial group as well as his or her status as living with both parents or in a big city. In Table 6, we symbolize each of the possible combinations with IRijkl where i = 1 if the individual is male, j = 1 if the individual lives with both parents, k = 1 if the individual is black and l = 1 if the individual lives in a big city. Next, we allowed the amount of television watching to vary over its range. We then predicted the probability of bullying for several combinations of variables. The effect of parental, teacher and emotional support, self-confidence, and school atmosphere on the probability of bullying at maximum and minimum values of television watching for each combination of the above characteristics are presented in Table 6.
[Table 6 here]
Table 6 shows that despite constancy across combinations of variables, the effect of television watching on bullying varies within levels of the predictor variables. Disregarding for the moment perceptions of school environment, level of self-confidence, hours of television watched per day, and the level of parental, teacher and emotional support, males consistently exhibit a higher probability of bullying than do females. Similarly, when examining the interaction between race and gender, it becomes clear that White males are more likely to be bullies than are Black males or females of any race, again disregarding other factors.
Clearly, television watching affects the likelihood of being a bully: the predicted probability of bullying is higher among individuals who watch several hours of television per day (i.e., television watching is high) and this holds across all categories and irrespective of socio-demographic characteristics. Conversely, children who claim to watch television infrequently (i.e., television watching is low) are less likely to self-report being a bully conditional on being school-integrated (i.e., school atmosphere is high) and possessing high levels of parental and teacher emotional support (i.e., parental and teacher emotional support is high). In addition, children who watch several hours of television per day but who have higher levels of teacher support are less likely to bully others irrespective of race, gender, place of residence and parental presence in the home. For students who watch television frequently, school atmosphere is the most important mediator of bullying behavior. When negative school atmosphere is at its max within high levels of television, the probability of bullying is higher than it is across any other category. In fact, the probability of bullying is highest for White males living with both parents in an urban area who frequently watch television and attend schools with adverse climates (IR1101 = .124).
Overall, the effect of television watching on the odds of being a bully is the largest for males living in urban areas who live with both parents [pic], who attend schools with adverse environments, and report having very little teacher support. For example, if a child who watches television frequently does not have support from his teachers, the probability he will bully increases from 5.1% to 8.0% for Blacks and from 5.3% to 8.2% for Whites who live in big cities with both parents present. Poor school environments similarly increase the probability of bullying from 3.3% to 12.0% for Blacks and 3.4% to 12.4% for Whites. Black females who feel supported by their teachers and who attend schools with positive environments have the lowest probability of bullying. For example, the predicted probability that a Black female living in single parent home in a rural area will bully is 0.9%. Interestingly, having low levels of self-confidence increases the likelihood of bullying when television is at its maximum but decreases the probability of bullying when television is at its minimum.
DISCUSSION
Our interpretation of the results followed naturally from an ecological perspective to understanding the nature of bullying behavior. This perspective moves from the microsystem to the mesosystem, exosystem and then to the macrosystem, emphasizing the set of patterned interactions that occur between them. At each contextual level, we found statistically significant effects, suggesting that our conceptual understanding of bullying behavior is advanced by using an ecological model as the theoretical lens in which we come to terms with this type of childhood aggression. In what follows, we summarize our main findings, note similarities and differences between the current research and previous studies, and offer our interpretation of the results.
Based on the pattern of responses, our unadjusted estimate is that approximately 4.9% of adolescents aged 11-14 are chronic and frequent bullies. Our adjusted estimate suggests that, on average, 3% of students aged 11-14 are chronic and frequent bullies. These are considerably low estimates in relation to Nansel et al.’s (2001) finding of a 19.3% prevalence; however, this is likely the result of methodological differences in the operationalization of “bullying.” Rather than including acute or infrequent acts, the present estimate relied on defining bullying as a frequent behavior that occurs several times per week. While measuring bullying as a dichotomous, or yes/no, variable produces a prevalence estimate of 42.6%, excluding only those who never or rarely bullied from the estimate yields a prevalence of 16.8%. Thus, our analyses suggest that researchers failing to account for frequency in their operationalization of bullying are likely to report a 3-8 fold overestimation (Zimmerman, 2005).
