How Diversity Approaches Affect Ethnic Minority and Majority Adolescents: Teacher–Student Relationship Trajectories and School Outcomes

This study aimed to relate school diversity approaches to continuity and change in teacher-student relationships, comparing Belgian-majority (N = 1,875, Mage  = 14.56) and Turkish and Moroccan-minority adolescents (N = 1,445, Mage  = 15.07). Latent-Growth-Mixture-Models of student-reported teacher support and rejection over 3 years revealed three trajectories per group: normative-positive (high support, low rejection) and decreasing-negative (moderate support, high-decreasing rejection) for both groups, increasing-negative (moderate support, low-increasing rejection) for minority, moderate-positive (moderate support, low rejection) for majority youth. Trajectories differed between age groups. Student and teacher perceptions of equality and multiculturalism afforded, and assimilationism threatened, normative-positive trajectories for minority youth. Diversity approaches had less impact on majority trajectories. Normative-positive trajectories were related to improved school outcomes; they were less likely, but more beneficial for minority than majority youth.

Schools are key social contexts for adolescent development (Eccles & Roeser, 2011), and school outcomes have long-lasting implications for future life chances (Heath & Brinbaum, 2014). Against the background of persistent inequalities in school outcomes between ethnic minority and majority students (Heath & Brinbaum, 2014), we focused on the quality of teacher-student relationships as a key protective factor (Sabol & Pianta, 2012). We asked how teacher-student relationship quality evolves throughout secondary school for ethnic minority and majority youth; how evolving relationships differ between individuals and schools; and how these relationships benefit later school outcomes. Building on Garc ıa Coll's et al. (1996) integrative model of minority child development and bridging theoretical approaches from ecological systems (Bronfenbrenner, 2005) and intergroup relations (Derks, Van Laar, & Ellemers, 2007), our study aimed to disentangle distinctive intergroup factors in the school environment for minority adolescents from protective factors for all adolescents.
Teacher-student relationship quality refers to students' positive as well as negative experiences with their teachers. Thus, youngsters can experience teacher support such as when they feel understood and teacher rejection such as when they feel treated unfairly. While supportive relationships promote achievement and adjustment (McGrath & Bergen, 2015;Sabol & Pianta, 2012), experiences of rejection or discrimination undermine these outcomes (Brown & Chu, 2012;Thomas, Caldwell, Faison, & Jackson, 2009). Moreover, early experiences can be transient or stable as relationships continue or change through adolescence. It is critical to look at change and continuity in these relationships because cumulative experiences of supportive relationships enable better school adjustment than do temporary experiences, whereas prolonged periods of relational difficulty are more harmful than are shorter periods (Ladd, Herald-Brown, & Reiser, 2008).
Our first objective was thus to identify distinct patterns of teacher-student relationship quality over time. We asked how trajectories of teacher support and rejection differed between ethnic majority and Turkish and Moroccan minority adolescents in Belgium. Turkish and Moroccan minority youth in Europe, most of whom are second-generation of immigrant origin, perform worse than their majority peers in school, even when controlling for parental and individual background (Heath & Brinbaum, 2014). They are also targets of anti-Muslim prejudice in Europe, as the majority of them are Muslim (Strabac & Listhaug, 2008). Our second objective was to predict distinct relationship trajectories from the way schools deal with cultural diversity as perceived by teachers and students. As critical components of diversity approaches in schools, we assessed perceived equality (i.e., equal treatment), multiculturalism (i.e., valuing diversity and combating discrimination), and assimilationism (i.e., expecting minorities to relinquish their minority culture for the majority culture). No previous research, to our knowledge, has examined how cultural-diversity approaches impact teacher-student relationships. Our third objective was to associate relationship trajectories with changes in school outcomes over time. We assessed "hard" outcomes such as grades and school compliance and "soft" outcomes such as school engagement and belonging.
Adolescence is the period when young people's views on equality and diversity are formed as they become more aware of, and susceptible to, discrimination or unfairness (Baysu, Celeste, Brown, Verschueren, & Phalet, 2016). At the same time, older adolescents are more at risk of experiencing negative relationships with teachers in secondary school (McGrath & Bergen, 2015). Our fourth aim therefore was to examine age differences in the relationship trajectories of adolescents and in the effects of cultural diversity approaches on these trajectories. In the following section, we will discuss continuity and change in teacher-student relationships, cultural diversity approaches as affordances, and school outcomes as consequences of these relationships.

