Reducing Bullying and Victimization: Student- and ...

J Abnorm Child Psychol DOI 10.1007/s10802-013-9841-x

Reducing Bullying and Victimization: Student- and Classroom-Level Mechanisms of Change Silja Saarento & Aaron J. Boulton & Christina Salmivalli

# Springer Science+Business Media New York 2014

Abstract This longitudinal study examines the mediating mechanisms by which the KiVa antibullying program, based on the Participant Role approach, reduces bullying and victimization among elementary school students. Both student-level mechanisms leading to reduced perpetration of bullying and classroom-level mechanisms leading to reductions in bullying and victimization are considered. Analyses are based on a sample of 7,491 students (49.5 % boys) nested within 421 classrooms within 77 schools. At the beginning of program implementation, the children were in Grades 4, 5, and 6 (mean age 11.3 years). Multilevel structural equation modeling was used to analyze whether changes in the hypothesized mediators accounted for later reductions in the outcomes. At the student level, antibullying attitudes and perceptions regarding peers’ defending behaviors and teacher attitudes toward bullying mediated the effects of KiVa on self-reported bullying perpetration. The effects on peer-reported bullying were only mediated by antibullying attitudes. At the classroom level, the program effects on both selfThe data used in this study were collected during a randomized controlled evaluation trial of the KiVa program developed at the University of Turku. The development of the program and the related research is funded by the Finnish Ministry of Education and Culture. The present study was also supported by the Academy of Finland Grants 121091 and 135577 to Christina Salmivalli and by the Turku University Foundation’s grant to Silja Saarento. We thank the whole KiVa project team, and especially Marita Kantola and Jonni Nakari, for their contributions to the data collection. S. Saarento (*) : C. Salmivalli Department of Psychology, University of Turku, 20014 Turku, Finland e-mail: [email protected]

and peer-reported bullying were mediated by students’ collective perceptions of teacher attitudes toward bullying. Also, perceived reinforcing behaviors predicted bullying but did not emerge as a significant mediator. Finally, bullying mediated the effects of the classroom-level factors on victimization. These findings enhance knowledge of the psychosocial developmental processes contributing to bullying and victimization and shed light on the key mechanisms by which school bullying can successfully be counteracted. Keywords Bullying . Victimization . Bystander behaviors . Antibullying program . Multilevel structural equation modeling . Mediation

Several school-based antibullying programs targeting multiple levels of the school context have been developed and evaluated in recent decades. Although the results of the programs have varied, a number of them have proven effective in reducing bullying and victimization (being bullied) among students (Smith et al. 2004; Ttofi and Farrington 2011). A critical question that remains to be empirically answered, however, is how these programs work. Recently, the need to examine the mediating causal mechanisms of antibullying programs, and the significant theoretical and practical benefits gained through such investigations, have been pointed out by many scholars (e.g., Eisner and Malti 2012; Kärnä et al. 2011b; Salmivalli et al. 2011). This study answers the call by empirically examining the mediating mechanisms of the KiVa antibullying program.

A. J. Boulton Department of Psychology, University of Kansas, Lawrence, KS, USA

Mediating Mechanisms in Intervention Research

C. Salmivalli Edith Cowan University, Joondalup, WA, Australia

Intervention programs usually aim to change the outcome of interest, such as bullying, by targeting certain mediating

J Abnorm Child Psychol

factors hypothesized to be causally related to the outcome. Oftentimes, the outcome is not easily changed directly but needs to be influenced indirectly by first targeting proximal factors such as cognitions, emotions, and behaviors that are contributing to, or preventing, changes in the outcome. Thus, mediators in intervention research are factors that transmit the effects of a given program to the ultimate outcome. Despite the expected gains associated with research on mediating mechanisms, few intervention studies have examined program effects on hypothesized mediators, and even fewer have tested the relations between the potential mediators and program outcomes (MacKinnon et al. 1989). To our knowledge, no studies to date have empirically tested the mediating mechanisms accounting for antibullying program effects. Longitudinal research on the mechanisms would not only generate information about the programs themselves but may also provide insight into possible targets for program improvement (Eisner and Malti 2012; MacKinnon 1994). By offering an ideal means to test the theories that have served as the backbone for program development, many of which have been based on cross-sectional studies, such research would also elaborate knowledge of the key psychosocial developmental processes contributing to bullying and victimization. Such knowledge needs to be the basis for evidencebased intervention programs (Eisner and Malti 2012). The KiVa Antibullying Program The KiVa program, developed as a collaborative effort of the Department of Psychology and the Center for Learning Research at the University of Turku, has been shown to be effective in reducing bullying and victimization in Finnish elementary schools both in a randomized controlled trial (RCT; Kärnä et al. 2011b; 2013) and during the first year of the nationwide roll-out of the program (Kärnä et al. 2011a). Despite the Finnish legislation concerning school safety, which requires schools to take action against bullying, no large-scale intervention programs specifically targeting bullying were being implemented in the schools prior to, or during, the roll-out of KiVa. By the end of 2012—that is, less than 5 years after the beginning of the RCT—approximately 90 % of all the schools providing basic education in Finland were registered users of the program. KiVa is comprised of several complementary components which are categorized into two general categories (see Salmivalli et al. 2010a; 2010b). Indicated actions, the purpose of which is to handle identified cases of bullying, include discussions with students involved in the incidents as well as encouraging prosocial students to support their victimized peers who are having a difficult time. Universal actions, on the other hand, are preventive actions targeted at the whole school community and mainly implemented at the classroom

