Article

Haynie, Nansel, Eitel, et. al (2001) found that more than half of the bullies in their research also reported being victimized. Ma (2001) has noted that very little attention has been given to the possible ‘bully-victim cycle’ in research, suggesting that methodological difficulties account for this oversight. His research clearly identified a reciprocal relationship between bully and victim.

The purpose of this study is to better understand, from a cross-cultural, systemic perspective, the interplay of factors that impact upon students’ experience of bullying and victimization. Although Japan and Australia are highly similar in relation to nationwide levels of social and economic development, the different cultural contexts suggest that cross-cultural comparisons will help us to better understand how the stresses and supports in the lives of these impact on their sense of safety and wellbeing at school.

Participants
 
The Japanese students in our sample are representative of their essentially mono-cultural society. The Australian students were from more diverse, but nevertheless predominantly Anglo-European, backgrounds. In both countries students’ ages ranged from 9-16 years ( = 12.9).

Twenty two Australian schools participated in the study, 11 primary and 11 secondary schools including, rural, regional and metropolitan areas in South Australia, and both the independent and government school sectors. Of the total survey population of 3145 Australian students (49% males; 51% females) 35% were primary (Grades 5, 6, and 7), and the remaining were secondary (Grades 8, 9 and 10) school students. Students from equivalent grade levels to the Australian students in 12 primary and 6 secondary schools in metropolitan Tokyo completed the Japanese version of the survey. The total Japanese sample comprised 5518 students (51% males; 49% females) with 48% from primary, and the remainder from secondary schools.

Procedure

Taki’s (2001b) survey instrument was collaboratively adapted to include 54 items, of which 43 are used for the analyses in this study. With the assistance of a Japanese interpreter, adjustments were made to the items by back translation to account for the different nuances in meaning between the two languages.

The procedure for administration of surveys was discussed by researchers from the two countries and dates set so that surveys were administered at the same stage of each country’s respective academic years. This involved a research assistant associated with the project in each country visiting the schools and supervising the administration and collection of the questionnaires.

Data from both Japanese and Australian surveys were entered into a common data base. In order to create comparable models for testing, preliminary model testing was conducted using Mplus (Muthén & Muthén, 1998) and items that did not contribute to the hypothesized constructs in either the Japanese or Australian data set were removed.

The development of the theoretical path model, presented as Figure 1, was informed by previous path analyses where models had been tested separately for each country and was based on common items designed to elicit students’ perceptions on a range of issues of joint concern to the researchers. In particular, the issue of stress (both sources of stress and symptomatic responses to stress) and its relationship to bullying others and being bullied (victimized) was of mutual interest. The path analysis (Sellin & Keeves, 1994) employed in this study enabled us to examine the relationships between the constructs under investigation (the latent variables) and the criterion variable (bullying), as well as the relationships among the latent variables themselves. The selection and construction of variables included in the path model is more fully discussed in the next section of the paper.

Each variable selected for inclusion in the path model was drawn from research where that variable, in its own way, was shown to be correlated with bullying and/or victimization. The manifest variables (MVs) (those on which data were gathered) chosen to represent the respective latent variables (LVs) have been the subject of discussions around students’ lives at school in Australia and Japan. Hence, variables were included which would allow comparisons between groups of students in both countries. The manifest variables that comprised the outer model are described in Table I.
 

 
Path estimations of relations between manifest and latent variables are recorded as factor loadings and these also are presented in Table I. From these variables, 10 latent variables were formed and these constituted the inner core of the model. The latent and manifest variable relationships are displayed in Figure 1.

The arrow scheme of the path model displayed in Figure 1 specifies its theoretical conceptual design giving information about the relationship of the manifest variables to the latent variables and the relationship between the latent variables in the model.

Latent Variable-Manifest Variable Relationships

The manifest variables (MVs) in the model were formed from items in the survey that were hypothesized to represent them. The items are listed in Table I. So, for example, items 37, 38, and 39 were regarded as common sources of stress from parents, and items 51, 54 and 57 were believed to represent the types of support that parents could offer students.

The following section explains the relationship between the latent and manifest variables.

1. Stressors. Three MVs represented possible sources of stress for students. They were family, teachers, and peers. (3 items for family and for peers, 2 items for teachers). Students were asked “have you experienced any of the following during the last term?” and to rate the frequency of each experience on a 4-point scale for each (1 = never to 4 = very often). Items within each set were summed to form a score for each source of stress.