Apart from methodological differences, our study differs from the Nansel et al. (2001) study in additional ways. We included measures of parental and friend emotional support, number of friends the bully has, media usage and measures of perceived teacher support. In addition, we focused on a smaller subset of adolescents, namely those between the ages of 11 and 14. Finally, using the same data set, we found a violation in the proportional odds model technique that provided the basis for their estimation. Therefore, we turned to statistical methods designed to account for this violation in the data.
At the level of risk factors associated with the individual, our findings are consistent with other studies reporting that bullying is more common among boys than girls (Natvig et al., 2001; Olweus, 1997). Anderson et al., (2001) reported that perpetrators of violence were more than twice as likely to report being bullied and on this basis concluded that bullying may be related to other types of school violence. We provide further evidence that victims of bullying are more likely to bully others (Haynie et al., 2001). Anecdotal evidence suggesting that race and/or ethnicity are not risk factors for bullying behaviors are contradicted by our results. The coefficients pertaining to two race variables in our model were significant and therefore we conclude that race can be added to the growing list of risk factors for bullying. In particular, Whites have a higher probability of being (chronic and frequent) bullies than do either Asians or African Americans.
Prior research has been unclear as to the effect of low levels of self-esteem on the proclivity to engage in bullying. For example, in their study of Irish school children, O'Moore and Hillery (1989) found that bullies tend to have lower levels of self-esteem (see also, Duncan, 1999, Rigby & Slee, 1991, Tritt & Duncan, 1997). Others have suggested that greater amounts of self esteem increase bullying (Natvig et al., 2001, Olweus 1997, Slee & Rigby, 1993). We did not have a specific measure of self-esteem, using instead a measure of self-confidence. Therefore, we cannot settle this ongoing debate either way. On the other hand, we found that, controlling for other factors, feeling helpless increases bullying behaviors. In this respect, our results are consistent with prior research finding that bullies feel more helpless than non-bullies (Duncan, 1999). It is possible that students who feel helpless engage in bullying as a means of empowering themselves. Feeling helpless may also arise from having inadequate social support systems.
In accord with Zimmerman et al.’s (2005) findings in children, our results suggest that bullying may arise out of deficits in social support among adolescents. We found that involvement in bullying is related both to interactions with parents, teachers and friends and to the type of support provided by each. Students with parental emotional support are less likely to be bullies. It seems intuitively plausible that students who feel strong emotional support from their parents will turn to discussion as a means of tempering any aggressive tendencies that might otherwise result in bullying. It is also likely that adolescents turn to their parents to discuss problems, issues and concerns, thereby increasing the salience of having parental emotional support.
Our results indicate that emotional alienation from friends decreases bullying. At first glance, this appears to be a counterintuitive finding. Previous research has suggested, however, that bullies tend to associate with other students who engage in bullying behavior, such that peer groups of bullies induce individual group members to engage in even more bullying (Espelage et al., 2003). Unfortunately, we were unable to assess the characteristics of the bully’s friends. However, this view that bullying is supported by association with other bullies is corroborated by two additional findings. First, we found that the probability of bullying increases with number of friends. Second, our results show that students who feel included in school activities (i.e., do not feel left out) are more likely to bully (this could also be because students who are more involved in school activities have more opportunity to bully). Clearly, bullying does not exist in a vacuum but is influenced by the broader environment, in which friendships play an important role.
Perceived social support from teachers was associated with lower probabilities of bullying. Our results suggest that the frequency of bullying depends on the extent to which teachers 1) take an active role in promoting student welfare, 2) are interested in helping students in need, 3) allow for the possibility of alternative forms of self-expression, 4) promote cooperation, and 5) create an equitable school environment. Furthermore, these results illustrate the critical relationship of the school context to the likelihood of bullying behavior, directly through supportive teacher behavior and indirectly in structuring the classroom setting.
