Comparing school-related risk factors of stereotypical bullying perpetration and cyberbullying perpetration

Abstract

Based on the combination of two representative surveys of ninth graders (N = 20,150) conducted in 2013 and 2015 in Lower Saxony, Germany, this article first examines the prevalence of cyberbullying perpetration and stereotypical bullying perpetration. Second, in addition to already existing research, the prevalence of the simultaneous perpetration of stereotypical bullying and cyberbullying is presented here as well. In order to overcome the limitations of existing research, General Strain Theory (GST; Agnew, 1992) is used to theoretically determine why it is expected that school characteristics are associated with the three forms of bullying. In the students’ preceding semester, 6.1 percent of the surveyed adolescents were perpetrators of stereotypical bullying, 1.1 percent bullied someone online, and 1.3 percent of the juveniles engaged in both bullying behaviors. Binary logistic multilevel analyses show that school-related strains are in fact related to the perpetration of stereotypical bullying, cyberbullying, and the perpetration of both behaviors simultaneously. The risk of engaging in the perpetration of stereotypical bullying, and the perpetration of cyberbullying and stereotypical bullying simultaneously, is especially increased by school-related strains. Thus, those respondents who carry out bullying behaviors in the school context seem to be especially affected by school-related strains.

Keywords

Cyberbullying perpetration stereotypical bullying perpetration General Strain Theory student survey

Introduction

Following Olweus (1993), bullying in the school context is often defined as ‘the intentional, repetitive or persistent hurting of one pupil by another (or several others)’, whereby a further condition is that ‘the relationship involves an imbalance of power’ (Lucia, 2016: 50). Bullying behavior can be verbal, physical, and/or psychological (Dooley et al., 2009). In contrast to stereotypical bullying, however, cyberbullying encompasses no form of physical violence but, instead, aims at psychological aggression (Baier et al., 2016). It is frequently based on Olweus’s (1993) initial definition of stereotypical bullying as ‘a repeated aggressive act by individuals or groups, intended to inflict harm or discomfort on others by means of contemporary information and communication technologies (i.e. cellular phones, smart phones, emails, social networks, chat rooms, instant messaging programs, etc.)’ (Athanasiades et al., 2016; Patchin and Hinduja, 2012; Smith et al., 2008: 376). Although school bullying and cyberbullying are both similar aggressive behaviors (Dooley et al., 2009), there are critical differences between the two. Hence, in contrast to stereotypical bullying, cyberbullying can take place regardless of time and space and can be witnessed by a larger audience (Baier et al., 2016; Del Rey et al., 2015; Slonje and Smith, 2008).

The prevalence of bullying and cyberbullying varies substantively depending on how these behaviors are defined and measured (Athanasiades et al., 2016; Dooley et al., 2009; Kowalski et al., 2014; Patchin and Hinduja, 2015). Modecki et al. (2014) conducted a meta-analysis, taking into account 80 studies concerning bullying and cyberbullying. They reported that the mean prevalence rate of bullying was 35 percent, whereas the mean prevalence rate of cyberbullying was 16 percent. Further, higher rates were found when a clear definition of bullying was provided and when terms that indicated a rather minor deviance such as ‘fun’ were included (Modecki et al., 2014). Other studies report a range of 15–25 percent of juveniles who engage in cyberbullying (Hinduja and Patchin, 2008; Patchin and Hinduja, 2006; Ybarra et al., 2007; Ybarra and Mitchell, 2004), and a range of 7–34 percent of juveniles who engage in stereotypical bullying (Finkelhor et al., 2005; Lucia, 2016; Patchin and Hinduja, 2011).

The existing extensive body of research shows that engagement in both stereotypical bullying and cyberbullying is associated with several negative outcomes, such as: social isolation, eating disorders, deterioration of academic performance, poorer psychosocial adjustment, and engagement in other deviant behaviors (for example, Earnshaw et al., 2017; Kowalski and Limber, 2013; Nansel et al., 2004; Staubli and Killias, 2011; Ttofi et al., 2012, 2016; Zych et al., 2015). Thus, bullying and cyberbullying are still a public concern.

Previous research on risk factors for both perpetration of school bullying and cyberbullying has focused mainly on individual factors (for example, Cook et al., 2010; Festl, 2016; Lucia, 2016). Several meta-analyses have condensed the influencing factors (Cook et al., 2010; Kowalski et al., 2014; Modecki et al., 2014). With regard to gender, it is frequently reported that girls tend to engage more often in indirect stereotypical bullying, whereas boys are more likely to use physical violence (Baier et al., 2016; Dooley et al., 2009). Lucia (2016), however, reports that boys were more likely to engage in psychological bullying. With respect to cyberbullying, the evidence is more contradictory. It is often argued that girls tend to be more likely than boys to communicate using electronic means (Dooley et al., 2009) and, thus, have more opportunity to engage in cyberbullying. In fact, however, as with stereotypical bullying, Sourander et al. (2010) report a higher cyberbully prevalence for boys. Nevertheless, other studies do not find a gender difference for cyberbullying perpetration (for example, Hinduja and Patchin, 2008; Smith et al., 2008; Ybarra and Mitchell, 2004).

Beside gender, psychosocial factors are related to both forms of bullying perpetration. Based on a sample of Greek and Spanish juveniles, Del Rey et al. (2015) showed that empathy negatively predicted stereotypical and cyberbully perpetration. Similar results have been reported by other scholars (Ang and Goh, 2010; Cook et al., 2010; Kowalski et al., 2014). Another psychosocial factor that was found to be related to bullying behavior is that of self-esteem (for example, Brewer and Kerslake, 2015). Patchin and Hinduja (2010), for instance, showed that low self-esteem is related to a higher likelihood of engagement in cyberbullying. With regard to stereotypical bullying perpetration, the results are less consistent (Patchin and Hinduja, 2010). Furthermore, it has been found that a positive child–parent relationship serves as a protective factor against cyberbullying perpetration (Ybarra and Mitchell, 2004). Similar results have been reported with regard to stereotypical bullying (Cook et al., 2010; Lucia, 2016). Moreover, academic performance is negatively associated with engagement in both forms of bullying (for example, Cook et al., 2010; Kowalski et al., 2014; Ybarra and Mitchell, 2004; Zych et al., 2015). An oft-reported risk factor for cyberbullying perpetration only is online behavior. First, time-spent-online is positively related to perpetration of cyberbullying (for example, Kowalski et al., 2014; Ybarra and Mitchell, 2004). Second, experiencing cyberbullying is found to be positively related to cyberbullying perpetration (Kowalski et al., 2014; Smith et al., 2008). Several studies have reported an overlap between stereotypical bullying perpetration and cyberbullying perpetration (for example, Baldry et al., 2016; Bergmann and Baier, 2018; Dooley et al., 2009; Kowalski et al., 2014; Smith et al., 2008).

Whereas the majority of studies examining bullying perpetration in school or online have focused on individual risk factors, there is evidence that school characteristics influence the likelihood of engaging in online and offline bullying behavior. Although cyberbullying does not occur directly in the school setting, up to one-third of juveniles report that cyberbullying affects them at school (Patchin and Hinduja, 2006). This pattern can be explained by the spillover that occurs between the school and the home context: when juveniles experience strain in one context, it also affects the other context (Betts et al., 2017). Furthermore, empirical evidence shows that cyberbullying can be an extension of stereotypical bullying since perpetrators target the same victims online and offline at school (Wegge et al., 2014).