level. Among other elements, the universal actions include a set of themed student lessons as well as the utilization of virtual learning environments. Hypothesized Mechanisms of KiVa Today, the peer group is viewed as playing a key role in the emergence and maintenance of bullying problems (for a review, see Salmivalli 2010). KiVa is by nature a wholeschool program which emphasizes that, in addition to directly tackling identified cases of bullying and the behavior of individual bullies, influencing the entire peer context is essential in counteracting the problem (Salmivalli et al. 2010a, 2010b). The goal of the program, which is based on the Participant Role approach to bullying (Salmivalli et al. 1996), is to influence the peer group in such a way that more students will express their disapproval of bullying and stand up for the victims, thereby eliminating the motivating social rewards often pursued, and gained, by bullies that are likely to reinforce their detrimental behavior. For this end goal, the different components included in the program aim first and foremost to affect students’ attitudes toward bullying, raise empathy for the victims, and develop bystanders’ efficacy to counteract bullying in safe ways. Bullying-related attitudes, often operationalized as moral judgments about the acceptability of bullying, are an important predictor of bullying involvement. Probullying attitudes are associated with bullying perpetration (Salmivalli and Voeten 2004). Most children and adolescents report attitudes that are opposed to bullying (e.g., Rigby and Slee 1991); however, it is relatively uncommon for peer bystanders to intervene in bullying episodes (e.g., Hawkins et al. 2001). The anticipation of negative responses by peers, based on perceptions of what is normative (i.e., common or accepted) in the classroom, might prevent the public expression of antibullying attitudes. Classroom norms have been found to predict students’ behavior in bullying situations over and above private attitudes (Saarento et al. 2011; Salmivalli and Voeten 2004), which speaks for the importance of influencing not only the attitudes of individual bullies but also the peer context. Despite some mixed evidence, the negative link between empathy and aggression, and bullying in particular, has been reported in the extant literature (for a short review, see Caravita et al. 2009). Recent studies suggest that it might be the ability to share others’ emotions (i.e., affective empathy) that plays the key role in inhibiting bullying and in catalyzing antibullying behaviors (Caravita et al. 2009; Pöyhönen, Juvonen, & Salmivalli 2010), whereas the cognitive understanding of other’s emotions (i.e., cognitive empathy) can be utilized for harmful purposes and may even promote bullying perpetration (Caravita et al. 2009). Therefore, it has been recommended that antibullying programs should emphasize


affective empathic skills rather than merely promoting cognitive empathy (Caravita et al. 2009). Moreover, recent studies have promoted examinations of the implications of target specific—rather than global—empathy such as empathy for the victimized peers (MacEvoy and Leff 2012; Pöyhönen et al. 2010) which many aggression intervention programs aim to increase. Although the role of bystanders (i.e., those witnessing bullying) has been recognized and their behaviors have increasingly been targeted in school-based antibullying programs, surprisingly little research has been done on the links between different bystander responses (e.g., reinforcing the bully and defending the victim) and bullying. In an observational study, peers providing support for the victim was often found to be effective in putting an end to a bullying episode (Hawkins et al. 2001). Furthermore, a cross-sectional study recently showed that the more classmates tend to reinforce the bully by cheering or laughing, for example, and the less they tend to support and defend the victim, the higher the frequency of bullying in the classroom (Salmivalli et al. 2011). The effect of reinforcing was found to be more pronounced than that of defending behaviors. These studies provide empirical support for the notion that bystanders are influential in bullying— either providing or withdrawing social rewards from the bullies—and should be emphasized in efforts to tackle the problem. Still, further empirical research on the effects of bystander behaviors is clearly warranted (Polanin et al. 2012; Salmivalli et al. 2011). It is not only the reactions of the peers that are likely to influence students’ involvement in bullying. Teachers, by their efforts to intervene in bullying, or by the lack thereof, might contribute to the perceived acceptability and frequency of bullying among students. Indeed, crosssectional studies recently demonstrated that perceived teacher attitudes contribute to both bullying and victimization; that is, students who perceive their homeroom teacher to clearly disapprove of bullying are less likely to bully others (Saarento et al. 2011), and being bullied is more common in contexts where students believe their teacher condones bullying (Saarento et al. 2013). Positive intervention effects of KiVa have already been reported not only on bullying and victimization but also on most of the key factors (e.g., antibullying attitudes, empathy, and bystander behaviors; Kärnä et al. 2011b) hypothesized to mediate program effects. Research on the associations between these potential mediators and bullying—most of which have utilized a cross-sectional design, however—as well as the results of the program evaluation lend preliminary support to the mediation hypotheses, but they do not suffice to conclude that mediated effects indeed exist. Whether changes in the hypothesized mediators in fact account for later reductions in bullying and victimization, and thereby for the effects of KiVa, is yet to be determined.