The results of the current study also provide support for previous findings that students’ perceptions of different aspects of the school environment are risk factors for bullying (Natvig et al., 2001). While Natvig et al. (2001) focused on the nature of school work as a measure of school stress, we considered an alternative aspect of the school atmosphere: the degree to which students experience a pleasant, welcoming and fair school environment. Quite possibly, it may be that different aspects of the school environment, as perceived by the student, are observed as effects rather than causes of bullying (Natvig et al., 2001). The strength of our measure is that we included the students’ own perceptions, namely whether they believe their parents and/or teachers expect too much from them at school. School-related stress in the form of unreasonable expectations placed on students by their parents and teachers was found to decrease bullying behaviors, contrary to our prediction. This finding is actually consistent with research implying that children who engage in bullying are more likely to come from homes in which parents are extremely permissive (Olweus, 1993).
Given that preferred leisure activities of adolescents include playing computer games (Cesarone, 1994) and watching television, it is surprising that only a handful of researchers have studied their relationship to bullying. In any event, the ceritus paribus effect of television watching on bullying frequency is of particular interest. Our analysis revealed evidence to support Zimmerman et al.’s (2005) finding of a significant relationship between television watching and bullying. We extended their results to show the separate and independent effects of television watching on subsequent bullying. We found that while the odds of bullying increase by 28.7% per standard deviation change in hours spent watching television, the odds of bullying increase by only 1.5% per standard deviation change in computer game playing. An explanation for this finding was offered by Zimmerman et al. (2005) who noted that a high percentage of television programs contain explicit or implicit violence which could serve as a model for children to engage in aggressive behaviors, including bullying. The causal mechanism is believed to be that media violence causes aggressive behaviors and/or cognitions, increased arousal, and an aggressive affective state that is internalized during psychosocial development (Bushman & Anderson, 2002). While in our study we were unable to differentiate between violent and non-violent television viewing and computer gaming, others have reported that approximately 60% of television programs and 60%-90% of computer games contain violent themes (National Television Violence Study, 1998). This is especially true for school-aged children who are exposed to a much broader range of television programs than adults. Thus, our measures of viewing and gaming hours are likely to provide reasonable measures of exposure to media violence.
A student with low self confidence and who watches a lot of television is more likely to take behavioral cues from television, which can be violent, leading to a greater chance of being a bully. A student with low self confidence who does not watch television is more likely to take behavioral cues from other, less violent sources (e.g. parents, teachers, etc.) and therefore less likely to bully. Conversely, students who watch very little television and/or have higher levels of self confidence are less likely to take behavioral cues from television and therefore less likely to bully. Overall, the probabilities discussed here demonstrate the importance of interaction effects between the four socio-demographic characteristics when seeking to develop an understanding of bullying behavior, television viewing hours and the mediating influence of social support systems. More importantly, many of the differences in predicted probabilities we presented illustrate the results found within the partial proportional logistic regression. Specifically, children who lack a strong social base due to low levels of teacher support or who attend schools that are unwelcoming and unfavorable have a higher probability of engaging in bullying, but the bullying behavior is mediated by other influences, such as the number of hours of television watching per day.
Our last macrosystem variable, urbanicity, was an attempt to measure the broader contextual environment in which the school is located. More specifically, since there are many schools in any particular geographic locale, all of which are presumed to suffer from the same kinds of social problems that aggravate bullying (Ortega & Lera, 2000), we included a variable to measure the differential effect of urbanization on bullying. Our variable was not significant so we were unable to conclude that place of residence significantly affects the propensity to bully.