2. Support. Three groups of people, parents, teachers and peers, comprised the MVs that represented possible sources of support for students. Students responded to the following request on the questionnaire: “The following statements are about how much support you get from parents, teachers and classmates. What do you think about the following statements?” Using a 4 point scale (1 = strongly disagree to 4 = strongly agree), for each group (parents, teachers and peers) separately, they then indicated their level of agreement on the three items: (a) “If I feel left out I am encouraged by “… (b) “If I express my troubles/problems I am listened to by…” (c) “These people usually try to understand my feelings…” A total score was calculated for each set of three items.

3. Psychological Health. The four MVs that comprised this LV were Apathy, Somatic Symptoms, Depression, and Aggression. These symptoms were interpreted as indicators of students’ psychological health. A score for each MV was calculated by summing the three component items. Students rated how well each of the 12 items described how they felt (1 = not at all like me to 4 = a lot like me).

4. Academic Performance. This unity variable was constructed from three MVs represented by students’ evaluations of their achievement for items related to; (a) getting low test results (b) not understanding lessons, and (c) not understanding questions asked by teachers (1 = very often, 2 = sometimes, 3 = a little, and 4 = never). These items were combined to create a miniscale for Academic Performance. Confirmatory Factor Analysis Weighted Omega values supported the formation of this miniscale (see Table II).

5. Belonging to School. This unity variable was constructed from three MVs considered indicative of students’ feelings about school. Students rated their feelings about school in terms of (a) enjoying my school life (b) getting along with other students and (c) proud of belonging to my school (1 = yes, very much; 2 = yes, a little; 3 = not much and 4 = not at all). These items were combined to create a miniscale for Belonging. Confirmatory Factor Analysis Weighted Omega values supported the formation of this miniscale (see Table II).

6. Victimization. Four MVs were associated with this latent variable, which on the basis of acceptable model fit statistics, was constructed as a unity variable. Victimization indicated frequency of having been bullied at school in the school term by being: (a) isolated, ignored, called names (b) picked on by others (c) pushed, hit, kicked on purpose (jokingly), and (d) robbed, kicked, hit harshly (on purpose): 1 = never; 2 = once/twice; 3 = 2-3 times per month; 4 = more than once a week. These items were combined to create a miniscale for Victimisation. Confirmatory Factor Analysis Weighted Omega values supported the formation of this mini-scale (see Table II).

7. Bullying. Four MVs formed this LV. Bullying indicated frequency of having bullied someone in the school term by : (a) isolating, ignoring, calling them names; (b) picking on others; (c) pushing, hitting, kicking on purpose (jokingly) ; and (d) stealing, kicking, hitting harshly (on purpose): 1 = never; 2 = once or twice 3 = 2-3 times /month; 4 = more than once per week. These items were combined to create a miniscale for Bullying. Confirmatory Factor Analysis Weighted Omega values supported the formation of this mini-scale (see Table II).
 

 
  Path Analysis

Partial Least Squares Path Analysis (PLSPATH) (Sellin, 1990) was employed to test the hypothesized model. PLSPATH is particularly well suited to exploring relationships with educational data that do not always meet the restrictive distributional assumptions required by SEM approaches; that is, that variables are normally distributed and independent of one another, as is the case with a number of the variables in this study. The main strength of PSLPATH is its use to suggest where relationships may or may not exist and provides the opportunity to move beyond correlational data. For information, latent variable correlations are provided in Table III.

Important variable relationships are presented and discussed in this section. Figure 1 provides a diagrammatic representation of the final model results and indicates that the amount of explained variance for Victimization is 28 per cent (R² = 0.28) and for Bullying 20 per cent (R² = 0.20) which is highly satisfactory for models of the kind encountered in educational research. The arrows indicate the important (non-trivial) LV relationships and their path coefficients (Beta weights β > .10). Indirect paths are shown with dotted lines. The path coefficients along with their Jackknife Standard Errors are provided in Table IV.
 

 
 
The first point to note is that all the manifest variables contributed strongly to their respective latent variable. With regard to the discussion that follows on Stressors and Support this is important; it shows that all three groups – teachers, peers and family, play their part in contributing to the support students need at school (indicated by factor loadings of.84, .64, and .81 respectively). Similarly, loading values for all three groups are equally represented as sources of stress (.72, .72, and .75 respectively) in students’ lives at school and suggests that considering relationships among students alone only partially represents the complex issue of bullying and victimization at school. Families and teachers as well need to be represented both as part of the problem and part of the solution.

