The Role of Intrapersonal-,Interpersonal-,Family-,and School-Level Variables in Predicting Bias-Based Cybervictimization

Abstract

This study investigated whether social position (e.g., gender, migration, family status), intrapersonal-level (e.g., online risk behaviors, motives of Internet use), interpersonal-level (e.g., victimization and bullying), family-level (e.g., parental mediation), and class-level (e.g., teachers’ mediation, ethnic diversity) variables predict bias-based cybervictimization. Self-report questionnaires were completed by 1,018 Austrian adolescents (52.3% girls), aged 12 to 17 years ( $\bar{X}$ = 13.55, SD = 0.88). The logistic part of a multilevel zero-inflated Poisson model showed that higher levels of offline victimization and a higher proportion of immigrants in classes were predictors for students reporting at least one form of bias-based cybervictimization. The Poisson part of the model showed that being a girl, higher levels of cybervictimization, lower levels of avoiding online risks, and more discussions about media use with teachers in classes were predictors for students reporting a higher number of bias-based cybervictimization. Implications for prevention are discussed.

Keywords

bias-based cybervictimization socio-ecological model stigma-based bullying online hate multilevel zero-inflated Poisson model

Bias-based cybervictimization covers offensive, mean, or threatening online actions devaluing, insulting, or harassing individuals or social groups in relation to their national origin, religion, ethnicity, gender, sexual orientation, disability, or some other characteristics and that are perpetrated through online posts, comments, text messages, videos, or pictures (Blaya, 2019; Wachs & Wright, 2018). Around 17% of adolescents, aged 12 to 17 years, reported that they have experienced cyberhate victimization at least once during their life in a recent study conducted in Germany (Wachs, Wright, Gámez-Guadix, Görzig and Schubarth, 2020; Wachs & Wright, 2019). Importantly, a four-country study conducted in the United States, Finland, Germany, and the United Kingdom showed that the number of social media users that have witnessed hateful or degrading writings or speech online during the last 3 months was even larger (Hawdon, Oksanen, & Räsänen, 2017). Because of the rapid spread of smartphones (Paus-Hasebrink, Kulterer, & Sinner, 2019), adolescents are increasingly using the Internet (Kowalski, Limber, & McCord, 2019). Therefore, studying bias-based cybervictimization in this age group is important. However, studies on this topic are largely absent. To fill this gap, the present study (a) combines the main ideas of the stigma-based ecological bullying framework (Earnshaw et al., 2018) and the socio-ecological hate speech framework (Wachs, Schubarth and Bilz, 2020), and (b) investigates whether social position (e.g., gender, migration, family status), intrapersonal-level (e.g., online risk behaviors, motives of Internet use), interpersonal-level (e.g., offline victimization), family-level (e.g., parental mediation), and class-level (e.g., ethnic diversity, teachers’ mediation) variables predict bias-based cybervictimization among adolescents.

Socio-Ecological Perspective on Bias-Based Cybervictimization

The socio-ecological framework assumes that the development of a person unfolds as result of ontogenetic characteristics and complex, interrelated interactions at the ontogenetic, micro-, meso-, exo-, macro-, and chronosystem levels (Bronfenbrenner, 1979). On the ontogenetic level, the socio-ecological model seeks to identify the biological and person-related factors influencing behavior and development. The second level consists of various microsystems surrounding the individual (e.g., family, peers, school). The third level, mesosystem, involves interactions between two or more microsystems (e.g., home and school). Surrounding this is the exo-system, which includes the social settings that affect the individual indirectly (e.g., the organization of the parents’ workplace). The macro-system influences the individual by means of culture, norms, belief systems, and material resources. Finally, the chronosystem level includes consistency or change (e.g., historical or life events) of the individual and the environment (Bronfenbrenner, 1994).

The socio-ecological framework has been applied to bullying and peer victimization in school (Hong & Espelage, 2012), to cyberbullying and cybervictimization (Gradinger & Strohmeier, 2018), and to bias-based bullying of sexual and gender minority youth (e.g., Newman & Fantus, 2015). Because ideologies that define devalued social groups are defined on different socio-ecological levels, it is important to apply a socio-ecological framework when studying bias-based victimization. This is especially true as the use of biased language when victimizing a peer (e.g., based on weight, intelligence, sexual orientation, religion, or ethnicity) was associated with higher levels of feeling sad, skipping school, avoiding school activities, or lost friends compared with non-biased victimization (Jones, Mitchell, Turner, & Ybarra, 2018). Nevertheless, the majority of studies focused on individual-level characteristics when studying bias-based victimization, which highlights the necessity for research on this topic. Thus, when applying the socio-ecological perspective to bias-based cybervictimization, the basic assumption is that victimization is a complex social phenomenon that is embedded in a number of interrelated systems, which may cause, reinforce, and maintain negative social interactions. Thus, bias-based cybervictimization is the result of the complex interplay between adolescents and their multilayered environments. Potentially relevant variables on different systemic levels have been identified in two theoretical models, the stigma-based ecological bullying framework and the socio-ecological hate speech framework (Earnshaw et al., 2018; Wachs, Schubarth and Bilz, 2020).

The stigma-based ecological bullying framework (Earnshaw et al., 2018) is a useful model to better understand bias-based cybervictimization, because it differentiates between the individual perpetrator level, the interpersonal level, the structural level, and the societal level. Although stigma-based bullying is a broader construct compared with bias-based cybervictimization, both behaviors are carried out because of stigma (or bias). Taking the perspective of the perpetrators, this model defines stigma-based bullying as a type of bullying that is motivated by stigma (or bias) and that is driven by distinct, stigma-related factors on different systemic levels. On the societal level, social stigma is constructed with the goal to define and maintain a devalued social position of certain social groups and their members. Importantly, multiple groups (e.g., gender, race/ethnicity, immigration, religion, disability, sexual orientation) that can change depending on historical time are the targets of socially constructed stigma. On the structural level, policies, institutions, and organizations create and maintain environments in which stigma-based bullying is more likely to occur. The ethnic segregation of neighborhoods and schools is one example of a mechanism on the structural level that is associated with higher levels of bullying for ethnic minority groups (Schumann, Craig, & Rosu, 2013). On the interpersonal level, teachers, parents, and peers act as important socialization agents and role models to transmit social dominance orientations, stereotypes, and prejudices. These constructs are important, because stigma-based bullying is understood as a targeted form of social dominance that create, maintain, and reinforce group hierarchies. Thus, social dominance, stereotypes, and prejudices are conceptualized as manifestations of socially constructed stigma on the individual perpetrator level.

As Blaya (2019) points out, the Internet and social media have become the privileged tools for bias-based victimization and hate speech during the last decade. The Internet not only contributes to the dissemination of hate and propaganda but also enables networking and contact of individuals sharing similar hate ideologies. The rapidly developing technological innovations that are changing the communication, networking, and participation patterns of adolescents create amplified opportunities for bias-based cybervictimization. Therefore, the socio-ecological hate speech framework (Wachs et al., 2020) integrated some ideas of the media effects model (Chen, Ho, & Lwin, 2017) and the stigma-based ecological bullying framework. On the intrapersonal level, Wachs and colleagues (2020) consider media competencies as key protective factors for bias-based victimization. On the interpersonal level, experiences of offline and cybervictimization are considered relevant risk factors, while on the school level, urbanization, school size, and higher levels of ethnic diversity are suggested to be relevant risk factors for bias-based victimization (Wachs, Schubarth and Bilz, 2020).