LIMITATIONS AND DIRECTIONS FOR FUTURE RESEARCH
By conceptualizing bullying as multidimensional and by considering many variables at multiple levels of analysis, the current study presents a clear advance on past research. Nevertheless, it is important to acknowledge several limitations that exist in our study. First, as noted above, the nature of the data poses some limitations. The data set is 10 years old and was not originally collected for the purpose of studying the concomitants of bullying behavior. However, the age of the data set only reinforces the need to collect more representative data on bullying behaviors of students in this age range. Also, due to the cross sectional nature of the HBSCS data, we are unable to establish any causal relationships and hence the time ordering between exposure and outcome is difficult to establish. Therefore, our analysis rests primarily on theoretical considerations that violence is a learned response to negative environmental stimuli such as media influences. Some researchers prefer instead to emphasize a different directional relationship by noting that individuals with violent or aggressive tendencies are inclined to play video games and watch television (Kuntsche, 2004). Second, given the rapid pace at which the content of both television and computer games changes, our results should be interpreted with caution. As such, the lack of effect for game playing in this study might not apply to the more violent games of today. Third, in the present study, the data on bullying are based on self-report, which may have implications for our results. Self-reported bullying behavior is bound to be an under-representation of its actual frequency given that social desirability effects are likely to exist in the reporting of negative behavior. Different methodologies used to report bullying across studies have resulted in conflicting findings. Observation of students’ behavior may be a more accurate way of ascertaining bullying because a bully may be unconscious or unaware that he is engaging in this type of behavior. Some researchers have argued, however, that teachers’ reports of bullying behavior may not accurately reflect the extent of bullying that takes place in settings away from the classroom, such as on the playground, where bullying occurs most frequently (Pellegrini & Bartini, 2000). Future research should attempt to replicate our findings with data that measures bullying by peer report since self-report probably underestimates the true amount of bullying in the population. Fourth, our findings could be strengthened via the use of additional data that is more targeted and specific. For example, understanding the precise mechanisms in which peer groups influence bullying would be more informative than our current measure “number of friends.“ Also, self-confidence, helplessness and feeling “left out” are in reality more complex and should be measured as multidimensional rather than as single item constructs. In this regard, we were constrained by the data. Consequently, our knowledge would greatly benefit from future studies that focus more specifically on the causal mechanisms of self-esteem and confidence, feelings of helplessness and peer group functioning such as popularity, group status in the broader social network as well as levels and approval of aggressive behavior in the group.
CONCLUSION
Bullying is a complex behavior with multiple causes and risk factors, ranging from individual characteristics to school settings to broader social contexts. In our view, the ecological perspective provides both a vehicle for better understanding the complex features of bullying and also for crafting sensitive and effective interventions at multiple contextual levels. For example, teaching students to “use their words” to resolve conflicts is unlikely to be successful if the school is perceived negatively by students (who may then tend to ignore the school’s recommendation on conflict resolution) and/or if the school condones or ignores bullying in other contexts such as during athletic programs. Additionally, interventions at both the individual and school level may be insufficient if parents use bullying and aggressive behavior at home or encourage children to solve problems physically. Finally, broader societal attitudes towards violence—as exemplified by the violent content of television shows—increase the likelihood of chronic and frequent bullying. Thus, interventions to reduce bullying are most likely to succeed if they address contributors to bullying behavior at every level from the individual to school and family to society.
ACKNOWLEDGEMENTS
We would like to thank the United States Department of Justice -- Office for Juvenile Justice and Delinquency Prevention and the Michigan Department of Human Services Grant #: 071B2001414 for funding this research.
REFERENCES
Anderson, C.A., Berkowitz, L., Donnerstein, E., Huesmann, L.R., Johnson, J.D.,Linz, D., Malamuth, N., and Wartela, E. (2003). The Influence of Media Violence on Youth. Psychological Science in the Public Interest. 4: 81-110.
Anderson, M., Kaufman, J., Simon, T., Barrios, L., Paulozzi, L., Ryan, G., Hammond, R., Modzeleski, W., Feucht, T., Potter, L., and the School-Associated Violent Deaths Study Group. (2001). School-associated violent deaths in the United States, 1994-1999. Journal of the American Medical Association, 286: 2695–702.
Bandura, A. (1997). Self-efficacy: The exercise of control. New York: Freeman.
Batsche, G. M. and Knoff, H. M. (1994). Bullies and their victims: Understanding a pervasive problem in the schools. School Psychology Review, 23: 165-174.
Borg, M. G. (1998). The emotional reactions of school bullies and their victims. Educational Psychology, 18: 433-444.
Bowers, L., Smith, P.K., and Binney, V. (1994). Perceived family relationships of bullies, victims and bully/victims in middle childhood. Journal of Social and Personal Relationships, 11: 215–232.
Bronfenbrenner, U., and Morris, P. (1998). The ecology of developmental process. The handbook of child psychology, 1: 993-1029.