With regard to Psychological Health, the high subscale loadings for Apathy, Somatic Symptoms, Depression and Aggression (.83, .79, .82, and .80, respectively) show that these dimensions exert much the same level of impact on Psychological Health. This point is worth highlighting because, in focusing on overt and direct negative student behaviors like apathy and aggression, it is easy to overlook that students who are not visibly acting out may also be troubled at school. The remaining discussion aims to shed light on the inner model, latent variable relationships.

Stressors and Victimization

The strongest relationship produced by the path analysis was between Stressors and Victimization (.44) with the sources of stress from family, teachers and peers exerting much the same influence, as indicated earlier, by similar loadings on this variable. Correlations were examined to check the possibility that the peer stressor variable was not in fact measuring a component of Victimization. While the correlation between peer stressor items and victimization items was strong (r = .61) this correlation was not unacceptably high, thus indicating that the peer items for Stressors could be regarded as conceptually different from the peer items for Victimization.

Stressors, Psychological Health and Belonging

The relationship between Stressors and Psychological Health is an important finding. The strength of this relationship (.50) highlights the influence that family, peer and teacher pressures exert on students’ Psychological Health indicated by Apathy, Depression, Aggression and Somatic Symptoms. And while there was no direct impact on Stressors on School Belonging, there was an indirect effect (-.11) of negative feelings about school through Psychological Health.

Support, Belonging and Psychological Health

The strength of the relationship between Support and Belonging (.33) provides clear evidence that when students feel supported by teachers and family foremost, and to a lesser, but also important extent, by peers, they also report a greater sense of belonging to school. As well, lower levels of Support were found to be related to reports of poorer Psychological Health (-.16).

Not surprisingly, poor Psychological Health was found to be highly related to a lack of Belonging to school (-.35). Stressors were also found to play a part in contributing to students’ negative feelings about belonging to school. This relationship, as indicated by an indirect path (-.32) between Stressors and Belonging through Psychological Health, draws attention to the value of considering variables in relationship to each other.

Psychological Health and Academic Performance

The path analysis provides a clear picture of relationships, not only between Psychological Health and Academic Performance (-.33) but also between Stressors and Academic Performance (-.30). The interacting nature of the Stressor-Psychological Health-Academic Performance relationships should not be ignored.

Stressors and Support

The strong connection between Stressors and Support (-.37) points to the need for a two-pronged approach in efforts to improve the quality of students’ lives at school. By this we mean that, the school community has joint tasks to undertake; one is to act to reduce stressors in students’ lives, the other is to build the capacity of the school community to provide the supportive interaction that students perceive to be important for their wellbeing and their sense of belonging to school.

Stressors and Bullying

In terms of the path analysis we can see that not only is there a relationship between Stressor and Support variables, and between Psychological Health and School Belonging; it is also evident that Bullying (.23) and Victimization (.44) are in some way connected to Stressors.

Psychological Health and Victimization

While poorer Psychological Health (.13) was found to be related to Victimization, there was an unexpected lack of association found between Psychological Health and Bullying. This lack of effect, despite the significant LV correlation (r = .27), is evidenced by the fact that most of the explained variance had been already accounted for in the strength of the relationships of the preceding variables.

Victimization and Bullying

An interesting finding from the present study concerned the relationship between bullying and victimization (.28). This finding lends support for an increasing awareness of the danger of presenting stereotypical views of individuals as either ‘bullies’ or ‘victims’ and is consistent with findings reported by Ma (2001) and (Pellegrini & Bartini, 2000) as well as with research into the different roles that individuals take on in a bullying situation (Salmivalli, Lagerspetz, Bjorkqvist, Osterman & Kaukiainen, 1996).

The justification for employing path analysis to gain a more comprehensive picture of issues related to school violence than can be portrayed by an examination of one-to-one associations derived from correlational analysis, is borne out in the results produced by these two different analyses. It can be seen for example, in the correlations between the latent variables (Table III) that Bullying is significantly correlated with several other latent variables; Stressors, Support, Academic performance, and Belonging. The path analysis however, provides a set of effects that take into account the relative influence of these variables in relationship to each other in their effect on Bullying.

What emerges now is that while poorer Academic Performance and lower levels of Belonging are significantly correlated with Bullying, these variables are not shown in the path analysis to directly influence Bullying (or Victimization). This is because Stressors, when seen in relationship to Academic Performance and Belonging exerts such a strong effect on Bullying that it accounts for most of the variance explained in the model. Likewise, while Stressors is significantly negatively correlated with Belonging (r = -.38), most of the variance explained on Belonging is captured by the strong effects of Psychological Health and Support on Belonging. However, Stressors does exert a sizeable indirect influence on Belonging through its own influence on Psychological Health and Support.