The Importance of Online Behaviors

To the best of our knowledge, no study to date investigated the associations between online behaviors and bias-based cybervictimization. However, according to meta-analyses, risky online behavior has been identified as the third major risk factor for cybervictimization, while experiences of offline victimization and offline bullying were the most important risk factors (Chen et al., 2017; Kowalski, Guimetti, Schroeder, & Lattanner, 2014). Adolescents who spend a lot of time online, who share their private information and photos online, and who use online social networks were consistently found to be at higher risk cybervictimization (Álvarez-García, Núnez Pérez, Dobarro González, & Rodríguez Pérez, 2015; Chen et al., 2017; Kowalski et al., 2014; Kowalski et al., 2019; Navarro, Serna, Martínez, & Ruiz-Oliva, 2013). In a recent study, the motives of Internet use were investigated in four groups of adolescents: non-victims, offline victims, cybervictims, and dual victims (Gini, Marino, Xie, Pfetsch, & Pozzoli, 2019). Adolescents who reported high levels of offline and online victimization (e.g., dual victims) were more likely than the other three groups to use the Internet because of social, coping, or enhancement motives. Navarro and colleagues (2013) also reported positive associations between the wish to communicate and cybervictimization. Thus, there is some evidence that online activities during which adolescents interact with others and share their private details might offer opportunities for potential offenders to attack them, while the ability to avoid and recognize online risks might operate as protective factors.

The Importance of Parents’ Behavior

How parents manage their children’s online behaviors is called parental mediation (Livingstone & Helsper, 2008). In principle, parents can actively monitor their children’s online activities, they can restrict what their children are doing online, or they can discuss options how their children can use the Internet safely (Sasson & Mesch, 2017). Although the associations between parental mediation strategies and cybervictimization have been investigated in many studies (for reviews, see Elsaesser, Russell, Ohannessian, & Patton, 2017; Kowalski et al., 2019; López-Castro & Priegue, 2019), the results are not conclusive. In some studies, parental monitoring operated as protective factor for cybervictimization, while in others no or even negative associations were found (e.g., Álvarez-García et al., 2015; Baldry, Sorrentino, & Farrington, 2019; Sasson & Mesch, 2017). Overall, it is most plausible to conclude that different mediation strategies are differently associated with cybervictimization (e.g., Elsaesser et al., 2017; Wright, 2017). However, as Baldry and colleagues (2019) point out, the direction of effects is still difficult to interpret, because without longitudinal studies it is impossible to decide whether parental mediation strategies are applied to prevent the emergence of online risks or are increased as reactions after cybervictimization already has occurred. Structural family variables like living with a single parent or in a divorced family were risk factors for cybervictimization in several studies (Kowalski et al., 2019).

The Importance of Teachers and Schools

Although there is a large literature how teachers handle bullying cases (Rigby & Bauman, 2010; van der Zanden, Denessen, & Scholte, 2015; Yoon & Bauman, 2014), the question how teachers manage their students’ online behaviors received much less attention. Teachers are not in the position to monitor and restrict the Internet use of their students after or out of school; however, they are still able to discuss how adolescents can use the Internet safely during their lessons. To the best of our knowledge, no study to date investigated whether this kind of active teacher mediation strategy is associated with cybervictimization. However, a recent study investigated teachers’ knowledge and intervention strategies to handle hate-postings (Strohmeier & Gradinger, 2021). It was found that teachers would most often alert other colleagues, followed by using victim-oriented rehabilitating strategies, working with the perpetrators’ parents, applying authority-based sanctions, and seeking help from external professionals, and they would rather not ignore the incident. Structural school variables like ethnic diversity and school size have also theorized to be relevant for bias-based cybervictimization (Wachs, Schubarth and Bilz, 2020), but have not been empirically investigated yet.

The Present Study

This study assumes that stigma (or bias) is constructed on the societal level with the goal to devalue the social position of multiple groups and their members (e.g., gender, migration, religion, disability, sexual orientation, income) and takes into account that these biases are often acted out on the Internet (Blaya, 2019). Therefore, (a) the prevalence of eight different forms of bias-based cybervictimization is investigated and (b) the importance of an exceptionally large number of predictors of bias-based cybervictimization on different socio-ecological levels is explored. Extending previous studies that used a global item (e.g., Hawdon et al., 2017; Wachs & Wright, 2019), eight different forms of bias were differentiated. Because of the large conceptual overlap, offline victimization and cybervictimization were also measured. The sample comprised a diverse group of adolescents, as mediating technologies have increasingly shaped the social relationships of this age group during the last years (Kowalski et al., 2019; Paus-Hasebrink et al., 2019). Differences in bias-based cybervictimization depending on gender, migration, and family status were investigated. An exceptionally wide range of possible predictors on the intrapersonal level, interpersonal level, family level, and class level was simultaneously tested applying a multilevel zero-inflated Poisson model (PM; Hox, Moerbeek, & van de Schoot, 2018).

Because this is the first study that empirically investigates bias-based cybervictimization from a socio-ecological perspective, the main analysis was exploratory. However, it was possible to formulate the following hypotheses based on results from studies on cybervictimization:

Hypothesis 1: Girls, first-generation immigrants, and adolescents living with a single parent report more forms of bias-based cybervictimization compared with boys, non-immigrants, and adolescents living with both parents (Kowalski et al., 2019).

Hypothesis 2: Offline and cybervictimization are moderately positively associated with bias-based cybervictimization (Kowalski et al., 2014).

Hypothesis 3: Online behavior, especially all activities during which adolescents interact with others and share their private details are associated with more forms of bias-based cybervictimization, because these activities offer opportunities for potential offenders (Gini et al., 2019; Navarro et al., 2013).

Hypothesis 4: Small associations between bias-based cybervictimization and parental and teacher mediation are expected; however, the direction of effects is unclear as previous studies are inconclusive (Baldry et al., 2019; Sasson & Mesch, 2017).

Method

Procedure

After all necessary ethical permissions were obtained and the directorates of education of the federal states of Upper and Lower Austria approved the study, a convenience sample of 17 vocational secondary schools were invited and agreed to participate. Three schools were located in a medium-sized town with around 200.000 inhabitants, two schools were located in a small town with around 60.000 inhabitants, the other 12 schools were located in villages with not more than 20.000 inhabitants. All Grade 7 and Grade 8 students enrolled in these schools were invited to participate. Participation was voluntary and confidential, and active parental consent was obtained. The parental consent rate was 83.5% and the participation rate of the consented adolescents was 96.6%. Trained research assistants collected the data with an Internet-based survey in the school’s computer labs between March and June 2019. To avoid any systematic order effect, items within scales were counterbalanced across participants.

Participants

In total, 1,018 adolescents (52.3% girls), aged 12 to 17 years ( $\bar{X}$ = 13.55, SD = 0.88), from 61 different classes participated. The majority of adolescents were non-immigrants (89.2%), and 10.8% were first-generation immigrants who were born in 42 different countries. The number and cultural diversity of immigrants sampled is representative for Upper Austria that mainly consists of rural places (Statistik Austria, 2020). The majority (78.4%) lived together with both of their parents, 18.8% lived together with their mother, 2.2% with their father, and 0.6% did not live with one of their parents. The majority had employed fathers (94.2%) and mothers (86.7%), and rated the financial situation of their family as good (43.8%) or very good (34.8%).

Measures

Demographic information

Gender, age, country of birth, living condition, father’s and mother’s occupational status, and socioeconomic status (SES) were measured with multiple-choice items. Living condition was measured with the question “With whom are you living together?” with the answer options “my mother,” “my father,” “my siblings,” “my grandmother,” “my grandfather,” “somebody else, namely . . .” Parental occupational status was measured with the question “Does your father/mother work?” with the answer options “yes, full-time,” “yes, part-time,” “no.” The subjective financial situation was measured with the question “How is the financial situation of your family?” with the answer options “very bad,” “bad,” “medium,” “good,” “very good.”

Victimization and bullying

After providing the following definition, offline victimization, cybervictimization, offline bullying, and cyberbullying were each measured with one item. Sometimes it happens that students who cannot easily defend themselves are intentionally treated in a mean way, they get insulted or hurt. The mean behavior can happen directly, for instance, when mean things are said, somebody is called names, embarrassed, threatened, shoved around, or kicked. The mean behavior can also happen indirectly, for instance, when somebody is given a mean look, is ignored, is excluded from a group, or is otherwise treated in an unfair way. The four items read as follows.

