Bystanders of ethnic victimization: Do classroom context and teachers' approach matter for how adolescents intend to act?

The study examined how adolescents' individual characteristics and class context are related to bystander behaviors in cases of ethnic victimization. The sample included 1065 adolescents in Sweden (Mage  = 13.12, SD = 0.42; 55%males). Female adolescents, adolescents of immigrant background, and adolescents with positive attitudes toward immigrants had greater intentions to defend and comfort victimized peers. Positive inter-ethnic contact norms in class were positively associated with intention to comfort the victim. Teachers' non-tolerance of ethnic victimization was positively related to adolescents' intentions to ask the perpetrator to stop and talk to teacher. The effects were the same across adolescents with different attitudes toward immigrants. Findings highlight the importance of class context and teachers in fostering adolescents' prosocial and assertive interventions in bias-based hostile behaviors.

intraclass correlation coefficient

Markov Chain Monte Carlo

Today's youth are growing up in increasingly diverse societies due to significant waves of immigration and globalization. The context provides young people with opportunities to engage in different perspectives and to become familiar with different cultural customs and practices. At the same time, it also poses challenges. Studies from immigrant‐receiving countries have shown that some young people discriminate against or victimize peers who differ from themselves in ethnic or cultural origin (Bayram Özdemir & Özdemir, 2020; Bayram Özdemir et al., 2020; Caravita et al., 2020). Such negative treatments inevitably have consequences for the psychological health and adjustment of minority and immigrant youth. Despite substantial evidence showing the harmful consequences of ethnicity‐based negative treatments on victims (see Benner et al., 2018 for a meta‐analytical review) and recent efforts to identify the characteristics of perpetrators (e.g., Bayram Özdemir et al., 2020), little attention has been paid to the problem from a bystander perspective. The perpetrator and the victim are the two main actors in a case of ethnic victimization. However, a large group of young people at school are neither perpetrators nor victims, but rather bystanders. Some of these bystanders may take action to support the victim, whereas others may prefer to stay passive or even provide support to the perpetrator, explicitly or implicitly. The action (or lack of action) of bystanders sends an implicit message to both perpetrators and victims about the acceptability of their behaviors. Thus, the reactions of bystanders may be a key to understanding the prevalence of, and changes in, ethnic victimization in a particular setting. Considering this important gap in knowledge, the current study focused on early adolescence (the age ranges for this period are 10–14 years) and examined the extent to which early adolescents' individual characteristics (i.e., gender, immigrant background, and attitudes toward immigrants) and their class context (i.e., positive inter‐ethnic contact norms and teachers' non‐tolerance of ethnic victimization) are related to their prosocial and assertive intentions (i.e., comfort the victim, talk to the teacher, and ask the perpetrator to stop) to intervene in ethnic victimization incidents at school. We also examined whether the possible effects of class context on youth's intended reactions vary across adolescents with different levels of tolerance toward immigrants within the Swedish cultural context.

Similar to other immigrant‐receiving countries in Europe, population transformations are also noticeable in Sweden. New ethnic and cultural groups are migrating to Sweden for various reasons, including escaping from war and political oppression, getting a better education, and finding a better job. From 2016 to 2020, more than half million people moved to Sweden from different parts of the world. The current rate of people with foreign background is 25.88%, and youth with foreign background comprises 26% of the students who are between the ages of 13 to 15 (http://www.scb.se). Sweden has been presented as one of the most successful European countries in terms of integration, including access to education and health services, labor market mobility, and the receipt of permanent residence (MIPEX, 2020). Despite its encouraging profile on migration policies, growing anti‐immigrant ideologies in Europe have led to increased polarization in Sweden. For example, the vote of a far‐right nationalist political party with an anti‐immigrant discourse has increased since 2006. Further, in a recent report by Sweden's public health agency, one in five school‐aged and recently‐arrived immigrant children (12–18 year‐olds) stated that they sometimes, often or very often had felt poorly treated because of their background (Folkhalsomyndigheten, 2019). Together, these findings indicate that polarization might also have consequences for how young people of diverse backgrounds interact with each other and how they respond to bias‐based negative treatments in schools. These issues require further investigation.

Bystander responses to ethnic victimization: state of current knowledge

The existing literature contains a variety of theoretical approaches to explaining how young people act when they witness peer victimization that targets ethnic or cultural background (Palmer & Abbott, 2018). The first theoretical approach emphasizes the role of intergroup contact (Allport, 1954). It maintains that the availability of social contact between people of different backgrounds may create a base for the formation of positive intergroup attitudes (Chen & Graham, 2015; Kelleghan et al., 2019) and shape the way children and adolescents reason about, and respond to, ethnicity‐based negative treatments and discriminatory behaviors (Abbott & Cameron, 2014; Gönültaş & Mulvey, 2020; Palmer et al., 2017). For example, Abbott and Cameron (2014) asked secondary school children (11–13 year‐old) in UK how they would respond if they witnessed bias‐based name calling in their school. They found that students with greater interethnic contact were more likely to take assertive action to intervene in the situation, e.g., by comforting the victim, by asking the bully to stop, or by reporting an incident to the teacher. A similar finding was reported in a more recent study. Specifically, Gönültaş and Mulvey (2020) showed that early adolescents and adolescents (11–16 year‐old) who had greater contact with immigrants were more likely to talk to the victim after a bias‐based bullying incident and to say something to the perpetrator. As highlighted by Abbott and Cameron (2014), it is likely that, when young people have more intergroup contact, they have greater opportunities to engage in diverse perspectives. Such opportunities may help them to develop a better understanding of the perspectives and experiences of peers of diverse background, and become more flexible in their approach to different views. Thus, these young people may be more inclined to take assertive and prosocial action when they witness ethnicity‐based negative treatments in their peer settings.

The second theoretical approach emphasizes the role of social group affiliation and intergroup processes, capitalizing mainly on the premises of self‐categorization (Turner et al., 1987) and social identity theory (Tajfel & Turner, 1986). This theoretical framework postulates that, starting from an early age, young people seek an answer to the question of who they are, and start forming their identities on the basis of different categories, including ethnicity and nationality. Relatedly, they also perceive others' identities in comparison with their own identity, and evaluate the similarities and dissimilarities between them. As a result, they develop a set of beliefs about and attitudes to others, which may form the motivational grounds for how young people of diverse backgrounds interact with each other (Bayram Özdemir et al., 2018, 2020), and also for how they act in relation to unpleasant incidents in peer settings (Abbott & Cameron, 2014; Gönültaş & Mulvey, 2020). In line with this reasoning, recent findings suggest that youth's social group affiliation (Gönültaş & Mulvey, 2020; Palmer et al., 2017), and also their views on outgroup members, are associated with their intended bystander responses to verbal racism and bias‐based bullying (Abbott & Cameron, 2014). For example, early adolescents and adolescents of non‐immigrant background have been found to be more passive in response to verbal racism at school compared with ethnic minority adolescents (Palmer et al., 2017), and to be more likely to intervene when the victim is a non‐immigrant peer (Gönültaş & Mulvey, 2020). By contrast, early adolescents and adolescents of immigrant background (11–16 year old) are more likely to actively respond (e.g., say something to the perpetrator, talk to the victim, or get help from someone else) in cases of bias‐based victimization compared with their peers of non‐immigrant background (Gönültaş & Mulvey, 2020). In addition to social group affiliation, youth's views on diversity in general and their beliefs about out‐group members in particular seem to play a key role in their bystander responses, such that early adolescents who are open to cultural differences and have low ingroup bias take stronger assertive action to intervene in cases of bias‐based name calling (Abbott & Cameron, 2014). Together, these findings suggest that early adolescents and adolescents of minority background or those with tolerant attitudes may perceive ethnicity‐based negative treatments or discriminatory behaviors as more severe and unacceptable, and thus are more willing to challenge these incidents.