Continuity and Change in Teacher Support and Rejection
Supportive relationships with teachers promote better school outcomes for minority and majority students alike (McGrath & Bergen, 2015;Roorda, Jak, Zee, Oort, & Koomen, 2017;Roorda, Koomen, Spilt, & Oort, 2011;Sabol & Pianta, 2012). According to Bronfenbrenner's (2005) ecological systems theory, at the level of microsystems (such as schools) proximal processes (such as interpersonal relationships between teachers and students) are the primary mechanisms that explain developmental outcomes. Although teacher-student relationship is important for all students (Roorda et al., 2017), minority students are at risk of experiencing less support or more discrimination from teachers (McGrath & Bergen, 2015). From an intergroup relations perspective, we associate teacher support and rejection with socialidentity affirmation and threat (Derks et al., 2007). As most teachers have a majority background, minority adolescents' relationships with them have a distinct intergroup dimension so that teacher support (vs. rejection) signals the (de)valuation of their minority identity in school (Verkuyten, Thijs, & Gharaei, 2019). In line with Garc ıa Coll's et al. (1996) integrative model of minority development, discrimination puts children at risk of poorer developmental outcomes through affecting their relationships and engagement with learning. Thus, minority experiences of teacher discrimination predict impaired school outcomes (Brown & Chu, 2012;Thomas et al., 2009). To disentangle distinctive minority experiences from those common to all adolescents, we investigated both minority and majority experiences of teacher support and rejection over time.
Longitudinal studies about teacher-student relationship quality generally document a decline with age (O'Connor & McCartney, 2007) and with transition from primary to secondary school (Hughes & Cao, 2018). Looking beyond general trends in relationship quality, five longitudinal studies modeled different trajectories of teacher-student relationship quality in primary school, using measures of teacher-reported closeness and conflict (Bosman, Roorda, van der Veen, & Koomen, 2018;O'Connor & McCartney, 2007;O'Connor, Collins, & Supplee, 2012;O'Connor, Dearing, & Collins, 2011;Spilt, Hughes, Wu, & Kwok, 2012). Most studied samples of largely ethnic majority children (11%-20% ethnic minority), with Spilt et al. (2012) as a notable exception using an ethnically diverse sample. These studies did not distinguish between trajectories of ethnic minority and majority children, but two (Bosman et al., 2018;Spilt et al., 2012) tested the effect of ethnic minority status on the trajectories.
Our study extends these studies in several ways. First, rather than reports of teachers, we focus on adolescents' own experiences of their relationships with teachers, which can affect school outcomes, regardless of their congruence with teacher perceptions (Wu, Hughes, & Kwok, 2010). Second, while our measure of teacher support (e.g., "teachers understanding you," "having attention for you") overlaps with closeness or warmth in teacher-student relationship, our measure of rejection differs from the conflict measures used in these studies. Although both rejection and conflict indicate negative relationships, conflict refers to disharmonious and coercive interactions between teacher and student (e.g., "This child and I always seem to be struggling with each other," in all five studies, "I often need to discipline this child," in Spilt et al., 2012), whereas rejection indicates when students feel that teachers treat them unfairly or make them feel unwelcome or incompetent. Such rejection experiences can undermine trust and perceived justice (Baysu et al., 2016;Yeager, Purdie-Vaughns, Hooper, & Cohen, 2017).
Third, we aimed to replicate previous findings for adolescents' experiences of teacher support and rejection. Because experiencing teacher support is normative and more likely than rejection, for both minority and majority youth, we expected that majority of ethnic minority and majority adolescents would follow a normative-positive trajectory, experiencing moderate-to-high support (Bosman et al., 2018;O'Connor & McCartney, 2007;O'Connor et al., 2011O'Connor et al., , 2012 and low rejection (Bosman et al., 2018;O'Connor et al., 2012;Spilt et al., 2012). We also expected smaller numbers of minority and majority adolescents to have non-normative or problematic relationships, either initially or increasingly over time. Problematic trajectories can take the form of worsening relationships over time (increasing teacher rejection, decreasing support or both) and of initially negative yet improving relationships over time (decreasing teacher rejection, increasing support or both; Bosman et al., 2018;O'Connor et al., 2011O'Connor et al., , 2012Spilt et al., 2012). One might also find a most negative relationship trajectory with low support and high rejection (O'Connor & McCartney, 2007;O'Connor et al., 2011O'Connor et al., , 2012Spilt et al., 2012).
Fourth, relationship trajectories have not been analyzed separately for ethnic majority and minority youth. Minority students may experience less teacher support or more unfair treatment than majority peers in school (McGrath & Bergen, 2015;Verkuyten et al., 2019). Research on trajectories of teacher-student relationship has shown that teachers report less closeness and more conflict with minority children (Bosman et al., 2018), who are more often represented in increasing-conflict trajectories than their majority peers (Spilt et al., 2012). We expected that minority adolescents might be more at risk of having problematic relationship trajectories than majority peers. Minority adolescents might either experience specific forms of problematic relationships with teachers or face the common problematic trajectories more often. To allow for group differences in both form and frequency of relationship trajectories, we identified different trajectories within minority and majority samples separately.
Fifth, we shifted the focus from primary-school children to adolescents and their teachers in secondary school. Adolescents generally experience less supportive teacher relationships as they get older (McGrath & Bergen, 2015). Adolescence is also a period when minority children in particular become more aware of, and vulnerable to, discrimination (Baysu et al., 2016). Thus, middle-to-late adolescents can expect more ethnic victimization and more readily infer ethnic motives even in ambiguous situations (Killen, Henning, Kelly, Crystal, & Ruck, 2007). Against this background, we explored whether older majority and minority adolescents were more likely to develop problematic relationship trajectories relative to younger adolescents.
Finally, we contextualized different relationship trajectories for majority and minority adolescents by focusing on cultural diversity approaches in school.

