1 Introduction

From 2005 to 2019, the proportion of school-aged children of immigrant descentFootnote 1 rose from 9.6 to 39.1% in Germany (German Federal Office for Migration and Refugees, 2008; German Federal Statistical Office, 2021). In major metropolitan areas (e.g., Hamburg, Frankfurt, the Ruhr Area), students of immigrant descent constitute the numerical majority (Blossfeld et al., 2016). The mixing of students from different ethnic groups has been linked to positive adjustment outcomes and pro-social behaviors, including an increase in creativity and cross-ethnic friendships (Schachner et al., 2015; Vezzali et al., 2016). Yet students in ethnically diverse schools might also face inter-personal tensions, such as being confronted with ethnic discrimination (Diehl & Fick, 2016). In a national report on discrimination based on ethnicity among 18,000 surveyed participants (Beigang et al., 2017), results indicated that school-based discrimination experiences (e.g., peer exclusion, unfair treatment by teachers) are not uncommon, with 23.7% reporting such incidents in the last two years. Specifically, the current study focused on students of Turkish and Arab descent because they are particularly exposed to high levels of ethnic discrimination in German schools (Kunyu et al., 2021; Vietze et al., 2023).

Experiences of school-based ethnic discrimination can be challenging to deal with because they may reduce feelings of inclusion and efficacy, leading students of immigrant descent to disengage from the school socially and academically (Verkuyten et al., 2019). School engagement represents a robust predictor of academic performance of all children and adolescents. Prior work on the relationships between school-based ethnic discrimination and school engagement has mainly relied on between-person associations, showing that students who report more ethnic discrimination are more likely to disengage from school than their peers who report less ethnic discrimination (Civitillo et al., 2021; Coutinho & Koinis-Mitchell, 2014; Leath et al., 2019). Typically, these studies have relied on cross-sectional data, wherein targets of discriminatory actions report their memories retrospectively (e.g., over the last year), and the assessment of school engagement is conducted at one point in time. However, studies that used experience sampling methodology (e.g., daily diaries) demonstrated that discrimination experiences vary from one day to next, uncovering within-person fluctuations in the same ethnic-racial group (Douglass et al., 2016; Seaton & Douglass, 2014; Seaton & Iida, 2019; Wang & Yip, 2020; Zeiders, 2017). Similarly, research on within-person school engagement does not exclude that on one day, a student may be attentive, interested, and attempt to use sophisticated learning strategies, while the next day the same student may exert limited effort, be bored, and use superficial learning strategies (Martin et al., 2015; Patall et al., 2018). Thus, the purpose of the current study was to extend prior work on between-person levels by examining the short-term, within-person daily links between school-based ethnic discrimination and school engagement in a sample of secondary school students of immigrant descent (i.e., Turkish and Arab origin) over a two-week period.

Even studies that have used traditional longitudinal designs (e.g., yearly assessments of school-based ethnic discrimination) typically conflate between- and within-person sources of variance (e.g., by using traditional cross-lagged panel models). Where between-person analysis uses individuals as the unit of analysis, examining associations between variables based on individual differences, within-person analysis relies on time-points within individuals as the unit of analysis (Murayama et al., 2017). The consequences of failing to distinguish between these two sources of variance are quite severe as research in other fields is starting to show. For instance, one study showed between-person but not within-person effects of intergroup contact when distinguishing these sources of variance (Friehs et al., 2023). Similarly, two studies failed to show any evidence for within-person effects of volunteering for individual wellbeing (Lühr et al., 2022a, 2022b), which contrasts with previous literature, in which this connection seemed well-established. Importantly, most theories make assumptions about between-person change (e.g., when a student makes discrimination experiences at school, this will lower their school engagement) and it is thus pivotal to show that associations indeed exist at the within-person level. In addition, it is important to understand the differences between intensive (e.g., daily diary) and more traditional (e.g., yearly assessments) longitudinal designs and which questions they can answer (cf. McNeish & Hamaker, 2020). Intensive longitudinal designs focus on processes where the main focus is on variability around a stable mean (e.g., day-to-day fluctuations in school engagement), and on variables that predict within-person variability (e.g., are students less engaged in school on days that they make discrimination experiences?). This is different from more traditional longitudinal designs where the primary interest is in mean-level change (e.g., growth in school engagement across one school year) and in explaining between-person variability (do students who are discriminated more display different growth in school engagement?). The primary focus of this paper is on within-person change.