Cook et al. (2010) was able to show, based on a meta-analysis of 153 studies, that contextual factors such as peer influence and community factors (that is, characteristics of neighborhood or communities such as socioeconomic indicators or crime rates) and school climate (that is, the degree of respect and fair treatment of students by teachers; attachment to school) are related to involvement in stereotypical bully perpetration. Consistently, results based on a sample of Swiss youth aged 13–16 years indicated that, beside individual characteristics such as gender and low self-control, school climate is one of the most important variables related to stereotypical bullying perpetration (Lucia, 2016). Students who perceived a negative school climate have a 2.2 times higher risk of engaging in stereotypical behaviors (Lucia, 2016). Similarities were found with regard to cyberbullying: results of a meta-analysis by Kowalski et al. (2014) indicated that a positive school climate and a feeling of greater security at school have a small but significant protective influence on cyberbullying perpetration. Likewise, based on survey of 2215 Finnish adolescents, Sourander et al.(2010) reported that a feeling of not being safe at school was a risk factor for cyberbullying perpetration. Furthermore, the educational track seems to be related to perpetration of stereotypical bullying. In a Swiss study, Eisner et al. (2000) reported that verbal and physical bullying was lowest in the highest educational track, and highest in the lowest educational track. A correlation between the level of the educational track and deviant behavior has been frequently reported (for example, Bergmann et al., 2017). However, it remains an open question as to whether this correlation holds for cyberbullying as well. All in all, the school environment seems to influence the risk of becoming a bully in school or online.

Despite the amount of research on stereotypical bullying and cyberbullying, limitations are still prevalent. Only a few studies have captured both forms of bullying behavior simultaneously, based on representative samples. This, however, seems crucial in order to systematically analyze which school-related factors influence the risk of becoming either a stereotypical or a cyberbully, or perhaps even both. A better understanding of the influence of school-related factors would certainly provide even more useful insights into the prevention of bullying behaviors. Furthermore, most studies conducted so far focus on psychological research and tend to lack a solid theoretical foundation (for example, Kowalski et al., 2014). This is surprising since studies show that bullying and cyberbullying are associated with other delinquent behaviors. Lucia (2016), for example, shows that bullying is related to truancy. Rigby (2003) finds that juveniles who bully are more likely to appear in court on delinquency-related charges than are juveniles who do not engage in bullying. Nansel et al. (2004) report that bullies more often report substance abuse and weapon carrying. Based on a meta-analysis of longitudinal studies, Ttofi, Farrington and Lösel (2012) indicate that bullying perpetration at school is related to an underlying long-term antisocial tendency. Beside this, bullying behavior itself can be understood as a delinquent behavior. In addition, bullying, independently of whether it is online or offline, can be classified as an antisocial and undesirable behavior. Thus, bullying perpetration goes alongside other delinquent behaviors, which justifies investigating bullying behavior in the light of theories applied to delinquent behaviors.

The purpose of this study

The first aim of this study is to show the prevalence of stereotypical and cyberbullying perpetration based on a large representative sample of German ninth- graders. In addition to the already existing research, I also present statistics on the prevalence of the simultaneous perpetration of stereotypical bullying and cyberbullying. The second aim is to analyze class-related predictors of stereotypical bullying perpetration, cyberbullying perpetration, and perpetration of both bullying behaviors simultaneously. In order to overcome the limitations experienced by existing research, General Strain Theory (Agnew, 1992) is used to theoretically determine why it is predictable that school characteristics are associated with the three forms of bullying.

Theory

Building on Merton’s Initial Strain Theory (1938), Robert Agnew developed the General Strain Theory (GST) in the early 1990s. At variance with previous versions of the strain theory that proposed that strain arises from the failure to achieve culturally encouraged goals, which concerned mostly the accumulation of wealth, Agnew defines strain as a ‘relationship in which others are not treating the individual as he or she would like to be treated’ (1992: 48); for example, exclusion by peers, negative experiences at school, or victimization (Hay and Meldrum, 2010). Three forms of strain were identified by Agnew: ‘Others may (1) prevent individuals from achieving their positively valued goals . . ., (2) remove or threaten to remove positively valued stimuli that individuals possess . . ., and (3) present or threaten to present individuals with noxious or negatively valued stimuli’ (Agnew et al., 2002: 44). Agnew further argues that strain results in negative affective states such as anger or other negative emotions, which in turn may pressure adolescents into delinquency (Agnew, 1992; Agnew et al., 2002). Thus, strain does not necessarily lead to delinquent behavior. There is a relatively large body of research supporting GST, at least to some extent (for a short list of studies, see Agnew, 2001; or Patchin and Hinduja, 2011).

In the body of research on GST, the school environment has not been the focus of interest. Only a few studies have analyzed bullying victimization as a source of strain (for example, Hay and Meldrum, 2010; Moon et al., 2009). Cullen et al. (2008), for example, focused on bullying victimization as a particular type of strain and were able to show that this type of strain was related to other delinquent behaviors. However, even fewer studies have analyzed bullying perpetration as an outcome of strain (Patchin and Hinduja, 2011). This is remarkable, because the school environment is often mentioned as a potential cause of strain (Agnew, 1992, 2001; Agnew et al., 2002). As elaborated earlier, strain is defined as a negative relationship with others in which the individual is not treated in the way that he or she would like to be treated. The social setting of the school might provide conditions in which strain, seen in the form of these unpleasant relationships, could easily be experienced (Lucia, 2016). Following Patchin and Hinduja’s (2011) argument, bullying others – regardless of whether this takes place in school or on the Internet – provides the bully with a feeling of power and superiority. This might help strained juveniles to cope with their school-related strain. With regard to the setting in which the bullying takes place, the authors further stress that it is those juveniles who are not willing or able to respond to perceived school-related strain in a face-to-face setting who are more likely to use cyberbullying in order to feel better.

Hypotheses

Most studies include in their analyses a composite measure of school climate or school safety. This does not allow for analyses of factors that might affect the school environment in a way that causes strain. Based on the three main types of strain already discussed above, the following school-related characteristics have been determined to be potential strains: academic performance; low social integration; the extent of perceived violence in school; a bad teacher–student relationship; being a victim of stereotypical bullying; the perceived fairness of teachers; the level of social control.

Academic performance

The rationale behind this association is that juveniles tend to compare themselves with their relevant reference group. Furthermore, achieving good grades can certainly be described as a positively valued goal. An academic performance that is lower than the average academic performance of this reference group might lead to frustration and disparagement, and thus the failure to achieve a positive valued goal. This in turn might lead to a higher likelihood of engagement in aggressive behavior, such as stereotypical bullying or cyber-bullying (Ybarra and Mitchell, 2004). Thus, it is hypothesized that lower academic achievement is associated with a higher likelihood of engaging in stereotypical bullying, cyberbullying, or both forms of bullying simultaneously (H1).