Cultural Diversity Approaches in Schools
The quality of relationships in schools and how well schools manage diversity are key to a positive school climate (Thapa, Cohen, Guffey, & Higgins-D'Alessandro, 2013;Wang & Degol, 2016). From an ecological systems theory approach (Bronfenbrenner, 2005), teacher-student relationships as proximal processes are afforded by the school environment as a microsystem. Thus, the school environment shapes how teachers deal with their students and how students see their teachers. No research, to our knowledge, has related a positive diversity climate, such as when schools value fairness or diversity, to individual experiences of teacher-student relationship quality. By bridging theoretical approaches from ecological systems and intergroup relations (Derks et al., 2007), we connected cultural diversity approaches to teacher-student relationships, particularly for minority youth. In line with a social identity approach of intergroup relations, we assume that minority youth feel more included and perform better in schools that value their minority identities than in identity-threatening contexts (Baysu et al., 2016;Celeste, Baysu, Meeusen, Kende, & Phalet, 2019;Walton & Cohen, 2007). Extending existing evidence on diversity approaches, we proposed that schools could make a difference in relationship quality with teachers by signaling identity affirmation to minority students, for example, by ensuring equal treatment or valuing cultural diversity. To the extent that cultural diversity approaches affirm or threaten majority identities, majority students may also be affected (Plaut, Garnett, Buffardi, & Sanchez-Burks, 2011). Because minority identities are more likely to be targets of unequal treatment or devaluation in schools, we expected diversity approaches to be most relevant for minorities, but we also explored whether majorities might be affected as well.
Three different approaches to diversity can be identified in European societies and schools: equality, multiculturalism, and assimilationism (Guimond, Sablionniere, & Nugier, 2014). While few studies relate diversity approaches to school outcomes (Celeste et al., 2019;Schachner, 2019;Schachner, Noack, Van de Vijver, & Eckstein, 2016), none predicts teacher-student relationship quality. Yet, these approaches can have different consequences for teacher-student relationship quality.
The equality approach refers to student perceptions of the general school climate, whether schools treat everyone equally and fairly (Baysu et al., 2016). In line with a social identity perspective, perceived fairness promotes various positive outcomes for minority youth such as well-being, engagement, and trust in the organization (Morin, Ma€ ıano, Marsh, Nagengast, & Janosz, 2013;Schachner et al., 2016;Yeager et al., 2017) because it conveys the message that their minority identity is equally valued in the school context (Baysu et al., 2016). Thus, when minority adolescents saw their school as more fair, they reported less discrimination and better school outcomes (Benner & Graham, 2011;Juvonen, Kogachi, & Graham, 2018). Perceived school fairness not only increased minority school outcomes, but it also buffered disengagement in the face of discrimination (Baysu et al., 2016). Extending these findings, we expected that perceived equality would enable minority adolescents to develop normative-positive (rather than problematic) relationship trajectories. We explored whether a fair school would also improve the relationship of majority adolescents with their teachers.
A multiculturalist approach values diversity (Schachner et al., 2016) and challenges racism and discrimination on grounds of race, ethnicity or religion (Thijs, Westhof, & Koomen, 2012;Zirkel, 2008). From a social identity perspective, minority adolescents could benefit from a multiculturalist approach when they feel that their cultural identity is valued, and growing evidence attests to the benefits of multiculturalism for minority students' achievement and adjustment (Celeste et al., 2019;Vedder & van Geel, 2012). When schools and teachers were seen to value diversity, minority students reported less discrimination (Brown & Chu, 2012;Vedder & van Geel, 2012). Thus, we expected that perceived multiculturalism would afford normative-positive trajectories of teacher-student relationship quality. We had no hypotheses for majority adolescents because the existing evidence is mixed: benefits depend on whether the majority feels that their identity is valued or whether they feel excluded by multiculturalism (Plaut et al., 2011).
Assimilationism requires minorities to prioritize the mainstream culture over their heritage cultures (Guimond et al., 2014). From a social identity perspective, assimilationism harms minority outcomes when minorities feel that their cultural identity is disregarded (Hornsey & Hogg, 2000). Minority students experience more peer rejection in classrooms with assimilationist peer norms (Celeste, Meeussen, Verschueren, & Phalet, 2016) and feel alienated and less belonging when schools enforce assimilationism (Celeste et al., 2019;Pulinx, Van Avermaet, & Agirdag, 2015). We likewise expected that minority adolescents in assimilationist schools would be less likely to have normative-positive relationship trajectories. We had no hypotheses for majority adolescents. Because assimilationism does not threaten the majority cultural identity, majorities may be unaffected (Celeste et al., 2019). Alternatively, they might indirectly benefit if they feel affirmed in their majority identity.
Cultural diversity approaches are critical during adolescence. From a developmental intergroup perspective (Killen & Rutland, 2011), adolescents explore group identities as part of their social development and they develop a deeper understanding of fairness in their moral judgments. In line with changes in social-cognitive and moral development during adolescence, perceived equal treatment buffered minority adolescents against discrimination in school and this buffer effect was stronger for older adolescents (Baysu et al., 2016). We thus explored age-related differences in the consequences of perceived equality, multiculturalism and assimilationism for minority and majority adolescents.

Changes in Minority and Majority School Outcomes
The quality of teacher-student relationships shapes school outcomes both longitudinally and cross-sectionally (Baysu & Phalet, 2012;McGrath & Bergen, 2015;Roorda et al., 2011Roorda et al., , 2017. Moreover, continued teacher support-or protection from adverse relationships with teachers-has more enduring effects than momentary perceptions of support or adversity (Ladd et al., 2008). Longitudinal studies of teacher-student relationship quality in primary school related normative-positive trajectories to higher achievement relative to negative (O'Connor & McCartney, 2007) and worsening relationship trajectories (Spilt et al., 2012). By combining the relationship trajectories of closeness, conflict and dependency into an overall risk measure, Bosman et al. (2018) found that those in the no-risk group (overall high-quality relationships) had higher motivation and achievement relative to either low or high-risk groups. We thus expected that normative-positive relationship trajectories would improve school outcomes for all adolescents.
From a social identity approach, minority students may be more responsive to the quality of relationships in school, which could protect them from negative consequences of identity threat (Walton & Cohen, 2007). Similarly, the academic risk hypothesis (Hamre & Pianta, 2001) states that (minority) students at risk of academic failure have more to gain from high-quality teacher-student relationships. Thus, supportive relationships with teachers enhanced minority students' school outcomes (Roorda et al., 2011). In other studies, however, teacher support (or lack of conflict) was no less important for majority students as well (Baysu & Phalet, 2012;Hughes, Luo, Kwok, & Loyd, 2008;Roorda et al., 2011). We explored whether teacher-student relationship trajectories were more consequential for the school outcomes of minority (vs. majority) adolescents.