A third conceptual approach focuses on the social‐cognitive processes underpinning bystanders' responses (Neto & Pedersen, 2013; Palmer & Abbott, 2018). The main element in this approach is that youth differ from one another with regard to their social‐cognitive skills, and consequently show variations in how they interpret social cues, in how they interpret the intentions of perpetrators, and in how they understand the emotions of victims. Such differences in social‐cognitive skills may result in variations in adolescents' responses to ethnicity‐based negative interactions in peer settings (Palmer & Abbott, 2018). For example, recent findings have shown that young people with greater emphatic skills or with a more advanced “theory of mind” (which is conceptualized as an ability to see the mental states of self and others) are more likely to take assertive action to intervene in cases of verbal racism (Abbott & Cameron, 2014) or bias‐based bullying (Gönültaş & Mulvey, 2020) by seeking help from others and talking to the victim. It is likely that greater emphatic capacity and a good theory of mind may help young people to understand and feel what their peers may be going through when they experience bias‐based negative treatments, and thus are more willing to stand up against the injustice.

These three prevailing conceptual arguments and associated empirical studies have provided valuable insights into how social group affiliation, intergroup processes, and social‐cognitive processes may promote or hinder young people's assertive and prosocial responses to ethnicity‐based negative treatments. However, they have all overlooked the possible effects of contextual factors. The social‐ecological model of peer victimization (Hong & Espelage, 2012) indicates that contextual factors (e.g., class climate), and also interactive associations between individual and contextual factors, may contribute to the formation of bystander behaviors. Supporting the premises of this model, studies in the bullying literature have shown that descriptive norms (i.e., what is commonly done) and injunctive norms (i.e., what is commonly approved) in class at school are related to bystanders' responses to it in early adolescence and adolescence (e.g., Pozzoli et al., 2012; Salmivalli & Voeten, 2004; Thornberg et al., 2017, 2018). For example, it has been found that when early adolescents are in a class where anti‐bullying norms are strongly emphasized (Pozzoli et al., 2012) or when children and early adolescents (9–13 year‐old) belong to a class where students have friendly, supportive, and respectful relationships with each other (Thornberg et al., 2017), they are more likely to defend the victim (Pozzoli et al., 2012). Together, these findings indicate that the norms in and climate of the classroom setting may contribute to how youth act when they witness their classmates being victimized by others. However, to our knowledge, no prior research has examined the role of class context in bystanders' responses to ethnicity‐based victimization. Using the social‐ecological model of peer victimization as a theoretical framework (Hong & Espelage, 2012), we aimed to address this important gap in knowledge. Specifically, we examined whether positive inter‐ethnic contact norms in class and teachers' non‐tolerance of ethnic victimization contribute to early adolescents' prosocial and assertive bystander behaviors, and whether these associations vary in accordance with their attitudes toward immigrants.

Positive inter‐ethnic contact norms in class and bystander responses to ethnic victimization

Young people differ from one another with regard to their views on diversity and their inter‐ethnic attitudes and behaviors. The norms emphasized in their social context (i.e., classroom) seem to play an essential role in how early adolescents (13 year‐old) view others who differ from themselves (e.g., Bayram Özdemir, Özdemir, & Boersma, 2021) and in how they interact with each other in diverse settings (e.g., Bayram Özdemir & Özdemir, 2020; Schachner et al., 2015). For example, it has been shown that perceived positive inter‐ethnic contact norms in class (e.g., being inclusive, respecting each other, cooperating in class activities) are associated with more openness to diversity (Schachner et al., 2021), greater intercultural competence (Schwarzenthal et al., 2019), and a higher likelihood of forming interethnic friendships (Schachner et al., 2015). These norms are also related to a lower likelihood of holding prejudiced beliefs (Molina & Wittig, 2006) and engagement in ethnic victimization among early adolescents (Bayram Özdemir & Özdemir, 2020). As Bayram Özdemir and Özdemir (2020) argue, when youth perceive that students in their classes are open to diverse views, respect each other's cultural values, and cooperate with each other in different class activities, they may be more hesitant to victimize or discriminate against their peers, so as to avoid social sanctions.

Positive inter‐ethnic contact norms in class may also facilitate the development of “we‐ness” and contribute to a common in‐group identity (Gaertner et al., 1993). Students may then not perceive classmates who are different from themselves as a threat, but rather as offering an opportunity to learn more about diverse perspectives. Relatedly, they may be more willing to be inclusive in their social interactions (Schachner et al., 2015). Further, such context may help youth feel more efficacious and confident in standing up to any negative treatments targeting their minority peers; they will probably sense that their actions are likely to be appreciated and supported by the others in their class. However, to our knowledge, no previous research has examined whether and how inter‐ethnic contact norms in class are related to early adolescents' bystander responses to ethnic victimization. More importantly, for whom the norms are most influential is unknown. Relying on the premises of the social‐ecological model of peer victimization (Hong & Espelage, 2012), we argue that an optimal social context (i.e., positive inter‐ethnic contact norms) may encourage early adolescents to adopt a position against bullies by confronting the bully, reporting the incident to the teachers, and comforting their victimized peer(s). Confronting the bully might be more socially detrimental for defenders than providing emotional support to the victim (Reijntjes et al., 2016), and thus this behavioral action might be more sensitive to contextual factors. However, a recent study focusing on children and early adolescents in the Netherlands (9–14 year old) showed that class norms similarly effected both bully‐oriented (e.g., confronting the bully) and victim‐oriented defending behaviors (e.g., supporting the victim). Specifically, in classrooms where bullies were more rejected/disliked, early adolescents both confronted the bully and supported the victim (Garandeau et al., 2019). Taking these findings into consideration, we decided not to propose differential hypotheses for the possible effect of class context on different bystander behaviors. We expect that early adolescents would be more likely to defend and comfort their ethnically victimized peers when they were in a class where positive inter‐ethic contact norms were emphasized. Further, we expect that this pattern of association may be especially true for those with high social capital (i.e., positive attitudes toward immigrants) because these youth might internalize and act in accordance with class norms to a greater extent partly due to the high degree of overlap between their own attitudes and class norms.

Teachers' non‐tolerance of ethnic victimization and bystander responses to ethnic victimization

Teachers have the opportunity to oversee interactions among their students on a day‐to‐day basis, and may foster positive interactions among children and adolescents of diverse backgrounds by creating a supportive and inclusive environment (Geerlings et al., 2017). They also have the opportunity to intervene in cases of negative interaction between students in general (Demol et al., 2020; Saarento et al., 2015; Yoon & Bauman, 2014), and between students of diverse backgrounds in particular (Bayram Özdemir & Özdemir, 2020; Bayram Özdemir, Özdemir, & Elzinga, 2021; Closson et al., 2014). For example, Saarento and colleagues (2014) showed that early adolescents (10–12 year‐old) who perceive their teacher as clearly disapproving of bullying are less likely to engage in such behavior. By contrast, early adolescents are more likely to engage in bullying in classrooms where teachers are perceived as paying no attention to bullying. Similar findings have also been reported in relation to ethnicity‐based peer victimization. Specifically, in a recent study focusing on seventh grade students in Sweden, Bayram Özdemir and Özdemir (2020) reported that when early adolescents perceived their teachers as not tolerating ethnic victimization, the students with a high tolerance of immigrants were less likely to engage in victimization of this kind. Taken together, these findings highlight the importance of teachers' approaches and behaviors in fostering positive interactions and in counteracting negative interactions among youth of diverse backgrounds.

Teachers' approaches and behaviors may not only impact how youth of diverse backgrounds interact with each other, but may also affect early adolescents' bystander responses to ethnicity‐based victimization. Specifically, when teachers clearly communicate an expected social behavior to students (e.g., non‐tolerance of ethnic victimization), the students are probably less inclined to justify the behavior, and more likely to take assertive action to prevent it. However, if teachers ignore or trivialize negative incidents, their behaviors may be perceived as an implicit acceptance of ethnic victimization. In such a classroom context, students may be more reluctant to defend or comfort an ethnically victimized peer (Demol et al., 2020; Yoon & Bauman, 2014). However, to our knowledge, no previous research has examined whether and how teachers' responses to ethnic victimization are related to early adolescents' bystander responses. To address this gap in knowledge, we aimed to examine whether teachers' non‐tolerance of ethnic victimization is related to early adolescents' prosocial and assertive responses to ethnic victimization. We expected that, when early adolescents are in classes where teachers do not tolerate ethnicity‐based victimization, they would have a stronger intention to defend and comfort their victimized peers. This would be especially true of early adolescents with positive attitudes toward immigrants, because these young people are probably more receptive to their teachers' messages given that their own personal views about immigrants largely match what the teachers convey to the class.

The current study

Early adolescence is marked by changes in physical, hormonal, social, and cognitive processes that might have a profound impact on later life. Specifically, young people (on average) start to develop enhanced social cognitive competencies (e.g., emphatic reasoning, perspective taking, and prosocial moral judgment; Eisenberg et al., 2005; Fabes et al., 1999). They also continue to explore the self and form identity on the basis of social categories (McLean & Syed, 2015), and demonstrate a greater sensitivity to peer relationships (Brown & Larson, 2009). Further, they become increasingly engaged in problem behaviors during this developmental period (Jennings & Reingle, 2012), partly due to experience of the maturity gap. These cognitive, social, and behavioral changes may have implications for how young people think and engage the world. More specifically, these changes may impact youth's views and reasoning about their peers' actions, and eventually contribute to how they respond to these actions. Thus, developing a comprehensive understanding of why early adolescents respond to ethnic victimization incidents in certain ways would be informative in identifying the means to promote prosocial and assertive bystander behaviors early on. Relatedly, the present study focused on the period of early adolescence, and aimed to advance our understanding of the factors that may contribute to early adolescents' bystander responses to incidents of verbal ethnic victimization (operationalized as making fun of or teasing another student on the ground of ethnic or cultural background). We posed three research questions.

Our first research question was to examine the extent to which early adolescents' individual characteristics (i.e., gender, immigrant background, and attitudes toward immigrants) are related to their bystander behaviors. Based on previous research, we expected that females (Palmer et al., 2017), adolescents of immigrant background (e.g., Gönültaş & Mulvey, 2020), and adolescents with positive attitudes toward immigrants (Abbott & Cameron, 2014) would have stronger intentions to defend and comfort their ethnically victimized peers. Our second research question was to examine the extent to which early adolescents' classroom context (i.e., the inter‐ethnic climate of the class and teachers' non‐tolerance of ethnic victimization) is related to their bystander behaviors. Relying on the premises of the social‐ecological model of peer victimization (Hong & Espelage, 2012) and previous research on inter‐ethnic relationships (e.g., Bayram Özdemir & Özdemir, 2020; Bayram Özdemir et al., 2019; Schachner et al., 2015) and bullying (e.g., Demol et al., 2020; Pozzoli et al., 2012; Saarento et al., 2015; Salmivalli & Voeten, 2004), we hypothesized that youth would be more likely to defend and comfort their ethnically victimized peers when they were in a class where positive inter‐ethic contact norms were emphasized, and/or where their teachers did not tolerate ethnicity‐based victimization. Our third research question was to examine whether the possible effects of classroom context on youth's intended actions vary across adolescents with different levels of positive attitudes toward immigrants. Synthetizing the premises of the social‐ecological model of peer victimization (Hong & Espelage, 2012) and social identity theory (Tajfel & Turner, 1986), we expected that the effects of positive inter‐ethnic class norms and teachers' non‐tolerance of ethnic victimization on adolescent's assertive bystander behaviors would be greater among those with high positive attitudes toward immigrants than among those with low positive attitudes.

METHOD

Participants

The sample for the present study comes from the first wave of an ongoing 3‐year longitudinal study, the Youth and Diversity Project, which examines whether and in which ways school context plays a role in the development of positive and negative relationships among youth of diverse background. The Youth and Diversity Project has been implemented in 55 classrooms across four medium‐sized cities in Sweden, and the target sample has included seventh grade students. Although there are some minor differences, students generally start seventh grade at age 13 in Sweden. Like in many other countries, they have a different teacher for each subject (on average a total of 10–15 different teachers). The sizes of classes vary across schools, but the classes typically include 20 to 30 students. The target sample of the current study was 1286 seventh grade students. Of the target sample, 17% did not participate in the study for various reasons, including parents' disapproval of participation, lack of consent from the adolescents, and non‐attendance during data collection. A total of 1065 adolescents participated in the study (M age = 13.12, SD = 0.42; 55% males). The participation rate across classes ranged from 62% to 96% (M = 84.62, SD = 8.01). Out of 55 classes, 49 had a participation rate of 75% or above. Among the participating adolescents, a majority (71%) came from intact families and were living with both parents (71%). Nearly all of the adolescents (97%) had siblings. More than two‐thirds of them reported that their parents were working (86% of mothers, and 92% of fathers). More than half (60%) had Swedish‐born parents, while the rest (40%) had at least one parent born outside Sweden and were defined as youth of immigrant background. Among the youth of immigrant background, 58% were born in Sweden (defined as second generation immigrants), and the rest (42%) were first generation immigrants. The parents of these youth had migrated to Sweden from around 60 different countries, including Iraq, Iran, Somalia, Syria, Pakistan, Turkey, Bosnia, Kosovo, Germany, India, Italy, and the Netherlands. Adolescents varied regarding use of the Swedish language at home with their parents. About a quarter of the immigrant youth (26%) reported speaking Swedish at home with their parents, while another quarter (26%) reported speaking another language at home. About half (48%) reported that they sometimes spoke Swedish and sometimes another language at home. More than one‐third (42%) reported that they attended a native language course inside or outside school.

Procedure

A research manager and trained research assistants oversaw the data collection in the fall of 2018. The data collection took place in class and took about 90 min. Before the data collection, a letter with information about the project was sent to parents, and parents were asked to sign and return a form if they refused to allow their children to participate in the study. Not returning the form in the information letter was interpreted as giving consent (i.e., passive consent). This procedure for obtaining consent is frequently used in developmental studies to increase participation and reduce sampling bias (Pokorny et al., 2001; Shaw et al., 2015). During the data collection, students were informed about the goals of the study, and were assured that their participation was voluntary, and that their responses would be confidential and not shared with anyone. Only the students whose parents did not decline their children's participation, and who themselves were willing to participate, took part in the study. The questionnaire was administered in Swedish, but children with language difficulties (less than 2%) received help from the research assistants in reading the questions. Most of these students received help from research assistants who spoke their language. In cases where there was no bilingual research assistant available, they received help in Swedish. A sum of 500 Swedish crowns was given to each class in recognition of participation, and the students were provided with snacks during data collection. The Regional Research Ethics Committee in Uppsala approved the study procedures.

Measures

Adolescents' positive attitudes toward immigrants

The Tolerance and Xenophobia Scale (van Zalk et al., 2013) was used to measure adolescents' positive attitudes toward immigrants. The scale consists of six items including: “Immigrants should have the same social rights as people born in Sweden” and “It is good for the Swedish economy that people move to Sweden”. Adolescents were asked to report on the extent to which they agreed or disagreed with these statements on a 5‐point Likert scale, ranging from “1” (strongly disagree) to “5” (strongly agree). The scale has been found to have high internal consistency and predictive validity (e.g., van Zalk et al., 2013). In the present study, Cronbach's alpha for the scale was .84.

Positive inter‐ethnic contact norms and cooperation in class

A revised version of the Classroom Cultural Diversity Climate scale was used to measure perceived positive inter‐ethnic contact norms and cooperation in class (Bayram Özdemir & Özdemir, 2020; Schachner et al., 2021). Adolescents were presented with five items and were asked to report on how true these statements were in their classroom environment on a 5‐point scale ranging from “1” (not true at all) to “5” (completely true). Sample items are: “Students in my class are open to viewpoints different from their own,” and “Students in my class respect each other's cultural values and customs.” Cronbach's alpha for the five items was .81 in the present study. Adolescents' responses on this scale were aggregated to measure classroom‐level positive inter‐ethnic contact norms.

Teachers' non‐tolerance of ethnic victimization

A two‐item measure was used to assess adolescents' perceptions of teachers' non‐tolerance of ethnic victimization (Bayram Özdemir & Özdemir, 2020). The items on the scale are: “Our teachers make it clear that no‐one can make negative comments about others because of their appearance, culture, or religion” and “Our teachers show their disapproval when they see/hear anyone making negative comments about another student because of her/his appearance, culture, or religion.” Students were asked to report on how true these statements were in their classroom environment on a 5‐point scale ranging from “1” (not true at all) to “5” (completely true). These two items were positively and strongly correlated with each other (r = .52, p < .001). Further, following the recommendation by Eisinga et al. (2013), we computed the Spearman‐Brown coefficient. The coefficient value was .87, which suggests that the scale is reliable. Adolescents' responses on this scale were aggregated to measure classroom‐level teachers' non‐tolerance of ethnic victimization.

Adolescents' responses to ethnic victimization

A revised version of the Participant Role Questionnaire (Salmivalli & Voeten, 2004) was used to assess how adolescents would respond if they observed an ethnic victimization incident at school. Adolescents were presented with a stem question (What would you do if one or more students at your school made fun of or teased another student because of her/his appearance, ethnic background, or religion?), and were asked how likely they would be to intervene in relation to the incident. They were provided with the following three statements: “I would try to comfort the student who is teased” “I would go and tell a teacher,” and “I would tell the others to stop making fun of him or her.” Then, they were asked to rate the likelihood of each item on a 5‐point scale ranging from “1” (not at all likely) to “5” (very likely). These three items were used separately as outcome variable in the analyses. Adolescents' responses on this scale were positively correlated with tolerant attitudes toward immigrants (rs range from .25 to .34) and negatively associated with engagement in ethnic victimization (rs range from −.18 to −.23), suggesting concurrent validity of the scale.

Analytic strategy

Multilevel modeling (Hox et al., 2018) at two analytic levels (Level 1: student, Level 2: classroom) was conducted using Mplus version 8.5 (Muthén & Muthén, 1998–2017) to test the hypotheses in the present study. Initially, a null model was fitted to examine the proportion of variance of the outcome variables (i.e., comfort the victim, talk to the teacher, and ask the perpetrator to stop) at student and classroom level (Model 0). Then, a series of models were estimated to test the hypotheses sequentially. First, individual level predictors (i.e., gender, immigrant status, and attitude toward immigrants) were included in the model (Model 1). Second, classroom predictors (i.e., inter‐ethnic contact norms in class, and teachers' non‐tolerance of ethnic victimization) were included (Model 2). Third, the random slope effect was tested for adolescents' attitudes toward immigrants at student level (Model 3). Fourth, cross‐level interactions between adolescents' attitudes toward immigrants and classroom context were included in the model to test for the moderating effect of attitudes toward immigrants on positive inter‐ethnic norms and teachers' non‐tolerance of ethnic victimization (Model 4). Group mean centering was used for all the predictors at the individual level, and grand mean centering for all the predictor variables at classroom level (Enders & Tofighi, 2007). As recommended, cluster means of the student level predictors were included as control variables at the classroom level (Rights et al., 2020).

The Bayesian Markov Chain Monte Carlo (MCMC) estimation method, based on non‐informative prior distributions according to the program's default settings, was applied. Convergence of the model estimation was assessed using the Gelman‐Rubin criterion, with a cut‐off value of 0.01 (see Hox et al., 2012). Mplus uses the Gelman‐Rubin method by default to detect the convergence of Bayesian estimates, which compare within and between chain variability of the parameter estimates (Gelman et al., 2004). Eight chains were requested for the Gibbs sampler, which is a MCMC technique that draws iteratively on a sequence of parameters, latent variables and missing observations to construct the posterior distribution, on the basis of the observed data and specifications of the parameters (Asparouhov & Muthén, 2010). A minimum number of 10,000 iterations were specified. Starting values were based on the maximum likelihood estimates of the model parameters. Trace plots for all parameters were manually inspected to check for convergence.

Missing data

In total, 7.64% of the data, stemming from 175 incomplete records, were missing at individual level, whereas there were no missing values at class level. The percentage of missing values across the six variables on the individual level ranged from 0.00% to 11.27%. Table S1 shows the results of the analysis of missingness. The results showed that missingness on positive attitudes toward immigrants was related to migration background ( = −.32), that missingness on comfort the victim was related to migration background ( = −.41) and talk to the teacher (d = −1.07), that missingness on talk to the teacher was related to gender ( = .29) and migration background ( = .42), and that missingness on ask the perpetrator to stop was related to migration background ( = −.40). Bayesian estimation was used to handle the missing data (Enders, 2010).

RESULTS

Descriptive statistics and preliminary analysis

Means, standard deviations and correlations among the study variables are presented in Table 1. The results of correlation analysis at the student level showed that female adolescents, compared with male adolescents, reported that they would be more likely to comfort the victim, to talk to the teacher, and to ask the perpetrator to stop when they witnessed ethnic victimization at school. Similarly, first generation adolescents of immigrant background reported that they would be more likely to comfort their ethnically victimized classmates and ask the perpetrator to stop than adolescents of non‐immigrant background, while second generation adolescents were more likely to talk to the teacher and ask the perpetrator to stop than adolescent of non‐immigrant background. As expected, positive attitudes toward immigrants were positively associated with a higher level of assertive and prosocial intentions to intervene in ethnic victimization incidents (see the correlation coefficients in the lower triangle in Table 1). Importantly, the results of correlation analysis at the class level also showed that there were significant associations between classroom social climate and adolescents' bystander behaviors. Specifically, inter‐ethnic contact norms in class were positively associated with comforting the victim. Finally, teachers' non‐tolerance of ethnic victimization was associated with adolescents' greater intention to talk to the teacher and to ask the perpetrator to stop (see the correlation coefficients in the upper triangle in Table 1).

TABLE 1

Descriptive statistics, bivariate correlation coefficients and ICCs

Variables	1	2	3	4	5	6	7	8	9	10
1. Gender (0 = boys, 1 = girls)		—	—	−.12	.04	−.08	.04	.13	.53	.31
2. Migration background (0 = Swedish, 1 = first generation immigrant)	.00		—	—	—	—	—	—	—	—
3. Migration background (0 = Swedish, 1 = second generation immigrant)	.00	−.28		—	—	—	—	—	—	—
4. Proportion of students of immigrant background	—	—	—		.45	−.06	−.05	.26	−.04	.12
5. Positive attitudes toward immigrants	.10	.04	.05	—		.29	−.05	.56	.20	.33
6. Inter‐ethnic contact norms in class	—	—	—	—	—		.38	.51	.20	.36
7. Teacher's non‐tolerance of ethnic victimization	—	—	—	—	—	—		.29	.44	.42
8. Comfort the victim	.26	.08	.01	—	.32	—	—		.36	.78
9. Talk to the teacher	.22	−.03	.07	—	.26	.45	—	—		.61
10. Ask the perpetrator to stop	.13	.07	.07	—	.28	.59	.32	—	—
M	0.45	0.24	0.17	0.42	3.61	3.87	3.98	3.75	3.76	3.74
SD					.34	.30	.39	.33	.31	.31
ICC(1)	.05	.03	.17	—	.11	—	—	.09	.06	.07
ICC(2)	.51	.35	.80	—	.69	—	—	.65	.56	.58

Note: N = 1065 students in 55 classes; correlation coefficients at the student level in the lower triangle, and correlation coefficients at the class level in the upper triangle; ICC(1) = intraclass correlation coefficient 1, i.e., proportion of between person variance to the total variance; ICC(2) = intraclass correlation coefficient 2, i.e., reliability of aggregated variable. STATISTICALLY significant results are shown in bold.

Intraclass correlation coefficients (ICCs) based on the null model among the multilevel models showed that between 6% and 9% of the variance in adolescents' responses to ethnic victimization was at classroom level. The variance at school level after accounting for the individual and the class level effects was negligible (ICCs ranged from .001 to .021); thus, school level effects were not estimated in the multilevel models.

Student characteristics and adolescents' bystander responses to ethnic victimization

Individual level predictors were included in Model 1. The results are reported in Tables 2, 3, 4 for the outcome variables: comfort the victim, talk to the teacher, and ask the perpetrator to stop.

TABLE 2

Multilevel modeling results: Comfort the victim

	Model 1		Model 2		Model 3		Model 4
	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.
Fixed effects
Level 1—student level
Gender (0 = boys, 1 = girls)	0.45 (0.06)	0.23	0.46 (0.06)	0.23	0.45 (0.06)	0.23	0.45 (0.06)	0.23
Migration background (0 = Swedish, 1 = first generation immigrant)	0.17 (0.07)	0.07	0.17 (0.07)	0.07	0.18 (0.07)	0.08	0.18 (0.07)	0.08
Migration background (0 = Swedish, 1 = second generation immigrant)	0.03 (0.09)	0.01	0.04 (0.09)	0.01	0.03 (0.09)	0.01	0.03 (0.09)	0.01
Positive attitudes toward immigrants	0.37 (0.04)	0.29	0.37 (0.04)	0.29	0.38 (0.05)	0.29	0.39 (0.05)	0.29
Level 2—class level
Intercept	3.74 (0.05)		3.75 (0.04)		3.75 (0.04)		3.75 (0.04)
Proportion of girls			0.30 (0.29)	0.16	0.33 (0.29)	0.17	0.32 (0.29)	0.17
Proportion of students of immigrant background			0.15 (0.24)	0.11	0.15 (0.24)	0.11	0.16 (0.24)	0.12
Average positive attitudes toward immigrants			0.40 (0.17)	0.43	0.41 (0.17)	0.43	0.41 (0.17)	0.42
Inter‐ethnic contact norms in class			0.36 (0.19)	0.33	0.35 (0.19)	0.31	0.34 (0.19)	0.31
Teacher's non‐tolerance of ethnic victimization			0.16 (0.14)	0.19	0.15 (0.14)	0.18	0.15 (0.14)	0.18
Inter‐ethnic contact norms × positive attitudes toward immigrants							−0.15 (0.25)	−0.16
Teacher's non‐tolerance × positive attitudes toward immigrants							−0.02 (0.18)	−0.03
Random effects
Level 1—student level	0.76 (0.04)		0.76 (0.04)		0.73 (0.04)		0.73 (0.04)
Level 2—class level	0.07 (0.02)		0.04 (0.02)		0.04 (0.02)		0.04 (0.02)
Slope for positive attitudes toward immigrants					0.06 (0.03)		0.06 (0.03)
Model summary
Deviance	7788.00		7721.41		7706.54		7699.33
Total R 2 measure
Rt2f_1	.09		.09		.09		.09
Rt2f_2	.00		.05		.05		.05
Rt2v	.00		.00		.02		.02
Rt2m	.09		.04		.04		.04
Rt2f	.09		.14		.14		.14
Rt2fv	.09		.14		.16		.17
Rt2fvm	.18		.19		.21		.21
Within‐cluster R 2 measure
Rw2f_1	.10		.10		.10		.10
Rw2v	.00		.00		.02		.02
Rw2f_1v	.10		.10		.12		.13
Between‐cluster R 2 measure
Rb2f_2	.00		.55		.54		.54
Rb2m	1.00		.45		.46		.46

Note: N = 1065 students in 55 classes; all class‐level variables were calculated based on students' reports; Est. = unstandardized Bayesian posterior median estimate; SD = standard deviation of the posterior distribution; 95% CI = 95% Bayesian credible interval; Std. Est. = standardized estimate; R 2 measures according to Rights and Sterba (2018) were computed in R (R Core Team) using the package r2mlm (Shaw et al., 2020);  = proportion of total outcome variance explained by level‐1 predictors via fixed slopes;  = proportion of total outcome variance explained by level‐2 predictors via fixed slopes;  = proportion of total outcome variance explained by level‐1 predictors via random slope variation/covariation;  = proportion of total outcome variance explained by cluster‐specific outcome means via random intercept variation;  = proportion of total outcome variance explained by all predictors via fixed slopes;  = proportion of total outcome variance explained by predictors via fixed slopes and random slope variation/covariation;  = proportion of total outcome variance explained by predictors via fixed slopes and random slope variation/covariation and by cluster‐specific outcome means via random intercept variation;  = proportion of within‐cluster outcome variance explained by level‐1 predictors via fixed slopes; ;  = proportion of within‐cluster outcome variance explained by level‐1 predictors via random slope variation/covariation;  = proportion of between‐cluster outcome variance explained by level‐2 predictors via fixed slopes;  = proportion of between‐cluster outcome variance explained by cluster‐specific outcome means via random intercept variation. Statistically significant results are shown in bold.

TABLE 3

Multilevel modeling results: Talk to the teacher

	Model 1		Model 2		Model 3		Model 4
	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.
Fixed effects
Level 1—student level
Gender (0 = boys, 1 = girls)	0.44 (0.07)	0.20	0.43 (0.07)	0.20	0.43 (0.07)	0.20	0.43 (0.07)	0.20
Migration background (0 = Swedish, 1 = first generation immigrant)	−0.07 (0.08)	−0.03	−0.07 (0.08)	−0.03	−0.07 (0.08)	−0.03	−0.08 (0.08)	−0.03
Migration background (0 = Swedish, 1 = second generation immigrant)	0.16 (0.10)	0.05	0.15 (0.10)	0.05	0.15 (0.10)	0.05	0.15 (0.10)	0.05
Positive attitudes toward immigrants	0.32 (0.04)	0.23	0.32 (0.04)	0.23	0.33 (0.06)	0.23	0.33 (0.06)	0.23
Level 2—class level
Intercept	3.76 (0.04)		3.76 (0.04)		3.76 (0.04)		3.76 (0.04)
Proportion of girls			0.83 (0.28)	0.51	0.92 (0.29)	0.53	0.93 (0.29)	0.52
Proportion of students with migration background			−0.07 (0.23)	−0.06	−0.03 (0.24)	−0.03	−0.03 (0.24)	−0.02
Average positive attitudes toward immigrants			0.19 (0.17)	0.23	0.13 (0.17)	0.15	0.14 (0.17)	0.15
Inter‐ethnic class norms			0.00 (0.19)	0.00	−0.02 (0.19)	−0.02	0.03 (0.20)	0.02
Teacher's non‐tolerance of ethnic victimization			0.33 (0.14)	0.45	0.30 (0.14)	0.38	0.33 (0.14)	0.41
Inter‐ethnic contact norms × positive attitudes toward immigrants							−0.26 (0.26)	−0.27
Teacher's non‐tolerance × positive attitudes toward immigrants							−0.15 (0.19)	−0.20
Random effects
Level 1—student level	0.96 (0.05)		0.96 (0.05)		0.93 (0.05)		0.93 (0.05)
Level 2—class level	0.05 (0.02)		0.02 (0.02)		0.03 (0.02)		0.03 (0.02)
Slope for positive attitudes toward immigrants					0.06 (0.03)		0.06 (0.03)
Model summary
Deviance	8804.41		7923.34		7864.37		7856.99
Total R 2 measure
Rt2f_1	.10		.10		.10		.10
Rt2f_2	.00		.02		.02		.02
Rt2v	.00		.00		.02		.02
Rt2m	.04		.02		.02		.02
Rt2f	.10		.12		.12		.13
Rt2fv	.10		.12		.14		.14
Rt2fvm	.14		.14		.16		.16
Within‐cluster R 2 measure
Rw2f_1	.10		.10		.10		.11
Rw2v	.00		.00		.02		.02
Rw2f_1v	.10		.10		.13		.13
Between‐cluster R 2 measure
Rb2f_2	.00		.55		.52		.54
Rb2m	1.00		.45		.48		.46

Note: N = 1065 students in 55 classes; all class‐level variables were calculated based on students' reports; Est. = unstandardized Bayesian posterior median estimate; SD = standard deviation of the posterior distribution; 95% CI = 95% Bayesian credible interval; Std. Est. = standardized estimate; R 2 measures according to Rights and Sterba (2018) were computed in R version 4.1.1 (R Core Team, 2021) using the package r2mlm version 0.3.0 (Shaw et al., 2020);  = proportion of total outcome variance explained by level‐1 predictors via fixed slopes;  = proportion of total outcome variance explained by level‐2 predictors via fixed slopes;  = proportion of total outcome variance explained by level‐1 predictors via random slope variation/covariation;  = proportion of total outcome variance explained by cluster‐specific outcome means via random intercept variation;  = proportion of total outcome variance explained by all predictors via fixed slopes;  = proportion of total outcome variance explained by predictors via fixed slopes and random slope variation/covariation;  = proportion of total outcome variance explained by predictors via fixed slopes and random slope variation/covariation and by cluster‐specific outcome means via random intercept variation;  = proportion of within‐cluster outcome variance explained by level‐1 predictors via fixed slopes; ;  = proportion of within‐cluster outcome variance explained by level‐1 predictors via random slope variation/covariation;  = proportion of between‐cluster outcome variance explained by level‐2 predictors via fixed slopes;  = proportion of between‐cluster outcome variance explained by cluster‐specific outcome means via random intercept variation. Statistically significant results are shown in bold.

TABLE 4

Multilevel modeling results: Ask the perpetrator to stop

	Model 1		Model 2		Model 3		Model 4
	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.	Est. (SD)	Std. Est.
Fixed effects
Level 1—student level
Gender (0 = boys, 1 = girls)	0.24 (0.07)	0.11	0.24 (0.07)	0.11	0.23 (0.07)	0.11	0.24 (0.07)	0.11
Migration background (0 = Swedish, 1 = first generation immigrant)	0.20 (0.08)	0.08	0.21 (0.08)	0.08	0.20 (0.08)	0.08	0.19 (0.08)	0.08
Migration background (0 = Swedish, 1 = second generation immigrant)	0.24 (0.11)	0.08	0.24 (0.11)	0.08	0.24 (0.11)	0.08	0.23 (0.11)	0.08
Positive attitudes toward immigrants	0.36 (0.05)	0.26	0.36 (0.05)	0.26	0.37 (0.06)	0.25	0.37 (0.06)	0.26
Level 2—class level
Intercept	3.73 (0.05)		3.73 (0.04)		3.73 (0.04)		3.73 (0.05)
Proportion of girls			0.53 (0.31)	0.30	0.56 (0.31)	0.31	0.55 (0.31)	0.29
Proportion of students with migration background			0.06 (0.26)	0.05	0.14 (0.25)	0.11	0.15 (0.26)	0.11
Average positive attitudes toward immigrants			0.22 (0.19)	0.25	0.22 (0.19)	0.24	0.22 (0.19)	0.24
Inter‐ethnic class norms			0.20 (0.21)	0.20	0.16 (0.21)	0.15	0.19 (0.22)	0.18
Teacher's non‐tolerance approach to ethnic victimization			0.28 (0.15)	0.36	0.23 (0.15)	0.29	0.27 (0.16)	0.33
Inter‐ethnic contact norms × positive attitudes toward immigrants							−0.16 (0.26)	−0.17
Teacher's non‐tolerance × positive attitudes toward immigrants							−0.22 (0.19)	−0.31
Random effects
Level 1—student level	0.99 (0.05)		0.99 (0.05)		0.97 (0.05)		0.97 (0.05)
Level 2—class level	0.06 (0.02)		0.04 (0.02)		0.05 (0.02)		0.05 (0.02)
Slope for positive attitudes toward immigrants					0.06 (0.03)		0.05 (0.03)
Model summary
Deviance	8022.75		7964.82		7914.25		7899.54
Total R 2 measure
Rt2f_1	.07		.07		.07		.07
Rt2f_2	.00		.02		.02		.02
Rt2v	.00		.00		.01		.01
Rt2m	.04		.02		.03		.03
Rt2f	.07		.09		.09		.09
Rt2fv	.07		.09		.10		.11
Rt2fvm	.11		.11		.13		.13
Within‐cluster R 2 measure
Rw2f_1	.07		.07		.07		.07
Rw2v	.00		.00		.01		.01
Rw2f_1v	.07		.07		.08		.09
Between‐cluster R 2 measure
Rb2f_2	.00		.05		.43		.47
Rb2m	1.00		.05		.57		.53

Note: N = 1065 students in 55 classes; all class‐level variables were calculated based on students' reports; Est. = unstandardized Bayesian posterior median estimate; SD = standard deviation of the posterior distribution; 95% CI = 95% Bayesian credible interval; Std. Est. = standardized estimate; R 2 measures according to Rights and Sterba (2018) were computed in R (R Core Team) using the package r2mlm (Shaw et al., 2020);  = proportion of total outcome variance explained by level‐1 predictors via fixed slopes;  = proportion of total outcome variance explained by level‐2 predictors via fixed slopes;  = proportion of total outcome variance explained by level‐1 predictors via random slope variation/covariation;  = proportion of total outcome variance explained by cluster‐specific outcome means via random intercept variation;  = proportion of total outcome variance explained by all predictors via fixed slopes;  = proportion of total outcome variance explained by predictors via fixed slopes and random slope variation/covariation;  = proportion of total outcome variance explained by predictors via fixed slopes and random slope variation/covariation and by cluster‐specific outcome means via random intercept variation;  = proportion of within‐cluster outcome variance explained by level‐1 predictors via fixed slopes; ;  = proportion of within‐cluster outcome variance explained by level‐1 predictors via random slope variation/covariation;  = proportion of between‐cluster outcome variance explained by level‐2 predictors via fixed slopes;  = proportion of between‐cluster outcome variance explained by cluster‐specific outcome means via random intercept variation. Statistically significant results are shown in bold.

Comfort the victim

The results showed that female adolescents were more likely to comfort the victim than male adolescents ( = .45, 95% CI [.36, ∞]). First generation immigrant youth were more likely to comfort the victim than adolescents of non‐immigrant background ( = .17, 95% CI [.05, ∞]). No significant difference was observed among second generation immigrant youth versus those of non‐immigrant background. Adolescents' positive attitudes toward immigrants positively predicted their intention to comfort the victim ( = .37, 95% CI [.31, ∞]).

Talk to the teacher

The results showed that female adolescents were more likely to talk to the teacher than male adolescents ( = .44, 95% CI [.33, ∞]). No significant difference was observed among adolescents with immigrant versus non‐immigrant background. Adolescents' positive attitudes toward immigrants positively predicted their intention to talk to the teacher ( = .32, 95% CI [.25, ∞]).

Ask the perpetrator to stop

The results showed that female adolescents were more likely to ask the perpetrator to stop than male adolescents ( = .24, 95% CI [.13, ∞]). Adolescents of immigrant background (both first generation and second generation) were more likely to ask the perpetrator to stop than adolescents of non‐immigrant background ( = .20, 95% CI [.07, ∞];  = .24, 95% CI [.07, ∞]; first and second generation immigrant youth respectively). Adolescents' positive attitudes toward immigrants predicted their intention to ask the perpetrator to stop ( = .36, 95% CI [.29, ∞]).

Classroom context and adolescents' bystander responses to ethnic victimization

The classroom level predictors were included in Model 2. The results are reported in Tables 2, 3, 4 for the outcome variables: comfort the victim, talk to the teacher, and ask the perpetrator to stop.

The results showed that the average of positive attitudes toward immigrants was positively related to comforting the victim at classroom level ( = .40, 95% CI [.12, ∞]). Likewise, inter‐ethnic contact norms in class ( = .36, 95% CI [.05, ∞]) were positively related to comforting the victim at classroom level. However, teachers' non‐tolerance of ethnic victimization did not predict adolescents' intention to comfort the victim ( = .16, 95% CI [−.07, ∞]).

The results showed that the proportion of girls in the classroom positively predicted intention to talk to the teacher at classroom level ( = .83, 95% CI [.37, ∞]). Similarly, teachers' non‐tolerance of ethnic victimization was positively related to intention to talk to the teacher at classroom level ( = .33, 95% CI [.10, ∞]). The effect of inter‐ethnic norms in class, however, was statistically non‐significant ( = .01, 95% CI [−.32, ∞]).

The results showed that the proportion of girls in the classroom positively predicted intention to ask the perpetrator to stop at classroom level ( = .53, 95% CI [.02, ∞]). Likewise, teachers' non‐tolerance of ethnic victimization ( = .28, 95% CI [.03, ∞]) was positively related to intention to ask the perpetrator to stop at classroom level. However, inter‐ethnic norms in class did not significantly predict adolescents' intention to ask the perpetrator to stop ( = .20, 95% CI [−.14, ∞]).

Cross‐level interaction between positive attitudes toward immigrants and classroom context

In Model 3 the random slope effect of positive attitudes toward immigrants was tested, so as to enable investigation of the cross‐level interaction between adolescents' attitudes toward immigrants and classroom context in Model 4. The results are reported in Tables 2, 3, 4 for the outcome variables: comfort the victim, talk to the teacher, and ask the perpetrator to stop.

The results of Model 3 showed that the random slope variance for the student level predictor “attitudes toward immigrants” was statistically significant ( = .06, 95% CI [.02, .14]). The results of Model 4, however, did not show a statistically significant cross‐level interaction for inter‐ethnic contact norms in class ( = −.15, 95% CI [−.55, ∞]) or for teachers' non‐tolerance of ethnic victimization ( = −.02, 95% CI [−.31, ∞]).

The results of Model 3 showed that the random slope variance for the student level predictor “attitudes toward immigrants” was statistically significant ( = .06, 95% CI [.02, .14]). The results of Model 4, however, did not show a statistically significant cross‐level interaction for inter‐ethnic contact norms in class ( = −.26, 95% CI [−.68, ∞]) or for teachers' non‐tolerance of ethnic victimization ( = −.15, 95% CI [−.46, ∞]).

The results of Model 3 showed that the random slope variance for the student level predictor “attitudes toward immigrants” was statistically significant ( = .06, 95% CI [.01, .14]). The results of Model 4, however, did not show a statistically significant cross‐level interaction for inter‐ethnic contact norms in class ( = −.16, 95% CI [−.57, ∞]) or for teachers' non‐tolerance of ethnic victimization ( = −.22, 95% CI [−.52, ∞]).

DISCUSSION

Adolescents may witness discrimination or victimization incidents that target their peers' ethnic, cultural, or religious background. They may differ from one another regarding how they act in such situations. Some youth may take action to support victims, whereas others may prefer to stay passive, or may even provide explicit or implicit support to the perpetrator. Obtaining a comprehensive understanding of the factors that may contribute to adolescents' bystander responses to ethnic victimization is vital to the development of strategies to intervene in relation to bias‐based hostile behaviors in diverse school settings. Yet, the factors that play a role in adolescents' bystander behaviors are less well understood. We addressed this gap in knowledge in the present study. Specifically, we examined the extent to which early adolescents' individual characteristics (i.e., gender, immigrant background, and attitudes toward immigrants) and their classroom context (i.e., inter‐ethnic contact norms, and teachers' non‐tolerance of ethnic victimization) are related to their prosocial and assertive intentions to intervene in relation to ethnic victimization incidents at school. We also investigated for whom positive inter‐ethnic contact norms in class and teachers' non‐tolerance of ethnic victimization matter most by focusing on adolescents' attitudes toward immigrants.

In line with our expectations, the findings show that female adolescents have stronger intentions to comfort the victim, to talk to the teacher, and to ask the perpetrator to stop. The observed gender difference is in line with previous research on bullying (Gini et al., 2008; Thornberg et al., 2018) and racism (Palmer et al., 2017), which suggests that female adolescents tend to have positive attitudes toward victims, and relatedly have a greater prosocial intention to comfort their victimized peers. One possible explanation for the observed gender difference may be related to the differences in perspective taking (Tucker Smith et al., 2016) and empathy between males and females (Butrus & Witenberg, 2013). Alternatively, the observed difference could be related to gender socialization. Females are often socialized into being the nurturers, and such socialization experiences might help females demonstrate higher moral sensitivity (e.g., by recognizing the harm caused by bullying and sympathizing with victims) than males (Thornberg & Jungert, 2013), and, in turn, adopt strategies to protect their victimized peers.

In line with conceptual arguments in the literature (Turner et al., 1987) and previous empirical research (Gönültaş & Mulvey, 2020; Palmer et al., 2017), our findings show that youth of immigrant background (especially first generation immigrant youth) also show a greater tendency to comfort the victim and to ask the perpetrator to stop. It is possible that adolescents of immigrant background associate themselves with ethically victimized peers due to their shared migration history. Thus, they can empathize with them to a greater extent than their peers of non‐immigrant background. Such emotional and cognitive awareness may shape their judgments about the acceptability of bias‐based negative treatments (Gönültaş & Mulvey, 2020), and motivate them to take action to comfort and defend their victimized peers.

As expected, and supporting previous research findings (Abbott & Cameron, 2014), we found that adolescents with positive attitudes toward immigrants are also more likely to defend and comfort their ethnically victimized peers. This finding extends the literature on ethnic victimization by providing evidence that views on immigrants may not only form the motivational base of adolescents' engagement in ethnic victimization (Bayram Özdemir et al., 2018, 2020; Caravita et al., 2020) but also play a critical role in how adolescents act when they witness situations of this kind at school. It is likely that adolescents with positive views on immigrants perceive cultural, ethnic, or religious differences as offering an opportunity to learn more about diverse perspectives rather than as a threat to themselves. Accordingly, they are probably less likely to have intergroup anxiety (Abbott & Cameron, 2014; Mähönen et al., 2011), and, in turn, to have fewer cognitive and motivational information‐processing biases and polarized evaluations of outgroup members. They may, instead, focus on fairness and social justice in their moral judgments on problematic social relationships. These adaptive socio‐cognitive processes may foster their ability to evaluate situations fairly, and, in turn, their willingness to defend their victimized peers. Taken together, these findings further highlight the importance of fostering positive attitudes toward immigrants as a target for programs aiming to preventing ethnic victimization at school.

A noteworthy finding of the present study is that it draws attention to the importance of classroom context in understanding adolescents' prosocial and assertive intentions to intervene in relation to ethnic victimization incidents at school. Specifically, our findings suggest that adolescents have a greater intention to comfort the victim when they are in a class where openness to diverse views is strongly emphasized, and where students respect each other's cultural values and cooperate with each other in different class activities. Importantly and interestingly, the effect of positive inter‐ethnic contact norms in class on adolescents' intention to comfort the victim is the same among youth with different attitudes toward immigrants. As previously highlighted, positive inter‐ethnic contact norms in class may facilitate the formation of a common in‐group identity in the class, and contribute to the development of cultural competence (Schwarzenthal et al., 2019) and inclusive behaviors (Schachner et al., 2015) among young people. In a socially cohesive class context, adolescents may be courageous enough to comfort their victimized peers (Thornberg et al., 2018), probably because their intended actions are likely to be well‐regarded by their classmates; thus, they may have less fear of being the target of bullies. Taken together, these findings support the social‐ecological perspective in the peer victimization literature (e.g., Hong & Espelage, 2012; Thornberg et al., 2017) and indicate that an inclusive and socially cohesive class climate may facilitate adolescents' prosocial bystander behaviors. Thus, the classroom climate should be regarded as a vital factor to target in preventing bias‐based hostile interactions at school and in fostering harmonious interactions among youth of diverse backgrounds.

Another important contribution of the present study lies in its examination of whether teachers' non‐tolerance of ethnic victimization impacts youth's bystander behaviors. In line with our expectation and previous research (Demol et al., 2020), the findings suggest that adolescents have a stronger intention to ask perpetrators to stop and to talk to their teachers about victimization incidents when teachers make it clear to their students that no‐one can make negative comments about others because of their background. Importantly, this finding holds across students with different attitudes toward immigrants. Two conceptual explanations are plausible in relation to this finding. First, as highlighted in social learning theory (Bandura & McClelland, 1977), teachers are important role models for children and adolescents at school. They have the ability to influence students' attitudes and behaviors through how they respond to non‐normative behaviors (Demol et al., 2020; Yoon & Bauman, 2014). For example, in their study of early adolescents, Demol et al. (2020) showed that students in a hypothetical condition where the teacher corrected a bully were more willing to report bullying incidents than students in a hypothetical condition where the teacher did not respond to what had happened. Applied to the context of this study, it is possible that when teachers adopt a non‐tolerance approach to ethnic victimization incidents, they act as a role model to students, showing that there is a need to stand up to perpetrators of victimization. By contrast, when they ignore or do not actively intervene in relation to such incidents, they may present a model of insensitive or uncaring behavior to their students. Accordingly, these adolescents may be less inclined to defend their ethnically victimized peers.

Alternatively, teachers' responses to ethnic victimization may affect students' socio‐cognitive processes, and thereby contribute to their judgments on the acceptability of incidents of this kind. More specifically, when teachers set clear expectations concerning the non‐tolerance of ethnic victimization, students may be less disposed mentally to justify this behavior. They may take action corresponding to their cognitive processes, partly due to the elimination of the experience of cognitive dissonance. Supporting this argument, Campaert et al. (2017) showed that when students perceive their teachers as offering comfort to victims, they are less likely to disengage from morality, and, in turn, less likely to engage in bullying. Taken together, it can be argued that teachers' non‐tolerance of ethnic victimization may contribute to adolescents' understanding of the moral essence of ethnic victimization, and, in turn, their engagement in ethnic bullying and related bystander behaviors. These possible explanations, however, need to be tested in future studies in order to draw firm conclusions about why teachers' non‐tolerance of ethnic victimization is related to adolescents' prosocial and assertive bystander behaviors.

As well as its important contributions to the literature, several limitations of the present study need to be acknowledged. First, the study was correlational by nature, and the data captured were from only one time point. This limits our ability to draw conclusions regarding the extent to which the observed associations hold over time. Thus, studies with multiple assessment points are needed to investigate whether and how adolescents' bystander behaviors change over time, and which individual and class level factors explain time‐related change. Second, we used adolescents' self‐reports in the assessments of all the study constructs. This approach may raise some concerns. For example, it is unknown whether adolescents' perceptions of teachers' behaviors accurately reflect the actual behaviors of teachers. Further, adolescents were asked to report on non‐tolerance of ethnic victimization among teachers in general, rather than reporting their views about every single teacher. Even though this measurement approach is more feasible and gives us an overall picture, it should be acknowledged that it may lead to a lack of specificity in measurement. That is, it may limit our ability; for example, to test how similar or different approach taken by teachers (e.g., having just one teacher who provides inclusive classroom norms versus having multiple teachers with similar approach) contribute bystander behaviors among early adolescents. Future studies that integrate adolescents' self‐reports and data from multiple teachers may advance the literature in this regard. Third, we examined adolescents' bystander behaviors using hypothetical scenarios, and the adolescents were asked to imagine a student who has been victimized without any further specification (i.e., as to whether the victim was a close friend or a classmate). As stated by the theory of reasoned action (Ajzen & Fishbein, 1980), behavioral intentions are strong determinants of actual actions, especially when the behavior in question is under the individual's voluntaristic control, and when the individual has the resources to engage in the behavior. Even though our measurement approach provides us with valuable information (Chapin & Brayack, 2016; Gönültaş & Mulvey, 2020), we should acknowledge that we can only draw conclusions concerning the adolescents' intended behaviors, not their actual behaviors per se. Further, we had just a single item to measure each intended behavior, i.e., comfort the victim, talk to the teacher, and ask the perpetrator to stop. The validity of the scale used might be questioned given that a single or two‐item scale limits our ability to validate a latent construct measurement (Hair et al., 2006). Qualitative studies where youth reflect on their behaviors as bystanders and motives for their actions, or retrospective studies where youth are asked to recall a recent incident that they have witnessed, would allow us to capture nuances and to examine the issue of bystanding more thoroughly. Alternatively, assessing victims' experiences of bystander interventions using multiple scale items would allow us to determine the characteristics of the interveners. Fourth, the stem question in measuring adolescents' bystander behaviors was restricted to verbal ethnic victimization, and the intersectionality between religion and ethnic identity was not fully captured. This measurement approach may be limited with regard to providing in‐depth information about the variation in the ways youth respond to other forms of ethnic victimization (e.g., social exclusion and physical coercion), and also the differences in adolescents' bystander behaviors between ethnic and religious groups (e.g., a Muslim Arab peer from Syria versus a Christian Arab peer from Syria). Further, the stem question includes information about appearance, ethnic background, or religion. Even though inspection of the pattern of association between this scale and the related constructs (i.e., tolerant attitudes toward immigrants and ethnic victimization) suggests the concurrent validity of the scale, we should acknowledge that appearance in the stem question might act as a possible confound. To sum, future research on adolescents' bystander behaviors across different forms of ethnic victimization that also considers intersectionality between religion and ethnic identity and that clearly defines appearance (e.g., skin color) is needed to develop a comprehensive understanding of bias‐based victimization. Fifth, we examined immigrant adolescents as a relatively homogenous group, which limits our ability to investigate variations in bystander behaviors based on adolescents' ethnic, religious, or socio‐economic background. Further, the proportion of students of immigrant background in each class was used as a proxy for the assessment of classroom ethnic composition in a similar manner to that found in previous research (e.g., Seuring et al., 2021). However, it should be noted that a higher or lower number of members of a specific immigrant group in a class (in addition to the proportion of immigrant students) and the rates and the quality of inter‐ethnic relationship among youth might determine the intention of a bystander, and thus require future investigation. Sixth, our main focus in this study was to examine whether class context contributes to how adolescents act after witnessing ethnic victimization incidents at school. Thus, our findings are limited with regard to explaining why class context plays a role in adolescents' bystander behaviors. As Darley and Latané (1968) highlighted, different psychological processes, including diffusion of responsibility, evaluation apprehension and pluralistic ignorance, might explain kinds of bystanding. It is possible, for example, that teachers' non‐tolerance of ethnic victimization counteracts the development of diffusion of responsibility (e.g., Another student will take action, I do not need to do anything) or pluralistic ignorance (e.g., No‐one does anything, so the situation is not that serious) among adolescents. Therefore, the adolescents might be more willing to defend and help their ethnically victimized peers. Future research is needed to empirically test these alternative explanations through the integration of contextual and social psychological perspectives. Finally, we focused on early adolescents in this study, and thus our findings are only applicable to this developmental period. Given that the effects of school (e.g., the nature of the student–teacher relationship) on young people's views and interactions may vary across different developmental stages, future research is needed to examine whether the current findings are generalizable to older adolescents across different cultural contexts.

Despite these limitations, the important implications of the present study are worth reiterating. The findings clearly suggest that class context and teachers matter in fostering adolescents' prosocial and assertive intentions to intervene in relation to bias‐based hostile behaviors in schools. More specifically, they highlight the importance of creating a class setting where diverse views and values are appreciated and respected in order to facilitate young people's intentions to support and comfort their victimized peers. The findings also suggest that teachers' clear messages of non‐tolerance of ethnic victimization have the potential to promote youth's intentions to seek help and to defend their victimized peers. In sum, these findings highlight the importance of focusing on diversity and inclusion in teacher training, and promoting cultural competence and conflict resolution skills among teachers. Such an approach may be vital in reducing the possibility of negative role modeling, or of condoning ethnic victimization in schools, and in the longer run in promoting cooperation among students of different backgrounds.

ETHICS STATEMENT

The Regional Research Ethics Committee in Uppsala approved the study procedures.

Table S1.

Click here for additional data file.