1.1 School engagement as a multidimensional and fluctuating construct

School engagement can be defined as the quality of children’s participation in or involvement with school activities (Wang et al., 2019). Students who report higher levels of school engagement tend to score better on standardized tests (Dotterer & Lowe, 2011; Wang & Eccles, 2013). Conversely, low levels of school engagement are associated with underachievement (Chase et al., 2014), youth behavioral problems (Hirschfield & Gasper, 2011; Li & Lerner, 2011), and increased risk of dropping out of school (Fall & Robert, 2012; Janosz et al., 2008). These results hinder societies to prosper by capitalizing on youth potential and to provide equal opportunities for all– independent of ethnic background.

We draw on Fredricks’ and colleagues’ (2004) holistic framework and description of behavioral, emotional, and cognitive engagement. Behavioral engagement entails positive conduct, including student’s degree of compliance with school and classroom rules, as well as manifesting disruptive behaviors such as skipping school and getting in trouble; emotional engagement reflects students’ positive and negative emotions such as happiness and anxiety attached to school, classmates, and teachers; and cognitive engagement entails students’ investment in learning activities, including how thoroughly they complete school and homework, or the time they invest in enrichment activities that relate to the topics being discussed. School engagement dimensions are correlated and may reciprocally influence each other (Fredricks et al., 2016; Wang & Eccles, 2013; Wang et al., 2011), yet the relationship between these dimensions and discrimination experiences is complex and not fully understood. Previous research has produced mixed findings (Dotterer & Lowe, 2011; Smalls et al., 2007), which deserve further investigation.

Longitudinal research on school engagement has consistently shown that students tend to disengage from school when they progress from primary to secondary school and from the start to the end of the school year (Chase et al., 2014; Li & Lerner, 2011). Notably, using experience sampling methodology, studies among secondary students have shown day-to-day fluctuations in school engagement. For instance, in a sample of 94 African American and Latinx ninth graders, Park and colleagues (2012) found that situational changes in students’ perceptions of competence and relatedness were associated with emotional school engagement changes over a 1-week period. In a study by Patall et al. (2018), 208 secondary school students were followed across six weeks. Results indicated that daily changes in interactions with other classmates and teachers were associated with subsequent daily changes in student’s behavioral, emotional, and cognitive engagement. Gillen-O’Neel (2019) found similar day-to-day fluctuations in emotional and behavioral engagement among 280 first-generation college students. Thus, prior work demonstrates that school engagement is best recognized as a construct fluctuating, largely depending on the relational environment at school.

Because of the interaction between the relational environment and the individual, there is likely to be variability within students’ engagement responding to the diverse stimuli encountered in each school day (Martin et al., 2015). This is in line with the transactional model of development (Sameroff, 2009), which underscores the dynamic interplay between the individual and the environment, influencing within-person changes over time. With a foundation in within- and between person heterogeneity, the transactional model views individual development in terms of possibilities shaped by the dynamic transactions between the person and the environment. Consequently, although sharing the same academic environment, every student experiences teachers, peers, and school activities differently. Thus, considering the way the academic environment is received and experienced by the individual, in the current study, we adopt a transactional rationale for anticipating daily-level variations in school engagement among students. Whereas traditional cross-sectional studies focus on group-level variance, intensive longitudinal designs (e.g., daily diary) with repeated measures in multilevel models represent an ideal data analytical approach to test transactional processes involving school engagement (Sameroff & MacKenzie, 2003).