The Present Study
Drawing on large-scale school-based longitudinal data (3 cohorts, 3 waves), our study aimed to elucidate what is distinctive about ethnic minority adolescents' schooling experiences compared to their majority peers (Garc ıa Coll et al., 1996). We presented the following expectations: (a) Across groups, many ethnic minority and majority adolescents would develop normative-positive relationship trajectories, but problematic trajectories, though less frequent, would be more common among minority than majority adolescents. (b) Equality and multiculturalism approaches of cultural diversity would afford more frequent positive -and assimilationism more frequent problematicrelationship trajectories in minority adolescents. We did not formulate any hypotheses for majority adolescents. We measured equality, multiculturalism and assimilationism at the individual-level of personal perceptions and at the school-level of shared student and teacher perceptions. (c) Positive relationship trajectories would improve and problematic trajectories would undermine school outcomes for all adolescents, and the consequences might be stronger for minority than majority adolescents (d) Older adolescents (both majority and minority) might be more at risk of developing problematic relationship trajectories. We also explored age-related variation in the associations of school diversity approaches with relationship trajectories.

Participants
This study was part of a large-scale longitudinal study (Children of Immigrants Longitudinal Study Belgium, Phalet, Meuleman, Hillekens, & Sekaran, 2018) with three waves of data collected 1 year apart in 70 randomly selected secondary schools in Flanders, Belgium (Time 1: February 2012-July 2013; Time 2: February 2013-July 2014; Time 3: February 2014-July 2015). Following consent from respective parties in line with the university ethical guidelines, students participated in the study during class hours in the presence of research assistants. Student nonresponse rate was 13.3% (Phalet et al., 2018). Schools were stratified by ethnic composition, ranging from low (< 10%) to moderate (10%-30% and 30%-60%) to high (> 60%) percentages of minority students, on the basis of administrative data on foreign languages spoken at home. Participants were in their first (28.2%), second (30.7%), or third (41.2%) year of secondary education. Ethnic majority and minority samples were selected based on self-reported parentage (i.e., all grandparents born in Belgium vs. one or more (grand)parents born in Turkey or Morocco). Majority adolescents (N = 1,875, 48.7% girls) were on average 14.56 years-old in Wave 1 (SD = 1.06, range = 12.21-18.43).

Measures
We identified trajectories using measures of teacher support and rejection as reported by students at each wave. Responses ranged from 1 (never) to 4 (always).
Teacher rejection (T1, T2, T3) was assessed with four items: "In your daily life at school how often do you experience that your teachers . . . 'treat you unfair or hostile', 'expect you cannot do anything right', 'talk to you as if you were stupid', 'let you know that you are not welcome'." This scale was adapted from the rejection subscale of Perceived Ethnic Discrimination Questionnaire (Brondolo et al., 2005;a = .88-.70). The original scale cited ethnicity as a motive, which was removed to measure general feelings of rejection for both groups (Minority, a T1-T2-T3 = .80, .83, .82; Majority, a T1-T2-T3 = .76, .76, .80).
Perceived diversity approaches (reported by students, teachers or both) were assessed as predictors at time 1 and rated on 1 (strongly disagree) to 5 (strongly agree).
Adolescents' Perceptions of Equality in School (T1) were measured with two items that were used in previous research (Baysu et al., 2016; two-item version, r = .56), adapted from the Experience of School Rules scale (Gregory, Cornell, & Fan, 2011; longer version a = .95): "The rules are applied equally to all students," "Some students are allowed more than others" (reversed). We used adolescents' own perceptions at the individual level (Majority a = .69, Minority a = .53), and aggregated perceptions of majority students and of Turkish and Moroccan minority students (a = .61) at the school level.
Adolescents' and Teachers' Perceptions of Multiculturalism in School (T1) were measured with the same items: "In my school" "different cultures and religions are treated with respect," "they take strong action against racism and discrimination," "teachers treat all students equally regardless of their religion or descent," "teachers say that you shouldn't discriminate students with another culture or origin" (from Teachers' Multicultural Attitudes Scale; Thijs et al., 2012; a = .70). We used individual-level student perceptions (Minority: a = .67; Majority: a = .68). We also aggregated majority and minority students' (a = .68) and teachers' perceptions (a = .73) at the school level.
Teachers' Perceptions of Assimilationism in School (T1) were measured with two items "In my school speaking another language than Dutch is not tolerated"; "it is forbidden to wear a headscarf or other religious clothing" (a = .40). The scale was based on the analysis of Belgian school policies (Celeste et al., 2019), which revealed assimilationism as a separate cluster with these two items (mentioned 25% of the time in the policies). A factor analysis of teacher reported multiculturalism and assimilationism confirmed that the two items load on a separate factor (factor loadings > .73). We aggregated teacher perceptions at the school level.
Several outcomes based on student reports were assessed at T3, controlling for scores at T1 on scales from 1 (strongly disagree) to 5 (strongly agree) unless otherwise indicated.
School engagement (T1, T3) was assessed separately for emotional engagement ("I like to learn new things in class," "I feel good in class," "I like to be in class"; minority a = .73-.70, majority a = .70-.67 in T1 and T3), behavioral engagement ("I work as hard as I can in class," "I listen carefully during the class," "I pay attention in class," minority a = .81-.82 majority a = .82-.83), and behavioral disengagement ("In class I am easily distracted," "I often think of other things during class," "In class I do not really do my best"; minority a = .59-.60; majority a = .70-.67). The scales were adapted from the Engagement versus Disaffection with Learning scale (Skinner, Kindermann, & Furrer, 2008; scale a = .61-.82, across constructs).
School belonging (T1, T3) was measured with four items (Wang et al., 2011; original scale a = .75): "I feel at home at this school"; "I am proud to be a student of this school," "I would prefer to go to another school (reverse item)," "I feel at home at this school," "I feel happy at this school." (Minority a = .84-.85; Majority a = .85-.86 in T1-T3).
Age (T1) was included as a predictor of the trajectories and as a moderator in the association between diversity approaches and the trajectories.
Control variables (T1) were gender and school track (vocational vs. nonvocational). Parental education as a proxy for socioeconomic status, ethnic school composition, and Turkish vs. Moroccanbackground were not significant and were dropped from further analysis.

