The way bullying works: How new ties facilitate the mutual reinforcement of status and bullying in elementary schools Rozemarijn van der Ploeg, Christian Steglich and René Veenstra
The self-archived postprint version of this journal article is available at Linköping University Institutional Repository (DiVA): http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-154080
N.B.: When citing this work, cite the original publication. van der Ploeg, R., Steglich, C., Veenstra, R., (2019), The way bullying works: How new ties facilitate the mutual reinforcement of status and bullying in elementary schools, Social Networks. https://doi.org/10.1016/j.socnet.2018.12.006
Original publication available at: https://doi.org/10.1016/j.socnet.2018.12.006 Copyright: Elsevier http://www.elsevier.com/
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
The Way Bullying Works: How New Ties Facilitate the Mutual Reinforcement
of Status and Bullying in Elementary Schools
Rozemarijn van der Ploega,b, Christian Steglicha,c, René Veenstraa
a Department of Sociology, University of Groningen, Groningen, the Netherlands
b Department of Pedagogy and Educational Science, University of Groningen, Groningen, the
Netherlands
c Institute for Analytical Sociology, Linköping University, Linköping, Sweden
Article in press
Social Networks: https://doi.org/10.1016/j.socnet.2018.12.006
Correspondence regarding this paper should be addressed to Rozemarijn van der Ploeg, University of
Groningen, Grote Rozenstraat 38, 9712 TJ Groningen, the Netherlands. Electronic mail may be addressed to [email protected].
Telephone: 0031 50 363 2486.
1
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Highlights
Younger children punished bullying by a refusal to attribute status to bullies.
Older children reward bullying with peer status.
High-status bullies seemed to avoid continual bullying of the same victims.
Abstract
This study addresses the puzzle how high-status bullies in elementary school are able to maintain high status among their classmates despite bullying (some of) them. The dynamic interplay between bullying and status was studied, focusing on how relational bullying affects the creation, dissolution, and maintenance of status attributions, and vice versa. Longitudinal round-robin peer nomination data were obtained from 82 school classes in 15 Dutch elementary schools (2055 students; 50% boys) followed over three yearly measurements, starting out in grades 2-5 when students were aged 8-11. An age-dependent effect of bullying on the creation of new status attributions was found. Whereas the youngest group punished bullying by a refusal to attribute status to the bully, this turned into a reward of bullying in the oldest groups. Unexpectedly, high-status bullies seemed to avoid continual bullying of the same victims, pointing to explanations of why their status can persist.
Keywords: bullying; peer status; creation and maintenance of ties
2
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
The way bullying works: How new ties facilitate the mutual reinforcement of status and
bullying in elementary schools
Bullying is commonly defined as repetitive and intentionally negative behavior against a victim who finds it difficult to defend him- or herself (Olweus, 1993). It can be formalized as a directed, negative social relation in which one social actor (the bully) behaves negatively against another (the victim). The effects of bullying are generally not limited to the bully- victim dyad. When bullying takes place in a social group and under conditions of high visibility (e.g., among students in a school class or among colleagues at a workplace), third actors become witnesses and can react to the bullying, either pro-socially (by defending the victim, or expressing disapproval to the bully), or anti-socially (by siding with the bully, or even joining in the bullying). Also a non-reaction in the face of bullying conveys the social cue of tacit approval. Bullying is a group process that affects all the group members
(Salmivalli et al., 1996; Salmivalli, 2010) and is characterized by complex interdependencies
(Huitsing et al., 2014; Fujimoto, Snijders, & Valente, 2017). From an organizational viewpoint, bullying is a challenge to the social order in a task-related group. If bullying plays too prominent a role in daily life, students cannot concentrate on school tasks and employees cannot work well. It is in the interest of management (in the school class: the teacher) to contain bullying, and if necessary intervene to reduce it. For such interventions to be successful the processes by which bullying spreads need to be understood in the first place.
A key ingredient of such an understanding is the social standing of the involved individuals. High-status students are more visible and more accepted in the peer group, and what these students do also gains visibility and acceptance. This way, they are in a position to set behavioral norms in the school class (Dijkstra & Gest, 2015). Students with high social status among their classmates were shown to be influential in their peer group and often serve as role models (Dijkstra, Cillessen, & Borch, 2013). If they act as bullies, there is the danger
3
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF that bullying becomes socially accepted. A high-status bully can potentially lead a whole school class into bullying. For this reason, many anti-bullying interventions focus on diminishing the social status of bullying by targeting group norms (Swearer, Espelage,
Vaillancourt, & Hymel, 2010), by encouraging others to disapprove of bullying.
Unfortunately, these interventions do not always work (e.g., Gaffney, Ttofi, & Farrington, in press; Nickerson, 2017). Especially bullies with high status tend to continue bullying and keep their high status among peers, thus defeating the intervention’s purpose (Garandeau, Lee, &
Salmivalli, 2014). In contrast to medium- or low-status bullies, these high-status bullies somehow succeed in convincing a sufficiently large number of their peers to attribute status to them despite their bullying. Given the sensitivity of group behavior to what high-status students do, it is essential to examine the interplay between bullying and status attribution more thoroughly.
While in the past, bullying was often considered an impulsive, uncontrolled outburst of aggression (Olweus, 1978), scientists and practitioners today tend to agree that it predominantly involves proactive, strategic, and goal-directed behavior (Reijntjes, Vermande,
Olthof, et al., 2013; Volk, Veenstra, & Espelage, 2017). Bullies are thought to bully in order to achieve dominance and high social status in the peer group (Salmivalli et al., 1996). Bullies were shown to be more strongly motivated by considerations of peer status than non-bullies
(Cillessen & Mayeux, 2004; Sijtsema et al., 2009). To achieve this high status, they strategically pick on easy victims, e.g., the physically weak or those who are rejected by other classmates (Sijtsema et al., 2009; Veenstra et al., 2007). This strategy seems effective for them: Bullying was repeatedly found to be positively associated with social status among peers in cross-sectional studies (Caravita, Di Blasio, & Salmivalli, 2009; De Bruyn, Cillessen,
& Wissink, 2010). Longitudinal studies reveal that the two concepts are entrained, referring to that they remain associated over time. High-status bullies tend to keep their high status, and
4
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF continue bullying (Cillessen & Borch, 2006; Reijntjes, Vermande, Olthof, et al., 2013; Sentse,
Kretschmer, & Salmivalli, 2015; see also for aggression instead of bullying: Juvonen, Wang,
& Espinoza, 2013). On the conceptual level, it remains a puzzle how bullying and high status can reliably co-occur, and how some bullies can successfully derive high status over longer time periods from the very group they victimize. Although the existence of such a
“controversial” group of children has been documented at least since the early 1980s (e.g.,
Coie, Dodge & Coppotelli, 1982), the literature remains inconclusive about the underlying mechanisms. We intend to examine how status and bullying co-determine each other, this way contributing to a solution of the puzzle and facilitating the development of more effective anti-bullying interventions.
We think that the lack of solid insights is rooted in inappropriate methodology.
Despite both concepts being defined on the dyadic level, the literature on adolescent development has traditionally taken an actor level approach in studies of bullying (e.g.,
Olweus, 1993) as well as peer status (e.g., Cillessen, Schwartz, & Mayeux, 2011). This choice of the level of analysis makes it impossible to look at the finer-grained dyadic and triadic patterns required for understanding the interplay between bullying and peer status. Whereas such studies on social relationships and status have recently started emerging (Betancourt,
Kovács, & Otner, 2018, for bullying and related behaviors see Appendix B), the specific puzzle of how bullying and status attribution affect one another has so far remained unaddressed. To address this puzzle, we put forward what might be termed a network understanding of the mechanisms that allow high-status bullies to keep their high status while staying bullies, even when bullying is disapproved of. For this aim, we investigate the determinants of status attribution over time, paying special attention to the bullying of status recipients. And we examine the determinants of bullying over time, paying special attention to
5
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF the peer status of the bully. Because both processes depend on one another, we examine them as an interdependent system.
Theory
Bullying can take different forms, including physical bullying (hitting, kicking), relational bullying (ignoring, gossiping), verbal bullying (calling names, insulting), material bullying (stealing or damaging things), or cyber bullying (via social networking sites). All these are negative interactions and, if engaged in voluntarily, should result in dislike and mutual avoidance (cf. Homans, 1950, chapter 5). Compared to positive networks, negative networks tend to generally have lower density and lower stability over time (for our own data, we can see this in Table 1). However, bullying deviates from Homans’ scenario of a voluntary interaction in two important ways. First, there is a power asymmetry between the bully and the victim. Bullying is an act of dominance, voluntary only from the bully’s point of view, not the victim’s (Salmivalli et al., 1996). With education being compulsory in almost all countries, opting out is not a realistic option for the victim. In short, being victimized is not a choice, but bullying is. Second, the bully and the victim may evaluate the interaction differently. Whereas for the victim, it is almost certainly experienced as negative, the case may lie different for the bully. Depending on whether this evaluation is positive or negative, the bully may decide to repeat or discontinue the bullying. From these considerations, we conclude that to understand bullying, we need to identify the conditions under which the bully decides to bully.