During the last 2 months, how often have you been insulted or hurt by your classmates with mean words or behavior? (offline victimization)

During the last 2 months, how often have you been insulted or hurt by your classmates with mean text messages, emails, videos, or photos on the Internet? (cybervictimization)

During the last 2 months, how often have you insulted or hurt a classmate with mean words or behavior? (offline bullying)

During the last 2 months, how often have you insulted or hurt a classmate with mean text messages, emails, videos, or photos on the Internet? (cyberbullying)

The 5-point response scale ranged from 1 (not at all), to 2 (once or twice), 3 (2 or 3 times a month), 4 (once a week) to 5 (nearly every day).

Bias-based cybervictimization

After the students filled in the general items on victimization and bullying, they were instructed as follows. Sometimes mean things on the Internet happen because of certain reasons, for instance, because of the target’s country of origin, religion, gender, disability, sexual orientation, appearance, age, or income. Please think about your own experiences on the Internet now. How often have you been insulted or hurt with mean text messages, emails, videos, or photos on the Internet, because

you are a boy or a girl?

your family migrated to Austria?

of your religion?

you have a disability?

you are in love with someone of the same gender?

others do not like your appearance?

you are too young for something?

your family does not have a lot of money?

The 5-point response scale ranged from 1 (not at all), to 2 (once or twice), 3 (2 or 3 times a month), 4 (once a week) to 5 (nearly every day).

Internet access

Access to the Internet was assessed with a multiple-choice item. “Do you have access to the Internet via a smartphone?” Answer options were “Yes, I have my own device”; “Yes, I can use the device from somebody else”; “No.” Nearly all adolescents had Internet access via a smartphone. Descriptive analyses revealed that 98.5% have their own device, 0.5% can use the device from somebody else, and only 1% (n = 10) stated that they do not have access to the Internet via a smartphone.

Media use motives

Sixteen items were developed to measure entertainment, identity management, social contact, and information search. “How often do you use digital media . . . to listen to the music (entertainment), . . . to express yourself (identity management), . . . to keep in touch with friends (social contact), and . . . to search for information (information search)?” The 5-point response scale ranged from 1 (never), to 2 (rarely), 3 (sometimes), 4 (often) to 5 (very often). To check the factor structure of the media use motives scale, exploratory factor analysis (EFA) was calculated. Five factors emerged with an Eigenvalue > 1.00. The five-factor structure was theoretically meaningful and explained 62.69% of the variance. The scales were (a) self-expression (three items, α = .77), (b) social contact (two items, α = .76), (c) playing (two items, α = .80), (d) information search (three items, α = .62), and (e) discussion (three items, α = .70). Three items had double loadings and were deleted. The items are displayed in Table S1 (Supplementary Materials).

Online risks

Fourteen items were developed to measure two aspects of online risks. Recognition was measured with seven items. “How easy or difficult is it for you to recognize the following things in the Internet?” for example, “fake photos.” Avoidance was measured with seven items. “How easy or difficult is it for you to avoid the following things in the Internet?” for example, “unwanted spread of your private information.” The 5-point response scale ranged from 1 (very difficult), to 2 (difficult), 3 (neutral), 4 (easy) to 5 (very easy). Two factors emerged with an Eigenvalue > 1.00. The two-factor structure was theoretically meaningful and explained 56.29% of the variance. The scales were (a) recognition (seven items, α = .84) and (b) avoidance (seven items, α = .89). The items are displayed in Table S2 (Supplementary Materials).

Revealing private information

A vignette was developed to measure this construct. Students were asked to imagine the following scenario: You want to play a cool online game with your smartphone. You clicked through the webpage and now there is only one final step left! An online form pops up. You can only play the online game after you filled in this form. Then they were asked the following question: “Which information would you enter into this form?” (a) your first name, (b) your family name, (c) your address, (d) your phone number, (e) your date of birth, (f) your favorite pet, (g) your hobbies, (h) your photo, and (i) your email address. The 5-point response scale ranged from 1 (certainly not), to 2 (rather not), 3 (I am unsure), 4 (rather yes) to 5 (certainly yes). The reliability of the scale was very good (nine items, α = .80).

Parental mediation

Seventeen items were developed to measure three aspects of parental mediation. Monitoring was measured with six items (e.g., “How often do your parents control which webpages you visit?). Advice was measured with five items (e.g., “How often do you ask your parents for their advice when you want to download something from the Internet?”). Discussion was measured with six items (e.g., “How often do your parents explain you why some webpages are good or bad?”). The 5-point response scale ranged from 1 (never), to 2 (rarely), 3 (sometimes), 4 (often) to 5 (very often). Three factors emerged with an Eigenvalue > 1.00. The three-factor structure was theoretically meaningful and explained 59.64% of the variance. The scales were (a) monitoring (six items, α = .86), (b) advice (five items, α = .85), and (c) discussion (six items, α = .82). The items are displayed in Table S3 (Supplementary Materials).

Teachers’ mediation

Six items were developed to measure “discussion.” These items were identical with the items developed for parents (e.g., “How often do your teachers explain you why some webpages are good or bad?”). The 5-point response scale ranged from 1 (never), to 2 (rarely), 3 (sometimes), 4 (often) to 5 (very often). One factor emerged with an Eigenvalue > 1.00 and explained 57.60% of the variance. The reliability of the scale (six items, α = .85) was very good.

Analytic Strategy

To explore gender, immigrant status, and family status differences, three multivariate analyses of variance (MANOVAs)—consisting of gender, immigrant status, and family status as independent variables, and bias-based cybervictimization, bullying and victimization, online behavior, parents’ behavior, and teachers’ behavior as dependent variables—were applied.

To explore the predictors of bias-based cybervictimization, multilevel count modeling (Hox et al., 2018) was conducted in Mplus 8.4 (L. K. Muthén & Muthén, 1998-2017) using the maximum likelihood estimation method to predict the number of discrimination incidents on the student level and class level. In order to determine the count model, the PM, negative binomial model (NBM), zero-inflated PM, and the zero-inflated NBM were compared using the Bayesian information criterion (BIC; B. O. Muthén et al., 2016). Results showed that the zero-inflated PM had the lowest BIC value indicating the best trade-off between model fit and model parsimony (see Table S4, Supplementary Materials). Therefore, this model was chosen for the present study. The zero-inflated PM is a mixture model comprising a logistic regression model for predicting the non-occurrence of an event and a PM for predicting the non-zero counts (Atkins & Gallop, 2007). In other words, the zero-inflated PM predicts the outcome variable in two parts: (a) occurrence versus non-occurrence of bias-based form of cybervictimization (i.e., logistic part of the model) and (b) the number of bias-based forms of cybervictimization (i.e., Poisson part of the model).

Models were built up from the null model by adding sets of student-level and class-level predictors to a previous model at each step. In the first model, demographic variables were added (Model 1), bullying and victimization were added in the second step (Model 2), online behavior was added in the third step (Model 3), parents’ and teachers’ behaviors were added in the fourth step (Model 4), and predictors on the class level were added in the last step (Model 5). The same predictors were used in the logistic and in the Poisson part of the model. All predictors on the student level were centered within classes, while predictors on the class level were centered at the grand mean (Enders & Tofighi, 2007).

Results

Missing Data

The data quality in this data set was exceptionally good. There were no missing values in none of the variables. Thus, all results are based on the answers of 1,018 adolescents.

Prevalence of Bias-Based Cybervictimization

Preliminary descriptive analyses revealed that the number of adolescents who reported that they were victims of one of the eight forms of bias-based cyberbullying was much lower than the number of adolescents who reported offline victimization (for proportion scores, see the first column of Table 1). For six out of eight forms, the prevalence rates ranged between 4% and 7%, while 52% stated that they experienced offline victimization and 15% experienced cybervictimization at least once during the last 2 months. A cumulative risk index of bias-based cybervictimization was calculated by summing up the numbers of the eight different forms. Descriptive analyses revealed that 61.8% (n = 629) of adolescents did not report any of the eight bias-based victimization forms, 20.6% (n = 210) reported one form, 10.8% (n = 110) reported two forms, and 6.8% (n = 69) reported three or more forms.