Low social integration

This period of adolescence is marked by an expansion of the peer network and an increased importance of friendships, which begin to replace the parents as the primary source of social support and well-being (for example, La Greca and Harrison, 2005). As research shows, the school is a context in which juveniles are most likely to make most of their friends (Haynie, 2001). Consequently, being integrated into a peer group can be seen as a positive stimulus, especially in the age group of school pupils. Thus, not being integrated into a peer group can be seen as the removal of a positive stimulus. Agnew argues that the loss, or the anticipated loss, of a positively valued stimulus might lead to delinquency as ‘the individual tries to . . . seek revenge against those responsible for the loss’ (Agnew, 1992: 57f). A form of revenge for not being integrated into a class could be bullying those classmates who are preventing integration. It is therefore hypothesized that the lower the social integration in a class, the higher the likelihood of engaging in stereotypical bullying, cyber-bullying, or both forms of bullying simultaneously (H2).

The extent of perceived violence in school

The perceived level of violence in a school can be linked to the third type of strain: the presentation of a negative stimulus, such as physical or psychological abuse (Patchin and Hinduja, 2011). It can be assumed that attending a school in which the overall level of violence is rather high might cause strain for a student, even if this student is not a victim of violence him/herself. Thus, it is hypothesized that the higher the perceived overall level of violence in a school, the higher the likelihood of engaging in stereotypical bullying, cyberbullying, or both forms of bullying simultaneously (H3).

Bad teacher–student relationship

Being treated by a teacher in an unpleasant way can likewise be associated with the third type of strain: the presentation of a negative stimulus. Thus, it is hypothesized that the more often degrading teacher interactions are experienced, the higher the likelihood of engaging in stereotypical bullying, cyberbullying, or both forms of bullying simultaneously (H4)

Victim of stereotypical bullying

Agnew (2002) states that experiencing physical or psychological abuse can be characterized as the presentation of a negative stimulus. In this light, studies show that there is indeed an association between the perpetration of bullying and being victimized by it, both online and offline (for example, Hollfeld and Mishna, 2018; Lucia, 2009; Olweus, 1993; Staubli and Killias, 2011). Although the direction of causation is unclear, one explanation for this association might be found to lie in inadequate coping strategies (Staubli and Killias, 2011). Victims of bullying might bully others so as to vent their frustration at having been bullied themselves. Becoming a victim of stereotypical bullying at school might be seen as a negative stimulus, leading to frustration and anger, which in turn leads to engagement in similar bullying behavior. Therefore, it is hypothesized that being a victim of aggression in school is associated with a higher likelihood of engaging in stereotypical bullying, cyberbullying, or both forms of bullying simultaneously (H5).

According to Agnew (2001), strains that are perceived as unjust, chronic, or severe are associated with a low level of social control and offer an incentive to engage in a delinquent way of coping with the problems (Agnew, 2001). Based on this specification of strain, two more hypotheses have been developed.

Perceived fairness of teachers

The lower the perceived fairness of teachers, the higher the likelihood of engaging in stereotypical bullying, cyberbullying, or both forms of bullying simultaneously (H6).

Level of social control

The lower the level of social control in the class, the higher the likelihood of engaging in stereotypical bullying, cyberbullying, or both forms of bullying simultaneously (H7).

Method

Sample and procedure

The analyses are based on the combination of two large-scale representative, self-reported school surveys, conducted in 2013 and 2015 in the German state of Lower Saxony. The school surveys were designed as repeated cross-sectional studies in which pupils in the ninth grade of every school type were randomly selected (exception: special schools with focal points other than learning). In total, 1311 classes with 30,298 students were randomly selected. Of the classes selected, 281 classes did not participate. Thus, the surveys were administered to 1030 classes with 20,150 students. This corresponds to a response rate of 66.5 percent. The surveys lasted 90 minutes each and were conducted in the students’ classrooms using paper-and-pencil questionnaires, usually in the presence of a teacher. The surveys were approved by the state education authority. The students’ parents were sent a one-page information sheet about the survey and were required to give their consent for their children to participate. At the beginning of the survey, the students were expressly reminded of the anonymity and voluntary nature of their responses. Accordingly, all students and parents gave their informed consent for inclusion before they participated in the study. The survey topics included, among others, self-reported delinquent behavior, including bullying perpetration and victimization online and in school, individual attitudes, and family characteristics (Bergmann et al., 2017).

Selected socio-demographic characteristics of the sample are: one in two pupils are male (50.4 percent), the average age is 14.89 years (minimum 13 years, maximum 19 years), and about one-quarter of the respondents have a migration background (24.2 percent). The largest migrant group consists of young people from countries of the former Soviet Union (6.8 percent), and the second-largest migrant group are Turkish respondents (4.4 percent); 10.8 percent of respondents are growing up in families receiving some form of public support. Almost one-third of the respondents are not living with both biological parents (29.9 percent).

Measures

Dependent variables

Three dependent variables were constructed: (1) stereotypical bullying perpetration, (2) cyberbullying perpetration, and (3) perpetration of both forms of bullying simultaneously.

(1) Engagement in stereotypical bullying was assessed by asking the students how often they engaged in one of the following three behaviors during the last school term: teased other students or said nasty things about them, asked friends to stop doing anything else with another student, deliberately treated a fellow student as if he or she weren’t there. Students were asked to identify the frequency of such occurrences on a scale of ‘1–never’ to ‘6–several times a week’. Since bullying behavior is characterized by the aspect of repetition (Solberg and Olweus, 2003), students were classified as perpetrators of relational stereotypical bullying if they indicated that they had engaged in one of the behaviors at least ‘4–several times a month’. Thus, stereotypical bullies were those who admitted to engaging in such behaviors ‘several times a month’, ‘once a week’, or ‘several times a week’. The decision to focus on only relational school bullying is based on two reasons. First, other studies have shown that this type of bullying behavior was adopted most often (see Lucia, 2016). Second, cyberbullying is likewise considered to be relational bullying (Bergmann and Baier, 2018). Thus, relational bullying in the school context and cyberbullying are more comparable.

(2) Engagement in cyberbullying was assessed by asking the students how often they committed one of the following behaviors during the last school term: ridiculed, insulted, abused, or threatened others online; spread rumors about other people or dissed them online; posted others’ private messages, confidential information, photos, or videos online in order to out them or ridicule them; excluded others from an online group. Again, students were asked to identify the frequency of such occurrences on a scale of ‘1–never’ to ‘6–several times a week’. Cyberbullying offenders are those respondents who have carried out at least one of these actions at least several times a month. For these two dependent variables, a respondent is classified as a perpetrator of either cyberbullying or stereotypical bullying only if he or she has performed only one of the two behaviors.

If a respondent indicated that he or she had committed both cyberbullying and stereotypical bullying behaviors at least several times a month, he or she is categorized as (3) a perpetrator of cyberbullying and stereotypical bullying simultaneously. Thus, all three dependent variables have a dichotomous nature. Accordingly, a ‘1’ is assigned to those who indicated that they (1) engaged only in stereotypical bullying; (2) engaged only in cyberbullying; and (3) engaged in both behaviors simultaneously.