Probably the most important aspect of this decision is social standing in the peer group. As an act of dominance, the main effect of bullying is a reduction of the rank of the victim and an increase of the rank of the bully in a group-internal pecking order or dominance hierarchy (Peets & Hodges, 2014). Additionally, there may be effects relative to third parties.
Because the bullying act attracts peers’ attention, which is a limited resource, the bully as well
6
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF as the victim gain visibility at the expense of third parties. Figure 1 illustrates the combined effects. If we assume a triad of actors of equal social standing (the black dot in the middle of the left panel), then the bullying act on the one hand increases visibility of bully B and victim
V while reducing it for third parties T. On the other hand, in a dominance hierarchy, it increases the rank of B and reduces the rank of V, while leaving the rank of T unaffected. If we operationalize social standing as a linear combination of visibility and rank in the dominance hierarchy, the resulting preference orders depend on the direction of the preference gradient (see right panel of the figure). If actors strive for dominance (up) as well as visibility
(right), they experience an incentive to bully.
In Figure 1, social standing is by way of simplification assumed to be a direct result of bullying. It is true that the bullying act is a display of dominance, and emphasizes a status difference. However, status among peers is a network-mediated consequence of bullying and endogenous to the peer system. Peers decide to attribute or withhold status. Their dyadic status attributions collectively determine whether the bullying act is socially accepted and transformed into social standing among peers (henceforth peer status). We conclude that if we are to understand the peer status of bullies, we need to identify the rules according to which third parties decide to attribute status to their peers. By collectively rewarding or punishing bullying, peers set an injunctive group norm, which in the past has been the receptacle for anti-bullying interventions (Salmivalli, 2014; Wölfer & Scheithauer, 2014). However, once a student has acquired a high status among peers, the very nature of status systems facilitates perpetuation of this high status. We know since the seminal works by De Solla Price (1965) and Merton (1968) that status and reputation systems have inherent dynamic tendencies to perpetuate and aggravate status differences. This implies that high-status individuals are much less vulnerable to incidental peer status withdrawal, whatever it is they are doing. If they are able to keep their high status simply because they had high status in the past, other
7
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF determinants of status become disposable. Whereas building up peer status may require specific behavior (e.g., acting pro-socially), from a certain status level on, this behavior is not required any more for maintaining high status. On the contrary, high-status individuals might even change the criteria of what it means to have high status (Dijkstra & Gest, 2015).
Thus, the center of our investigation needs to be the dynamic interplay of the bully’s decision to bully a victim, and a third party’s decision to attribute status to the bully. This is schematically depicted in Figure 2. The central panel illustrates the problematic situation in which bullying is being rewarded by status attribution.
If we are to explain why this dysfunctional configuration persists over time, we need to assess how incoming status attributions inform the bully’s decision to continue vs. stop bullying a given victim (comparison between middle and left panel of Figure 2), and we need to assess how the bullying of a given victim informs a third party’s decision whether or not to attribute status to the bully (comparison between middle and right panel of Figure 2). The arrows in the figure indicate what we might call “micro action”: one actor – the bully or the third party – deciding for or against a specific change. It is these micro actions that we need to understand and assess evidence for. Whereas the two “loops” starting and ending in the dysfunctional triad help stabilize the association between peer status and bullying, the two incoming arrows build up this association, whereas the two outgoing arrows reduce it. As indicated in the textual labels of these arrows, these tendencies can be understood, formalized, and ultimately analyzed from the involved actors’ perspective.
The Student’s Decision to Bully
Past research has shown that students with high status tend to be status-sensitive. They feel competition to maintain their high status and can resort to bullying to achieve their goal
(Fujimoto et al., 2017; Garandeau, Lee, et al., 2014; Pattiselanno, Dijkstra, Steglich,
Vollebergh, & Veenstra, 2015; Volk et al., 2015). For instance, high-status students were
8
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF shown to use verbal or physical bullying to intimidate ‘fellow competitors’ who threaten their social standing (Cillessen & Rose, 2005), and to overtly demonstrate their superiority over competitors by exercising control over others (Kolbert & Crothers, 2003). High-status students who challenge their high status peers are likely to increase their influence and power over others (Peets & Hodges, 2014; Volk et al., 2015). This can foster relational bullying as high-status students are in an ideal position to exclude peers and spread gossip (Faris, 2012;
Faris & Felmlee, 2011; Garandeau, Lee, et al., 2014; Reijntjes, Vermande, Olthof, et al.,
2013). In addition, high-status students may be less likely to abide by peer norms of pro- sociality if they already entered a cycle of self-reinforcing status perpetuation.
In sum, we hypothesize that having higher peer status will facilitate bullying
(Hypothesis 1). More precisely, with regard to Figure 2, we expect that high status will facilitate the bullying of new victims over time (creation bullying tie, Hypothesis 1a), and that existing bullying is continued (maintenance bullying tie, Hypothesis 1b).
The Nature of Peer Status
Whether in a school class or in a workplace context, it is useful to carefully differentiate between social acceptance or being liked by one’s peers on the one hand, and the peers’ perceptions of how popular one is in the whole group on the other hand (e.g., Dijkstra et al., 2015; Wolters, Knoors, Cilessen, & Verhoeven, 2014). The latter is revealed in the peers’ dyadic status attributions as studied in this paper, whereas the former is revealed in the peers’ dyadic friendship nominations, which we do not examine here. Acknowledging that both concepts are treated under the same label “popularity” in different branches of the literature, we have added a clarifying note on terminology in Appendix A. In particular, and in contrast to friendship, it is possible that an individual attributes status to someone whose behavior he or she personally disapproves of, simply because the individual thinks there are sufficiently many others who do appreciate the behavior in question. Status attributions reflect
9
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF dominance, prestige, and visibility in the peer group as a collective entity (Bellmore &
Cillessen, 2006). Conceptually, a status attribution does not require any personal relationship to the nominated person. As such, peer status is not bound by the same resource constraints as personal relationships are. Whereas there is a natural limit in the amount of time and attention devoted to maintaining a personal relationship, the same is not true for mere acknowledgements such as status attribution. Because of this, and again in contrast to friendship, status attributions can show strong evidence for cumulative advantage processes
(De Solla Price, 1965; Merton, 1968).
The Student’s Decision to Attribute Status
During late elementary and middle school years, discussions about who is popular or
‘cool’ are widespread (Lease, Kennedy, & Axelrod, 2002; Shoulberg, Sijtsema, & Murray-
Close, 2011), indicating that in this age range, status considerations play an important role in the social life of students. In particular in early adolescence, antisocial and ‘tough’ behavior, such as physical and verbal bullying, tends to be perceived as ‘cool’ (Reijntjes, Vermande,
Goossens, et al., 2013; Rodkin, Farmer, Pearl, & Van Acker, 2006; Salmivalli & Peets, 2009).
Students sharing these values are also inclined to socially reward fellow students for displaying corresponding behaviors by attributing status to them. This way, they are setting a social incentive to bully (see Figure 1). That in almost all bullying situations witnesses are present (e.g., Salmivalli, 2010) emphasizes that bullies depend on their classmates to gain peer status. Why do they collectively set this incentive? As argued by Faris (2012), peer status in any group is relative, and as such any loss of status for victims is balanced by small status gains for everyone else. This is most clear when considering status rankings. If a bullying act occurs in a bully-victim-tertius triad among status-equals (Figure 1), and if the small visibility gains for the victim are outweighed by the losses in dominance, then tertius ends up on a status rank higher than the victim without having done anything. Tertius therefore is doubly
10
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF motivated to reward the bully. First as a personal gratification because the bullying act allowed tertius to increase in status (essentially, for picking somebody else as victim), and second as a reification of the ranking itself, which one feels more comfortable with, the higher one’s own rank is.
In sum, we hypothesize that engagement in bullying behavior will lead to higher peer status in the classroom (Hypothesis 2). More precisely referring to the transitions depicted in
Figure 2, we expect that students who bully will attract more new peer status nominations than students who do not bully (creation status attribution tie, Hypothesis2a), and they will also be able to maintain more of their existing incoming status attributions than students who do not bully (maintenance status attribution tie, Hypothesis 2b).
Interfering Network Mechanisms
Our hypotheses are roughly in line with theories that link the sending or receiving of positive or negative ties to latent individual traits and/or postulate a self-reinforcing,
Matthew-effect type of mechanisms. Addressing the former case of exogenous explanatory variables, in particular the motivational traits to strive for high social status in the classroom or to avoid losing status have been argued to constitute a latent dimension explaining adolescents’ differences in terms of bullying and peer status (e.g., Gangel, Keane, Calkins,
Shanahan, & O’Brien, 2017; Puckett, Aikins, & Cillessen, 2008; Cillessen & Mayeux, 2004;
Sijtsema et al., 2009). Unfortunately, our data set does not allow us to test this underlying explanation. Concerning the latter case of endogenous explanatory variables, we will make sure that self-reinforcing mechanisms are assessed and controlled for when testing our hypotheses. Specifically, we will test to what degree there is evidence for students (a) bullying a victim because other students already are bullying the same victim (Matthew effect in bullying), (b) attributing status to a fellow student because other students already do the
11
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF same (Matthew effect in status attribution), (c) withdrawing status from students who get bullied, and (d) bullying those who are low in status.