Brown, M. (2003). Bully-proof your school. Today’s School (3) 5. Dayton, OH: Peter Li Education
Group.
Bushman, B. J. and Anderson C. A. (2002). Violent video games and hostile expectations: A test of the general aggression model. Personality and Social Psychology Bulletin, 28: 1679-1686.
Cesarone, B. (1994). Video games and children. ERIC Digest ED365477.
Crick, N. R. and Dodge, K.A. (1999). ‘Superiority’ is in the eye of the beholder: A comment on Sutton, Smith, Swettenham. Social Development, 8: 128 - 131.
DeLara, E. (2006). Bullying and violence in American schools, In N. E. Dowd, D.G. Singer and R.F. Wilson (eds.), Handbook of Children, Culture and Violence, Thousand Oaks, Ca.: Sage Publications.
Duncan, R. D. (1999). Peer and sibling aggression: An investigation of intra- and extra-familial bullying. Journal of Interpersonal Violence 14: 871-886.
Duncan, R. D. (2004). The impact of family relationships on school bullies and their victims. In Social-Ecological Model 23 D. L. Espelage and S. M. Swearer (Eds.), Bullying in American schools: A Social Ecological Perspective on Prevention and Intervention (pp. 227-244). Mahwah, NJ: Lawrence Erlbaum.
Elsea, M. and Mukhtar, K. (2000). Bullying and racism among Asian schoolchildren in Britain. Educational Research, 42: 207-217.
Espelage, D.L., Holt, M.K., and Henkel, R.R. (2003). Examination of peer-group contextual effects on aggression during early adolescents. Child Development, 74: 205-220.
Espelage, D. and Swearer, S. (2004). Bullying in American Schools: A Social-Ecological Perspective on Prevention and Intervention. Mahwah, N.J.: Lawrence Erlbaum.
Garbarino, J. and deLara, E. (2002). And Words Can Hurt Forever. New York: The Free Press.
Gentile, D.A. and Walsh, D.A. (2002). A Normative Study of Family Media Habits. Journal of Applied Psychology, 23: 157-178.
Graham, S., and Juvonen, J. (2002). Ethnicity, peer harassment, and adjustment in middle school: An exploratory study. Journal of Early Adolescence, 22: 173-199.
Hartup, W., and Stevens, N. (1997). Friendships and adaptation in the life course. Psychological Bulletin, 121: 355-370.
Haynie, D.L., Nansel, T., Eitel P., Davis- Crump, A.., Saylor, K., Yu, K., and Simons-Morton, B. (2001). Bullies, victims, and bully/victims: Distinct groups of at-risk youth. Journal of Early Adolescence, 21: 29-49.
Johnson, J.G., Cohen, P., Smailes, E.M., Kasen, S.,and Brooks, J.S. (2002). Television viewing and aggressive behavior during adolescence and adulthood. Science, 295: 2468-2471.
Kaltiala-Heino, R., Rimpela, M., Marttunen, M., Rimpela, A., and Rantenan, P. (1999).
Bullying, depression, and suicidal ideation in Finnish adolescents: School survey. British Medical Journal, 319: 348–351.
Kuntsche, E.N. (2004). Hostility among adolescents in Switzerland? Multivariate relations between excessive media use and forms of violence. Journal of Adolescent Mental Health, 34: 230-236.
Lall, R., Campbell, M.J., Walters, S.J., and Morgan, K. (2002). A review of ordinal regression models applied on health-related Quality of Life assessments. Statistical Methods in Medical Research, 11: 1-19.
Lickel, B., Schmader, T., and Hamilton, D.L. (2003). A case of collective responsibility: Who else was to blame for the Columbine High School shootings? Personality and Social Psychology Bulletin, 29: 194-204.
Limber, S.P. (2006). Peer victimization: The nature and prevalence of bullying among children and youth,. In N. E. Dowd, D.G. Singer and R.F. Wilson (eds.), Handbook of Children, Culture and Violence, Thousand Oaks, Ca.: Sage Publications.