1.2 Theoretical and empirical framework linking ethnic discrimination and school engagement

Although school engagement is seen as a positive experience, researchers have suggested that negative experiences, such as school-based ethnic discrimination, may drive the within-person variability in school engagement (Baysu et al., 2016; Buhs et al., 2006; Wang et al., 2019). The present study draws upon the conceptual model put forth by Verkuyten et al. (2019), which highlights the interrelations between ethnic discrimination, social identity needs, and school engagement (Motti-Stefanidi & Masten, 2013; Reeve, 2012; Wang & Eccles, 2013; Wong et al., 2003). Grounded in identity process theory (Breakwell, 1986) and self-determination theory (Ryan & Deci, 2000), this conceptual model posits that individuals who are exposed to ethnic discrimination may experience social identity threat because discrimination can convey to individuals that they are devalued due to their ethnic group membership (Umaña-Taylor, 2016). Identity threat is problematic because group identities tend to satisfy a range of social identity needs (Breakwell, 1986). More specifically, the identification with a particular group fulfils the needs of control (sense of efficacy) and belongingness (closeness to others). If fulfilment of these needs is threatened, individuals may react by psychologically disengaging from a particular domain such as school (Tajfel & Turner, 1979). Nevertheless, we cannot assume that a student who is exposed to discrimination experiences is singularly disengaged throughout the school year. The salience of individual experiences in a given day and the way these events are processed through the student’s perspective and orientation may account for the variability in school engagement. This may be particularly true during adolescence, when youth make sense of ethnic tensions and conflicts due to individuals’ growth in cognitive and social maturity that occurs around this time (Umaña-Taylor, 2016).

Discrimination is a complex phenomenon, and not all discriminatory experiences affect social identity needs in the same manner. Verkuyten and colleagues (2019) consider two key aspects of discrimination — behavior and source — and highlighted their specific role in shaping school engagement. The first aspect, characteristics of behavior, refers to the different forms that discrimination can take, for example, overt or subtle. Overt discrimination is the blatant differential treatment of ethnic minorities, based on certain personal characteristics or group affiliations that are deemed inferior to those of the dominant group (Dovidio & Gaertner, 2004). Examples of overt discrimination include being insulted, called names, being hit, or handled roughly (Noh et al., 2007). Such episodes may escalate and push the target of discriminatory actions to react as a way to re-establish a sense of control (Hirschfield & Gasper, 2011). This, in turn, may limit the time and energy spent in classroom participation or lead students to violate school rules on the day of the discriminatory event (Wang & Fredricks, 2014), but could continue the subsequent day. Hence, we hypothesized that overt discrimination by peers would be associated with lower levels of behavioral school engagement on the same day and the next day (Hypothesis 1a).

Previous evidence from the US suggests that overt discrimination occurs relatively infrequently in the school context (1–2 times a week, see Civitillo & Jugert, 2023, for a review of 18 studies). Yet researchers have highlighted that subtle forms of interpersonal discrimination are more pervasive in the school ecology (Steketee et al., 2021; Sue et al., 2007). Subtle discrimination manifests through behaviors or verbal comments that convey insensitivity or demean a person’s ethnic identity (‘You speak German well’; ‘Oh, this is really great workespecially for a Turk’; ‘We don’t do this in Germany’; Colak et al., 2023; Moffitt et al., 2019). Subtle discrimination reiterates, on a daily basis, who belongs and who does not, presuming a permanent foreign status (Armenta et al., 2013; Juang et al., 2021), and thus harming a sense of school belonging. From a motivational perspective, the feeling that one belongs to and is a member of the community has consistently been found to be related positively with emotional engagement in primary and secondary school students (Fredricks et al., 2004; Lawson & Lawson, 2013). However, the emotional dimension of school engagement taps into positive and negative affective reactions, and it is likely that affective states are not stable but change from moment-to-moment (Murayama et al., 2017).