Method
The association between bullying and peer status is the result of a dynamic interplay of the actions of bullies on the one hand and an audience of status attributors on the other hand.
Any empirical investigation into these relational processes therefore requires a longitudinal approach. We will apply stochastic actor-based modelling (Snijders, Van de Bunt, & Steglich,
2010) to a unique data set on bullying in elementary schools. By linking antecedents to consequences in a longitudinal study and by incorporating feedback between the two variables of interest on a fine-grained, dyadic level, we are able to differentially assess evidence for all network mechanisms.
Data and Sample
Data stem from the evaluation of the Dutch implementation of the KiVa anti-bullying program (Kaufman, Kretschmer, Huitsing, & Veenstra, 2018). An overview of published studies using these network data is given in Appendix B. Data collection covered 99 Dutch primary schools and took place in three waves: May 2012 (pre-intervention), October 2012, and May 2013. Prior to the pre-assessment in grades 2-5 in May 2012 – and for new students prior to the other assessments – schools sent information on the study and permission forms to parents. Passive rather than active parental consent was sought in order to obtain the high response rates that are necessary for meaningful social network analysis (note that observational research does not fall within the ambit of the Dutch Act on research on human subjects). Parents who did not want their child to participate in the assessment were asked to return a form. Students were informed at school about the research and gave oral consent.
Both parents and students could withdraw from participation at any time. Participation rates were high: at most two students per classroom did not have consent.
12
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
When the pre-assessment was finished, schools were randomly assigned by the
Netherlands Bureau for Economic Policy Analysis (CPB) to the control condition (33 schools) or the intervention condition (66 schools). Control schools were asked to continue their “care as usual” anti-bullying approach until their participation in the KiVa program began in June 2014.
In the Netherlands, especially in large schools, it is not uncommon to change the classroom composition each year. However, our aim was to longitudinally investigate developments in relatively stable peer groups. We therefore used data only from single-grade classrooms (N = 83). Moreover, we needed classrooms networks present at all three waves with less than 20% missing cases to perform social network analyses (Ripley, Snijders, Boda,
Vörös, & Preciado, 2018). As in one classroom 92% of the students did not participate in wave 3, 82 classrooms from 15 schools (11 KiVa schools and 4 control schools) were suitable for the analyses. The total number of students was 2,055 (Mage = 9.71, range 6.76 – 13.06 in wave 1; 50% boys). Missing data were handled the default way (Rigby et al., 2018, see section 4.3.2). All students were included, despite the possibility of having missing values for the variables at one of the waves, for instance caused by absence during the assessment or non-consent (wave 1: 1%, wave 2: 1%, wave 3: 3%). These absent students could nevertheless be nominated by others and were thus included in the networks.
Procedure
Students completed online questionnaires on the schools’ computers during regular school hours, under supervision of their classroom teachers who were supplied with detailed instructions before the data collection started. Teachers were present to answer questions and, if needed, help students in such a way that it would not affect their answers (e.g., by asking them questions such as “Which words are unclear to you?”). The order of questions, items, and scales used in this study were randomized to prevent any systematic order effects.
13
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Difficult topics were explained in several instructional videos. In one video at the start of the questionnaire, students were told that their answers would remain confidential, but that their teacher might be given general feedback to improve the classroom climate. Before answering the questions about bullying and victimization, the term bullying was defined as formulated in Olweus’ Bully/Victim questionnaire in another video (Olweus, 1996). Several examples covering different forms of bullying were given, followed by an explanation emphasizing the intentional and repetitive nature of bullying and the power imbalance between bullies and victims.
Measures
Bullying. At each time point, students were first asked to indicate whether they were being victimized, using 11 bully/victims items covering the various types of bullying
(Bully/Victims Questionnaire, Olweus, 1996). Those who indicated that they were victimized at least once on any of the items were asked to nominate the classmates who were victimizing them ("Who in your class always starts bullying you?"). A roster with the names of all the children in class was presented on the computer screen. Bully nominations were coded 1 and non-nominations 0. As our study aimed to investigate active bullying behavior and not being nominated as a bully, the resulting network matrix was transposed so that the presence of a relation indicates a bully-victim relation instead of a victim-bully relation. It is likely that a different subgroup of bullies would have been identified if we had asked students to indicate whom they bully. Dutch students seem more inclined to admit that they are a victim of bullying than to confess that they bully as the prevalence of self-proclaimed bullies was found to be lower compared to the prevalence of self-proclaimed victims (Veenstra et al., 2007).
Peer status attribution. Students could nominate an unlimited number of classmates they perceived as popular (“Who is popular in your class?”) at each time point. Similar to the bullying network, peer status nominations were coded 1 and non-nominations 0, resulting in
14
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF status attribution network matrices consisting of directed peer status nominations for each classroom. Please consult Appendix A for understanding the use of the term popularity in this context.
Analytical Strategy
Our hypotheses were tested using longitudinal social network analysis with the RSiena software (Ripley et al., 2018). This software estimates the parameters of stochastic actor- based models of the (co-)evolution of (multiple) social networks over time (Snijders et al.,
2010; Snijders, Lomi, & Torló, 2013). Under this approach, the first observation of a networked social system is conditioned upon, and subsequent observations are interpreted as the cumulative outcome of an unobserved series of small changes applied to the preceding observation. These unobserved small changes are modelled as decisions made by individuals in the network about maintaining, dissolving, or creating the ties they have to others.
Parameter estimates are obtained through simulation-based inference, which is common for fitting models with intractable likelihood functions to complex data sets (e.g., Gourieroux &
Monfort, 1996).
Model specification. There are two main model parts, one for each dependent network variable (referring to bullying and status attribution). Because our hypotheses are about the effects of the two networks on each other, we will first describe how these hypotheses are operationalized. We then give a detailed description of the other effects used for modeling the dynamics of status attribution, and finish with a sketch of the corresponding model for bullying dynamics.
Stochastic actor-based models of a single network distinguish between effects modeling the speed of the change process (rate effects) and effects modeling the nature of the network changes (jointly contributing to the objective function). In our case of two co- evolving networks, there are rate and objective function effects for the status attribution
15
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF network on the one hand, and for the bullying network on the other hand. We used an intercept model for the rate functions, referring to that the rate function was not allowed to differ according to actor attributes or network position. To test our hypotheses about peer status (referring to status attribution indegree) going together with bullying, in Model 1 we estimated two effects. The status to bullying effect indicates whether high indegree in status attribution (referring to being considered high in status by more others) implied high outdegree in the bullying network (referring to bullying more others, Hypothesis 1).
Conversely, the bullying to status effect indicates whether a higher outdegree in the bullying network implied a higher indegree in the status attribution network (Hypothesis 2). In Model
2, these two effects are further nuanced according to whether they explain the creation of new ties (Hypotheses 1a and 2a) or the maintenance of existing ties (Hypotheses 1b and 2b). We tested, as a second type of network-crossing effects, whether a high indegree in bullying
(referring to being a victim) implied a lower indegree in status attribution (victimization to status effect), or whether the converse was the case (status to victimization effect).
In addition to these cross-network effects, we included univariate, structural effects of network change for both networks, which capture the tendencies of individuals to form and maintain relationships under specific network-structural conditions. These effects also serve to optimize the goodness of fit of the model (Huitsing et al., 2014; Rambaran et al., 2015;
Snijders et al., 2010).
The following univariate, structural effects were added to explain the dynamics of status attribution. The outdegree effect expresses the overall tendency of individuals i to attribute status to other individuals j in the network (notation: ij). The reciprocity effect models the tendency to reciprocate a status nomination (ij facilitates ji). Two effects of triangular closure (referring to group formation) were included. The first is the transitive triplets effect, which reflects the tendency of individual i to attribute status to peers k who
16
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF received status nominations from peers j that i also attributes status to (transitive closure: ij and jk together facilitate ik). This group formation effect is in line with the assumption of a status hierarchy inside the group, which can be seen from a simple tie count: k receives two ties, but sends none (high status), i sends two ties, but receives none (low status), and j sends and receives one tie (middle rank; Snijders & Steglich, 2015). The second group formation effect is the three cycles effect, which investigates the tendency of individuals i, j, and k to form a non-hierarchical group (cyclical closure: ij and jk together facilitate ki).
In order to differentiate between individuals who received or sent many ties, three degree-related effects were included. The degree-related effects were all measured with the square roots of the degrees instead of the raw degrees (Huitsing et al., 2014; Snijders et al.,
2010). Indegree attractiveness reflects the tendency for those who receive many status attributions to receive even more over time – known as the Matthew effect on status reputation
(Merton, 1968). This effect expresses status differences that are (exclusively) captured in standardized peer status measures (Cillessen & Rose, 2005), and accordingly we expect it to be strong in the data. Outdegree activity is about the tendency for those who attribute status to many others to send even more attributions over time. Finally, indegree activity models the tendency to attribute status to others when being attributed status often oneself. One more effect we included was sex similarity, accounting for whether individuals were more likely to attribute status to others of the same sex than to the opposite sex.