Independent variables

Agnew proposed building a combined strain scale, since it is not the exact type of strain but rather the accumulated effect of all strains that is relevant (Agnew, 1992: 62–3). However, in order to gain a detailed understanding of which class characteristics might be most problematic, it seems logical to analyze the characteristics separately.

In order to record the respondents’ academic performance, they were asked about the last report card grades they had received in the subjects of mathematics, German, history, and biology (on the usual German grading scale where 1 is excellent and 6 is failing). The mean for all four subjects was 3.01 (Cronbach’s alpha = .73). A high value on this scale represents a low academic performance.

Social integration was captured by presenting the respondents with the following two statements: ‘I am popular with my classmates’ and ‘I have many friends in school’. The respondents had to indicate how strongly they agree or disagree with the statements on a scale ranging from ‘1 – does not agree at all’ to ‘4 – strongly agree’. The overall mean score of 3.03 indicates that the respondents rated their social integration in school as rather good. The reliability of the scale was acceptable, with a Cronbach’s alpha of .69.

The perceived level of violence at school was likewise assessed using two items: ‘There is a lot of violence at my school’ and ‘At my school there are often arguments and aggravations among the students’. Again, the respondents had to indicate how strongly they agree or disagree with the statements on a scale ranging from ‘1 – does not agree at all’ to ‘4 – strongly agree’. The overall mean of the scale is 2.06. Accordingly, respondents were more likely to disagree with the statements (Cronbach’s alpha = .71).

Teacher–student relationship was measured again using two items: ‘The teachers treat us fairly and with respect’ and ‘Teachers speak openly about problems’. The response categories ranged from ‘1 – does not agree at all’ to ‘4 – strongly agree’. The overall mean score of the scale is 2.81; thus respondents rated the teacher–student relationship as relatively good (Cronbach’s alpha = .66).

The dummy variable victim of stereotypical bullying is based on three behaviors: ‘Other students teased me or said nasty things about me’, ‘I was excluded from joint ventures because other students wanted me to be excluded’, ‘Other students treated me like air and purposely ignored me’. Students were asked to identify the frequency of such occurrences on a scale of ‘1–never’ to ‘6–several times a week’. Since bullying behavior is characterized by the aspect of repetition (Solberg and Olweus, 2003), students were classified as victims of stereotypical bullying if they indicated that they had experienced at least one of the behaviors at least ‘4–several times a month’. Accordingly, 7.8 percent of the respondents had experienced at least one of these forms of aggression in the previous semester and can be classified as victims of stereotypical bullying.

In order to assess the perceived fairness of teachers, students were asked to indicate how fair the teachers of German, mathematics, biology, sport and history are to the pupils (on the usual German grading scale where 1 is excellent and 6 is failing). From these assessments, a mean value was calculated across all teachers, which reflects the average assessments of the teachers who teach a student. A high value on this scale represents a low perceived fairness of teachers. The mean of the scale is 2.43 (Cronbach’s alpha = .71).

Level of social control is measured by two items: ‘The teachers intervene when violence occurs among students’ and ‘Teachers prefer to look away when there are fights between pupils’. The response categories ranged from ‘1 – does not agree at all’ to ‘4 – strongly agree’. The latter item was recoded in a way that higher values indicate a higher level of social control by teachers. The overall mean value of the scale is 3.39; thus respondents rate the level of social control by teachers rather high (Cronbach’s alpha = .58).

Control variables

At the class level, the educational level of the school is controlled for. Owing to the structure of the school system in Lower Saxony, the schools were classified as either low-level education (Hauptschulen), medium-level education (Gesamtschulen, integrierte Haupt- und Realschulen, Realschulen), or high-level education (Gymnasium). On the individual level, migrant background is controlled for. Students were classified as having a migration background if at least one parent, or the student him/herself, was born in a country other than Germany. Further, the survey year is controlled for.

Descriptive statistics of the dependent and independent variables are shown in Table 1.

Table 1.

Descriptive statistics of the dependent and independent variables.

Variable	N	Mean/Percent	Std. dev.	Range
Dependent variables
Stereotypical bullying perpetration	19,204	6.1		0–1
Cyberbullying perpetration	19,204	1.1		0–1
Stereotypical and cyberbullying perpetration	19,204	1.3		0–1
Independent variables
Academic performance	19,695	3.02	0.69	1–6
Social integration	19,916	3.02	0.66	1–4
Perceived level of violence at school	19,852	2.06	0.66	1–4
Teacher–student relationship	19,523	2.81	0.65	1–4
Perceived fairness of teachers	19,169	2.43	0.88	1–6
Level of social control of teachers	19,690	3.39	0.63	1–4
Stereotypical bullying victim	19,918	7.80		0–1
Control variables – school level
School level – low	20,150	9.4		0–1
School level – medium	20,150	57.0		0–1
School level – high	20,150	33.6		0–1
Control variables – individual level
Migrant background	19,688	24.2		0–1
Survey year (1 = 2015)	20,150	52.8		0–1

Results

Table 1 presents the descriptive statistics of the three dependent variables of exclusively stereotypical bullying, perpetrator of exclusively cyberbullying, and perpetrator of both behaviors simultaneously. It can be seen that stereotypical bullying was executed most often: 6.1 percent of the respondents (N = 1183) indicated that they had engaged in stereotypical bullying behaviors during the last school term, whereas only 1.1 percent of the respondents (N = 216) admitted that they had engaged in cyberbullying behavior, and 1.3 percent of the respondents (N = 253) admitted to having engaged in both bullying behaviors simultaneously. For all three behaviors, significant gender differences are found. Boys are more often the perpetrator of stereotypical bullying (boys: 7.5 percent; girls: 4.7 percent; χ²(1, N = 19,223) = 66.00, p < .001), cyberbullying (boys: 1.5 percent; girls: 0.8 percent; χ²(1, N = 19,223) = 18.90, p < .001), and both bullying behaviors simultaneously (boys: 1.7 percent, girls: 0.9 percent; χ²(1, N = 19,223) = 22.81, p < .001).

In Table 2, the results of the multivariate analyses are shown. Owing to the dichotomous nature of the three dependent variables (engaged in the corresponding behavior at least several times per month: no vs. yes) and the clustered data structure, binary logistic multilevel regressions were run. In order to analyze potential differences between offenders of exclusively stereotypical bullying, offenders of exclusively cyberbullying, and offenders of both behaviors simultaneously, separate models for each behavior have been fitted. In addition, since existing research reports contradictory evidence regarding gender differences in bullying perpetration, separate models for boys and girls were calculated for each dependent variable. Owing to listwise deletion, 18,179 students in 963 classes were included in the models. However, in each model, only those respondents were included who either did not engage in any bullying behavior, or who exercised the corresponding bullying behavior only and were not perpetrators of the other bullying behavior. This procedure prevents potential underlying factors that generally differentiate between perpetrators of bullying and non-perpetrators, which could be covered up.

Table 2.

Correlates of the prevalence of stereotypical bullying perpetration, cyberbullying perpetration, and simultaneous perpetration of stereotypical and cyberbullying: Binary logistic multilevel regressions (shown: unstandardized coefficients).