The Present Study The goal of this study is to unravel the chain-link processes by which the KiVa program leads to reduced rates of bullying and victimization among fourth- to sixth-grade students. Consistent and considerably large program effects have been reported for this age group (Kärnä et al. 2011a; Kärnä et al. 2011b; Williford et al. 2012), which makes it a plausible target sample for examining the mediating mechanisms accounting for the effects. For instance, at the end of the RCT the odds of being a victim were about 1.5 to 1.8 times higher, and the odds of being a bully about 1.2 to 1.3 times higher, for a control school student than for a student in a school implementing KiVa (Kärnä et al. 2011b). The mechanisms of intervention programs have often only been examined at the individual level probably due to modest sample sizes and practical limitations of the available data analytic methods. However, because KiVa is a whole-school program that aims to prevent and reduce bullying problems by changing the psychosocial contexts shared by students, it is essential that attention be paid not only to student-level but also to group-level mechanisms of change. More specifically, we use a longitudinal design to test whether KiVa reduces bullying and victimization by changing the student- and classroom-level factors hypothesized in the program’s theoretical grounds to be the key mediating mechanisms (Fig. 1). Firstly, we hypothesize that KiVa leads to reduced perpetration of bullying at the student level by increasing students’ private antibullying attitudes and affective empathy for the victim. Moreover, we hypothesize that the program effects are mediated by changes in the way students perceive their peers and teachers to view and react to bullying. Looking into the effects of one’s perceptions of the social context is essential because the perceptions of other people’s judgments often are of great importance in determining behavior (Miller and McFarland 1987; Prentice 2008). Specifically, we explore whether perpetration of bullying decreases as a result of a child observing that the relative number of classmates engaging in defending the victim has increased whereas the relative number of those reinforcing the bully has decreased. Traditionally, the implications of perceived bystander behaviors at the classroom level have been examined (Kärnä, Voeten, Poskiparta, & Salmivalli 2010; Salmivalli et al. 2011) whereas, to our knowledge, no previous study has explicitly addressed the question of whether the way an individual student perceives his or her classmates’ responses in bullying situations predicts perpetration of bullying. In the present study, we utilize both the more traditional as well as this novel approach to examining the effects of perceived bystander behaviors. Based on findings from a previous study (Salmivalli et al. 2011), we anticipate the effect of perceived reinforcing behavior to be stronger than that of perceived defending. Regarding students’ perceptions of their teacher’s attitudes


Fig. 1 Theoretical model of the effects of the KiVa program

toward bullying, we examine whether perpetration of bullying decreases as a consequence of a student evaluating that the homeroom teacher takes bullying more seriously than before the beginning of program implementation. Secondly, we hypothesize that the reduction in bullying at the classroom level is explained by prior changes in students’ bystander behaviors (i.e., the tendency of classmates to defend the victim or reinforce the bully) and collective perceptions of teacher attitudes toward bullying. Finally, we hypothesize that at the classroom level bullying concurrently mediates the effects of the hypothesized key mediators on victimization, such that the reduction in bullying explained by the mediators accounts for the reduction in victimization. It is worthwhile to point out that we did not test hypotheses about the mediators of the effects of KiVa on victimization at the student level. The focus of the KiVa program is not on reducing victimization directly; rather, it is to change the social context in which victimization takes place. In addition to using longitudinal data and bringing together the set of hypothesized student- and classroom-level mediators, focus on which has been lacking from research on antibullying programs, this study utilizes both self- and peer reports of bullying and victimization. Although they have rarely been included in the same study, both self- and peer reports have their unique strengths and limitations and may provide complementary information (for a review, see Cornell and Bandyopadhyay 2010). Confirmation of associations with both types of reports would enable the identification of key mediators that are associated with bullying and victimization regardless of source of information.

Method Participants In this study, analyses are based on a target sample of 8,248 students nested in 429 classrooms in 78 schools. The data were collected during a randomized controlled evaluation trial

of the KiVa program. Stratified random sampling of schools providing basic education that had volunteered to participate in the evaluation was used to include schools from all the provinces of mainland Finland (for a detailed description, see Kärnä et al. 2011b). Whereas half of the participating schools were randomly assigned to the intervention condition, the others were offered an opportunity at their discretion to initiate the KiVa program after 1 year of serving as control schools, that is, after the data collection was completed. Active consent forms, which were translated into 15 languages according to the largest immigrant groups in Finland, were distributed to the parents of all the students in the target sample. At the beginning of program implementation, the students were in Grades 4, 5, and 6 and were approximately 10–12 years old. Due to the exclusion of students who did not receive parental consent, and one control school dropping out before the beginning of the data collection, the final sample consisted of 7,491 students (90.8 % of the target sample; 49.5 % boys; mean age 11.3 years; 2.0 % immigrants) nested in 421 classrooms in 77 schools. In order to improve the reliability of the received peer nomination scores, the student-level mediated effects on peer reported bullying as well as the classroomlevel mediated effects were analyzed using a sample from which the smallest classrooms (N