Table 1.

Gender, Immigrant Status, and Family Status Differences (Means and Standard Devitations are displayed).

	Whole sample (n = 1,018)	Girls (n = 532)	Boys (n = 468)	F(1, 1016)	Non-immigrants (n = 908)	First-generation immigrants (n = 110)	F(1, 1016)	Both parents (n = 798)	Single parent (n = 220)	F(1, 1016)
Bias-based cybervictimization^a
Cumulative risk	0.63 (0.93)	0.71 (0.99)	0.54 (0.85)	9.00**	0.58 (0.89)	0.98 (1.11)	18.54**	0.59 (0.90)	0.77 (1.01)	6.68**
Gender	0.05 (0.22)	0.06 (0.24)	0.04 (0.19)	3.17^ns	0.04 (0.20)	0.14 (0.34)	18.04**	0.05 (0.22)	0.05 (0.22)	0.03^ns
Migration	0.07 (0.25)	0.07 (0.26)	0.06 (0.24)	0.57^ns	0.04 (0.19)	0.28 (0.45)	102.86**	0.06 (0.24)	0.08 (0.27)	1.17^ns
Religion	0.06 (0.24)	0.06 (0.24)	0.05 (0.23)	0.50^ns	0.05 (0.21)	0.15 (0.35)	16.89**	0.05 (0.23)	0.07 (0.26)	0.96^ns
Disability	0.06 (0.23)	0.05 (0.22)	0.07 (0.25)	1.06^ns	0.05 (0.22)	0.13 (0.33)	10.95**	0.06 (0.24)	0.05 (0.23)	0.06^ns
Homophobic	0.04 (0.19)	0.04 (0.18)	0.04 (0.20)	0.61^ns	0.03 (0.18)	0.07 (0.26)	3.97*	0.04 (0.19)	0.04 (0.19)	0.03^ns
Look	0.24 (0.43)	0.30 (0.36)	0.11 (0.31)	18.37**	0.23 (0.42)	0.30 (0.46)	2.38^ns	0.22 (0.42)	0.31 (0.46)	7.23**
Age	0.13 (0.34)	0.15 (0.36)	0.11 (0.31)	5.05*	0.13 (0.33)	0.15 (0.36)	0.68^ns	0.13 (0.33)	0.15 (0.35)	0.62^ns
Poverty	0.05 (0.22)	0.05 (0.22)	0.05 (0.22)	0.01^ns	0.05 (0.21)	0.09 (0.29)	3.78**	0.04 (0.19)	0.11 (0.31)	18.80**
Bullying and victimization^a
Offline victimization	0.52 (0.50)	0.52 (0.50)	0.50 (0.50)	0.42^ns	0.51 (0.50)	0.56 (0.50)	1.18^ns	0.51 (0.50)	0.53 (0.50)	0.33^ns
Cybervictimization	0.15 (0.35)	0.19 (0.39)	0.10 (0.30)	17.12**	0.13 (0.34)	0.25 (0.43)	9.77**	0.14 (0.35)	0.15 (0.36)	0.03^ns
Offline bullying	0.42 (0.49)	0.36 (0.48)	0.48 (0.50)	13.85**	0.41 (0.49)	0.45 (0.50)	0.77^ns	0.41 (0.49)	0.42 (0.49)	0.06^ns
Cyberbullying	0.09 (0.29)	0.09 (0.29)	0.09 (0.29)	<0.01^ns	0.08 (0.28)	0.19 (0.39)	13.63**	0.08 (0.27)	0.14 (0.34)	5.84*
Online behavior^b
Motive information search	3.38 (0.80)	3.44 (0.79)	3.31 (0.81)	7.01**	3.34 (0.79)	3.65 (0.83)	14.90**	3.38 (0.77)	3.37 (0.89)	0.03^ns
Motive self-expression	2.59 (1.00)	2.84 (0.99)	2.33 (0.95)	70.13**	2.56 (0.99)	2.87 (1.03)	9.77**	2.57 (0.99)	2.68 (1.02)	2.01^ns
Motive social contacts	4.35 (0.84)	4.52 (0.75)	4.17 (0.89)	46.64**	4.34 (0.83)	4.39 (0.88)	0.31^ns	4.34 (0.84)	4.36 (0.83)	0.02^ns
Motive playing	3.21 (1.28)	2.57 (1.07)	3.92 (1.10)	396.67**	3.20 (1.26)	3.28 (1.39)	0.42^ns	3.21 (1.26)	3.22 (1.33)	<0.01^ns
Motive discussion	2.47 (0.99)	2.54 (0.98)	2.40 (1.01)	4.70**	3.41 (0.96)	3.00 (1.08)	35.43**	2.45 (0.98)	2.57 (1.05)	2.41^ns
Avoidance of online risks	3.64 (0.92)	3.62 (0.90)	3.65 (0.95)	0.37^ns	3.66 (0.91)	3.40 (0.99)	8.41**	3.62 (0.93)	3.70 (0.86)	1.25^ns
Recognition of online risks	3.38 (0.79)	3.28 (0.74)	3.48 (0.84)	17.87**	3.37 (0.78)	3.40 (0.89)	0.13^ns	3.36 (0.80)	3.44 (0.76)	1.79^ns
Revealing private information online	2.55 (0.83)	2.63 (0.79)	2.47 (0.87)	9.07**	2.54 (0.82)	2.66 (0.94)	1.84^ns	2.54 (0.82)	2.62 (0.88)	1.72^ns
Parents’ behavior^b
Monitoring	1.71 (0.82)	1.78 (0.89)	1.64 (0.74)	7.01**	1.67 (0.79)	2.03 (1.01)	19.20**	1.72 (0.82)	1.67 (0.83)	0.62^ns
Advice	2.04 (1.00)	2.25 (1.05)	1.81 (0.90)	50.96**	2.04 (1.00)	2.04 (1.03)	<0.01^ns	2.06 (0.99)	1.97 (1.05)	1.37^ns
Discussion	2.30 (0.86)	2.34 (0.81)	2.26 (0.92)	2.69^ns	2.27 (0.84)	2.49 (0.99)	5.86*	2.35 (0.86)	2.13 (0.85)	11.35**
Teachers’ behavior^b
Discussion	2.22 (0.82)	2.34 (0.82)	2.10 (0.81)	20.85**	2.21 (0.81)	2.37 (0.88)	3.70^ns	2.25 (0.83)	2.13 (0.79)	3.55^ns

Note. ns = not significant.

The cumulative risk index ranged between 1 and 3; for the single items, the proportion scores ranged between 0 and 1.

Rating scales ranged from 1 to 5.

p < .05. **p < .01.

Gender, Immigrant Status, and Family Status Differences

The means and standard deviations of the study variables depending on gender, immigrant status, and family status are displayed in Table 1. To check for main effects, three MANOVAs were performed with gender, immigrant status, and family status as the independent variables and the different indicators of bias-based cybervictimization and bullying, online behavior, and parents’ and teachers’ behavior as the dependent variables.

According to multivariate tests, mean differences in bias-based cybervictimization depending on sex, F(9, 1008) = 3.04, p < .01, η² = .03; migration status, F(9, 1008) = 12.83, p < .01, η² = .10; and family status, F(9, 1008) = 3.04, p < .01, η² = .03, were found. The results of the univariate tests are displayed in Table 1. Confirming Hypothesis 1, girls, first-generation immigrants, and adolescents living with a single parent reported more forms of bias-based cybervictimization compared with boys, non-immigrants, and adolescents living with both parents.

According to multivariate tests, mean differences in bullying and victimization depending on sex, F(4, 1013) = 10.23, p < .01, η² = .04, and migration status, F(4, 1013) = 4.45, p < .01, η² = .02, were found, while family status was not significant, F(4, 1013) = 1.70, p = .15, η² < .01. The results of the univariate tests are displayed in Table 1. Boys reported higher levels of offline bullying and girls reported higher levels of cybervictimization, while first-generation immigrants reported higher levels of cybervictimization and cyberbullying compared with non-immigrants.