Dependent variable		Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
		Stereotypical bullying	Stereotypical bullying	Cyberbullying	Cyber bullying	Stereotypical & cyberbullying	Stereotypical & cyberbullying
		Girls	Boys	Girls	Boys	Girls	Boys
H1	Academic performance (z)	1.010	0.934	1.461	1.585^**	1.976^***	1.720^***
H1	Academic performance (z)	(0.13)	(−1.05)	(1.92)	(3.28)	(3.60)	(4.02)
H2	Social integration (z)	1.232^**	1.407^***	1.009	1.294	2.730^***	3.116^***
H2	Social integration (z)	(2.61)	(4.77)	(0.05)	(1.69)	(5.18)	(6.99)
H3	Perceived level of violence at school (z)	1.409^***	1.466^***	1.616^*	1.128	2.139^***	1.644^***
H3	Perceived level of violence at school (z)	(4.02)	(5.68)	(2.37)	(0.85)	(4.17)	(3.80)
H4	Teacher-student relationship (z)	0.757^**	0.707^***	0.822	0.623^**	0.825	0.500^***
H4	Teacher-student relationship (z)	(−3.01)	(−4.94)	(−0.91)	(−3.25)	(−0.96)	(−5.05)
H5	Traditional bullying victim	4.259^***	5.587^***	2.030^*	3.120^***	5.907^***	9.593^***
H5	Traditional bullying victim	(10.21)	(13.64)	(1.96)	(3.91)	(5.78)	(9.24)
H6	Perceived fairness of teachers (z)	1.123	1.143^**	1.398^*	0.979	1.337^*	1.120
H6	Perceived fairness of teachers (z)	(1.70)	(2.71)	(2.23)	(−0.20)	(2.07)	(1.19)
H7	Level of social control (z)	0.800^**	1.024	1.023	0.847	0.994	1.044
H7	Level of social control (z)	(−2.59)	(0.35)	(0.11)	(−1.23)	(−0.03)	(0.33)
Control variables – individual level
Migrant background		1.062	1.203	1.377	2.146^***	0.895	2.016^***
Migrant background		(0.51)	(1.89)	(1.22)	(4.15)	(−0.42)	(3.84)
Survey year 2015		0.620^***	0.654^***	1.167	1.281	0.572^*	1.192
Survey year 2015		(−4.13)	(−4.64)	(0.59)	(1.38)	(−2.24)	(0.96)
Control variables – class level
Educational level – low		(reference)	(reference)	(reference)	(reference)	(reference)	(reference)
Educational level -–medium		1.156	0.846	1.276	0.599	0.614	0.802
Educational level -–medium		(0.59)	(−0.99)	(0.43)	(−1.71)	(−1.26)	(−0.67)
Educational level – high		1.158	0.895	1.531	0.848	0.518	1.056
Educational level – high		(0.56)	(−0.60)	(0.70)	(−0.51)	(−1.43)	(0.15)
N pupils		9030	8704	8663	8147	8676	8172
N classes		960	959	960	959	960	959
R² (McKelvey & Zavoina’s)		.101	.114	.099	.139	.244	.252

Notes: Exponentiated coefficients; t statistics in parentheses; (z) – variables were centered at the grand mean.

p < .05, **p < .01, ***p < .001

All independent and control variables were entered into the regression analyses simultaneously for each model. As hypothesized, the failure to achieve a positive goal (H1), operationalized as academic behavior, has a risk-increasing effect. The variable academic performance is coded according to the German grading scale. Higher values represent a poorer performance. However, this holds true only for cyberbullying perpetration by boys (Model 4) and perpetration of stereotypical and cyberbullying behavior by both genders (Models 5 and 6). Thus, hypothesis 1 is partially supported.

Hypothesis 2 regarding the removal of a positive stimulus, however, is not empirically supported. Instead of the hypothesized negative effect of social integration on the likelihood of engaging in the different forms of bullying, a positive effect is found for stereotypical bullying perpetration (Models 1 and 2), and perpetration of stereotypical and cyberbullying (Models 3 and 4) for both genders.

Hypotheses 3, 4, and 5 postulate a risk-enhancing effect of the presentation of a negative stimulus on the probability of engaging in bullying perpetration. Hypothesis 3 operationalizes the presentation of a negative stimulus as the perceived level of violence at school. As hypothesized, a negative effect was found for the perceived level of violence on the likelihood of engaging in stereotypical bullying, cyberbullying, and both behaviors simultaneously. Thus, the higher the level of violence at school is perceived, the higher is the likelihood of engaging in one of the three distinct bullying behaviors. This applies to all genders, except for male perpetrators of cyberbullying (Model 4). As assumed in hypothesis 4, the teacher–student relationship is negatively associated with the probability of committing stereotypical bullying, cyberbullying, and both behaviors at the same time. However, this does not apply to female perpetrators of cyberbullying (Model 3), or to female perpetrators of both behaviors (Model 5). Thus, hypotheses 3 and 4 are partially supported. Hypothesis 5 operationalizes the presentation of a negative stimulus as experiencing stereotypical bullying in school. This hypothesis is empirically supported in all models. Thus, as expected, being a victim of stereotypical bullying increases the risk of engaging in stereotypical bullying, in cyberbullying, and in both behaviors simultaneously.

Hypothesis 6 is only partially empirically supported. The expected negative effect of the perceived fairness of teachers is found only for male perpetrators of stereotypical bullying, female perpetrators of cyberbullying, and female perpetrators of both behaviors (Models 2, 3, and 5). The variable perceived fairness of teachers is coded according to the German grading scale. Thus, higher values represent a lower level of fairness.

The least empirical support was found for hypothesis 7 regarding the level of social control. The level of social control is negatively related to the likelihood of engaging in stereotypical bullying only for girls (Model 1). For all other models, no significant effects are found for the level of social control. Since the Cronbach’s alpha for the scale was relatively low – even when it is kept in mind that Cronbach’s alpha is affected by the number of items (Cortina, 1993) – we also included both variables on which the scale is based separately in all models (not shown in Table 2). There were no significant effects for the item ‘The teachers intervene when violence occurs among students’. For the item ‘Teachers prefer to look away when there are fights between pupils’, which was recoded so that high values represent a high level of social control, the same effect as for the scale is found: a negative effect only in Model 1.

Regarding the control variables, mixed results were found. Male respondents with a migrant background are more likely to engage in cyberbullying and in stereotypical and cyberbullying simultaneously. Male and female respondents who were interviewed in 2015 were less likely to engage in stereotypical bullying. The same holds true for female respondents with regard to the perpetration of cyberbullying and stereotypical bullying simultaneously. No significant effects were found with respect to the control variables on class level. Thus, the educational level of the school has no effect on the likelihood of engaging in bullying behavior.

The amount of explained variance differs among the models. Models 5 and 6 explain 24.4 percent (Model 5, girls) and 25.2 percent (Model 6, boys) of the variance in perpetration of stereotypical and cyberbullying. In contrast, Models 1 and 2 explain only 10.1 percent (Model 1, girls) and 11.4 percent (Model 2, boys) of the variance in stereotypical bullying perpetration. Models 3 and 4 explain 9.9 percent (Model 3, girls) and 13.9 percent (Model 4, boys) of the variance in cyberbullying perpetration.