Analytic Strategy
We used Mplus version 7.31 (Muth en & Muth en, 1998 for all analyses and handled missing data using full information maximum likelihood (FIML). FIML uses all available data without imputing missing data, which may introduce randomness in the data. Thus, it is unbiased and preferable to other methods (Dong & Peng, 2013). Data analysis involved three parts. First, we identified different patterns of teacher support and rejection via multivariate Latent Growth Mixture Models. Second, we ran multilevel multinomial logistic regression analyses with trajectories as outcomes in a stepwise fashion, separately for majority and minority groups: (a) control variables; (b) age and individual-level perceptions of equality and multiculturalism and school-level aggregates of shared (majority and minority) student perceptions of equality and multiculturalism, (c) school-level aggregates of teacher perceptions of multiculturalism and assimilationism, (d) age interactions with equality and multiculturalism. Only significant interactions were kept in the model (see Supporting Information for model specifications). Third, we ran separate multilevel regression analyses with trajectories as predictors (by using dummy-coding) with T3 grades, noncompliance, engagement, and belonging as outcomes, controlling for T1 outcomes. We included T1 diversity approaches as covariates (see Supporting Information for model specifications). We also ran additional analyses with the two low-reliability scales (a < .60, school noncompliance and assimilationism), which confirmed the results reported here (see Supporting Information for details).

Results
For attrition analysis, we compared those who participated in all waves (minority: 40.3%; majority: 47.9%) to those who missed at least one wave. The details can be found in Supporting Information. Table 1 shows descriptive statistics of key variables (see Supporting Information for all study variables).

Trajectories of Teacher Support and Rejection
For both majority and minority adolescents, we decided that a three-class solution fitted the data best (      As shown in Figure 1, a positive relationship trajectory, labeled "normative-positive" was most frequent (77.7%). This subgroup combined high teacher support with a negligible decrease over time (I = 2.86, S = À0.04, p = .010), and low, stable teacher rejection (I = 1.38, S = À0.03, p = .099). We also found two problematic relationship trajectories. Adolescents in a "decreasing-negative" trajectory (13.8%) had moderate teacher support with a small increase over time (I = 2.32, S = 0.13, p = .030), and initially high teacher rejection that decreased over time (I = 2.90, S = À0.61, p < .001). A third subgroup followed an "increasing-negative trajectory" (8.5%) with moderate, stable teacher support (I = 2.46, S = À0.03, p = .667), with initially low teacher rejection that increased over time (I = 1.82, S = 0.60, p < .001).
Among majority adolescents, the three-class model also yielded the best fit. It had better fit statistics (lower BIC and AIC) and significantly improved model fit (better BLRT) over a two-class model Comparing the three-class to a four-class model, although the BLRT suggested significant improvement, log-likelihood, AIC, BIC values, and entropy showed little improvement; and the percentage of students was too small to analyze. The fourth class resembled the increasing-negative trajectory among minorities, but it was not statistically and meaningfully differentiated as a separate trajectory based on model fit indices.
Comparing numbers of adolescents in the two common trajectories, majority adolescents were more often in the normative-positive trajectory, minority adolescents in the "decreasing-negative" trajectory, v 2 (1) = 43.52, p < .001.

Cultural Diversity Approaches and Age as Predictors of Trajectories
Minority adolescents (Table 3). When these students perceived the school as more equal or more multicultural, they were more likely to have a normative-positive trajectory (vs. others). At the school level, when both minority and majority youth perceived the school as more equal, minority adolescents were more often in the normative-positive trajectory (vs. decreasing-negative). When teachers perceived the school as more multicultural or less assimilationist, minority youth were again more often in the normative-positive trajectory (vs. increasing-negative). Additionally, older minority adolescents were less likely to be in the normativepositive trajectory (vs. increasing-negative) than Note. Likelihood ratio tests compare the solution with k trajectories to a solution with k À 1 trajectories. BIC = Bayesian information criterion; AIC = Akaike information criterion; BLRT = bootstrap likelihood ratio test. younger ones. A significant interaction between age and multiculturalism ( Figure 3) showed that multiculturalism did not make a difference for younger adolescents; for older adolescents, lower multiculturalism made the increasing-negative trajectory more likely (vs. normative-positive). As for control variables, girls were more often in a normative-positive trajectory (vs. other trajectories). Majority adolescents (Table 4). We found some similar results for majority students. Those who perceived their school as more equal or more multicultural were more often in the normative-positive trajectory (vs. others). These effects were not replicated at the school-level: shared student perceptions of equality and multiculturalism had no significant effects, but teacher perceptions were inversely related to majority trajectories. When teachers reported more multiculturalism or less assimilationism, majority adolescents were less likely to be in the normative-positive trajectory (vs. moderate-positive). Older majority adolescents were also less likely to be in the normative-positive trajectory (vs. moderate-positive). Despite a significant interaction between age and perceived equality, there were no meaningful differences in simple slopes (see Supporting Information with estimated probabilities for this interaction). As for gender, girls were more likely to have a normative-positive trajectory.