Country, Sex and Year Level

Australian students reported higher negative evaluations of their academic performance than Japanese students (-.10). They also reported that they perceived teachers, family and peers to be a greater source of stress (Stressors) than the Japanese students (-.17) who, in turn, reported that they perceived their teachers, peers and family to be less supportive (-.16).

Girls reported poorer Psychological Health than boys (.14). Students in higher grades at school perceived they were less supported by family, teachers and peers (-.17) and this was aligned with their reports that these people were a greater source of stress (.14). Consistent with the literature, students in the lower grades reported higher levels of victimization (-.20). There were no effects of country or sex on either Bullying or Victimization.

There is no doubt that there are personal, relationship and structural variables that impact on students’ lives at school. It is important to recognize however, that these variables do not exist independently of each other. Because of this, efforts to create a safe, positive experience for students at school requires a multi-level approach to addressing school violence; an approach that shifts the focus away from intervention merely at the level of the individual and moves towards giving due attention to the importance of building positive, supportive relationships at school and at home.

The importance of the relationship between Stressors and Support and Psychological Health and their influence on feelings of Belonging to school and Academic Performance appears to far outweigh the contribution of these variables to Bullying and Victimization. This suggests that our focus of attention needs to be less on reactive interventionist approaches to managing bullying and victimization and more on the need for systemic change towards creating schools where students feel they belong and where they are academically successful.

Cross-cultural research has provided the opportunity to examine issues of mutual concern. Rather than seeking comparisons between students or countries, the cross-cultural focus of this research has been at a conceptual level, to further understand how the issue, in this case bullying and victimization, manifests itself in different cultural contexts. While there is recent evidence of a shift in emphasis, nevertheless, much of the literature describes and explains bullying and victimization at the level of the individual. This is typified by attempts to characterize attributes of the bully and the victim. In Western individualistic societies like Australia, deviant behavior is constructed as a problem that resides within the individual; thus leading us to seek solutions that focus more on changing individuals that seeking explanations and solutions at a broader, systemic level. In contrast, a highly westernized, but also less individualistic society like Japan, allows for social explanations of problems. The persistence of psychological explanations will continue to limit our ability to move toward systemic approaches to dealing with the problem. We have evidence from research across nations, that the definitions of bullying characterize Japanese bullying (ijime) are equally applicable to other countries, and as in Japan, where bullying incidents are seen to be a reflection of complex, multiple and synergic factors, this is just as likely to be the case in other societies.

This is not to say that we should ignore bullying and victimization issues. Rather, that prevention and intervention for bullying and victimization won’t go far if we don’t address the other, apparently more powerful social and institutional influences. The findings of this research support Taki’s (2001) theory that indicators of stress in students’ lives (apathy, somatic symptoms, depression and aggression) are predictive of ijime to the extent that stressors put students at risk and supports buffer against the symptoms of stress that predispose students to bully others. Our research findings also resonate with those of Soen (2002) who pointed out that we should expect failure of intervention/prevention programs if they do not include significant members of a school community. In other words, without a systemic approach to countering violence, it is unlikely that there will be significant, sustained change in student attitudes or behavior.

Even in the school described earlier in this paper, where there is a relatively low incidence of bullying, for those affected, life at school is likely to be at best unpleasant, and at worst, absolutely miserable. Because the sites of violence are often not located within classrooms it is easier for teachers to disregard or deny destructive behavior amongst students, and harder for them to intervene or take positive action. Intervention is however, only one approach to addressing these commonplace acts of violence in schools. And if we pay attention to what students are saying, intervention alone often has limited positive, and sometimes even escalating, effects on the problem. The findings of this research suggest that attempts to address violence need to consider the impact of antecedent support and stress variables. The interactive, mediating effects of these support and stress variables suggest that preventative approaches that work at the level of building positive, supportive networks within the school community also need to be a focus of attention. A reactive interventionist approach combined with a proactive preventative program is going to be more effective than focusing only on one approach. In the words of students themselves (Soen, 2002, p. 14) “proactive programs contribute to the improvement of the school climate as well as to changing the circumstances contributing to violence.”

Life at school for students needs to be understood as a complex network of interacting relationships between students, their teachers and parents, and their peers. An appreciation of the intersecting nature of these relationships suggests the need for a multi-level approach to addressing school violence that takes into account this complexity.

The authors extend their thanks to Judith Saebel for her statistical support with the analysis of data for the research reported in this paper.