According to multivariate tests, mean differences in online behavior depending on sex, F(8, 1009) = 74.64, p < .01, η² = .37, and migration status, F(8, 1009) = 6.84, p < .01, η² = .05, were found, while family status was not significant, F(8, 1009) = 0.86, p = .55, η² < .01. The results of the univariate tests are displayed in Table 1. The motives for using the Internet strikingly differed between girls and boys. While girls use the Internet more often to search for information, to express themselves, to have social contacts, and to discuss something, boys use the Internet more often to play. Girls are less likely to recognize online risks but more likely to reveal private information online compared with boys. There were also some differences between first-generation immigrants and non-immigrants. First-generation immigrants use the Internet more to search for information and to express themselves, but less to discuss something compared with non-immigrants. First-generation immigrants are less likely to avoid online risks compared with non-immigrants.

According to multivariate tests, mean differences in parents’ and teachers’ behavior depending on sex, F(4, 1013) = 16.67, p < .01, η² = .06; migration status, F(4, 1013) = 6.38, p < .01, η² = .03; and family status, F(4, 1013) = 3.26, p = .01, η² = .01, were found. The results of the univariate tests are displayed in Table 1. Girls reported higher levels of parental monitoring, parental advice, and discussions with teachers compared with boys. First-generation immigrants reported higher levels of parental monitoring and parental discussions compared with non-immigrants, and adolescents who live with a single parent reported lower levels of parental discussions compared with adolescents who live with both parents.

Correlations of Study Variables

When inspecting the bivariate correlations, the bias-based cybervictimization was moderately positively associated with offline and cybervictimization, followed by offline and cyberbullying (see Table 2). Bias-based cybervictimization correlated weakly positive with the motive information search, self-expression, and discussion. Bias-based cybervictimization correlated weakly negative with avoidance of online risks, but weakly positive with recognizing online risks and revealing private information online. Parental monitoring, parental discussion, and teachers’ discussion were also weakly positive associated with bias-based cybervictimization.

Table 2.

Correlations Between Study Variables.

	2.	3.	4.	5.	6.	7.	8	9.	10.	11.	12.	13.	14.	15.	16.	17.
1. Cumulative bias-based cybervictimization	.37**	.31**	.23**	.23**	.08**	.17**	.06	–.02	.21**	–.10**	.07*	.13**	.12**	.01	.07*	.09**
2. Offline victimization	—	.23**	.44**	.16**	.01	.10**	.00	.00	.04	–.02	.01	.02	.00	–.03	–.01	–.00
3. Cybervictimization		—	.16**	.34**	.06*	.23**	.07*	–.05	.15**	–.06*	.07*	.11**	.08**	–.04	.03	.05
4. Offline bullying			—	.29**	–.03	.06	.03	.14**	.10**	.01	.01	.03	–.08**	–.13**	–.04	–.01
5. Cyberbullying				—	.00	.14**	.07*	.07*	.21**	–.00	.09**	.11**	–.02	–.11**	–.05	.05
6. Motive information search					—	.22**	.20**	.02	.23**	.02	.11**	.04	.10**	.08*	.18**	.25**
7. Motive self-expression						—	.44**	–.06	.51**	–.06*	.16**	.25**	.04	–.08*	.10**	.11**
8. Motive contact with friends							—	.04	.39**	.03	.15**	.23**	–.01	–.08*	.08**	.11**
9. Motive playing with friends								—	.16**	.04	.13**	.04	–.08*	–.17**	–.04	–.04
10. Motive discussion									—	–.06	.19**	.25**	.01	–.16**	.01	.08*
11. Avoidance of online risks										—	.41**	–.05	–13**	–.08*	–.05	.03
12. Recognition of online risks											—	.05	–.05	–.17**	.03	.07*
13. Revealing private information online												—	–.03	–.15**	–.03	–.00
14. Parental monitoring													—	.42**	.49**	.29**
15. Parental advice														—	.47**	.29**
16. Parental discussion															—	.38**
17. Teachers’ discussion																—

p < .05. **p < .01.

Predictors of Bias-Based Cybervictimization

Intraclass correlations

The intraclass correlation coefficient (ICC) for the logistic part of the model was .069, while the ICC of the Poisson part of the model was .097, indicating that 6.9% of the variance of the occurrence versus non-occurrence of bias-based cybervictimization and 9.7% of the variance in the forms of bias-based cybervictimization are on the class level.

Class-level predictors

The ICCs of potentially relevant predictors were estimated. For gender, parental status, offline victimization, cybervictimization, offline bullying, and cyberbullying, the ICCs ranged between .000 and .053, indicating that these variables are nearly constants on class level. The ICC for immigrant status was .289 and the ICC of teachers’ discussion was .105, indicating that 28.9% and 10.5% of the variance of these variables are on the class level. Therefore, only class-level aggregated immigrant status (i.e., proportion of immigrant students) and teachers’ discussion (i.e., average teacher discussion), and class size were included as class-level predictors.

Results of the logistic part of the model

As shown in Table 3, offline victimization was the only significant predictor of not being a victim of bias-based cybervictimization on the individual level. That is, students with lower levels of offline victimization are less likely for being a victim of bias-based cybervictimization. On the class level, the proportion of first-generation immigrant students was a significant predictor of not being a victim of bias-based cybervictimization. That is, classes with a lower proportion of first-generation immigrant have a lower probability for students for being a victim of bias-based cybervictimization.

Table 3.

Multilevel Zero-Inflated Poisson Regression Model Results of the Logistic Part Predicting the Probability of Students Not Being Bias-Based Cybervictims.

	Model 1		Model 2		Model 3		Model 4		Model 5
	Est. (SE)	OR	Est. (SE)	OR	Est. (SE)	OR	Est. (SE)	OR	Est. (SE)	OR
Intercept	–0.60 (0.17)		–1.07 (0.32)		–1.40 (0.41)		–1.63 (0.552)		–1.17 (0.39)
Demographic variables
Gender (0 = girls, 1 = boys)	0.01 (0.35)	1.01	–0.31 (0.48)	0.73	–1.12 (0.59)	0.33	–1.12 (0.64)	0.33	–0.97 (0.61)	0.38
Age in years	–0.10 (0.21)	0.90	0.07 (0.27)	1.07	0.02 (0.32)	1.02	–0.09 (0.31)	0.92	0.04 (0.28)	1.04
Country (0 = Austria, 1 = other country)	–0.10 (0.43)	0.90	–0.10 (0.59)	0.91	0.39 (0.65)	1.47	0.71 (0.66)	2.03	0.66 (0.83)	1.93
Parents (0 = single parent, 1 = two parents)	0.27 (0.38)	1.31	0.72 (0.53)	2.05	0.57 (0.54)	1.76	0.10 (0.68)	1.11	0.95 (0.62)	2.58
Victimization and bullying
Offline victimization			–2.21 (0.64)	0.11	–2.13 (0.73)	0.12	–1.60 (0.60)	0.20	–2.76 (0.81)	0.06
Online victimization			–0.76 (0.66)	0.47	–0.87 (0.88)	0.42	–0.71 (0.83)	0.49	–0.57 (0.78)	0.57
Offline bullying			–0.75 (0.48)	0.47	–1.36 (0.59)	0.26	–1.62 (0.60)	0.20	–1.04 (0.50)	0.35
Online bullying			–0.35 (0.78)	0.71	0.09 (0.88)	1.09	–0.44 (1.07)	0.65	–0.18 (1.07)	0.84
Online motives
Information search					–0.20 (0.27)	0.82	–0.36 (0.31)	0.70	–0.25 (0.29)	0.78
Self-expression					–0.46 (0.30)	0.63	–0.49 (0.31)	0.61	–0.07 (0.28)	0.94
Social contact					0.06 (0.29)	1.06	0.04 (0.32)	1.04	–0.09 (0.28)	0.92
Playing					0.16 (0.22)	1.18	0.17 (0.24)	1.19	0.26 (0.21)	1.30
Discussion					–0.45 (0.27)	0.64	–0.31 (0.29)	0.73	–0.55 (0.28)	0.58
Online risks
Avoidance					0.10 (0.34)	1.11	–0.11 (0.39)	0.89	0.05 (0.28)	1.05
Recognition					–0.02 (0.38)	0.98	0.18 (0.42)	1.20	–0.10 (0.33)	0.91
Revealing private information					–0.42 (0.30)	0.66	–0.62 (0.38)	0.54	–0.34 (0.27)	0.71
Parents’ behavior
Monitoring							–0.14 (0.29)	0.87	–0.26 (0.34)	0.77
Advice							0.03 (0.25)	1.03	–0.14 (0.26)	0.87
Discussion							0.24 (0.34)	1.27	–0.10 (0.31)	0.90
Teachers’ behavior
Discussion							0.62 (0.36)	1.86	0.46 (0.30)	1.58
Class variables
Proportion of immigrant students									–5.00 (1.79)	0.01
Average teachers’ discussion									0.03 (0.65)	1.03
Class size									0.01 (0.07)	1.01
Model summary
Log likelihood	–1,070.28		–980.56		–957.91		–951.77		–934.69
BIC	2,216.73		2,092.70		2,158.22		2,201.34		2,201.23
McFadden’s pseudo-R²	.01		.09		.11		.12		.14