Discussion

Based on two combined large-scale representative school surveys from Lower Saxony, this study compared school-related risk factors for stereotypical bullying perpetration and cyberbullying perpetration. The design of the survey permits differentiation between those juveniles who engage exclusively in stereotypical bullying, those who engage exclusively in cyberbullying, and those who engage in both forms of bullying simultaneously. So far, the latter group has been neglected by the current research. The results of the study show that exclusively stereotypical bullying perpetration (6.1 percent) is more common than exclusively cyberbullying perpetration (1.1 percent) – a finding that had been reported previously (for example, Kowalski et al., 2014: 1110; Modecki et al., 2014). Furthermore, the prevalence of simultaneous perpetration of cyberbullying and stereotypical bullying (1.3 percent) is approximately comparable to the prevalence of the perpetration of cyberbullying exclusively. These prevalence rates are lower than the prevalence rates reported in other studies (for example, Finkelhor et al., 2005; Hinduja and Patchin, 2008; Lucia, 2016; Modecki et al., 2014; Patchin and Hinduja, 2011). This can be partially explained by the relatively short time frame (‘in the last school term’) that is used in the operationalization of bullying behavior in this study. Most studies show lifetime prevalence, which is naturally higher than prevalence rates based on a shorter period. Furthermore, differences in prevalence rates could be due to different operationalization and wording (Modecki et al., 2014). Nevertheless, the prevalence rates, which are based on representative data, show that bullying perpetration is a substantial problem in Germany, both at school and online.

Regarding gender differences in the prevalence of bullying perpetration, contradictory findings have been reported, especially with respect to cyberbullying perpetration (for example, Bergmann and Baier, 2018; Dooley et al., 2009; Lucia, 2016; Smith et al., 2008). In the present study, it was found that boys are more often perpetrators of bullying than girls, regardless of the contexts in which the perpetrations occur. Similar results have been previously reported, for example by Lucia (2016) and Sourander et al. (2010). These findings contradict the argument that girls are more likely to engage in indirect verbal bullying and that boys are more likely to engage in physical violence (Dooley et al., 2009).

In order to theoretically determine school-related risk factors for bullying perpetration, General Strain Theory (Agnew, 1992) was applied. As a general theory, it can be expected that GST will predict involvement in various types of delinquent behavior (Mazerolle et al., 2000). Therefore, the theory is applicable to engaging in bullying behavior at school and online. The multivariate analyses show that GST does indeed provide a useful framework to analyze school-related risk factors of bullying perpetration. This is an addition to previous research, and it shows that bullying belongs in the field of criminology because it is related to other deviant behaviors (for example, Lucia, 2016; Ttofi et al., 2012) and is itself a deviant behavior. The fact that GST is well applicable shows that this allocation also makes sense from a theoretical perspective and provides a fruitful framework to further analyze bullying behaviors offline and online in general.

Agnew (1992) describes three different types of strain that could eventually lead to delinquent behavior: the failure to achieve a positive valued goal, the removal of a positive stimulus, and, lastly, the presentation of a negative stimulus. According to these strains, hypotheses were derived. The failure to achieve a positive valued goal was operationalized by a low academic performance (H1). Academic performance is negatively associated with bullying perpetration. Thus, as expected, a lower academic performance increases the risk of engaging in bullying perpetration. However, significant effects were found only for cyberbullying perpetration by boys and the simultaneous perpetration of both forms of bullying by both genders. No effect was found for male or female perpetrators of stereotypical bullying.

The removal of a positive stimulus was operationalized as low social integration in the school classroom (H2). However, no empirical support was found with respect to low social integration. Instead, higher social integration is associated with a higher risk of engaging in stereotypical bullying, and for the perpetration of both behaviors simultaneously (for both genders). These findings are in line with the research on the effect of popularity on aggressive behavior (for example, Cillessen and Rose, 2005; Kreager, 2007; Weerman and Bijleveld, 2007), where it has been reported that minor deviant behavior, such relational violence, has a status-enhancing effect (Weerman and Bijleveld, 2007). Here it is argued that relational violence can be used in order to demonstrate power and to establish a central position in the social system (Estell et al., 2009; Wegge et al., 2016). Low social integration might be too much of an indirect measurement of the removal of a positive stimulus. Future research should focus on more direct measures, such as being suspended from school (Agnew, 1992).

The presentation of a negative stimulus was operationalized as the extent of perceived violence at school (H3), a bad teacher–student relationship (H4), and being a victim of stereotypical bullying (H5). These hypotheses were empirically supported, thus, as was assumed, it was found that the presentation of a negative stimulus in the school context is associated with a higher risk of engaging in bullying behaviors. However, with regard to the perceived level of violence at school, no significant effect was found for male perpetrators of cyberbullying. A possible explanation could be that those boys who are stressed by a high level of violence at their school are more likely to engage in bullying behaviors in the direct context in which the strain is experienced. Thus, those who engage in exclusively cyberbullying might not be stressed by a higher level of violence at school. Furthermore, the teacher–student relationship had no effect for female cyberbullying perpetrators and female perpetrators of both behaviors.

Only weak empirical support was found for the specification of strains as perceived as unjust or as characterized by a low level of social control (H6 and H7). This might be due to the indirect measurement of both specifications. Since the surveys captured a wide array of topics, it was not possible to assess an evaluation of each strain tested in the models. Thus, future research should include better measurements in order to properly test the assumptions of GST.

Regarding the gender differences, which can be found with respect to the correlates of the different behaviors, no clear pattern is discernible. Based on the contradictory findings of previous research, no hypotheses were derived regarding potential gender differences. However, it seems that girls and boys differ in reference to school-related risk factors for engaging in bullying behavior. Future research should therefore address gender differences in individual and contextual risk and protective factors for bullying perpetration and theoretically derive explanations for those differences.

To sum up, it can be stated that school-related strains are related to the perpetration of stereotypical bullying, of cyberbullying, and of both behaviors simultaneously. Especially in reference to the risk of engaging in the perpetration of stereotypical bullying and the risk of engaging in the perpetration of cyberbullying and stereotypical bullying simultaneously is increased by school-related strains. Thus, those respondents who carry out bullying behavior in the school context seem to be especially affected by school-related strains. Some scholars suggest that cyberbullying and stereotypical bullying reflect different means of engaging in a similar behavior (for example, Modecki et al., 2014). The findings of the present study regarding the perpetration of stereotypical bullying, and the perpetration of cyber- and stereotypical bullying simultaneously, are in agreement with that notion. However, the findings regarding those juveniles who engage only in cyberbullying indicate that there are indeed differences between engaging in exclusively cyberbullying or stereotypical bullying, at least when school-related correlates are considered. This differs from the argument of Patchin and Hinduja (2011), who maintained that those juveniles who would not be willing, or able, to respond to perceived school-related strains in a face-to-face setting might be more likely to use cyberbullying in order to make themselves feel better. Instead, school-related strains seem to be less important to those who engage only in cyberbullying.