Unlike overt discrimination, subtle discrimination is more challenging to interpret, which may make students to question whether their ethnic background was the cause of the experience. This uncertainty can lead them to engage in rumination, spending time trying to figure out if the experience was genuine and if the perpetrator’s intention was discriminatory or not (Yip et al., 2022a, b). As a result, when students experience subtle discrimination, they may feel an immediate impact, but there could also be a slight delay (1-day lag) in processing the emotions associated with their experiences in the educational settings. Thus, we expected that subtle discrimination by peers would be related to lower emotional school engagement on the same day and on the next day (Hypothesis 2).

The second aspect of discrimination, characteristics of source, encompasses the perpetrator of discrimination. There is systematic evidence that perpetrators of ethnic discrimination in the school context are peers, classmates, as well as adult figures such as teachers (Civitillo et al., 2023). Source-specific experiences of discrimination (peer- vs. teacher-based) may have a differential impact on school achievement. For example, Benner and Graham (2013) found that greater discrimination from teachers was associated with poorer academic performance, whereas greater discrimination from peers predicted higher levels of anxiety, loneliness, and depressive symptoms. Thus, examining the unique contribution of source-specific experiences of discrimination may help to better understand how students react to discrimination and how this relates to their engagement in school.

Outside of their home environment, teachers represent the first authority figures that youth come across. Prior work has shown that ethnic discrimination by teachers is particularly detrimental for cognitive and behavioral engagement because it is associated with low levels of student’s sense of control (D’Hondt et al., 2016). Students who are discriminated against by teachers, for example, believe that they have no control over educational success and feel that the school system is working against them (Agirdag et al., 2012; Van Houtte & Demanet, 2016). In line with social identity threat, both cross-sectional and longitudinal studies have shown that minority students who experienced discrimination from teachers put less cognitive effort into classroom learning (Baysu et al., 2016; Heikamp et al., 2020) and reported higher levels of behavioral misconduct (D’Hondt et al., 2017). By using a daily diary approach, we aimed to expand this evidence by capturing how potential negative interactions with teachers varied from day to day, and whether the consequences could be immediate or short-term. Thus, we expected that following discrimination by teachers, students would report lower levels of cognitive and behavioral engagement on the same day and on the next day (Hypothesis 3a and 3b).

1.3 The present study

There are several indications in research showing stressors that minority youth experience within a particular context (e.g., school-based ethnic discrimination) influence their engagement within that setting, but evidence on temporal directionality of these associations is limited. Further, there are some limitations in the present body of knowledge on daily experiences of discrimination that the present study aimed at reducing. Most daily diary studies examining the negative consequences of ethnic discrimination have focused on psychological adjustment, including depression, anxiety, and sleep quality among US youth (Douglass et al., 2016; Seaton & Douglass, 2014; Seaton & Iida, 2019; Wang & Yip, 2020; Zeiders, 2017). In this study, we assessed the potential negative consequences of school-based ethnic discrimination on an academic domain (i.e., school engagement), allowing to identify barriers to educational success of ethnic minority students in a highly heterogenous context (i.e., Germany). Moreover, only one daily diary study (i.e., English et al., 2020) distinguished types of discrimination (overt vs. subtle), and different sources of discrimination (peer- vs. teacher-based discrimination), illustrating the complexity of these experiences. Finally, the current study is further distinguished by examining the consequences of discrimination on multiple dimensions of school engagement, as restricting the focus on one sphere of school engagement may provide an incomplete picture of school experiences (Dotterer et al., 2009). To sum up, we formulated the following hypotheses (all of which were preregistered on the Open Science Framework (OSF) project page, https://osf.io/rhaqt):

Hypothesis 1

On days when students report overt discrimination by peers, they would also report lower levels of behavioral school engagement the same day and the following day.

Hypothesis 2

On days when students report subtle discrimination by peers, they would also report lower levels of emotional school engagement the same day and the following day.