The effects used to explain bullying dynamics are generally the same as for status attribution dynamics. However, instead of the indegree activity effect, we included the outdegree attractiveness effect which reflects the tendency to being victimized for those who bully others (we expect a negative effect). Moreover, because of the low density of the bullying network, the effects of reciprocity, transitive triplets, and three cycles could not be identified in most classrooms. The group formation effects therefore were dropped from the 17
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF model specification, whereas the reciprocity effect was not estimated, but score tested (we tested whether the model lacked fit, compared to an enriched model including the effect).
Also the direct tie-level effects that examine the main effects of dyad-level status attribution on bullying, and vice versa, were included in the score tests.
Model building. The co-evolution of the bullying and status attribution networks was analyzed in two steps. The first model included the main effects to test hypotheses 1 and 2.
We added endowment and creation parameters in the second model so that the effects for the maintenance and formation of ties could be distinguished (hypotheses 1a,b and 2a,b). The two models were estimated separately for each classroom, using all three time points, and then combined in a meta-analysis (Snijders & Baerveldt, 2003).
Results
Descriptive Results
Table 1 presents descriptive statistics of the bullying and status attribution networks.
The average degree shows that students bullied on average one to two classmates, and they on average attributed status to around three classmates. Status attributions tend to rise somewhat over time, whereas bullying nominations decrease. For both bullying and status attribution, reciprocation tends to increase in the year under study, and there was evidence for transitive closure. In both networks, the majority of nominations occurred between students of the same sex (57-69%), but this level is far below the age-specific, almost complete sex segregation in friendship networks.
The Jaccard index indicates the proportion of stable relations among the total number of new, lost, and stable ties between observed time-points. It needs to be sufficiently high if one wants to obtain reliable estimates of endogenous network effects (social ties explaining other social ties). For the status attribution networks the Jaccard indices were sufficiently high for RSiena analysis to proceed without problems (referring to larger than .30; Ripley et al.,
18
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
2018). For bullying, the indices were low (.18 and .19), but this is common across negative tie studies and the indices were not critically low (Ripley et al., 2018). We therefore estimated model specifications with reduced complexity, for which the analyses gave no numerical problems.
We investigated how changes in bullying activity from one observation moment to the next were associated with changes in dyadic status attributions while controlling for initial level of bullying. The results are depicted in Figure 3. We see an overall pattern of regression to the mean: below-average bullies (white bars) tend to increase their bullying whereas above- average bullies (grey bars) tend to strongly decrease it. The differences between the four dyadic status attribution change patterns are completely analogous for the two bully groups.
Stable status attribution goes together with more bullying than lost status attribution. For the above-average bullies, this means they reduce their bullying less under stable status attribution than they do under lost status attribution. For the below-average bullies, it means that they increase their bullying more under stable status attribution than they do under lost status attribution. When comparing dyads without an initial status attribution in terms of whether there still is none at the next time point (pattern “none”) or whether there is one
(pattern “new”), we see the same differences: new status attribution goes together with more bullying than continued absence of status attribution. Taken together, these descriptive results are consistent with our hypotheses about higher peer status being associated with more bullying (Hypotheses 1).
We also examined how changes in bullies’ peer status from one observation moment to the next were associated with changes in their dyadic bullying behavior, while controlling for the initial level of peer status (Figure 4). We again see the overall pattern of regression to the mean: students of below-average status (white bars) tend to increase their peer status strongly, whereas students of above-average status (grey bars) tend not to. There are some
19
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF interesting modulations on these general change patterns. Let us first compare initial bully- victim dyads in terms of whether the bullying continues at the next time point (pattern
“stable”) or stops (pattern “lost”). For the low-status bullies, there are only negligible differences, but for the high-status bullies, the difference is remarkable: they gain, on average, peer status under continued bullying of the same victims, whereas they lose peer status when bullying is stopped. Second, we compare initial non-bullying dyads in terms of whether non- bullying continues (pattern “none”) or bullying is observed at the next time point (pattern
“new”). For the low-status as well as the high-status non-bullies, there is a clear status gain advantage in favor of starting a new bully-victim relation. Taken together, these descriptive results are in line with Hypotheses 2 about engagement in bullying being associated with increased peer status.
Structural Network Effects
Table 2 presents mean estimates and their standard errors obtained from meta-analyses of RSiena-results for all school classes. To facilitate interpretation of the results, we calculated the exponential function of the estimates (odds ratios, presented in text). Model 1 shows that students tended to be selective in nominating classmates as a bully (outdegree, OR
= 0.002, p < .001) and in attributing status to classmates (outdegree, OR = 0.005, p < .001).
The positive reciprocity parameter in the status attribution network (OR = 1.27, p < .001) indicated that status nominations were likely to be reciprocated (referring to that students attributed status to those who provided status to them too). Additionally, when students attributed status to one of their classmates and this classmate attributed status to a third classmate, students were inclined to also attribute status to this third classmate over time
(transitive triplets, OR= 1.08, p < .001). The three cycles effect was negative (OR = 0.90, p <
.001), which indicates that status attributions led to local hierarchies.
20
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
The positive indegree attractiveness effects for both bullying (OR = 3.56, p < .001) and status attribution (OR = 2.46, p < .001) showed that often-bullied students attracted more being bullied nominations over time (Matthew effect in victimization), and that high-status students attracted more peer status nominations over time (Matthew effect in status attribution). Moreover, students who bullied many others or who attributed status to many others, tended to increase this tendency further (outdegree activity, OR = 1.99 for bullying;
OR = 1.77 for status attribution, p < .001).
Lastly, the positive sex similarity effects in both networks indicated that bullying relationships and status nominations were more likely to occur between students of the same sex (OR = 1.48 for bullying; OR = 1.84 for status attribution, p < .001).
The Interplay between Bullying and Peer Status
The “between-network” effects in Model 1 revealed that a high indegree in status attribution increased the likelihood of a high outdegree in bullying over time (status to bullying, OR = 1.27, p < .001). This is in line with our hypothesis that having higher peer status is associated with engagement in bullying over time (Hypothesis 1). Moreover, it was shown that a high outdegree in bullying is associated with receiving more status attributions over time (bullying to status, OR = 1.09, p = .004). In other words: being a bully makes you high in status. This outcome is consistent with our hypothesis that bullying was a way to gain peer status (Hypothesis 2).
Model 2 unraveled the interplay between bullying and peer status by distinguishing effects for the dissolution, maintenance, and formation of ties. The outcomes concerning the development of bullying demonstrated that students with high status discontinue bullying their former victims (maintenance bullying, OR = 0.66, p = .006) and start bullying classmates whom they did not bully before (creation bullying, OR = 2.27, p < .001). Whereas the latter result is in line with our expectation that high status is associated with the formation
21
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF of new bully relations over time (Hypothesis 1a), the former, negative result is in contradiction to our expectation that high-status bullies would continue bullying their former victims (Hypothesis 1b); in fact they discontinued bullying them.
For the status attribution network it turned out students who bullied gained new status attributions over time (creation status, OR = 1.28, p < .001). Put differently, students who bullied received status from classmates who did not provide them status before. This finding is consistent with what we expected (Hypothesis 2a). The maintenance status effect was positive, but not statistically significant (OR = 1.14, p = 0.065). Whereas the sign of the coefficient is in line with our expectation (Hypothesis 2b), there was no significant evidence for existing status attributions to be affected by bullying. Among classmates who already considered you high in status, your bullying seems not to make a decisive difference in their decision whether or not to continue providing you status.
The mean results of the score tests for the two omitted model parameters in the bullying network were significant for bullying reciprocity (M = 0.61, p < .001) and the direct tie-level effect of bullying on peer status (M = -0.15, p = .013). Inclusion of these effects in the model led to instability of the estimation algorithm, indicating that they refer to small fractions of the data in a few classrooms only. Because of this, we consider it unlikely that their omission challenges our main results.
Supplementary Analyses: Grade Differences
We additionally tested whether the results were homogeneous across grades (see
Appendix C). For the higher grades patterns, results were comparable to the findings in Table
2. Students who bullied tend to gain peer status among classmates and high-status bullies tend to choose new victims. However, for students in the lower grades, bullying led to less peer status (creation status, OR = 0.81, p = .001 in grade 2) or did not contribute to peer status
(creation status, OR = 1.17, p = .079 in grade 3).
22
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Discussion
This study focused on bullying as strategic behavior linked to high social status in the classroom. The aim was to unravel the complex interplay between bullying and one’s peer status on the dyadic level over time, which fills a serious gap in the existing literature on bullying. We argued that bullying and peer status reinforce each other, and we used longitudinal multiplex network analysis to get more insight into the relational patterns of bullying and peer status. Our study is the first that investigated how bullying ties affect the creation, dissolution, and maintenance of status attribution ties, and vice versa, in a social network framework.