Finally, the limitations of the present study have to be addressed. First, due to the data, it was not possible to include measurements for anger or frustration in the models; these are critical aspects of GST because negative affective states link exposure to strain with delinquency (Agnew, 1992; Mazerolle et al., 2000). However, this aspect of GST in particular is only partially supported empirically, especially when it comes to non-violent delinquency (for example, Mazerolle et al., 2000; Moon et al., 2009). Nevertheless, for the most part, the results of the present study provide support for a central hypothesis of GST, namely that exposure to strain is associated with delinquency. Second, another limitation of the study is related to the instruments used. Third, as stated, the survey was not primarily about stereotypical bullying perpetration and cyberbullying perpetration. This is reflected in the operationalization of the stereotypical bullying and cyberbullying, which do not cover the aspect of the imbalance of power –one of the defining aspects of bullying. However, with regard to cyberbullying, not capturing the imbalance of power might be less serious because power manifests itself differently online. With the given technological possibilities that facilitate cyberbullying, any perpetrator is at least in that moment of posting, for example, embarrassing content online in a position of power relative to the victim of the attack (Patchin and Hinduja, 2015).

To sum up, the present article adds to the literature in the following ways. Firstly, based on representative data it is possible to differentiate between those juveniles who engage only in either stereotypical bullying or cyberbullying and those who engage in both behaviors simultaneously. The latter group has been mostly neglected so far. The results show that there are indeed differences in the risk factors for juveniles who engage only in cyberbullying compared with those engaging in bullying in the school context. Secondly, applying one of the most important recent criminological theorie to derive risk factors of bullying behavior strengthens the argument that bullying is indeed a criminologically relevant behavior.

Footnotes

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: I would like to thank the Lower Saxony Ministry of Science and Culture for funding the survey.

ORCID iD

Marie Christine Bergmann

References

Agnew

(1992) Foundation for a general strain theory of crime and delinquency. Criminology 30(1): 47–87.

Agnew

(2001) Building on the foundation of General Strain Theory: Specifying the types of strain most likely to lead to crime and delinquency. Journal of Research in Crime and Delinquency 38(4): 319–361.

Agnew

Brezina

Wright

, et al. (2002) Strain, personality traits, and delinquency: Extending general strain theory. Criminology 40(1): 43–71. DOI: 10.1111/j.1745-9125.2002.tb00949.x.

Ang

Goh

(2010) Cyberbullying among adolescents: The role of affective and cognitive empathy, and gender. Child Psychiatry and Human Development 41(4): 387–397. DOI: 10.1007/s10578-010-0176-3.

Athanasiades

Baldry

Kamariotis

, et al. (2016) The ‘net’ of the Internet: Risk factors for cyberbullying among secondary-school students in Greece. European Journal on Criminal Policy and Research 22(2): 301–317. DOI: 10.1007/s10610-016-9303-4.

Baier

Krenz

Bergmann

(2016) Verbreitung und Einflussfaktoren des Cyberbullying. Ergebnisse einer Repräsentativbefragung in Niedersachsen. Zeitschrift für Soziologie der Erziehung und Sozialisation 36(3): 227–245.

Baldry

Farrington

Sorrentino

(2016) Cyberbullying in youth: A pattern of disruptive behaviour. Psicología Educativa 22(1): 19–26. DOI: 10.1016/j.pse.2016.02.001.

Bergmann

Baier

(2018) Prevalence and correlates of cyberbullying perpetration. Findings from a German representative student survey. International Journal of Environmental Research and Public Health 15(2): 274. DOI: 10.3390/ijerph15020274.

Bergmann

Baier

Rehbein

, et al. (2017) Jugendliche in Niedersachsen. Ergebnisse des Niedersachsensurveys 2013 & 2015. KFN-Forsch. Hannover: Kriminologisches Forschungsintitut Niedersachsen e.V.

10.

Betts

Spenser

Gardner

(2017) Adolescents’ involvement in cyber bullying and perceptions of school: The importance of perceived peer acceptance for female adolescents. Sex Roles 77(7–8): 471–481. DOI: 10.1007/s11199-017-0742-2.

11.

Brewer

Kerslake

(2015) Cyberbullying, self-esteem, empathy and loneliness. Computers in Human Behavior 48: 255–260. DOI: 10.1016/j.chb.2015.01.073.

12.

Cillessen

AHN

Rose

(2005) Understanding popularity in the peer system. Current Directions in Psychological Science 14(2): 102–105.

13.

Cook

Williams

Guerra

, et al. (2010) Predictors of bullying and victimization in childhood and adolescence: A meta-analytic investigation. School Psychology Quarterly 25(2): 65–83. DOI: 10.1037/a0020149.

14.

Cortina

(1993) What is coefficient alpha? An examination of theory and applications. Journal of Applied Psychology 78(1): 98–104. DOI: 10.1037/0021-09010.78.1.98.

15.

Cullen

Unnever

Hartman

, et al. (2008) Gender, bullying victimization, and juvenile delinquency: A test of General Strain Theory. Victims & Offenders 3(4): 346–364. DOI: 10.1080/15564880802338468.

16.

Del Rey

Lazuras

Casas

, et al. (2015) Does empathy predict (cyber) bullying perpetration, and how do age, gender and nationality affect this relationship? Learning and Individual Differences 45: 275–281. DOI: 10.1016/j.lindif.2015.11.021.

17.

Dooley

Pyżalski

Cross

(2009) Cyberbullying versus face-to-face bullying. Zeitschrift für Psychologie / Journal of Psychology 217(4): 182–188. DOI: 10.1027/0044-3409.217.4.182.

18.

Earnshaw

Elliott

Reisner

, et al. (2017) Peer victimization, depressive symptoms, and substance use: A longitudinal analysis. Pediatrics 139(6): e20163426.

19.

Eisner

Manzoni

Ribeaud

(2000) Opfererfahrungen und selbst berichtete Gewalt bei Schülerinnen und Schülern im Kanton Zürich. Aurau: Sauerländer.

20.

Estell

Farmer

Irvin

, et al. (2009) Students with exceptionalities and the peer group context of bullying and victimization in late elementary school. Journal of Child and Family Studies 18(2): 136–150. DOI: 10.1007/s10826-008-9214-1.

21.

Festl

(2016) Perpetrators on the internet: Analyzing individual and structural explanation factors of cyberbullying in school context. Computers in Human Behavior 59: 237–248. DOI: 10.1016/j.chb.2016.02.017.

22.

Finkelhor

Ormrod

Turner

, et al. (2005) The victimization of children and youth: A comprehensive, national survey. Child Maltreatment 10(1): 5–25. DOI: 0.1177/1077559504271287.

23.

Hay

Meldrum

(2010) Bullying victimization and adolescent self-harm: Testing hypotheses from general strain theory. Journal of Youth and Adolescence 39(5): 446–459. DOI: 10.1007/s10964-009-9502-0.

24.

Haynie

(2001) Delinquent peers revisited: Does network structure matter? American Journal of Sociology 106(4): 1013–1057. DOI: 10.1086/320298.

25.

Hinduja

Patchin

(2008) Cyberbullying: An exploratory analysis of factors related to offending and victimization. Deviant Behavior 29(2): 129–156. DOI: 10.1080/01639620701457816.

26.

Hollfeld

Mishna

(2018) Longitudinal associations in youth involvement as victimized, bullying, or witnessing cyberbullying. Cyberpsychology, Behavior, and Social Networking 21(4): 234–239. DOI: 10.1089/cyber.2017.0369.