Note. Statistically significant results at α = .05 are boldface. Est. = unstandardized parameter estimate; OR = odds ratio (i.e., change factor for the odds for not being discriminated when the predictor variable increases one unit); BIC = Bayesian information criterion.

Results of the Poisson part of the model

As shown in Table 4, gender, cybervictimization, and avoiding online risks were significant predictors on the individual level. That is, girls compared with boys, students with higher level of cybervictimization, and lower levels of avoiding online risks experience more bias-based cybervictimization. On the class level, discussions with the teachers was significantly associated with being a victim of a higher number of bias-based cybervictimization forms. That is, more discussions with the teacher was associated with more forms of bias-based cybervictimization in classes.

Table 4.

Multilevel Zero-Inflated Poisson Regression Model Results of the Poisson Part of the Model Predicting the Number of Bias-Based Cybervictimizations.

	Model 1		Model 2		Model 3		Model 4		Model 5
	Est. (SE)	IRR	Est. (SE)	IRR	Est. (SE)	IRR	Est. (SE)	IRR	Est. (SE)	IRR
Intercept	–0.11 (0.09)		–0.27 (0.11)		–0.34 (0.10)		–0.40 (0.12)		–0.19 (0.09)
Demographic variables
Gender (0 = girls, 1 = boys)	–0.31 (0.14)	0.73	–0.32 (0.12)	0.73	–0.40 (0.14)	0.67	–0.37 (0.14)	0.69	–0.36 (0.14)	0.70
Age in years	0.04 (0.09)	1.04	0.07 (0.08)	1.08	0.05 (0.08)	1.05	0.06 (0.08)	1.06	0.08 (0.08)	1.08
Country (0 = Austria, 1 = other country)	0.34 (0.18)	1.40	0.28 (0.17)	1.32	0.29 (0.16)	1.03	0.32 (0.16)	1.37	0.22 (0.18)	1.24
Parents (0 = single parent, 1 = two parents)	–0.16 (0.14)	0.85	–0.09 (0.12)	0.91	–0.11 (0.12)	0.90	–0.21 (0.13)	0.81	–0.10 (0.12)	0.91
Victimization and bullying
Offline victimization			0.34 (0.18)	1.41	0.48 (0.19)	1.62	0.63 (0.21)	1.89	0.25 (0.18)	1.29
Cybervictimization			0.36 (0.13)	1.43	0.30 (0.13)	1.34	0.30 (0.13)	1.34	0.32 (0.14)	1.38
Offline bullying			0.02 (0.12)	1.02	–0.05 (0.12)	0.95	–0.07 (0.12)	0.94	–0.02 (0.12)	0.98
Cyberbullying			0.19 (0.14)	1.20	0.16 (0.14)	1.17	0.11 (0.14)	1.11	0.13 (0.15)	1.13
Online motives
Information search					–0.03 (0.07)	0.97	–0.07 (0.07)	0.93	–0.05 (0.07)	0.95
Self-expression					–0.05 (0.06)	0.95	–0.06 (0.06)	0.94	–0.01 (0.07)	0.99
Social contacts					–0.07 (0.08)	0.94	–0.05 (0.08)	0.95	–0.09 (0.08)	0.91
Playing					0.01 (0.05)	1.01	0.01 (0.05)	1.01	0.02 (0.05)	1.02
Discussion					0.12 (0.07)	1.12	0.15 (0.07)	1.16	0.10 (0.07)	1.10
Online risks
Avoidance					–0.16 (0.07)	0.86	–0.18 (0.08)	0.84	–0.15 (0.07)	0.86
Recognition					0.11 (0.09)	1.11	0.15 (0.09)	1.16	0.10 (0.09)	1.11
Revealing private information					0.02 (0.07)	1.02	–0.01 (0.07)	1.00	0.02 (0.07)	1.02
Parents’ behavior
Monitoring							0.11 (0.07)	1.11	0.07 (0.07)	1.08
Advice							0.00 (0.07)	1.00	–0.04 (0.07)	0.96
Discussion							0.06 (0.08)	1.06	0.01 (0.08)	1.01
Teachers’ behavior
Discussion							0.14 (0.08)	1.15	0.12 (0.07)	1.13
Class variables
Proportion of immigrant students									0.25 (0.23)	1.29
Average teachers’ discussion									0.45 (0.17)	1.56
Class size									0.01 (0.02)	1.10
Model summary
Log likelihood	–1,070.28		–980.56		–957.91		–951.77		–934.69
BIC	2,216.73		2,092.70		2,158.22		2,201.34		2,201.23
McFadden’s pseudo-R²	.01		.09		.11		.12		.14

Note. Statistically significant results at α = .05 are boldface. Est. = unstandardized parameter estimate; IRR = incidence rate ratio (i.e., change factor for the expected counts when the predictor variable increases one unit); BIC = Bayesian information criterion.

Discussion

To the best of our knowledge, this is the first study that investigated an exceptionally large number of potential predictors of bias-based cybervictimization on different socio-ecological levels. The present study produced several novel findings that are discussed below.

Prevalence of Bias-Based Victimization

The first important insight from this study is that adolescents experience different forms of bias-based cybervictimization. Compared with cybervictimization and offline victimization, the prevalence of each form of bias-based cybervictimization was rather low. However, when cumulating the eight forms that were measured in the present study, 38.2% of adolescents reported at least one form of bias-based cybervictimization during the last 2 months. This number is much higher compared with the most recent prevalence rates reported in the literature (Wachs, Wright, Gámez-Guadix, Görzig and Schubarth, 2020; Wachs & Wright, 2019), probably because these studies used a global item and did not separately ask for different forms of bias. When looking on each social position variable separately, girls, first-generation immigrants, and adolescents living with a single parent reported more forms of bias-based cybervictimization compared with boys, non-immigrants, and adolescents living with both parents. This result was hypothesized and it confirms previous studies (for a review, see Kowalski et al., 2019). When looking at the cumulative index, the majority of adolescents reported one form of bias (20.6%). However, 6.8% reported three or more forms. These adolescents are polyvictimized (Finkelhor, Ormrod, & Turner, 2007), most likely because several of their minority group memberships intersect. Because social and mental health outcomes of polyvictimized adolescents are much worse compared with their peers (Russell, Sinclair, Poteat, & Koenig, 2012), future studies should focus on this group of adolescents.