In line with previous studies employing an actor-level framework (Cillessen & Borch,
2006; Cillessen & Mayeux, 2004; Reijntjes, Vermande, Olthof, et al., 2013; Sentse et al.,
2015), we found that bullying generally increases peer status among classmates. We also found that high status individuals tend to bully, but here the results are more nuanced than what actor-level analyses can reveal. When differentiating between the creation and maintenance of specific ties in our two networks, we could demonstrate that unlike low-status bullies, high-status bullies tended to discontinue bullying the same victims, but replaced those with new victims. Furthermore, when looking into the age-dependency of these effects, we could show that bullying is rewarded by peer status attributions only for the higher grades of elementary school (grades 4 and 5). For the youngest age group (grade 2) classmates even sanctioned bullying by a withdrawal of status attributions. Taken together, our findings reiterate that bullying is a complex group phenomenon in which obtaining and maintaining high social standing plays a decisive role (e.g., Salmivalli, 2010; Volk, Dane, & Marini,
2014), and that a network perspective deepens our understanding of the interplay between bullying and peer status.
23
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
We think that our approach to analyze peer status not as actor-level construct, but on the dyad level as a network of status attributions has broader potential in adolescence research. This is illustrated by our unexpected finding of high-status bullies replacing their victims over time, referring to discontinue existing bullying ties, but establish new ones.
Already previous social network analyses on bullying (Huitsing et al., 2014) had indicated that, on the relational level, bullying is not stable over time, whereas aggregated on the actor level (referring to the number of victims of a bully) it is. By design, an actor-level analysis overlooks these dyad-level changes in bullying pattern. Thus, such dynamic patterns become only visible to the researcher by adoption of a longitudinal network perspective.
This study raises new theoretical questions into how bullying and peer status are intertwined. If high-status students replace their victims on a regular basis, it would be interesting to investigate whether they start bullying particular fellow students who threaten their social standing in the classroom (referring to other high-status classmates, see also Peets
& Hodges, 2014), or whether they seek new, low status targets. In other words: is bullying behavior driven by the motivation to keep high status, opportunity, or both? Empirically, this could be investigated by further describing the differences in the dynamics of bullying and status attribution. For instance, by distinguishing bullies with reciprocated versus unreciprocated bullying relationships and test how these networks evolve over time.
Another important question is to what extent the classroom context influences the interplay between bullying and peer status, as bullying and peer status by definition both depend on the peer context (Salmivalli, 2010; Salmivalli et al., 1996). Several anti-bullying interventions aim to change group norms such that bullies are less supported by bystanders and that their antisocial behavior is less rewarded among peers (Salmivalli, 2014; Wölfer &
Scheithauer, 2014). Future research should examine whether the interplay between bullying
24
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF and peer status is different in classrooms with strong anti-bullying norms and investigate among which students these differences occur.
Various ramifications of our analyses can be thought of. First, the bullying and peer status networks in our study were examined in stable, single-grade classrooms only. In many countries, classroom composition is homogenous and remains the same during the students’ entire elementary school career. In the Netherlands, multi-grade elementary school classrooms are common and the composition of the classroom is likely to change between school years (Veenman, 1995). Especially status differences are very strong in these multi- grade classrooms. Now that we have developed a framework to investigate the relational processes behind bullying and peer status, it would be interesting to examine whether these processes differ in age-heterogeneous, unstable classrooms. Second, seeing that a substantial share of the bullying and victimization takes place outside the classroom (Huitsing et al.,
2014; Van der Ploeg, Steglich, Salmivalli, & Veenstra., 2015), it might be fruitful to move beyond the own classroom and analyze cohort-wide or even school-wide networks. Third, the focus of our study was on bullying in general. Nevertheless, bullying behavior can occur in several forms (Salmivalli, Kärnä, & Poskiparta, 2011): physical, verbal, material, relational, and cyberbullying. Some types are more visible or more rewarded than others, which makes it likely that the different types of bullying are also differently related to peer status.
Practical Implications
Increasingly, researchers are collecting rich data on relations between children and adolescents through network questions (Veenstra, Dijkstra, Steglich, & Van Zalk, 2013; even from 6-years on: Verlinden et al., 2014). Such network analyses provide insights in selection processes, referring to that bullies or victims flock together, and influence processes, referring to that friends’ bullying behavior or friends’ victimization is contagious (Lodder, Scholte,
Cillessen, & Giletta, 2016; Sentse, Dijkstra, Salmivalli, & Cillessen, 2013; Turanovic &
25
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Young, 2016). Our results illustrate that peer status plays an essential role in involvement in bullying behavior. The use of longitudinal social network data helps to better understand the relational patterns of bullying and peer status. The findings imply that bullying is an effective strategy to obtain and maintain high status in the classroom, particularly in higher grades.
However, our study also reveals that not all high-status students engage in bullying and that not all bullies gain peer status. This might reflect that other factors or behaviors explain status attributions. In previous studies, prosocial behavior has been linked to peer status (Slaughter,
Imuta, Peterson, & Henry, 2015). Next to interventions aimed at changing group norms and enhancing defending behavior (Salmivalli, 2014), teaching bullies in higher grades prosocial ways to gain or maintain high status therefore is probably essential to effectively intervene in school bullying (Ellis, Volk, Gonzalez, & Embry, 2015; Swearer et al., 2010).
Our finding that bully-victim relations in the higher grades of elementary school seem to be unstable over time might be an important finding for the design of new anti-bullying interventions or the evaluation of existing ones. Indicated anti-bullying interventions that aim to solve existing bullying situations (Garandeau, Poskiparta, & Salmivalli, 2014; Van der
Ploeg, Steglich, & Veenstra, 2016) should acknowledge that bullies tend to switch victims and refine their strategies to effectively target the complexity of bullying. Rather than solving specific bullying situations or helping particular victims, it is the norm that bullying is a way to achieve peer status that should be altered.
26
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
References
Adler, P. A., & Adler, P. (1998). Peer power: Preadolescent culture and identity. New
Brunswick, NJ: Rutgers University Press.
Bellmore, A. D., & Cillessen, A. H. N. (2006). Reciprocal influences of victimization,
perceived social preference, and self-concept in adolescence. Self and Identity, 5(3), 209–
229. doi:10.1080/15298860600636647
Betancourt, N., Kovács, B., & Otner, S. M. (2018). The perception of status: How we infer
the status of others from their social relationships. Network Science, 6(3), 319-347. doi:
10.1017/nws.2018.13
Caravita, S. C. S., Di Blasio, P., & Salmivalli, C. (2009). Unique and interactive effects of
empathy and social status on involvement in bullying. Social Development, 18(1), 140–163.
doi:10.1111/j.1467-9507.2008.00465.x
Cillessen, A. H. N., & Borch, C. (2006). Developmental trajectories of adolescent popularity:
A growth curve modelling analysis. Journal of Adolescence, 29(6), 935–959.
doi:10.1016/j.adolescence.2006.05.005
Cillessen, A. H. N., & Mayeux, L. (2004). From censure to reinforcement: Developmental
changes in the association between aggression and social status. Child Development, 75(1),
147–163. doi:10.1111/j.1467-8624.2004.00660.x
Cillessen, A. H. N., & Rose, A. J. (2005). Understanding popularity in the peer system.
Current Directions in Psychological Science, 14(2), 102–105. doi:10.1111/j.0963-
7214.2005.00343.x
Cillessen, A. H. N., Schwartz, D., & Mayeux, L. (Eds.). (2011). Popularity in the peer system.
Guilford Press.
Coie, J. D., Dodge, K. A., & Coppotelli, H. (1982). Dimensions and types of social status: A
cross-age perspective. Developmental psychology, 18(4), 557.
27
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
De Bruyn, E. H., Cillessen, A. H. N., & Wissink, I. B. (2010). Associations of peer
acceptance and perceived popularity with bullying and victimization in early adolescence.
The Journal of Early Adolescence, 30(4), 543–566. doi:10.1177/0272431609340517
De Solla Price, D. J. (1965). Networks of scientific papers. Science, 510-515.
Dijkstra, J. K., Cillessen, A. H. N., & Borch, C. (2013). Popularity and adolescent friendship
networks: Selection and influence dynamics. Developmental Psychology, 49(7), 1242–1252.
doi:10.1037/a0030098
Dijkstra, J. K., Cillessen, A. H. N., Lindenberg, S., & Veenstra, R. (2010). Same-gender and
cross-gender likeability: Associations with popularity and status enhancement: The TRAILS
study. The Journal of Early Adolescence, 30(6), 773-802. doi: 10.1177/0272431609350926
Dijkstra, J. K., & Gest, S. D. (2015). Peer norm salience for academic achievement, prosocial
behavior, and bullying: Implications for adolescent school experiences. The Journal of
Early Adolescence, 35(1), 79-96. doi:10.1177/0272431614524303
Dijkstra, J. K., Kretschmer, T., Lindenberg, S., & Veenstra, R. (2015). Hedonic, instrumental,
and normative motives: Differentiating patterns for popular, accepted, and rejected
adolescents. The Journal of Early Adolescence, 35(3), 308-328.
doi:10.1177/0272431614535092
Eder, D. (1985). The cycle of popularity: Interpersonal relations among female adolescents.
Sociology of Education, 58, 154–165.
Ellis, B. J., Volk, A. A., Gonzalez, J. M., & Embry, D. D. (2015). The meaningful roles
intervention: An evolutionary approach to reducing bullying and increasing prosocial
behavior. Journal of Research in Adolescence, 1–16. doi:10.1111/jora.12243
Faris, R. (2012). Aggression, exclusivity, and status attainment in interpersonal networks.