27.

Kowalski

Limber

(2013) Psychological, physical, and academic correlates of cyberbullying and traditional bullying. Journal of Adolescent Health 53(1 Suppl): S13–S20. DOI: 10.1016/j.jadohealth.2012.09.018.

28.

Kowalski

Giumetti

Schroeder

, et al. (2014) Bullying in the digital age: A critical review and meta-analysis of cyberbullying research among youth. Psychological Bulletin 140(4): 1073–1137. DOI: 10.1037/a0035618.

29.

Kreager

(2007) When it’s good to be ‘bad’: Violence and adolescent peer acceptance. Criminology 45(4): 893–923. DOI: 10.1111/j.1745-9125.2007.00097.x.

30.

La Greca

Harrison

(2005) Adolescent peer relations, friendships, and romantic relationships: Do they predict social anxiety and depression? Journal of Clinical Child and Adolescent Psychology 34(1): 49–61. DOI: 10.1207/s15374424jccp3401_5.

31.

Lucia

(2009) Multi-dimensional approach to bullying. PhD thesis, Lausanne.

32.

Lucia

(2016) Correlates of bullying in Switzerland. European Journal of Criminology 13(1): 50–66. DOI: 10.1177/1477370815608880.

33.

Mazerolle

Burton

Cullen

, et al. (2000) Strain, anger, and delinquent adaptations specifying general strain theory. Journal of Criminal Justice 28(2): 89–101. DOI: 10.1016/S0047-2352(99)00041-0.

34.

Merton

(1938) Social structure and anomie. American Sociological Review 3: 672–682. DOI: 10.2307/2084686.

35.

Modecki

Minchin

Harbaugh

, et al. (2014) Bullying prevalence across contexts: A meta-analysis measuring cyber and traditional bullying. Journal of Adolescent Health 55(5): 602–611. DOI: 10.1016/j.jadohealth.2014.06.007.

36.

Moon

Hays

Blurton

(2009) General strain theory, key strains, and deviance. Journal of Criminal Justice 37(1): 98–106. DOI: 10.1016/j.jcrimjus.2008.12.011.

37.

Moon

Morash

McCluskey

Hwang

(2009) A comprehensive test of general strain theory. Journal of Research in Crime and Delinquency 46(2): 182–212.

38.

Nansel

Craig

Overpeck

, et al. (2004) Cross-national consistency in the relationship between bullying behaviors and psychosocial adjustment. Archives of pediatrics & adolescent medicine 158(8): 730–736. DOI: 10.1016/S0084-3970(08)70012-4.

39.

Olweus

(1993) Bullying at School. Oxford: Blackwell.

40.

Patchin

Hinduja

(2006) Bullies move beyond the schoolyard: A preliminary look at cyberbullying. Youth Violence and Juvenile Justice 4(2): 148–169. DOI: 10.1177/1541204006286288.

41.

Patchin

Hinduja

(2010) Cyberbullying and self-esteem. Journal of School Health 80(12): 614–621. DOI: 10.1111/j.1746-1561.2010.00548.x.

42.

Patchin

Hinduja

(2011) Traditional and nontraditional bullying among youth: A test of general strain theory. Youth & Society 43(2): 727–751. DOI: 10.1177/0044118X10366951.

43.

Patchin

Hinduja

(2012) Cyberbullying. An update and synthesis of the research. In: Patchin

Hinduja

(eds) Cyberbullying and Prevention and Response. New York and London: Routledge, 13–35.

44.

Patchin

Hinduja

(2015) Measuring cyberbullying: Implications for research. Aggression and Violent Behavior 23(May): 69–74. DOI: 10.1016/j.avb.2015.05.013.

45.

Rigby

(2003) Consequences of bullying in schools. Canadian Journal of Psychiatry. Revue canadienne de psychiatrie 48(9): 583–90.

46.

Slonje

Smith

(2008) Cyberbullying: Another main type of bullying? Personality and Social Sciences. Scandinavian Journal of Psychology 49(2): 147–154. DOI: 10.1111/j.1467-9450.2007.00611.x.

47.

Smith

Mahdavi

Carvalho

, et al. (2008) Cyberbullying: Its nature and impact in secondary school pupils. Journal of Child Psychology and Psychiatry and Allied Disciplines 49(4): 376–385. DOI: 10.1111/j.1469-7610.2007.01846.x.

48.

Solberg

Olweus

(2003) Prevalence estimation of school bullying with the Olweus Bully/Victim Questionnaire. Aggressive Behavior 29(3): 239–268. DOI: 10.1002/ab.10047.

49.

Sourander

Brunstein Klomek

Ikonen

, et al. (2010) Psychosocial risk factors associated with cyberbullying among adolescents. Archives of General Psychiatry 67(7): 720. DOI: 10.1001/archgenpsychiatry.2010.79.

50.

Staubli

Killias

(2011) Long-term outcomes of passive bullying during childhood: Suicide attempts, victimization and offending. European Journal of Criminology 8(5): 377–385. DOI: 10.1177/1477370811415761.

51.

Ttofi

Farrington

Lösel

(2012) School bullying as a predictor of violence later in life: A systematic review and meta-analysis of prospective longitudinal studies. Aggression and Violent Behavior 17(5): 405–418. DOI: 10.1016/j.avb.2012.05.002.

52.

Ttofi

Farrington

Losel

, et al. (2016) School bullying and drug use later in life: A meta-analytic investigation. School Psychology Quarterly 31(1): 8–27. DOI: 10.1037/spq0000120.

53.

Weerman

Bijleveld

CCJH

(2007) Birds of different feathers: School networks of serious delinquent, minor delinquent and non-delinquent boys and girls. European Journal of Criminology 4(4): 357–383. DOI: 10.1177/1477370807080718.

54.

Wegge

Vandebosch

Eggermont

(2014) Who bullies whom online: A social network analysis of cyberbullying in a school context. Communications 39(4): 415–433. DOI: 10.1515/commun-2014-0019.

55.

Wegge

Vandebosch

Eggermont

, et al. (2016) Popularity through online harm: The longitudinal associations between cyberbullying and sociometric status in early adolescence. Journal of Early Adolescence 36(1): 86–107. DOI: 10.1177/0272431614556351.

56.

Ybarra

Mitchell

(2004) Online aggressor/targets, aggressors, and targets: A comparison of associated youth characteristics. Journal of Child Psychology and Psychiatry, and Allied Disciplines 45(7): 1308–16. DOI: 10.1111/j.1469-7610.2004.00328.x.

57.

Ybarra

Espelage

Mitchell

(2007) The co-occurrence of Internet harassment and unwanted sexual solicitation victimization and perpetration: Associations with psychosocial indicators. Journal of Adolescent Health: Official Publication of the Society for Adolescent Medicine 41(6 Suppl): S3141. DOI: 10.1016/j.jadohealth.2007.09.010.

58.

Zych

Ortega-Ruiz

Del Rey

(2015) Systematic review of theoretical studies on bullying and cyberbullying: Facts, knowledge, prevention, and intervention. Aggression and Violent Behavior 23: 1–21. DOI: 10.1016/j.avb.2015.10.001.