Limber, S.P. (2004). Implementation of the Olweus bullying Prevention Program: Lessons Learned from the Field. In D. Espellage and S. Swearer (Eds.), Bullying in American Schools: A Social-Ecological Perspective on Prevention and Intervention.(pp. 351-363). Mahwah, N.J.: Lawrence Erlbsaum.
Long, J. S. (1997). Regression models for categorical and limited dependent variables. Sage Publications.
Merton, D. E. (1996). Visibility and vulnerability: Responses to rejection by nonaggressive junior high school boys. Journal of Early Adolescence, 16: 5-26.
Nansel, T. R., Overpeck, M., Ruan, W.J., Scheidt, P., Simons-Morton, B. and Pilla, R. (2001). Bullying behaviors among U.S. youth: Prevalence and association with psychosocial adjustment. Journal of the American Medical Association, 258: 2094-2100.
National Television Violence Study. (1998). Santa Barbara, Center for Communication and Social Policy, University of California.
Natvig, G. K., Albrektsen, G., and QvarnstrØm, U. (2001). School-related stress experience as a risk factor for bullying behavior. Journal of Youth and Adolescence, 30: 561-575.
Olweus, D. (1991). Bully/victim problems among school children: Basic facts and effects of a school-based intervention program. In K. Rubin and D. Pepler (Eds.), The development and treatment of childhood aggression (pp. 411-488). Hillsdale, NJ: Lawrence Erlbaum.
Olweus, D. (1993). Bullying at school: What we know and what we can do. Oxford: Blackwell Publishers.
Olweus, D. (1997). Bully/victim problems at school: Knowledge base and an effective intervention program. The Irish Journal of Psychology, 18: 170-190.
Olweus, D., Limber, S.P. and Mihalic, S. (1999). The Bullying prevention program: Blueprints for Violence Prevention, Vol. 10. Center for the Study and Prevention of Violence: Boulder, Boulder: CO.
O'Moore, A. M. and Hillery, B. (1989). Bullying in Dublin schools. Irish Journal of Psychology, 10: 426-441.
O'Moore M. and Kirkham, C. (2001). Self-esteem and its relationship to bullying behaviour. Aggressive Behavior, 27: 269-283.
Ortega, R.and Lera, M.J. (2000). The Seville anti-bullying in school project. Aggressive Behavior, 26: 113-123.
Pearce J. B., and Thompson, A. E. (1998). Practical approaches to reduce the impact of bullying. Archives of Disease in Childhood, 79: 528–531.
Pellegrini, A., and Bartini, M. (2000). A longitudinal study of bullying, victimization, and peer affiliation during the transition from primary school to middle school. American Educational Research Journal, 37: 699-725.
Perry, D. G., Kusel, S. J., and Perry, L. C. (1988). Victims of peer aggression. Developmental Psychology, 24: 807-814.
Rigby, K. and Cox, I. (1996). The contribution of bullying at school and low self-esteem to acts of delinquency among Australian teenagers. Personality and Individual Differences, 21: 609-612.
Roberts, D.F., Henriksen, L., and Foehr, U.G. (2004). Adolescents and Media. In R. Lerner and L. Steinberg (Eds.), Handbook of Adolescent Psychology (second edition), Hoboken, N.J.: John Wiley, 487-521.
Rigby, K., and Slee, P.T. (1991). Dimensions of interpersonal relation among Australian children and implications for psychological well-being. The Journal of Social Psychology, 133: 33-42.
Rodkin, P. C., and Hodges, E. V. (2003). Bullies and victims in the peer ecology: Four questions for psychologists and school professionals. School Psychology Review, 32: 384-400.
Seals, D., and J. Young (2003). Bullying victimization: prevalence and relationship to gender, grade level, ethnicity, self-esteem, and depression. Adolescence, 38: 735-747.
Slee, P. T., and K. Rigby (1993). The relationship of Eysenck's personality factors and self-esteem to bully-victim behavior in Australian schoolboys. Personality and Individual Differences, 14: 371-373.
Smith, P., and Sharp, S. (Eds). (1994). School bullying: Insights and perspectives. London: Routledge. Vail, K. (2000). Words that wound. American School Board Journal, 186(9): 26-28.
Smith, P., and Brain, P. (2000). Bullying in Schools: Lessons from two decades of research. Aggressive Behavior, 26: 1 - 9.
Smith, P. K. (2004). Bullying: Recent developments. Child and Adolescent Mental Health, 9: 98-103.
StataCorp (2003). Stata Statistical Software, Stata Coorporation College Statio, TX.
Swearer, S. M., and Espelage, D. L. (2004). Introduction: A social-ecological framework of bullying among youth. In D. L. Espelage and S. M. Swearer (Eds.), Bullying in American schools (pp. 1–12). Mahwah, NJ: Erlbaum.
Tritt, C., and Duncan, R. D. (1997). The relationship between childhood bullying and young adult self-esteem and loneliness. Journal of Humanistic Education and Development, 36: 35-44.
Underwood, M. K. (2002). Sticks and stones and social exclusion: Aggression among girls and boys. In P.K. Smith and C.H. Hart (Eds.) Blackwell handbook of childhood social development Oxford, Blackwell: 533-548.
Veenstra, R., Lindenberg, S., Oldehinkel, A., DeWinter, A., and Verhulst, F.C. (2005). Bullying and Victimization in Elementary Schools: A Comparison of Bullies, Victims, Bully/Victims and Uninvolvled Preadolescents. Developmetal Psychology, 41(4): 672-682.
Whitney, I., and Smith, P. K. (1993). A survey of the nature and extent of bullying in junior/middle and secondary schools. Educational Research, 35: 3-25.
WHO Organization, W. H. (2003). "Health Behavior in School-Aged Children, 1997-1998." [UNITED STATES] [Computer file]. ICPSR version. Calverton, MD: Macro International [producer], 2002. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor].
Williams, R. (2005). Gologit2: Generalized ordered logistic regression/partial proportional odds models for ordinal dependent variables. Working Paper, University of Notre Dame. (online at ).
Wolke, D., Woods, S., Bloomfield, L., and Karstadt, L. (2001). Bullying involvement in primary school and common health problems. Archives of Disease in Childhood, 85: 197-201.
Wolke, D., and Samara, M. (2004). Bullied by siblings: Association with peer victimization and behavior problems in Israeli lower secondary school children. Journal of Child Psychology and Psychiatry, 45: 1015-1029.
Zimmerman, F.J., Glew, G.M., Christakis, D.A., and Katon, W. (2005). Early cognitive stimulation, emotional support, and television watching as predictors of subsequent bullying among grade-school children. Archives of Pediatric Adolescent Medicine, 159: 384-388.
-----------------------
1 When proportionality is unrealistic, the partial proportional ordinal logistic model appropriately allows for the possibility that some of the estimated coefficients are the same for all values of j while others are not. As Lall et al. (2002) contend, “strikingly” different results can be obtained when using alternative ordinal models. An additional benefit is that this model is much more parsimonious than other models frequently employed when the proportional odds assumption is deemed invalid. Accordingly, the following was fit to the data [pic]where [pic]are the set of covariates, q of which are known to have proportional odds and p – q do not. In the parameterization of the partial proportional odds model used in this paper, each X has a constant component associated with it (see Table 4). In addition, each X can have M – 2 gamma coefficients (entitled “Increment at cut-off points” in Table 4) where M = 5 (the number of categories in Y) (See Lall, 2002 for further explanation). The gamma coefficients represent deviations from the proportionality assumption, in other words if the gammas for a variable are all 0 then the variable meets the proportional odds assumption (Williams, 2005).
................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.
Related download
- differences between statistical software sas spss and
- coverage edu
- logistic regression using sas
- relations between media effects social support and self
- auburn university
- directed reading in statistics fall quarter 2002
- sas commands for logistic regression
- cambridge university press
- logistic regression portland state university
- chapter xyz logistic regression for classification and
Related searches
- words of support and strength
- social support in recovery handouts
- importance of social support systems
- letters of support and encouragement
- words of support and strength during illness
- building social support worksheet pdf
- sample letter asking for support and help
- poems of support and encouragement
- difference between social science and natural
- no social support icd 10
- social work and social development
- social support in recovery worksheets