Social Forces, 90(4), 1207–1235. doi:10.1093/sf/sos074
28
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Faris, R., & Felmlee, D. (2011). Status struggles. American Sociological Review, 76(1), 48–
73. doi:10.1177/0003122410396196
Fujimoto, K., Snijders, T. A. B., & Valente, T. W. (2017). Popularity breeds contempt: The
evolution of reputational dislike relations and friendships in high school. Social Networks,
48, 100–109. doi:10.1016/j.socnet.2016.07.006
Gaffney, H., Ttofi, M. M., & Farrington, D. P. (in press). Evaluating the effectiveness of
school-bullying prevention programs: An updated meta-analytical review. Aggression and
Violent Behavior. doi:10.1016/j.avb.2018.07.001
Gangel, M. J., Keane, S. P., Calkins, S. D., Shanahan, L., & O’Brien, M. (2017). The
Association Between Relational Aggression and Perceived Popularity in Early Adolescence:
A Test of Competing Hypotheses. The Journal of Early Adolescence, 37(8), 1078–1092.
doi:10.1177/0272431616642327
Garandeau, C. F., Lee, I. A., & Salmivalli, C. (2014). Differential effects of the KiVa anti-
bullying program on popular and unpopular bullies. Journal of Applied Developmental
Psychology, 35(1), 44–50. doi:10.1016/j.appdev.2013.10.004
Garandeau, C. F., Poskiparta, E., & Salmivalli, C. (2014). Tackling acute cases of school
bullying in the KiVa anti-bullying program: A comparison of two approaches. Journal of
Abnormal Child Psychology, 42(6), 981–991. doi:10.1007/s10802-014-9861-1
Gourieroux, C., & Monfort, A. (1996). Simulation-based econometric methods. Oxford
university press.
Homans, G. C. (1950). The human group. New York, NY: Harcourt, World, and Brace, Inc.
Huitsing, G., Snijders, T. A. B., Van Duijn, M. A. J., & Veenstra, R. (2014). Victims, bullies,
and their defenders: A longitudinal study of the coevolution of positive and negative
networks. Development and Psychopathology, 26(3), 645–659.
doi:10.1017/S0954579414000297
29
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Huitsing, G., & Monks, C. P. (2018). Who victimizes whom and who defends whom? A
multivariate social network analysis of victimization, aggression, and defending in early
childhood. Aggressive behavior, 44, 394–405. doi:10.1002/ab.21760
Huitsing, G., & Veenstra, R. (2012). Bullying in classrooms: Participant roles from a social
network perspective. Aggressive behavior, 38(6), 494-509. doi:10.1002/ab.21438
Juvonen, J., Wang, Y., & Espinoza, G. (2013). Physical Aggression, Spreading of Rumors,
and Social Prominence in Early Adolescence: Reciprocal Effects Supporting Gender
Similarities? Journal of Youth and Adolescence, 42(12), 1801–1810. doi:10.1007/s10964-
012-9894-0
Kaufman, T. M., Kretschmer, T., Huitsing, G., & Veenstra, R. (2018). Why does a universal
anti-bullying program not help all children? Explaining persistent victimization during an
intervention. Prevention science, 19, 822-832. doi:10.1007/s11121-018-0906-5
Kolbert, J. B., & Crothers, L. M. (2003). Bullying and Evolutionary Psychology. Journal of
School Violence, 2(3), 73–91. doi:10.1300/J202v02n03_05
LaFontana, K. M., & Cillessen, A. H. N. (1999). Children’s interpersonal perceptions as a
function of sociometric and peer-perceived popularity. Journal of Genetic Psychology, 160,
225-242. doi:10.1080/00221329909595394
Lease, A. M., Kennedy, C. A., & Axelrod, J. L. (2002). Children’s social constructions of
popularity. Social Development, 11(1), 87–109. doi:10.1111/1467-9507.00188
Lodder, G. M. A., Scholte, R. H. J., Cillessen, A. H. N., & Giletta, M. (2016). Bully
Victimization: Selection and influence within adolescent friendship networks and cliques.
Journal of Youth and Adolescence, 45(1), 132–144. doi:10.1007/s10964-015-0343-8
Merton, R. K. (1968). The Matthew Effect in Science: The reward and communication
systems of science are considered. Science, 159(3810), 56–63.
doi:10.1126/science.159.3810.56
30
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Nickerson, A. B. (2017). Preventing and Intervening with Bullying in Schools: A Framework
for Evidence-Based Practice. School Mental Health, 1–14. doi:10.1007/s12310-017-9221-8
Oldenburg, B., Van Duijn, M. A. J., & Veenstra, R. (2018). Defending one's friends, not one's
enemies: A social network analysis of children's defending, friendship, and dislike
relationships using XPNet. PloS one, 13(5), doi:e0194323.
Olweus, D. (1978). Aggression in the schools: bullies and whipping boys. Washington, DC:
Hemisphere (Wiley).
Olweus, D. (1993). Bullying at school: What we know and what we can do. Malden, MA:
Blackwell Publishing.
Olweus, D. (1996). The revised Olweus bully/victim questionnaire. Bergen, Norway:
Research Center for Health Promotion (HEMIL Center), University of Bergen.
Parkhurst, J. T., & Hopmeyer, A. (1998). Sociometric popularity and peer-perceived
popularity: Two distinct dimensions of peer status. Journal of Early Adolescence, 18, 125–
144. doi:10.1177/0272431698018002001
Pattiselanno, K., Dijkstra, J. K., Steglich, C., Vollebergh, W., & Veenstra, R. (2015).
Structure Matters: The role of clique hierarchy in the relationship between adolescent social
status and aggression and prosociality. Journal of Youth and Adolescence, 44(12), 2257–
2274. doi:10.1007/s10964-015-0310-4
Peets, K., & Hodges, E. V. E. (2014). Is popularity associated with aggression toward socially
preferred or marginalized targets? Journal of Experimental Child Psychology, 124, 112–
123. doi:10.1016/j.jecp.2014.02.002
Puckett, M. B., Aikins, J. W., & Cillessen, A. H. N. (2008). Moderators of the association
between relational aggression and perceived popularity. Aggressive Behavior, 34(6), 563–
576. doi:10.1002/ab.20280
31
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Rambaran, J. A., Dijkstra, J. K., Munniksma, A., & Cillessen, A. H. N. (2015). The
development of adolescents’ friendships and antipathies: A longitudinal multivariate
network test of balance theory. Social Networks, 43, 162–176.
doi:10.1016/j.socnet.2015.05.003
Reijntjes, A., Vermande, M. M., Goossens, F. A., Olthof, T., Van de Schoot, R., Aleva, E. A.,
& Van der Meulen, M. (2013). Developmental trajectories of bullying and social dominance
in youth. Child Abuse and Neglect, 37(4), 224–234. doi:10.1016/j.chiabu.2012.12.004
Reijntjes, A., Vermande, M. M., Olthof, T., Goossens, F. A., Van de Schoot, R., Aleva, E. A.,
& Van der Meulen, M. (2013). Costs and benefits of bullying in the context of the peer
group: A three wave longitudinal analysis. Journal of Abnormal Child Psychology, 41(8),
1217–1229. doi:10.1007/s10802-013-9759-3
Ripley, R., Snijders, T. A. B., Boda, Z., Vörös, A., & Preciado, P. (2018). Manual for RSiena.
Oxford, UK: University of Oxford, Nuffield College.
Rodkin, P. C., Farmer, T. W., Pearl, R., & Van Acker, R. (2006). They’re cool: Social status
and peer group suppports for aggressive boys and girls. Social Development, 15(2), 175–
204. doi:10.1046/j.1467-9507.2006.00336.x
Sainio, M., Veenstra, R., Huitsing, G., & Salmivalli, C. (2011). Victims and their defenders:
A dyadic approach. International Journal of Behavioral Development, 35(2), 144–151.
doi:10.1177/0165025410378068
Salmivalli, C. (2010). Bullying and the peer group: A review. Aggression and Violent
Behavior, 15(2), 112–120. doi:10.1016/j.avb.2009.08.007
Salmivalli, C. (2014). Participant roles in bullying: How can peer bystanders be utilized in
interventions? Theory Into Practice, 53(4), 286–292. doi:10.1080/00405841.2014.947222
32
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Salmivalli, C., Kärnä, A., & Poskiparta, E. (2011). Counteracting bullying in Finland: The
KiVa program and its effects on different forms of being bullied. International Journal of
Behavioral Development, 35(5), 405–411. doi:10.1177/0165025411407457
Salmivalli, C., Lagerspetz, K., Björkqvist, K., Österman, K., & Kaukiainen, A. (1996).
Bullying as a group process: Participant roles and their relations to social status within the
group. Aggressive Behavior, 22(1), 1–15. doi:10.1002/(SICI)1098-2337(1996)22:1<1::AID-
AB1>3.0.CO;2-T
Salmivalli, C., & Peets, K. (2009). Bullies, victims, and bully-victim relationships in middle
childhood and early adolescence. In K. H. Rubin, W. M. Bukowski, & B. Laursen (Eds.),
Handbook of peer interactions, relationships, and groups (pp. 322–340). New York:
Guilford.