Bivariate Associations

The inspection of the bivariate associations reveals the second novel insights of the present study. As expected, offline and cybervictimization were associated with bias-based cybervictimization confirming the assumed conceptual overlap between these constructs. However, the moderate correlations also illustrate that the empirical overlap is far away from being perfect. Thus, it makes sense to differentiate these three constructs and not to subsume them under one umbrella as suggested by some scholars (e.g., Olweus, 2012; Olweus & Limber, 2018). Adolescents who use the Internet because of discussions, self-expression, and information search reported higher levels of cybervictimization, while adolescents who are more able to avoid online risks and who are less likely to reveal private information online report lower levels of bias-based cybervictimization. These correlations are small, but in line with previous studies on cybervictimization (e.g., Gini et al., 2019; Kowalski et al., 2014; Navarro et al., 2013). A reversed and therefore unexpected pattern was found for recognizing online risks. Although the small associations should not be over-interpreted, a plausible explanation is that youth who are better able to recognize online risks might also have a higher awareness of stereotypes, prejudice, and dominance orientations that makes it easier for them to also recognize bias-based forms of cybervictimization. Contrary to the findings of a meta-analysis on cybervictimization (Kowalski et al., 2014), parental monitoring, and parental and teachers’ discussions were positively associated with bias-based cybervictimization. In line with studies reporting the same directions of effects (Baldry et al., 2019; Sasson & Mesch, 2017), these results might indicate parental and teachers’ mediation are also used as reactive strategies after bias-based cybervictimization has already been occurred. Of course, this interpretation remains purely speculative as long as longitudinal studies on this topic are lacking.

Multilevel Prediction of Bias-Based Cybervictimization

The results of the multilevel regression models revealed the most innovative insights of the present study. Guided by the socio-ecological model of development (Bronfenbrenner, 1979) that assumes that development unfolds because of the complex interplay between adolescents and their multilayered environments, the relevance of an exceptionally large number of variables on the intrapersonal level, interpersonal level, family level, and class level for bias-based cybervictimization was investigated simultaneously. Only two variables were able to predict whether adolescents would experience at least one versus no form of bias-based cybervictimization. These variables were on the interpersonal level (e.g., offline victimization) and on the class level (e.g., proportion of immigrants). Four variables were able to predict whether an adolescent would experience one, two, or three or more forms of bias-based cybervictimization. Relevant predictors were gender and variables on the intrapersonal level (e.g., avoidance of online risks), on the interpersonal level (e.g., cybervictimization), and on the class level (e.g., teachers’ mediation). Thus, none of the different motives for Internet use, nor revealing private information online, nor any form of parental mediation was associated with bias-based cybervictimization when applying a multilevel analysis. It is certainly necessary to replicate these findings in future studies, because their generalizability is limited to one Austrian federal state and vocational secondary schools. The results might still help developing the socio-ecological perspective on bias-based cybervictimization further by demonstrating the importance of individual-level and class-level factors.

Limitations

There are plenty of avenues for future studies that might be able to overcome some of the evident limitations of this study. To begin with, most of the scales used were developed for the purpose of the present study. Although we developed them based on good reasoning and adapted them from existing studies wherever possible, construct validity was established with EFAs only. The second limitation is that all constructs were measured with self-reports at one wave of measurement only. Future studies could use longitudinal and experimental designs to enable causal conclusions.

Future Research and Implications for Prevention

The present study shows that the most important environment that shapes bias-based cybervictimization experiences of adolescents is the school. When schools prevent offline and cybervictimization and when they avoid segregating immigrant students in certain classes, there is a good chance to prevent bias-based cybervictimization as well. However, this is only one part of the story, because stigma is constructed on the societal level (Earnshaw et al., 2018). As the present analyses also reveal, girls experience more forms of bias-based cybervictimization compared with boys. This is most likely the case, because for girls memberships in devalued social groups intersect. Future studies should focus on the impact of the intersection of different devalued group memberships, especially related to gender. Future studies should therefore focus on the intersection of gender with other socially devalued group memberships to shed further light on this topic. Teacher mediation was a relevant variable on class level, and the direction of effects might suggest that teachers start discussing the media use of their students as a reaction to cases of bias-based cybervictimization. As suggested by Wachs, Schubarth and Bilz (2020), media competence seem to matter as well. Adolescents who are more able to recognize online risks reported fewer forms of bias-based victimization. Thus, combining stigma-based anti-bullying programs with the training of media competence could be the most promising approach for schools to also tackle bias-based cybervictimization. Teachers should be encouraged to proactively reflect and discuss the media use of their students and to inform them about how to avoid possible online risks like, for example, the uncontrolled spread of personal information, photos, or videos. Likewise, teachers should discuss the issue of bias with their students and explain them how social stigma is constructed on the societal level, reinforced on the structural level, and enacted by prejudiced perpetrators on the individual level.

Supplemental Material

sj-docx-1-jea-10.1177_02724316211010335 – Supplemental material for The Role of Intrapersonal-, Interpersonal-, Family-, and School-Level Variables in Predicting Bias-Based Cybervictimization

Supplemental material, sj-docx-1-jea-10.1177_02724316211010335 for The Role of Intrapersonal-, Interpersonal-, Family-, and School-Level Variables in Predicting Bias-Based Cybervictimization by Dagmar Strohmeier, Petra Gradinger and Takuya Yanagida in The Journal of Early Adolescence

Footnotes

Acknowledgements

We would like to thank Dr. Rainer Schmidbauer and the students of the master program “Addiction and Violence Prevention in Educational Settings” (University of Education Upper Austria, Linz) for their invaluable support in realizing this study. We are also very grateful to the schools and teachers who participated in this study.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Dagmar Strohmeier

Supplemental Material

Supplemental material for this article is available online.

Author Biographies

Dagmar Strohmeier, PhD, is a professor at the University of Applied Sciences Upper Austria, Linz, and Professor II at the Norwegian Centre for Learning Environment and Behavioural Research in Education at the University of Stavanger, Norway. She is interested in peer relations among children and youth, with a cross-national and cross-cultural perspective and a special focus on immigrant youth.

Petra Gradinger, PhD, is a lecturer at the University of Applied Sciences Upper Austria, Linz. She is interested in media competence, cyberbullying, and evidence-based prevention. Her publications focus on bullying and cyberbullying, learning and motivation, and methodological issues.

Takuya Yanagida, PhD, is a research associate at the University of Vienna. He is interested in statistical conclusion validity in program evaluation. More specifically, he has been focusing on application of various latent variable models, psychometric models, and missing data handling techniques in psychological science.

References

Álvarez-García

Núnez Pérez

J. C.

Dobarro González

Rodríguez Pérez

(2015). Risk factors associated with cybervictimization in adolescence. International Journal of Clinical and Health Psychology, 15, 226-235. doi:10.1016/j.ijchp.2015.03.002

Atkins

D. C.

Gallop

R. J.

(2007). Rethinking how family researchers model infrequent outcomes: A tutorial on count regression and zero-inflated models. Journal of Family Psychology, 21, 726-735. doi:10.1037/0893-3200.21.4.726

Baldry

A. C.

Sorrentino

Farrington

D. P.

(2019). Cyberbullying and cybervictimization versus parental supervision, monitoring and control of adolescents’ online activities. Children and Youth Services Review, 96, 302-307. doi:10.1016/j.childyouth.2018.11.058

Blaya

(2019). Cyberhate: A review and content analysis of intervention strategies. Aggression and Violent Behavior, 45, 163-172. doi:10.1016/j.avb.2018.05.006

Bronfenbrenner

(1979). Ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard University Press.

Bronfenbrenner

(1994). Ecological models of human development. In Husen

Postlethwaithe

T. N.

(Eds.), International encyclopedia of education (2nd ed., pp. 1643-1647). New York, NY: Elsevier Sciences.