Associations of Trajectories With Minority and Majority School Outcomes
Minority adolescents (Table 5). Relationship trajectories were unrelated to T3 school grades    .02 Note. (controlling for T1 grades). Those in the normativepositive trajectory showed higher T3 school compliance and higher emotional engagement (vs. the other trajectories), lower behavioral disengagement and higher belonging (vs. the increasing-negative trajectory; controlling for T1 outcomes). Additionally, diversity approaches (T1) were directly related to school outcomes (T3): When minority students perceived their school as more equal, their school compliance and belonging increased over time. At the school level, however, student perceptions of multiculturalism predicted more behavioral disengagement for minorities. As for gender, girls reported more compliance, but less emotional engagement than boys. Majority adolescents (Table 6). Relationship trajectories were unrelated to T3 grades and compliance. Majority adolescents in the normative-positive trajectory reported higher T3 emotional engagement (vs. the decreasing-negative trajectory) and belonging (vs. the other two trajectories). Additionally, school-level diversity approaches at T1 also directly predicted T3 outcomes: Shared student perceptions of equality predicted higher school compliance, whereas teacher perceptions of lower multiculturalism or higher assimilationism predicted higher engagement over time. As for gender, girls reported higher grades, more compliance, higher behavioral engagement and less disengagement than boys.

Discussion
Against the background of persistent ethnic disparities in European schools, we examined the quality of evolving teacher-student relationships during adolescence as a protective proximal process in an ecological systems approach to development (Bronfenbrenner, 2005). Drawing on a large-scale multilevel (students within schools) and longitudinal (three waves, three cohorts) data, our study adds a unique comparative dimension by comparing Turkish and Moroccan minority and Belgian majority adolescents. Our goal was to disentangle generic change for all adolescents (e.g., a general decline in relationship quality with age) from distinctive challenges that minority adolescents face (e.g., increased risk of developing problematic relationships with teachers). Thus, we add to an emergent research stream on distinctive minority experiences in developmental science (Garc ıa Coll et al., 1996;Juang et al., 2018;Syed, Santos, Yoo, & Juang, 2018). Rather than amalgamating minority students, within both ethnic minority and majority groups, we differentiated the most common "normative-positive" trajectories of teacher-student relationship quality from specific problematic trajectories and assessed age-related changes in trajectories through adolescence. Moreover, we not only looked at developmental consequences (i.e. various school outcomes) of these trajectories (Bosman et al., 2018;Spilt et al., 2012) but also for the first time showed which cultural diversity approaches in school afforded normative-positive trajectories across minority and majority adolescents. In the following section, we summarize our contributions and consider limitations and implications for future research.