Schwartz, D., Tom, S. R., Chang, L., Xu, Y., Duong, M. T., & Kelly, B. M. (2010).
Popularity and acceptance as distinct dimensions of social standing for Chinese children in
Hong Kong. Social Development, 19(4), 681-697. doi:10.1111/j.1467-9507.2009.00558.x
Sentse, M., Dijkstra, J. K., Salmivalli, C., & Cillessen, A. H. N. (2013). The dynamics of
friendships and victimization in adolescence: A longitudinal social network perspective.
Aggressive Behavior, 39(3), 229–238. doi:10.1002/ab.21469
Sentse, M., Kretschmer, T., & Salmivalli, C. (2015). The longitudinal interplay between
bullying, victimization, and social status: age-related and gender differences. Social
Development, 24(3), 659–677. doi:10.1111/sode.12115
Shoulberg, E. K., Sijtsema, J. J., & Murray-Close, D. (2011). The association between valuing
popularity and relational aggression: The moderating effects of actual popularity and
physiological reactivity to exclusion. Journal of Experimental Child Psychology, 110(1),
20–37. doi:10.1016/j.jecp.2011.03.008
33
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Sijtsema, J. J., Veenstra, R., Lindenberg, S., & Salmivalli, C. (2009). Empirical test of bullies’
status goals: Assessing direct goals, aggression, and prestige. Aggressive Behavior, 35(1),
57–67. doi:10.1002/ab.20282
Slaughter, V., Imuta, K., Peterson, C. C., & Henry, J. D. (2015). Meta-analysis of Theory of
Mind and peer popularity in the preschool and early school years. Child Development,
86(4), 1159–1174. doi:10.1111/cdev.12372
Snijders, T. A. B., & Baerveldt, C. (2003). A multilevel network study of the effects of
delinquent behavior on friendship evolution. Journal of Mathematical Sociology, 27(2–3),
123–151. doi:10.1080/00222500305892
Snijders, T. A. B., Lomi, A., & Torló, V. J. (2013). A model for the multiplex dynamics of
two-mode and one-mode networks, with an application to employment preference,
friendship, and advice. Social Networks, 35(2), 265–276. doi:10.1016/j.socnet.2012.05.005
Snijders, T. A. B., & Steglich, C. E. G. (2015). Representing micro-macro linkages by actor-
based dynamic network models. Sociological Methods & Research, 44(2), 222–271.
doi:10.1177/0049124113494573
Snijders, T. A. B., van de Bunt, G. G., & Steglich, C. (2010). Introduction to stochastic actor-
based models for network dynamics. Social Networks, 32(1), 44–60.
doi:10.1016/j.socnet.2009.02.004
Swearer, S. M., Espelage, D. L., Vaillancourt, T., & Hymel, S. (2010). What can be done
about school bullying ? linking research to educational practice. Educational Researcher,
39(1), 38–47. doi:10.3102/0013189X09357622
Turanovic, J. J., & Young, J. T. N. (2016). Violent offending and victimization in
adolescence: Social network mechanisms and homophily. Criminology, 54(3), 487–519.
doi:10.1111/1745-9125.12112
34
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Van der Ploeg, R., Steglich, C., Salmivalli, C., & Veenstra, R. (2015). The intensity of
victimization: Associations with children’s psychosocial well-being and social standing in
the classroom. PLoS ONE, 10(10). doi:10.1371/journal.pone.0141490
Van der Ploeg, R., Steglich, C., & Veenstra, R. (2016). The support group approach in the
Dutch KiVa anti-bullying programme: effects on victimisation, defending and well-being at
school. Educational Research, 58(3), 221–236. doi:10.1080/00131881.2016.1184949
Veenman, S. (1995). Cognitive and noncognitive effects of multigrade and multi-age classes:
A best-evidence synthesis. Review of Educational Research, 65(4), 319–381.
doi:10.3102/00346543065004319
Veenstra, R., Dijkstra, J. K., Steglich, C., & Van Zalk, M. H. W. (2013). Network-behavior
dynamics. Journal of Research on Adolescence, 23(3), 399–412. doi:10.1111/jora.12070
Veenstra, R., Lindenberg, S., Zijlstra, B. J. H., De Winter, A. F., Verhulst, F. C., & Ormel, J.
(2007). The dyadic nature of bullying and victimization: Testing a dual-perspective theory.
Child Development, 78(6), 1843–1854. doi:10.1111/j.1467-8624.2007.01102.x
Verlinden, M., Veenstra, R., Ringoot, A. P., Jansen, P. W., Raat, H., Hofman, A., …
Tiemeier, H. (2014). Detecting bullying in early elementary school with a computerized
peer-nomination instrument. Psychological Assessment, 26(2), 628–641.
doi:10.1037/a0035571
Volk, A. A., Della Cioppa, V., Earle, M., & Farrell, A. H. (2015). Social competition and
bullying: An adaptive socioecological perspective. In V. Zeigler-Hill, L. Welling, & T.
Shackelford (Eds.), Evolutionary Perspectives on Social Psychology. Springer Internation
Publishing Switzerland.
Volk, A. A., Dane, A. V., & Marini, Z. A. (2014). What is bullying? A theoretical
redefinition. Developmental Review, 34(4), 327–343. doi:10.1016/j.dr.2014.09.001
35
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Volk, A. A., Veenstra, R., & Espelage, D. L. (2017). So you want to study bullying?
Recommendations to enhance the validity, transparency, and compatibility of bullying
research. Aggression and Violent Behavior, 36(July), 34–43. doi:10.1016/j.avb.2017.07.003
Wölfer, R., & Scheithauer, H. (2014). Social influence and bullying behavior: Intervention-
based network dynamics of the fairplayer.manual bullying prevention program. Aggressive
Behavior, 40(4), 309–319. doi:10.1002/ab.21524
Wolters, N., Knoors, H., Cillessen, A. H., & Verhoeven, L. (2014). Behavioral, personality,
and communicative predictors of acceptance and popularity in early adolescence. The
Journal of Early Adolescence, 34(5), 585-605. doi:10.1177/0272431613510403
36
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Figures and Tables
Figure 1. Dominance and visibility in a bully-victim-tertius triad.
B>T>V B T>B>V B>V>T T T>V>B V>B>T dominance V V>T>B
visibility
Note. Left panel: Differentiation in visibility and dominance after bully B bullies victim V (the three actors had initially identical social standing). Right panel: Resulting rankings of the three actors depending on the preference gradient after linear combination of preferences for visibility and dominance.
37
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Figure 2. The dysfunctional bully-victim-tertius triad (middle) and the surrounding network mechanisms.
B continues bullying V, given status attribution by T to B
B stops bullying V, T starts attributing given status status to B, given B’s attribution by T to B bullying of V B T B T B T
V B starts bullying V, V T stops attributing V given status status to B, given B’s attribution by T to B bullying of V
T continues attributing status to B, given B’s bullying of V
38
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Figure 3. Changes in bullying (referring to outdegree in the bullying network) over time by initial bullying and dyadic status attribution pattern over time.
Note. Unit of analysis is the tertius-bully (T-B) dyad; each actor contributes to multiple dyads. Stable: T attributes status to B at two subsequent observations; Lost: T attributes status to B at the first, but not at the second observation; New: T attributes status to B at the second, but not at the first observation; None: T does not attribute status to B neither at the first nor at the second observation.
39
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Figure 4. Changes in peer status (referring to indegree in the status attribution network) over time, by initial status and dyadic bullying pattern over time.
Note. Unit of analysis is the bully-victim (B-V) dyad; each actor contributes to multiple dyads. Stable: B bullies V at two subsequent observations; Lost: B bullies V at the first, but not at the second observation; New: B bullies V at the second, but not at the first observation; None: B does not bully V neither at the first nor at the second observation.
40
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Table 1. Class-level descriptive statistics for bullying and status attribution networks per wave.
Bullying network Status attribution network Wave 1 Wave 2 Wave 3 Wave 1 Wave 2 Wave 3 M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) Average degree 1.94 1.38 1.17 3.11 3.16 3.37 (1.06) (0.79) (0.80) (1.25) (1.39) (1.69) SD indegree 3.06 2.48 2.23 3.51 3.73 4.00 (1.24) (1.11) (1.22) (1.64) (1.67) (1.76) SD outdegree 1.96 1.58 1.41 3.16 3.18 3.27 (0.78) (0.71) (0.75) (1.23) (1.19) (1.51) Reciprocity .16 .17 .19 .19 .19 .21 (.10) (.10) (.17) (.08) (.09) (.15) Same sex nominations .57 .62 .68 .69 .68 .66 (.14) (.17) (.19) (.12) (.10) (.11) Transitivity .52 .47 .58 .53 .55 .57 (.18) (.22) (.23) (.14) (.14) (.14) Average class size 25.2 25.1 25.0 25.2 25.1 25.0 (4.2) (4.4) (4.4) (4.2) (4.4) (4.4) Non-respondents 1% 1% 3% 1% 1% 3%
Wave 1 2 Wave 2 3 Wave 1 2 Wave 2 3 Hamming distance a 53.6 43.4 77.3 77.6 Jaccard index b .19 (.10) .18 (.12) .32 (.12) .34 (.12) Note a Hamming distance is the number of tie changes . b Jaccard index is the fraction of stable ties among the new, lost, or stable ties N = 82 classrooms in 15 schools
41
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Table 2. Meta-analyses of multiplex network analysis: bullying and status attribution.
Model 1 Model 2 Est. SE Est. SE Bullying network Structural network effects Outdegree (density) -6.02 0.19*** maintenance -4.12 0.31*** creation -7.50 0.29*** Indegree attractiveness 1.27 0.05*** 1.19 0.04*** Outdegree attractiveness 0.16 0.14 -0.11 0.09 Outdegree activity 0.69 0.06*** 0.57 0.05*** Individual effects Sex similarity 0.39 0.09*** 0.35 0.06*** Between networks effects Status victimization -0.09 0.06 -0.10 0.07 Status bullying 0.24 0.04*** maintenance bullying -0.41 0.15* creation bullying 0.82 0.19***
Status attribution network Structural network effects Outdegree (density) -5.34 0.16*** maintenance -2.89 0.18*** creation -7.25 0.15*** Reciprocity 0.24 0.05*** 0.25 0.04*** Transitive triplets 0.08 0.02*** 0.07 0.01*** Three cycles -0.10 0.03*** -0.06 0.01*** Indegree attractiveness 0.90 0.03*** 0.88 0.02*** Indegree activity 0.03 0.03 0.04 0.03 Outdegree activity 0.57 0.03*** 0.57 0.03*** Individual effects Sex similarity 0.61 0.04*** 0.60 0.03*** Between networks effects Victimization status -0.02 0.02 -0.03 0.02 Bullying status 0.09 0.03** maintenance status 0.13 0.07 creation status 0.25 0.07*** Note. Rate of change effects were omitted from the table. All effects, except for status attribution reciprocity, show significant variation over classrooms. N = 74 classrooms * p < .05; ** p < .01; *** p < .001 (two-tailed tests) 42
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Appendix A: A Clarifying Note on the Notion of Popularity
Because of diverging definitions in the two different disciplinary communities that constitute our intended readership, we decided to avoid using the notion of popularity. These two disciplines are adolescent development (AD), and social network analysis (SNA). As one aim of our study is to provide a bridge between these disciplines, a clarifying note is expedient. In SNA, the notion of popularity is essentially a synonym for indegree in any directed network, referring to the number of ties received (Wasserman & Faust, 1994). It is commonly contrasted with expansiveness or activity, defined as outdegree, referring to the number of ties sent – and it is in particular agnostic to the content of the network. You could, for example, be SNA-popular in a bullying network, which simply means that you are bullied by many classmates.
In the sociological and psychological literature on AD, however, popularity is very narrowly defined by relationship content, namely as peer-acknowledged social status
(popularity among peers; Eder, 1985; Adler & Adler, 1998; Cillessen, Schwartz & Mayeux,
2011). It is typically assessed as indegree in a binary network of peer nominations in response to the question, “Who is popular in your class?” and it is commonly contrasted with similar constructs based on other relationship content, such as social acceptance or likeability, which are assessed as indegree in networks obtained with other name generators (Whom do you hang out / play with?; Whom do you like?; Who is your friend?). The AD literature identified important differences between these concepts (e.g., Parkhurst & Hopmeyer, 1998; LaFontana
& Cillessen, 1999; De Bruyn et al., 2010; Dijkstra et al., 2010; Schwartz et al., 2010), which is why researchers in this tradition strongly insist on not using the word popularity for the indegree in networks other than the one assessing popularity in the narrow sense.
Whereas AD-popularity satisfies the formal definition of SNA-popularity, the opposite obviously is not the case. When you are bullied by many classmates (referring to when you
43
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF are popular as a victim in SNA terminology), it certainly does not mean you have high social status among them (AD-popularity). However, if you are SNA-popular in a status attribution network, then the definition coincides with AD-popularity. As this little excursion illustrates, using the word popularity is ambiguous and would obfuscate our message to the reader.
Because of this, we chose to use the words indegree (with an explicit reference to the network on which it is calculated) when we mean SNA-popularity, and (peer) status when we mean
AD-popularity. And we speak of status attribution when we refer to the directed network of
AD-popularity nominations.
Appendix B: Studies on the Same Data as Ours or Taking a Similar Approach
So far, only a few studies investigated bullying and related behaviors, such as defending, as a directed dyadic relationship. One of the first studies (Huitsing & Veenstra,
2012) used dyadic information on bullying and defending to obtain an impression of the social structure of a classroom. It turned out that ingroup and outgroup effects were important in explaining the group process of bullying. In general, most bullies were boys, with boys bullying both boys and girls (Huitsing et al., 2014; Huitsing & Veenstra, 2012). Defending occurred mostly among same-sex classmates (Huitsing et al., 2014; Huitsing & Monks, 2018;
Sainio, Veenstra, Huitsing, & Salmivalli, 2011).
From two studies focusing on defending behavior (Oldenburg, Van Duijn, & Veenstra,
2018; Sainio et al., 2011) we know that aspects of liking, peer status, and friendship play a role. Victims liked their classmates who defended them and perceived them as popular
(Sainio et al., 2011). In addition, victims were likely to be defended by classmates who they perceive as friends or who perceive them as friends, whereas, it was unlikely that victims were defended by classmates they disliked or by whom victims were disliked. Moreover,
44
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF defending was more likely when the victim and (potential) defender shared friends or disliked the same classmates (Oldenburg et al., 2018).
In a study on the coevolution of defending, bullying and victimization networks, it was found that over time victims with the same bullies defended each other, defenders were at risk of becoming victimized by the bullies of the victims they stood up for, bullies with the same victims defended each other, and defenders of bullies started to bully those victims too
(Huitsing et al., 2014). These findings show that using a social network approach can be fruitful for investigating complex relationships such as bullying and peer status.
45
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Appendix C
Meta-analyses of multiplex network analysis per grade: bullying and status attribution Grade 2 Grade 3 Grade4 Grade 5 Est. SE Est. SE Est. SE Est. SE Bullying network Structural network effects Outdegree (density) maintenance -4.27 0.58*** -3.47 0.61*** -4.86 0.86*** -3.56 0.24*** creation -7.76 0.50*** -7.01 0.53*** -7.38 0.46*** -7.69 0.80*** Indegree attractiveness 1.19 0.12*** 0.98 0.04*** 1.19 0.07*** 1.38 0.07*** Outdegree attractiveness 0.08 0.15 -0.16 0.16 -0.11 0.18 -0.27 0.27 Outdegree activity 0.66 0.04*** 0.60 0.05*** 0.60 0.08*** 0.43 0.17** Individual effects Sex similarity 0.17 0.12 0.22 0.10* 0.50 0.11*** 0.57 0.15*** Between networks effects Status victimization -0.27 0.25 -0.24 0.14 0.04 0.07 0.01 0.10 Status bullying maintenance bullying -0.22 0.37 -0.13 0.22 -0.96 0.27*** -0.29 0.29 creation bullying 0.32 0.40 0.79 0.43 1.29 0.38** 0.88 0.24*** (continued)
46
ACCEPTED MANUSCRIPT / UNCORRECTED PROOF
Meta-analyses of multiplex network analysis per grade: bullying and status attribution (continued) Grade 2 Grade 3 Grade 4 Grade 5 Par. Est. SE Par. SE Par. Est. SE Par. SE Est. Est. Status attribution network Structural network effects Outdegree (density) maintenance -1.81 0.19*** -2.69 0.32*** -3.25 0.29*** -3.91 0.55*** creation -6.82 0.27*** -6.57 0.27*** -7.30 0.23*** -8.37 0.31*** Reciprocity 0.27 0.07*** 0.32 0.08*** 0.25 0.09*** 0.18 0.07* Transitive triplets 0.11 0.02*** 0.06 0.03* 0.07 0.02*** 0.04 0.02* Three cycles -0.11 0.05* -0.03 0.03 -0.07 0.03* -0.06 0.02* Indegree attractiveness 0.78 0.05*** 0.83 0.04*** 0.90 0.04*** 0.99 0.05*** Indegree activity 0.05 0.09 -0.01 0.05 0.06 0.04 0.06 0.03* Outdegree activity 0.44 0.03*** 0.54 0.04*** 0.62 0.05*** 0.70 0.08*** Individual effects Sex similarity 0.60 0.07*** 0.58 0.07*** 0.63 0.07*** 0.59 0.05*** Between networks effects Victimization status -0.01 0.04 -0.01 0.03 -0.02 0.03 -0.07 0.05 Bullying status maintenance status 0.22 0.13 0.18 0.15 0.13 0.16 0.02 0.15 creation status -0.21 0.06** 0.16 0.09 0.37 0.10*** 0.53 0.15*** Note. Rate of change effects were omitted from the table. Grade refers to wave 1, in waves 2 and 3 students were in grades 3-6 All effects, except for status attribution reciprocity, show significant variation over classrooms. * p < .05; ** p < .01; *** p < .001 (two-tailed tests)
47