Chen

S. S.

Lwin

M. O.

(2017). A meta-analysis of factors predicting cyberbullying perpetration and victimization: From the social cognitive and media effects approach. New Media & Society, 19, 1194-1213. doi:10.1177/1461444816634037

Earnshaw

V. A.

Reisner

S. L.

Menino

D. D.

Poteat

V. P.

Bogart

L. M.

Barnes

T. N.

Schuster

M. A.

(2018). Stigma-based bullying interventions: A systematic review. Developmental Review, 48, 178-200. doi:10.1016/j.dr.2018.02.001

Elsaesser

Russell

Ohannessian

C. M.

Patton

(2017). Parenting in a digital age: A review of parents’ role in preventing adolescent cyberbullying. Aggression and Violent Behavior, 35, 62-72. doi:10.1016/j.avb.2017.06.004

10.

Enders

C. K.

Tofighi

(2007). Centering predictor variables in cross-sectional multilevel models: A new look at an old issue. Psychological Methods, 12, 121-138. doi:10.1037/1082-989X.12.2.121

11.

Finkelhor

Ormrod

Turner

(2007). Poly-victimization: A neglected component in child victimization. Child Abuse & Neglect, 31, 7-26. doi:10.1016/j.chiabu.2006.06.008

12.

Gini

Marino

Xie

J.-Y.

Pfetsch

Pozzoli

(2019). Associations of traditional and peer cyber-victimization with adolescents’ Internet use: A latent profile analysis. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 13(4), 1. doi:10.5817/CP2019-4-1

13.

Gradinger

Strohmeier

(2018). Cyberbullying prevention within a socio-ecological framework: The ViSC Social Competence Program. In Campbell

Bauman

(Eds.), Reducing cyberbullying in schools: International evidence-based best practices (pp. 189-202). London, England: Elsevier Academic Press.

14.

Hawdon

Oksanen

Räsänen

(2017). Exposure to online hate in four nations: A cross-national consideration. Deviant Behavior, 38, 254-266. doi:10.1080/01639625.2016.1196985

15.

Hong

J. S.

Espelage

D. L.

(2012). A review of research on bullying and peer victimization in school: An ecological system analysis. Aggression and Violent Behavior, 26, 311-322. doi:10.1016/j.avb.2012.03.003

16.

Hox

Moerbeek

van de Schoot

(2018). Multilevel analysis: Techniques and applications (3rd ed.). New York, NY: Routledge.

17.

Jones

L. M.

Mitchell

K. J.

Turner

H. A.

Ybarra

M. L.

(2018). Characteristics of bias-based harassment incidents reported by a national sample of U.S. adolescents. Journal of Adolescence, 65, 50-60. doi:10.1016/.adolescence.2018.02.013

18.

Kowalski

R. M.

Guimetti

G. W.

Schroeder

A. N.

Lattanner

M. R.

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

19.

Kowalski

R. M.

Limber

S. P.

McCord

(2019). A developmental approach to cyberbullying: Prevalence and protective factors. Aggression and Violent Behavior, 45, 20-32. doi:10.1016/j.avb.2018.02.009

20.

Livingstone

Helsper

E. J.

(2008). Parental mediation of children’s Internet use. Journal of Broadcasting & Electronic Media, 52, 581-599. doi:10.1080/08838150802437396

21.

López-Castro

Priegue

(2019). Influence of family variables on cyberbullying perpetration and victimization: A systematic literature review. Social Sciences, 8, 98. doi:10.3390/socsci8030098

22.

Muthén

B. O.

Muthén

L. K.

Asparouhov

(2016). Regression and mediation analysis using Mplus. Los Angeles, CA: Muthén & Muthén.

23.

Muthén

L. K.

Muthén

B. O.

(1998-2017). Mplus user’s guide (8th ed.). Los Angeles, CA: Author.

24.

Navarro

Serna

Martínez

Ruiz-Oliva

(2013). The role of Internet use and parental mediation on cyberbullying victimization among Spanish children from rural public schools. European Journal of Psychology of Education, 28, 725-745. doi:10.1007/s10212-012-0137-2

25.

Newman

P. A.

Fantus

(2015). A socio-ecology of bias-based bullying of sexual and gender minority youth: Toward a conceptualization of conversion bullying. Journal of Gay and Lesbian Social Services, 27, 46-63. doi:10.1080/10538720.2015.988315

26.

Olweus

(2012). Cyberbullying: An overrated phenomenon? European Journal of Developmental Psychology, 9, 520-538. doi:10.1080/17405629.2012.682358

27.

Olweus

Limber

S. P.

(2018). Some problems with cyberbullying research. Current Opinion in Psychiatry, 19, 139-143. https://dx-doi-org.web.bisu.edu.cn/10.1016/j.copsyc.2017.04.012

28.

Paus-Hasebrink

Kulterer

Sinner

(2019). Social inequality, childhood and the media: A longitudinal study of the mediatization of socialization. Cham, Switzerland: Springer.

29.

Rigby

Bauman

(2010). How school personnel tackle cases of bullying: A critical examination. In Jimerson

S. R.

Swearer

S. M.

Espelage

D. L.

(Eds.), Handbook of bullying in schools: An international perspective (pp. 455-467). New York, NY: Routledge/Taylor & Francis Group.

30.

Russell

S. T.

Sinclair

K. O.

Poteat

V. P.

Koenig

B. W.

(2012). Adolescent health and harassment based on discriminatory bias. American Journal of Public Health, 102, 493-495. doi:10.2105/AJPH.2011.300430

31.

Sasson

Mesch

(2017). The role of parental mediation and peer norms on the likelihood of cyberbullying. The Journal of Genetic Psychology, 178, 15-27. doi:10.1080/00221325.2016.1195330

32.

Schumann

Craig

Rosu

(2013). Minority in the majority: Community ethnicity as a context for racial bullying and victimization. Journal of Community Psychology, 41(8), 959-972. doi:10.1002/jcop.21585

33.

Statistik Austria. (2020). Migration & integration. zahlen. daten. indikatoren 2019 [Migration & integration. numbers. data. indicators 2019]. Vienna, Austria: Author.

34.

Strohmeier

Gradinger

(2021). Teachers’ knowledge and intervention strategies to handle hate-postings. European Journal of Developmental Psychology. Advance online publication. doi:10.1080/17405629.2021.1877130

35.

van der Zanden

P. J.

Denessen

E. J.

Scholte

R. H.

(2015). The effects of general interpersonal and bullying-specific teacher behaviors on pupils’ bullying behaviors at school. School Psychology International, 36, 467-481. doi:10.1177/0143034315592754

36.

Wachs

Schubarth

Bilz

(2020). Hate speech als Schulproblem? Erziehungswissenschaftliche Perspektiven auf ein aktuelles Phänomen [Hate speech as school problem? Educational science perspectives on an actual phenomenon] In van Ackeren

(Ed.), Bewegungen. Beiträge zum 26. Kongress der Deutschen Gesellschaft für Erziehungswissenschaft [Movements. Contributions to the 26th congress of the German society for educational science] (pp. 223-236). Leverkusen, Germany: Verlag Barbara Budrich.

37.

Wachs

Wright

M. F.

(2018). Associations between bystanders and perpetrators of online hate: The moderating role of toxic online disinhibition. International Journal of Environmental Research and Public Health, 15, 2030. doi:10.3390/ijerph15092030

38.

Wachs

Wright

M. F.

(2019). The moderation of online disinhibition and sex on the relationship between online hate victimization and perpetration. Cyberpsychology, Behavior, and Social Networking, 22, 300-306. doi:10.1089/cyber.2018.0551

39.

Wachs

Wright

M. F.

Gámez-Guadix

Görzig

Schubarth

(2020). How do adolescents cope with cyberhate? Psychometric properties and socio-demographic differences of a coping with cyberhate scale. Computers in Human Behavior, 104, 106-167. doi:10.1016/j.chb.2019.106167

40.

Wright

M. F.

(2017). Parental mediation, cyberbullying, and cybertrolling: The role of gender. Computers in Human Behavior, 71, 189-195. doi:10.1016/j.chb

41.

Yoon

Bauman

(2014). Teachers: A critical but overlooked component of bullying prevention and intervention. Theory Into Practice, 53, 308-314. doi:10.1080/00405841.2014.947226

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB