The Impact of Personal Values, Gender Stereotypes, and School Climate on Homophobic Bullying: a Multilevel Analysis

Schools are among the most homophobic social contexts, where students who do not conform to gender norms are at high risk of stigma and discrimination. Using a multilevel approach, the aim of the current was to examine whether adolescents’ engagement in homophobic bullying behavior was associated with personal values and stereotyped victim-blaming attributions at individual level, and perceptions of school as a community and frequency of teachers’ reaction to bullying incidents at classroom level. Data were collected in 2010. The sample consisted of 2718 Italian middle and high school students (53.2% females; mean age = 15.36, SD = .85) from 144 classrooms. Results showed that self-transcendence values reduced the risk of engaging in homophobic bullying, whereas both self-enhancement values and stereotyped victim-blaming attributions were positively associated with homophobic bullying. At classroom level, only negative perceptions of school as a community had a unique positive contribution on homophobic bullying, over and above other individual and contextual factors. Two cross-level interactions were found, indicating that self-transcendence values had a significant effect in decreasing homophobic bullying in classrooms where teacher support was perceived as low, whereas stereotyped victim-blaming attributions had a significant effect in increasing homophobic bullying in classrooms where teacher support was perceived as high. These findings provide further support to the social-ecological perspective as a useful guiding framework for understanding the complexity of factors predicting homophobic bullying. Efforts should be made to develop clear anti-bullying school policies explicitly dealing with the issue of homophobic bullying.

School bullying is an ongoing concern worldwide. It describes a specific type of aggression involving intentional and repetitive abuses against peers, aimed at causing harm to the victim and operated within an imbalance of power between bullies and victims (Olweus, 2010;Smith, 2004). Bullying can be either direct or indirect and can occur through a variety of actions including physical contact (hitting, pushing), verbal abuse (teasing, name calling), spreading rumors, and social exclusion. Bullying is widespread throughout the world, involving young children and adolescents, as well as boys and girls; its negative consequences on child adjustment have been widely demonstrated both in bullies (Ttofi, Farrington, & Lösel, 2012) and in victims (Esposito, Bacchini, & Affuso, 2019;Hawker & Boulton, 2000).
Discriminatory bullying is a specific type of bullying motivated by a social stigma against individuals because of their physical characteristics (e.g., obesity, disability), ethnicity, religion, and/or their belonging to sexual minority groups (Elamé, 2013). Homophobic bullying (hereafter: HB) is defined as a deliberate action aimed at denigrating or discriminating peers because of their sexual orientation, either perceived or real (Meyer, 2008). As for general bullying, it manifests through direct forms (such as verbal injuries or physical attacks) or indirect modalities (such as spreading rumors or exclusion from group activities), or electronic devices (Hong & Garbarino, 2012). Research found that HB is highly frequent in school contexts, and sexual minority youth experience bullying two times more than their heterosexual peers (Earnshaw et al., 2018). Overall, HB has a prevalence ranging from 22 to 87% (Rodríguez-Hidalgo & Hurtado-Mellado, 2019). Males are generally reported to be more frequently involved than females in HB (Orue & Calvete, 2018;Poteat & DiGiovanni, 2010). However, also girls are highly involved in HB. For instance, in a study carried out in Switzerland (Weber & Gredig, 2018), 76% of female students reported having used anti-gay language in the previous 12 months.
HB and general bullying are only in part overlapping. Both reflect forms of peer aggression, with the former being specifically supported by the adherence to gender stereotyped norms, homophobic attitudes, and discriminatory beliefs against sexual minorities. Homophobia emerges from heteronormative cultural contexts (Butler, 2011) which consider heterosexuality as normal and those who do not conform to traditional gender roles and behaviors as deviant. Thus, peers whose behavior or appearance do not fit into the culturally dominant models of masculinity and femininity are perceived as deviating from socially accepted gender norms, and deserve, for this reason, to be punished and stigmatized . As being perceived as deviating from gender norms is enough for being stigmatized in a context where homophobia is deeply rooted, not only homosexual youth are at risk of homophobic bullying but also those who are thought to be so, or have homosexual or gender non-conforming friends, or are perceived as different in some ways (Rodríguez-Hidalgo & Hurtado-Mellado, 2019). Regardless of their actual sexual identity, youth that experience harassment based on actual or perceived sexual orientation find it particularly distressing (Swearer, Turner, Givens, & Pollack, 2008). According to minority stress theory (Meyer, 2003), indeed, individuals from stigmatized groups are exposed to unique stressors as a result of their marginalized social position. Thus, being target of homophobic victimization emphasizes the social marginalized position of sexual minority youth, but also causes minority stress among heterosexual youth, due to a marginalized minority identity that others have conferred on them (Tucker et al., 2016).
Although heteronormative dynamics characterize many social contexts, schools represent critical sites in which hegemonic beliefs about sexuality and gender role are enforced (Espelage, Hong, et al., 2018;Poteat, 2007;Rivers, 2001). Following a multilevel and ecological approach, the aim of the present paper is to investigate the role of individual and contextual factors as well as their interaction in predicting the engagement in HB behavior. More specifically, we investigate the role of human basic values and gender stereotyped attributions at individual level, and perceptions of school as a community and the frequency of teacher's intervention in support of victimized students at classroom level. To our knowledge, only a few studies, mainly carried out in the North American context, have considered both the individual and contextual perspective using a multilevel approach (e.g., Birkett & Espelage, 2015), and no study has specifically examined the role of personal values as risk or protective factors against HB behavior.

Individual Values and Beliefs Supporting Homophobic Bullying
Research on individual beliefs motivating stigma-based bullying mainly focused on stereotypes and prejudices concerning weak or minoritarian social groups as well as on the role of a typical masculine trait, the social dominance orientation (Earnshaw et al., 2018). As concerns the engagement in HB behavior, only a few studies have systematically investigated the role of individual factors in predicting the involvement in HB, mostly focusing on value attributions to masculine and feminine characteristics and on prejudice concerning sexual orientation (Poteat & Russell, 2013). Some authors have found that these aspects related to traditional gender role ideology are unique predictors of homophobic behavior, over and above other individual factors such as empathy, perspective taking, and identity importance (Poteat & Russell, 2013).
Several studies have suggested that HB is a behavior that serves as a mean to reinforce traditional gender-norms among peers (Tucker et al., 2016). In a recent study based on a focusgroup methodology (Romeo, Chico, Darcangelo, Bellinger, & Horn, 2017), the authors found that the use of misogynistic and homophobic language solved the function to ensure peers about their conformity to social norms related to gender and sexuality. In a meta-analysis carried out by Whitley (2001), the most influential factor of homophobic harassments was represented by the adherence to traditional gender role. Subsequent studies seem to confirm this finding (Parrott, 2009;. , for instance, found that those who endorsed gender stereotypes were more likely to judge homophobic harassment as not completely wrong when victims were males, thus concluding that homophobic harassment is considered by young people as an acceptable response to "feminine" behavior from a man but not to "masculine" behavior from a woman. Other studies have suggested the role of masculinity traditional values and traits in predicting homophobic behavior and bullying (Epstein, 1997;Phoenix, Frosh, & Pattman, 2003), such as social dominance orientation and conservative values (Weber & Gredig, 2018), lack of interpersonal expressivity, and lack of adequate relationships with peers based on trust (D'Urso & Pace, 2019;D'Urso, Petruccelli, & Pace, 2018).
However, the question that still remains open is whether the association between individual characteristics and HB remained significant when accounting for the role of group and contextual factors. Indeed, when Merrin and colleagues  analyzed the impact of individual factors on HB accounting for the influence of group factors, they found that masculinity, femininity, and dominance attitudes did not more predict the involvement in HB when grouping factors such as peer selection and peer influence were considered, suggesting that enhancing our understanding of HB requires considering it in a more comprehensive framework including the conjoint examination of both individual and contextual factors.

School-Level Factors Supporting Homophobic Bullying
Schools are among the most homophobic social contexts. A rigid adhesion to conforming gender roles is highly promoted in middle and high schools, where students who do not conform to gender norms are at high risk of stigma and discrimination (Rivers, 2011). The impact of peer group dynamics on HB has been evidenced by several studies adopting a multilevel approach, suggesting that HB is not only determined by individual attitudes against sexual minorities but also by contextual factors, such as the peer group climate and, overall, the broader school climate. In the last two decades, an increasing amount of research has shown that a wide range of behaviors and attitudes can be influenced by adolescent peers (Brechwald & Prinstein, 2011), especially in early adolescence when the resistance to peer influence is weaker (Steinberg & Monahan, 2007). A first important finding from research carried out in the first decade of the millennium highlighted the reinforcement of homophobic prejudice within the peer groups. For instance, Poteat (2007) found that homophobic prejudice increases over time when it was shared within the friendship network, with the consequence of an increasing in HB behavior. Moreover, as members of peer groups tend to influence each other, climate within the peer group has been found to reinforce stigmatization and bullying against sexual minorities, over and above individual attitudes (Poteat, 2008). Birkett and Espelage (2015) confirmed the Poteat's previous findings, showing that not only adolescent peer groups differ in their levels of homophobic perpetration but also that peer group levels of homophobic perpetration influenced individual levels of homophobic perpetration over time, even accounting for previous levels of individual's homophobic behavior. The influence of peers in perpetrating HB was confirmed in a subsequent and longitudinal study by Merrin et al. (2018), where friendship selection was a relevant factor in determining an increasing or decreasing involvement in HB: Individuals preferred to befriend peers with similar rates of homophobic teasing, and changes in adolescent homophobic teasing were predicted by a concurrent homophobic teasing behavior of their friends. What these studies suggest is that group factors seem to be more predictive of HB than individual factors do (Birkett & Espelage, 2015), especially for adolescent males (Espelage, Basile, Leemis, Hipp, & Davis, 2018). Similarly, values and norms that regulate informal social relationships within a school, and the extent to which students share these values and norms with each other, have been found to affect students' behavior and their attitudes towards schoolmates (Foà, Brugman, & Mancini, 2012). Factors related to school climate, such as sense of belonging to the school community and school connectedness, are believed to be protective in the face of general as well as homophobic bullying (Diaz, Kosciw, & Greytak, 2010;Espelage et al., 2019). More in detail, previous research has showed a bidirectional relationship between school climate and bullying behavior, with instances of bullying behavior contributing to the perception of a negative school climate, and a positive school climate, characterized by mutual help and care, positive relationships, feelings of personal commitment, common goals, and shared purposes, reducing the likelihood of school bullying episodes and buffering its harmful effects on victims (Toomey, McGuire, & Russell, 2012).

Teachers-Related Factors Supporting Homophobic Bullying
Beyond individual characteristics of perpetrators and beyond the peer group climate that shapes the everyday school experience, also teachers' attitudes contribute to the occurrence of HB episodes. Social support from teachers and other adults from school personnel has been found to be a crucial factor in preventing and counteracting HB (DePalma & Jennett, 2010;Gastic & Johnson, 2009). In their systematic review, Hong and Garbarino (2012) argued that teachers might foster a climate that is intolerant of homophobic bullying, supported by evidence highlighting that in schools where teachers and school personnel actively intervene in bullying situations and provide social support for sexual minority students, there are less HB incidents. In a multi-informant and multilevel study involving US students and school-staff, Rinehart and Espelage (2016) found that when teachers perceived their schools to be actively committed in anti-bullying strategies, students reported less sexual harassment perpetration. Furthermore, school climate promoting gender equity or intolerance of sexual harassment was the main factor associated with a decrease of homophobic name-calling perpetration and a decreased sexual harassment perpetration. These findings support the thesis that when teachers are less tolerant for sexual harassment, there are lower levels of gendered harassment. However, previous research also suggest that although teachers might be aware of the existence of sexual stigma (Douglas, Warwick, Whitty, Aggleton, & Kemp, 1999), it does not mean that teachers are always prone to intervene, because they often perceive themselves unable to face the incidents involving sexual minority students (Warwick, Aggleton, & Douglas, 2001). A large-scale survey of teachers in Irish secondary schools on their awareness and understanding of homophobic bullying revealed that 87% of the teachers had witnessed homophobic bullying more than one occasion; 41% found it more difficult to deal with homophobic bullying than other types of bullying. Overall, research indicates that most of teachers report several barriers to intervene in homophobic bullying situations, such as negative reactions from parents, staff, and students (Norman, 2004). Two studies carried out in the Italian contexts examined teachers' attitudes and coping strategies toward homophobic prejudice and bullying. Analyzing interviews of staff members in 24 Italian secondary schools, Zotti and colleagues (Zotti, Carnaghi, Piccoli, & Bianchi, 2019) found that the higher the sexual prejudice in teachers and the lower the contact with lesbian and gay individuals, the higher the legitimization of homophobic bullying. In addition, the perception that other colleagues have a tolerant attitude against HB episodes reduced the probability of personal intervention in these circumstances. Nappa and colleagues (Nappa, Palladino, Menesini, & Baiocco, 2018), instead, examined the reactions of teachers in response to HB episodes distinguishing between feelings of powerlessness and positive activation. They found that higher homophobic attitudes predicted both higher feelings of powerlessness and lower tendency to actively intervene, whereas higher levels of perceived self-efficacy as a teacher predicted a stronger activation toward victimized students.

The Present Study
In th e r ec en t ye ar s, foll owin g t he l es son fr om Bronfenbrenner's (1979) social-ecological model, there has been a growing consensus among researchers in considering HB a multidetermined phenomenon where individual characteristics, contextual factors, and their interaction concur to increase its occurrence (Basile, Espelage, Rivers, McMahon, & Simon, 2009;Espelage et al., 2019). Following the approach suggested by Espelage et al. (2019), in the present paper we aim to investigate the association between HB and a series of individual-and contextual-level factors.
At the individual level, we examine the role of human values, as theorized within the Schwartz's theory (1992), and stereotyped victim-blaming attributions in incidents of HB. Personal values have been defined as desirable and abstract goals that apply across situations and serve as guiding principles motivating individuals to behave in a certain way and evaluate people or events (Schwartz, 1992). According to the theory, values influence perceptions, attitudes, goals, and all intentional behavior (Schwartz, 2012). In this study, we specifically focus on human values of self-transcendence, represented by values of universalism and benevolence, and selfenhancement, reflecting values of achievement and power. According to Schwartz's theorization, self-transcendence and self-enhancement represent two opposite poles of the circumplex model of values, representing the extent to which they motivate individuals to concern for the welfare and interests of others (self-transcendence) versus the extent to pursue one's own interests and relative success and dominance over others. To date, only a few studies have investigated the role of personal values in influencing general bullying (Knafo, 2003;Menesini, Nocentini, & Camodeca, 2013). Menesini et al. (2013), for instance, found that only self-enhancement had a positive effect on bullying behavior, although it was mediated by immoral and disengaged behavior. No study has been specifically carried out in relation to HB, but there is some evidence supporting the role of low self-transcendence and high self-enhancement values in predicting anti-homosexuality attitudes (Donaldson, Handren, & Lac, 2017). The second individual-level factor concerns victim-blaming attributions based on gender and sexual stereotypes in incidents of HB. Starting from previous research supporting the influence of gender and sexual stereotypes and sexual prejudice (Poteat & Russell, 2013), we are interested in investigating the impact of stereotyped victim-blaming attributions on engaging in HB. In terms of contextual-level factors, based on the literature above described (Hong & Garbarino, 2012), we analyze the effects of students' perceptions of school as a community and teacher's intervention in support of victimized students when bullying incidents occur. More specifically, we use students' classroom as a unit of peer clustering, since the Italian education system considers the class as basic unit at all school levels. Differently from other school systems, indeed, Italian students stay in their classroom with the same peer group every day, all day, and for the entire duration of the school cycle, while the teachers move from classroom to classroom. Thus, classroom represents the most eligible context for bullying episodes, and classmates are of paramount importance in the Italian school system.
At the individual level, we hypothesize that HB will be negatively associated with the value of self-transcendence (H1) and positively associated with the value of selfenhancement (H2) and stereotyped victim-blaming attributions in incidents of HB (H3). At the classroom level, we hypothesize that HB will be positively associated with a low perception of school as a community (H4) and low levels of teacher's intervention in support of victimized students (H5). In addition to investigating the main associations between HB and the above-mentioned variables, we will test the interaction between student-and classroom-level variables, hypothesizing that the association between HB with personal values and stereotyped victim-blaming attributions could be stronger in the presence of low perceptions of school as a community and low levels of teacher support (H6).

Participants
This study was carried out in Campania Region, a large and populated area in the south of Italy, the most important city of which is Naples. The research was commissioned by the Regional School Office in order to investigate the phenomenon of bullying within schools. The funding agencies had no influence over the choice of research tools and procedures used. Regarding the socio-economic composition of the studied population, indices from the National Institute for Statistics, that is the main supplier of official statistical information in Italy, were used (ISTAT, 2010), considering the average level of education and income of the population in relation to the area of residence.
The sample comprised 2867 participants aged between 11 and 18 years, who were middle and high school students from 34 schools and 144 classrooms. In each school, all 7th, 10th, and 13th grade classes were included in the study. Ninety-two adolescents refused to participate, and 57 questionnaires had more than 80% of incomplete data. Thus, the final sample consisted of 2718 adolescents (1271 males and 1447 females), of whom 1105 were enrolled in the second year of middle school (7th grade; mean age = 12.36, SD = .73), 842 were enrolled in the second year of high school (10th grade; mean age = 15.58, SD = .80), and 771 were enrolled in the fifth year of high school (13th grade; mean age = 18.16, SD = 1.01). The sample was relatively well distributed in terms of socioeconomic status. Approximately 40% of the fathers and mothers had a low level of education (middle school diploma or less), 37% had a high school diploma, and approximately 23% had a university degree.

Homophobic Bullying
We assessed the frequency of engaging in homophobic bullying behavior in the last 3 months using an ad hoc questionnaire consisting of four items describing four specific bullying behaviors: threatening, spreading rumors or lies, homophobic teasing, and physical assault. For each item, we specifically asked to refer to episodes targeting peers because of their sexual orientation, either perceived or real. Participants were asked to respond on 5-point Likert scale ranging from 1 = never to 5 = more than once a week. The confirmatory factor analysis supported the psychometric structure of the scale, YBχ 2 (2) = 5.346, p = .07; CFI = .99, RMSEA = .02 90% C.I. [.00, .05], SRMR = .02.

Human Values
To measure values, adolescents completed the short-form version of the Portrait Value Questionnaire (PVQ; Schwartz et al., 2001), a self-reporting scale based on Shalom Schwarz's theory and validated in Italy by Capanna, Vecchione, and Schwartz (2005). The PVQ includes verbal portraits of different people who describe a person's goals, aspirations, or wishes that point implicitly to the importance of a value. For each portrait, respondents answer the question, "How much like you is this person"? on a 6-point Likert scale (1 = not at all like me to 6 = very much like me). We specifically considered two higher order values dimensions: self-transcendence (five items; sample item: He or she thinks it is important that every person in the world be treated equally; Cronbach's alpha = .77) and self-enhancement (4 items; sample item: He or she wants to have a lot of money and expensive things; Cronbach's alpha = .74).

Victim-Blaming Attributions Based on Gender and Sexual Stereotypes
For measuring victim-blaming attributions based on gender and sexual stereotypes in incidents of HB, we asked participants to indicate whether they considered a series of homophobic prejudices as valid reasons to bullying peers (no/yes). Prejudices concerned behaving like the opposite sex, preferring to stay with friends of the same sex, and having homosexual friends. Participants responded no or yes to each statement. Cronbach's alpha was .60.

Perception of School as a Community
To evaluate students' perceptions of school as a community, we used the questionnaire "Questions about you and the school" comprised in the Secondary School Moral Atmosphere Questionnaire (SMAQ; Høst, Brugman, Tavecchio, & Beem, 1998;Italian validation: Mancini & Fruggeri, 2003). The questionnaire assesses several subdimensions of school atmosphere. A principal component analysis (PCA) executed on all subscales' scores resulted in a two-components solution (68% explained variance). The first component, generally called "connectedness with school," included three subscales: sense of community, enthusiastic identification with school, and positive social relations within school, for a total of 17 items. Sample items were "Most of the students like to attend this school because they have a lot of friends" and "Most students of this school think that they can trust the other students." The second component, called "constraint," contained two subscales: negation of community and rejection of the school, for a total of 13 items. Sample items were "One has to watch what one does; otherwise other students will make fun of you," and "Most of the students who attend this school believe it's a waste of time." Participants were asked to indicate their agreement with each statement using a 5-point Likert scale (from completely disagree to completely agree). For the purpose of the present study, the items of the scales measuring positive aspects were reversed, in order to direct the scale to the negative pole, with higher scores reflecting negative perceptions of school as a community. Cronbach's alpha was .84, showing an adequate reliability.

Teachers' Reactions to Bullying Incidents
To measure teachers' intervention in support of victimized students, we asked participants to indicate the frequency with which their teachers responded to bullying incidents that occurred within the classroom during the last 3 months. Items concerned five kinds of teachers' interventions, consisting of supporting behaviors (i.e., comforting the victim; blaming the perpetrator; discussing incidents in the classroom) and nonsupporting behaviors (i.e., ignoring, laughing). Participants were asked to indicate their agreement with each statement using a 5-point Likert-type scale (from never true to always true). For the purpose of the present study, responses to items assessing supporting behaviors were reversed. Then, a composite score of the scale was computed, with higher values indicating higher lack of support by teacher (Cronbach's alpha = .81).

Procedure
The study was realized in the spring of 2010. The questionnaires were administered in the classroom by trained psychologists who had not previously met the students. During the administration of the questionnaires, which took place over a period of 1 month, teachers were not present in the classroom. All questionnaires were anonymous. All parents of the students involved signed an informed consent form that specified the goals of the study. At the end of the study, the local government received a copy of the research report.

Statistical Analysis
We used multilevel regression models to analyze two-level data from a sample of students (individual level) nested within classrooms (classroom level). In multilevel regression models, the outcome variable is measured at the lowest level, whereas explanatory variables might be measured at all existing levels. The usefulness of this approach is in that it assumes that data observations from the same group (classroom, in this study) are more similar to each other than the observations from different groups, allowing to control for and test withingroup and between-group dependencies (Hox, 2010). In this study, on the student-level, we measured the outcome variable (i.e., homophobic bullying), two control variables (i.e., gender and school grade), and three explanatory variables: personal values of self-transcendence and self-enhancement, and victim-blaming attributions based on gender and sexual stereotypes. On the classroom level, we used two explanatory variables: lack of teacher support of victimized students and negative perception of school as a community. The analyses were performed using Mplus 8 (Muthén & Muthén, 2017). We fit five models, in order to compare the relative proportion of variance explained by the addition of covariates and predictors. We began with the estimation of an unconditional model (or intercept-only model; model 1), containing no explanatory variables at both individual and classroom level, in order to assess the intraclass correlation coefficient. This coefficient indicates the proportion of the variance explained by the grouping structure in the sample. The intraclass correlation in this study was 0.019, indicating that 1.9% of variance in homophobic bullying was due to between-class differences. Furthermore, we found a statistically significant random class intercept variance, indicating that there was significant variation in homophobic bullying perpetration between classrooms. In model 2, we added individual covariates to the model (i.e., gender, school grade). In model 3 and model 4, we added predictors at student level and class level, respectively. Finally, in model 5, we tested cross-level interactions to estimate the effects of individual values and stereotyped attributions depending on lack of teacher support and negative perception of school as a community at the classroom level. To interpret the moderating effects, we used the pick a point procedure (Aiken, West, Luhmann, Baraldi, & Coxe, 2012), estimating the conditional effect of individual level variables at low (− 1 SD from the mean) and high levels (+ 1 SD from the mean) of classroom-level variables.
Variables at both levels were centered in order to isolate and differentiate within and between relationships (Bell, Fairbrother, & Jones, 2019). Individual-level variables were centered at the group mean, by subtracting each individual's raw score from their respective class mean score for each variable; classroomlevel variables were grand-mean centered, by subtracting each class mean from the grand mean. Model fit was assessed using reductions in deviance (− 2 Log Likelihood), with smaller values indicating a better fitting model.

Frequency of Involvement in Homophobic Bullying
Overall, participants who reported to have been engaged in homophobic bullying behavior at least one or two times in the last 3 months were 19.1% of the total sample. The chi-square test revealed no significant differences across school grade groups, χ 2 (2) = 3.79, p = .15. As regards participants' gender, 28.2% of male participants reported to have perpetrated homophobic bullying, whereas the percentage of female participants was 11%. Of the total sample, 13.2% of those who perpetrated homophobic bullying were males, and only 5.9% were females, with a statistically significant difference, χ 2 (1) = 130.63, p < .001.

Descriptive Statistics and Correlations
Descriptive statistics and correlations among study's variables are displayed in Table 1. Overall, all variables showed significant associations with homophobic bullying. At the individual level, values and stereotyped attributions were positively correlated with homophobic bullying. At class level, both teacher's lack of support and negative perceptions of school as a community were positively linked to bullying behavior, with students in classrooms with lower values of teacher support and higher values of negative perceptions of school as a community being more likely to engage in homophobic bullying behavior.

Multilevel Model of Homophobic Bullying
Fixed and random effects from multilevel regression models are displayed in Table 2. Model 2 shows the effects of control variables on homophobic bullying. As can be observed, only gender was a significant predictor, with males scoring higher than females. School grade was not significantly associated with homophobic bullying: neither being in grade 7 nor 10 had a significant effect, compared with being in grade 13. In model 3, we estimated the effects of variables at individual level. Specifically, we found that high self-enhancement and stereotyped attributions had a positive effect on homophobic bullying, whereas high levels of self-transcendence were negatively associated with homophobic bullying. That is, on average, students who reported higher levels of self-enhancement and stereotyped attributions also reported higher average rates of homophobic bullying, compared to other students in the same class. Similarly, those who reported higher average levels of self-transcendence also reported less homophobic bullying perpetration.
The effects of variables at class-level were estimated in model 4 (i.e., lack of teacher support and negative perceptions of school as a community). We found only a significant effect of negative climate on homophobic bullying, with students from classrooms with higher average levels of negative perceptions of school as a community showing higher average rates of homophobic bullying.

Cross-Level Interactions
Finally, we tested all possible cross-level interactions (model 5), finding two significant interactions: one between individual self-transcendence and classroom lack of teacher support ( Fig. 1), and the other one between individual stereotyped attributions and classroom lack of teacher support (Fig. 2). More specifically, we found that in classrooms where teacher support was high, individual self-transcendence had no effect on homophobic bullying (b = − .04, p = .08), whereas in classrooms where teacher support was low, high levels of individual self-transcendence had a decreasing effect on homophobic bullying (b = − .07, p < .01). Furthermore, we found that in classrooms where teacher support was low, homophobic bullying did not depend on levels of individual stereotyped attributions (b = .00, p = .99), whereas in classrooms where teacher support was high, high levels of individual stereotyped attributions increased the likelihood to engage in homophobic bullying behavior (b = .03, p < .001).

Discussion
In this study, we examined whether the involvement in homophobic school bullying as perpetrator was associated with personal values, stereotyped victim-blaming attributions, perceptions of school as a community, and teacher's reaction to bullying incidents. Following a social-ecological approach (Bronfenbrenner, 1979), according to which individual behavior is determined by both personal and contextual factors as Gender was coded as 0 for male and 1 for female. For school grade, we used grade 13 as the reference category †p < .10; *p < .05; **p < .01; ***p < .001 1 1.05  Fig. 1 The moderating effect of teacher support in the relationship between individual selftranscendence values and engagement in homophobic bullying well as by their interaction, we performed a multilevel analysis at individual and classroom level, with personal values and stereotyped victim-blaming attributions representing individual predictors of HB, and perceptions of school as a community and teacher's reactions as contextual predictors. In general, findings of this study support our hypotheses, evidencing the contribution of individual values and stereotyped victimblaming attributions on HB, and the contribution of contextual factors also when controlling for the effects of individual-level factors. Our findings are consistent with Poteat (2008) and Birkett and Espelage (2015), expanding our knowledge in considering for the first time in a research the impact of individual values on HB and in performing this kind of comprehensive study beyond the north American context. The study involved a large sample of 7th, 10th, and 13th graders representative of a region in the southern Italy. It evidenced that almost 20% of participants reported to have being engaged in HB in the last 3 months of school. Despite this percentage of HB perpetrators is lower than in other studies (Weber & Gredig, 2018), it is difficult to make a comparison since the majority of research refer only to the use of homophobic epithets, whereas we specified to participants that we sought to investigate homophobic bullying episodes, characterized by several types of aggressive behavior against victims that are not able to defend themselves. In terms of gender differences, consistent with previous studies, we found that males were more likely to be perpetrators of HB than females. One possible reason for this result could lie in masculinity attitudes that males, compared with females, are generally more prone to hold, and that have been found to be strongly predictive of an individual's homophobic perpetration (e.g., Weber & Gredig, 2018). However, much more research should be conducted to better understand and expand our knowledge on females' involvement in HB, and values and beliefs that are associated with this kind of behavior.

Homophobic bullying
Our first research questions concerned whether individual values were associated to HB. Findings confirmed our hypotheses, evidencing significant effects of selftranscendence (H1) and self-enhancement (H2) values on HB, also when accounting for other individual and contextual variables. Values are enduring goals that refer to what people consider as important (Roccas, Sagiv, Schwartz, & Knafo, 2002). They serve as guiding principles in people's lives, as criteria for selecting and justifying actions and for evaluating people and events (Schwartz, 2012). Previous research has found that self-transcendence is a protective factor against general bullying (Menesini et al., 2013). Since selftranscendence is a value that gives high importance to justice, equality, and protecting all people, enhancing this specific value could be a strong antidote in order to prevent stigma and discrimination, as well as their related behavioral manifestations.
Conversely, the value of self-enhancement gives a positive independent contribution to being involved in HB. According to the Schwartz's theory (Schwartz, 1992), self-enhancement values concern the goals to pursue high social status and prestige, and to exert control or dominance over people and resources. It is not surprising that people who adhere to value of power exert their dominance against minorities (such as sexual minorities) or those who are perceived as weaker because not conforming to gender stereotypes. It is plausible that individual who aspire to have power over others harass people perceived as weak in order to enhance their own social status within the peer network remarking the distance between their high social status as perpetrators and other people. This finding is consistent with previous research finding social dominance orientation increasing adolescents' involvement in homophobic behaviors Weber & Gredig, 2018), even if in some research it becomes non-significant when accounting for other variables . Overall, the role played by personal values seem to confirm the inner nature of HB (and bullying in general), that could not be intended as a reactive behavior or as uniquely driven by a lack of control. Conversely, it is a deliberate act as supposed  Fig. 2 The moderating effect of teacher support in the relationship between stereotyped victimblaming attributions and engagement in homophobic bullying by the Social Cognitive Theory (Bandura, 1986) and by the Reputation Enhancement Theory (Emler & Reicher, 1995), which support the thesis that hostile behavior is also the result of a deliberate choice that aims to project to others a certain image of one's self and status. Our result highlighting the central role of individual values in predicting HB, accounting for contextual factors, sheds light on the need to increase the importance that adolescents attribute to values, fostering their motivation to help and care for others and mitigating their desire of power over others. Although values represent a central aspect of the self-concept (Brewer & Roccas, 2001) and remain relatively stable across the overall lifespan and, in particular, across adolescence (Aquilar, Bacchini, & Affuso, 2018), it is also argued that they can change through specific interventions and the use of facilitators (Arieli, Grant, & Sagiv, 2014). Such interventions might be especially efficacious during adolescence, when the chance for changing is higher than in other period of the life (Steinberg, 2014). Stereotyped victim-blaming attributions resulted to be unique predictors of HB, confirming our hypothesis (H3). We conceived this dimension as a marker of gender stereotypes and sexual prejudice, but at the same time it is a proxy of the mechanism of blaming the victim that has been largely described as a mechanism of moral disengagement (Bandura, 2002) or cognitive distortions (Gibbs, 2003), whose importance in the bullying phenomenon has been documented in a large number of research (Dragone, Bacchini, Esposito, De Angelis, & Affuso, 2020;Gini, Pozzoli, & Hymel, 2014). Blaming attributions to victims of HB because nonconforming to typical gender norms could act as a defense mechanism against feeling guilts and, therefore, to legitimate the perpetration persuading aggressors that victims deserve being offended (Camodeca, Baiocco, & Posa, 2019).

Homophobic bullying
The fourth research question concerned whether perceptions of school as a community at classroom level had an impact on HB. Findings from the multilevel analysis confirmed our hypothesis (H4): the individual involvement in HB was higher in classrooms where perceptions were more negative than in others. This is consistent with previous research that evidenced how a negative school climate enhanced the occurrence of HB (Espelage et al., 2019). In their studies, Poteat (2008) and Birkett and Espelage (2015), for instance, found that the aggressive and homophobic social climate of individuals' peer groups was associated with an increased use of homophobic epithets, over and above individuals' own bullying behavior.
With respect to our fifth hypothesis, even if bivariate correlations showed a significant association between teacher's intervention in support of victimized students and individual HB, when we considered this variable accounting for other measures, it did not have an independent contribution on HB (H5). This finding is contrary to our expectations and previous literature underlying the protective role of teacher support in incidents of HB (Hong & Garbarino, 2012). However, there are several factors that might have biased our results and that need to be considered. First, the measure we used was not specific of HB, as tested in other research; second, differently from other studies investigating the role of teachers in preventing and contrasting homophobia at school, we considered the effect of teacher support also accounting for other individual-and classroom-level measures. Thus, it could be reasonable that teacher support does not have an independent contribution on HB when taking into account other sources of influence. Findings from several studies, for instance, suggest that students feel greater school connection and safety when teachers actively take measures to prevent and contrast bullying behavior (Rinehart & Espelage, 2016), thus supporting the hypothesis that teachers' deterrent effect on bullying could be totally mediated by their ability to foster a positive school climate. Other interesting insights come from our moderation analysis, showing that teacher's intervention in support of victimized students moderated the effects of individual selftranscendence and stereotyped victim-blaming attributions, even if this effect was not in the specific direction we hypothesized (H6). More specifically, the engagement in HB was independent of individual self-transcendence in classrooms where teacher support was perceived as high, whereas in classrooms where teacher support was low, high levels of individual self-transcendence had a decreasing effect on HB. This result suggests the importance of teacher's concern for victimized students in fostering a climate where homophobic behavior is considered as unacceptable, thus playing a crucial role in deterring youth from engaging in HB, over and above their own individual values and motivations. This would be in line with studies reporting sexual minority students experiencing less harassment and assault when teachers intervened in incidents of homophobic bullying and emphasizing teachers' intervention as the most effective measure to contrast HB (Kosciw, Greytak, Zongrone, Clark, & Truong, 2018). Looking at the interaction of teacher's intervention with individual stereotyped attributions, we found that in classrooms where teacher support was low, this condition was enough for students' involvement in HB, and individual stereotyped attributions had no significant effect on this. However, in classrooms where teacher support was high, the engagement in HB increased at increasing values of individual stereotyped attributions. Taken together, these findings from the interaction analysis support the hypothesis of a key role played by teachers in contrasting HB, but only to a certain degree. Indeed, individual protective factors did not have a decreasing effect in the context of a high teacher support; however, when teacher support was low, individual self-transcendence become crucial in decreasing the likelihood to engage in HB. Similarly, individual risk factors had no increasing effect in the context of a low teacher support, but when teacher support was high, having stereotyped victim-blaming attributions increased the risk to engage in HB. Thus, while perceiving school as a community seems to have an independent contribution on HB behavior, over and above individual beliefs and attributions, further studies should deepen the role of teachers in preventing HB, independent of and in interaction with individual dispositions.

Strengths and Limitations
Our findings are promising and particularly relevant if we consider that this is the first study that, using a multilevel approach, investigates the effects of both individual and contextual factors in relation to HB behavior within the Italian school context. Furthermore, to our knowledge, no other study has examined the role of personal values in predicting HB, despite their importance in motivating attitudes and behavior has been well documented (Schwartz, 1992(Schwartz, , 2012. Analyzing data at different levels of influence, the findings of this study expands previous knowledge on protective and risk factors associated with HB, suggesting an independent contribution of school sense of community, and a complex interplay between individual beliefs and attitudes and teacher's interventions in support of victimized students in predicting HB. Despite its strengths, however, there are several limitations that need to be acknowledged when interpreting the results of this study. First, all measures used in the study were based on students' self-reports, then results might be contaminated by shared variance associated with the use of a unique source of information. In addition, the use of a cross-sectional design does not allow us to identify likely causal links between individual and contextual factors analyzed in the study with involvement in HB. A further limitation concerns the measure of teacher's intervention in support of victimized students, which was not specifically developed to assess teacher's interventions in incidents of HB. Teachers' reactions to homophobic bullying might be different from those in incidents of bullying not motivated by homophobic bias, given the number of both individual (e.g., homophobic attitudes) and contextual factors (e.g., school commitment against homophobia) that can lead teachers to intervene or not in such cases (Poteat, Slaatten, & Breivik, 2019). Future studies should deepen the role of teachers in preventing and contrasting HB over and above other individual and contextual factors, by using a more specific measure that could allow to better explain its complex association with HB.

Conclusion
The findings of the current study highlight the importance of considering both individuals and social factors when examining predictors of HB behavior, as well as the complexity of their interplay in predicting HB. Given the strong association between HB and stereotyped victim-blaming attributions at the individual level, school anti-bullying policies should explicitly deal with issues related to bullying and discrimination based on sexual orientation and gender identity. School-based programs should target not only bullying behavior (e.g., through socio-emotional learning) but also individual bias that support this kind of behavior. Specific actions might aim at increasing students' access to appropriate information about issues related to sexual and gender identity, and at raising awareness of the unique consequences that this behavior has on youth who are targeted with. Of importance, effective prevention and intervention efforts should consider the protective role played by supportive school environments, where all members of the school community (teachers, staff, and peers) feel the responsibility for promoting a safe, welcoming, and respectful climate in school, such that all students have the opportunity to successfully achieve personal and academic goals, regardless of sexual orientation, gender identity, or gender expression.
provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.