Trajectories of Teacher-Student Relationship Quality
Our findings extend the limited evidence on trajectories of teacher-student relationship quality (Bosman et al., 2018;O'Connor & McCartney, 2007;O'Connor et al., 2011O'Connor et al., , 2012Spilt et al., 2012) by comparing how ethnic minority and majority adolescents' relationship quality with teachers changes through secondary school. The novelty in our comparative approach was to identify the trajectories within minority and majority samples separately, which allows for group differences in both form and frequency of relationship trajectories. We were also the first to look at trajectories of teacher-student relationships during adolescence. Looking at the common trajectories, we identified a normativepositive trajectory (with high teacher support and low rejection) for most adolescents, replicating earlier findings with primary-school children (Bosman et al., 2018;O'Connor & McCartney, 2007;O'Connor et al., 2011O'Connor et al., , 2012Spilt et al., 2012). Thus, highquality relationships with teachers appear to be the norm among both ethnic minority and majority Figure 3. Interaction between age and multiculturalism for minority adolescents. * p < .05; ** p < .01. Table 4 Predicting Teacher-Student Relationship Trajectories by Cultural Diversity Approaches for the Majority Sample .02 Note. Table presents unstandardized regression coefficients with standard errors in parenthesis and odds ratios (OR) for multinomial logistic regression. Normative-positive trajectory is the reference category. The nonsignificant age interaction that was dropped from the final analysis is indicated by ns.
Intraclass correlations are based on the models without any predictors. † p = .05. *p < .05. **p < .01. ***p < .001. Table 5 Predicting  .07 .03 .07 .05 Note.   .01 .09 Note. adolescents. We also identified a decreasing-negative trajectory (with moderate teacher support and initially high, but decreasing rejection) for both minority and majority adolescents, replicating another common trajectory of improving relationship quality found in earlier work (Bosman et al., 2018;O'Connor et al., 2011O'Connor et al., , 2012Spilt et al., 2012). We also found group-specific trajectories. An increasing-negative trajectory (with moderate teacher support and initially low, but increasing rejection), showing worsening relationship quality was found for minority adolescents only, replicating Spilt et al. (2012) findings with minority children. However, other studies have found this trajectory in majority samples as well (Bosman et al., 2018;O'Connor et al., 2011O'Connor et al., , 2012. Possibly, increasing teacher rejection, as distinct from the conflict assessed in earlier studies, is more common among minorities (vs. majorities). Our finding of increasing rejection resonates with previously reported increases in school discrimination for ethnic minority children (Benner & Graham, 2011). Finally, we found a moderate-positive trajectory for majority adolescents only (with moderate support and low rejection), replicating earlier findings with majority samples (Bosman et al., 2018;O'Connor & McCartney, 2007;O'Connor et al., 2012). Although some earlier studies found a primarily negative relationship trajectory for very small subgroups (3%-5%; O'Connor et al., 2011O'Connor et al., , 2012Spilt et al., 2012), this pattern did not emerge in our data.
Earlier studies either analyzed closeness and conflict measures separately (Bosman et al., 2018;Spilt et al., 2012) or aggregated them as one construct (O'Connor & McCartney, 2007;O'Connor et al., 2011); in contrast, our trajectories of relationship quality combined changes in both teacher support and rejection. Across different trajectories, teacher support varied less (high vs. moderate) than rejection (low, high-decreasing and low-increasing). This finding resonates with earlier studies, which reported more variation in teacher-reported conflict than in warmth (Bosman et al., 2018;O'Connor et al., 2012;Spilt et al., 2012). Thus, evidence on teacher-student relationship trajectories suggests more action on the negative side than on the positive side of relational experiences. More research is needed to clarify this asymmetry.
Comparing minority and majority trajectories, problematic relationships were more likely for minority adolescents. Only 5% of majority vs. 22% of minority adolescents were in problematic trajectories (increasing-negative or decreasing-negative trajectories). Even when we restrict the comparison to the common trajectories, minority adolescents were less often in a normative-positive trajectory. This finding extends the existing evidence on teacher-student relationship trajectories, which showed that ethnic minority children were underrepresented in low conflict (vs. increasing conflict) trajectories (Spilt et al., 2012), and that teachers reported lower closeness and higher conflict with ethnic minority children (Bosman et al., 2018). Overall, our findings suggest distinctive relational experiences for ethnic minority adolescents in school McGrath & Bergen, 2015;Sabol & Pianta, 2012), in terms of both the form of relationship trajectories (more problematic relationship trajectories for ethnic minorities) and their frequency (a higher risk of being in problematic trajectories).

Cultural Diversity Approaches
Our findings showed, for the first time, that school diversity approaches made a difference in trajectories of teacher-student relationship quality. When students perceived their school as more equal and multiculturalist, both minority and majority adolescents were more often in the normative-positive trajectory (vs. other trajectories). Because minority adolescents are at risk of experiencing unfair treatment in school (Bottiani, Bradshaw, & Mendelson, 2016), our finding that the perception of an inclusive diversity approach helps both majority and minority adolescents to form and maintain positive relationships with teachers is promising. As reverse causation is less likely for longitudinal relationship trajectories, our findings strengthen recent evidence linking equality and multiculturalism to school outcomes (Celeste et al., 2019;Schachner et al., 2016). For minority youth only, we also replicated positive effects at the school level so that shared student perceptions of equality made normative-positive relationship trajectories more likely. Perceived fairness affords normative-positive relationships with teachers, particularly for minorities, in line with findings of enhanced minority school outcomes in 'fair' schools (Baysu et al., 2016;Schachner et al., 2016).
Interestingly, teacher perceptions of multiculturalism and assimilationism affected minority and majority trajectories differently. When teachers perceived schools as more multicultural and less assimilationist, minority adolescents were more likely, but majority adolescents less likely, to form normative-positive relationships with their teachers. The minority group results resonate with earlier findings Pulinx et al., 2015). In contrast, majority students seem to benefit more from assimilationism than multiculturalism (as reported by teachers) as the latter makes them less likely to be in normative-positive trajectories. Reasoning from social identity threat versus affirmation, a multiculturalist approach may not be seen as all-inclusive by majorities (Plaut et al., 2011), whereas assimilationism might affirm the majority identity in highly culturally diverse schools.
To conclude, in schools with egalitarian or multiculturalist diversity approaches, minority adolescents were more likely to experience normativepositive relationships and less likely to experience problematic relationships. However, for majority adolescents, relationship quality was less consistently associated with school diversity approaches.

School Outcomes
Positive relationship trajectories were associated with longitudinal changes in affective school outcomes (i.e., emotional engagement and school belonging) for both minority and majority adolescents, in line with the protective function of supportive relationships with teachers for all students (Baysu & Phalet, 2012;Sabol & Pianta, 2012). We further asked whether relationship trajectories would affect the school outcomes of minority students more strongly if, in line with an intergroup relations perspective (Derks et al., 2007;Verkuyten et al., 2019), teacher support can affirm, and experiences of teacher rejection threaten, the distinct identity of minority adolescents in schools as an intergroup context. Consistent with this reasoning, relationship trajectories were most consequential for minority adolescents: a normative-positive trajectory uniquely protected them from behavioral problems so that they reported less behavioral disengagement and noncompliance. Confirming the academic risk hypothesis (Hamre & Pianta, 2001), minority adolescents seem to benefit more supportive relationships with teachers, which can promote their engagement in school despite migration-related adversities .
Within both groups, the normative-positive group had significantly better outcomes compared to the most problematic trajectory in that group. Minorities who were in increasing-negative trajectories were most vulnerable, in line with the findings by Spilt et al. (2012). For majorities, a decreasingnegative trajectory had the worst outcomes, in line with the findings of O' Connor and McCartney (2007) in their largely majority sample. Unlike with these studies, however, we did not find any association between trajectory patterns and achievement (grades). This is not inconsistent with evidence from two recent meta-analyses (Roorda et al., 2011(Roorda et al., , 2017 that found weak associations between teacher-student relationships and achievement. Moreover, the impact of teacher support on achievement in secondary school can be limited by structural inequalities in hierarchical tracking systems in Europe (Eccles & Roeser, 2011). Highly tracked educational systems stream students at an early age into different schools to follow vocational or academic secondary education; decisions determine a student's achievement and are often irreversible (Baysu, Alanya, & de Valk, 2018;Eccles & Roeser, 2011). Even when teachers are supportive in already tracked secondary schools, it may be "too little, too late" to counteract existing discrepancies in achievement.

Age and Gender Differences
The finding that older adolescents were less often in normative-positive trajectories is suggestive of general developmental processes and replicates other findings of declining relationship quality with teachers during adolescence (McGrath & Bergen, 2015;Roorda et al., 2011). Protective effects of multiculturalism were also age-dependent for minority adolescents: older (vs. younger) adolescents in less multiculturalist schools were more at risk of increasingly negative relationships with teachers. From a developmental intergroup perspective, older minority adolescents are more vulnerable to discrimination (Baysu et al., 2016). Because they are more aware of their minority status (Umaña-Taylor et al., 2014) and the related risks of ethnic victimization (Killen et al., 2007), they may be more apprehensive about teacher rejection when cultural diversity is not valued in their school.
We also found several gender differences. Both minority and majority girls were more likely than boys to have normative-positive relationships with their teachers (Bosman et al., 2018;Hamre & Pianta, 2001;Spilt et al., 2012). In terms of outcomes, majority girls did better than boys on five out of seven outcomes (including grades), in line with reported gender advantage in school outcomes favoring girls over boys (Baysu & de Valk, 2012). However, ethnic minority girls did better than boys on only one out of seven outcomes (and did worse in one outcome). Baysu and De Valk (2012) similarly found that the gender advantage of girls did not hold for the minorities when they compared the school careers of majority versus Turkish and Moroccan minorities in four European countries. More research is needed on the intersection of ethnicity with gender for school outcomes.

Limitations
We see diversity approaches as characteristics of the school environment, and teacher-student relationships and affective outcomes such as belonging as individual experiences. However, the quality of relationships can also be a school characteristic such that certain schools can be defined as having higher or lower quality relationships or connectedness (Morin et al., 2013;Thapa et al., 2013;Wang & Degol, 2016). From an ecological systems approach, these different layers are interrelated due to the shared experience of an ecological system (Thapa et al., 2013). We should thus acknowledge the possible reciprocal influences between cultural diversity approaches, teacher-student relationship quality, and school outcomes. Future research should look more closely at bi-directionality in these associations.
Our study had other limitations. First, teacher reports of relationship quality could complement our focus on adolescents' own experiences of relationship quality. Second, student-reported noncompliance and teacher-reported assimilationism had low reliability, but the additional analyses with these measures confirmed the results reported here. Although minority students' perceptions of equality also indicated low reliability, majority and shared student perceptions of equality had better reliabilities. Moreover, positive associations of minority perceptions of equality with trajectories were replicated at the school level with shared student perceptions, which strengthens the conclusions regarding this measure. Third, teacher and student perceptions of cultural diversity approaches were incompletely assessed (e.g., we had teacher, but not student perceptions of assimilationism), so we cannot fully separate out differences in perspectives from different approaches. Fourth, teacher data were available for only two thirds of the schools. This reduction may limit the generalizability of our findings on teacher perceptions (however, the FIML estimation in Mplus was robust). Fifth, school-level contextual effects should be qualified in light of low intraclass correlations. Finally, minority adolescents had a wider age range and were overrepresented in vocational tracks compared to majority adolescents. This is because ethnic minority students in Belgium are more often held back a year and referred to vocational education by their teachers (Baysu & de Valk, 2012;Baysu et al., 2016Baysu et al., , 2018. Still, we advise caution in inferring causal factors.

Applied Implications
Despite these limitations, our research findings have important applied implications. Our integrative models reveal continuity and change in adolescents' relationship quality with teachers as critical processes connecting early school environment to later school outcomes. Positive relationships with teachers are generally protective, but minority adolescents are more likely to benefit from positive relationships with teachers and at the same time more likely to lack such support. From an applied perspective, these findings suggest that schools can develop specific processes to protect at-risk minority adolescents from adverse outcomes. Our findings thus highlight the long-term benefits of practicing and communicating fairness and the value of diversity in building enduring positive teacher-student relationships, as well as the need for interventions to remedy worsening relational experiences.
Although majority adolescents benefitted from a positive diversity climate, we also found that in schools with stronger multiculturalist policies (as reported by teachers), majority adolescents were more likely to have moderate-quality rather than high-quality relationships with their teachers. Future research should ask whether actual teacher practices vary in schools with different cultural diversity approaches and which practices can make majority adolescents feel excluded. Thus, it is critical to work toward developing all-inclusive diversity approaches in which both ethnic minority and majority adolescents feel welcome in order to afford positive relationships and outcomes for all adolescents.