Hypothesis 3

On days when students report teacher-based discrimination, they report lower levels of behavioral (3a) and cognitive (3b) engagement the same day and the following day.

We investigated our hypotheses after controlling for participants’ family heritage and ethnic self-identification because these variables have helped to explain disparities in academic achievement and school engagement (Vietze et al., 2023). Additionally, given that prior research has produced mixed findings regarding the relationship of ethnic discrimination with gender and family cultural capital (English et al., 2020; Ríos-Salas & Larson, 2015), we also controlled for these variables.

2 Methods

2.1 Participants and procedure

During preregistration, the target sample size (150) was determined based on a Monte Carlo simulation. Due to time and cost limitations, however, this was not achieved. A post hoc power analysis was calculated using Murayama and colleagues’ (2022) shiny-app for mixed-effects modelling. This was performed using data from the current study (N = 87), with alpha set at 0.05, t = 3.00 (d = 0.32), and a power of 0.80. Results suggested a sample of 78 at Level 2 would be required to detect an effect with 80% probability.

Participants were 87 students (62% female, ranging from 14 to 17 years, Mage = 15) from all types of secondary schools (i.e., low-, medium-vocational, and academic track) (44.7% comprehensive school). Based on participants’, their parents’, and grandparents’ birth countries, participants were grouped into two family heritage categories: Turkey (40%), and predominantly Arab countries (60%), which included for example Syria, Iraq, and Lebanon. Students of Turkish as well as of Arab origin represent the highest share of students of immigrant descent in schools in the Ruhr Area, Germany (Ministerium für Kinder, Familie, Flüchtlinge und Integration des Landes Nordrhein-Westfalen, 2018). Informed consent was obtained from both students and their parents. The study was approved by the ethics committee of the department of psychology (University of Duisburg-Essen).

Two recruitment strategies were implemented. First, using a convenience sampling, recruitment of participants occurred in secondary schools and student support centers. Second, recruitment of participants occurred in youth clubs (e.g., sport organizations, religious associations) located in the same region. For the duration of the study, participants completed diaries via smartphones through the movisensXS experience sampling application (Movisens GmbH, Karlsruhe, Germany; movisens.com). Participants could use their own smartphone or got one on loan (free of charge) by the research team. During an orientation session in small groups, the researcher team described the purpose of the study, the procedure, and possible risks to participation. Information regarding age, gender, family heritage, ethnic self-identification, generational status, school grade, and school type (low-, medium-vocational, or academic track) were collected. In addition, all participants completed a baseline survey of ethnic discrimination and school engagement. The baseline measurements assessed ethnic discrimination and school engagement in the previous 30 days.

The daily-diary data consisted of 10 daily reports (Monday through Friday) over a 2-week period using an interval-contingent design (Bolger & Laurenceau, 2013). A 2-week period is the most common sampling period in studies assessing the consequences of daily ethnic discrimination during adolescence (Civitillo & Jugert, 2023). A 2-week period is not only common but also sufficient to capture the experiences of interest (i.e., daily school-based ethnic discrimination and daily school engagement). This period also accounts for attrition and the potential burden associated with an excessive number of observations. In an interval-contingent design, participants report experiences at regular and predetermined intervals of time set a priori by the researcher. The interval-contingent design is suitable for students because they have school daily schedules that repeat on a regular basis. To prompt diary completion, participants were notified to fill out daily entries (i.e., 21 questions) every day from 4.00 pm to 10.00 pm. If a response was not recorded within 10 min after the signal, a reminder was sent in the next time frame (i.e., 5.00 pm) and subsequently every hour in the case of non-compliance, for a total of seven daily prompts. It took around 3 min to complete a daily questionnaire. Participants received a 25 Euro voucher, as compensation for their time spent participating in the study. An additional voucher to the value of 25 Euro was given to participants who complete more than 80% of the assessments (the baseline measurement and eight daily surveys). The data collection commenced in March 2022 and was completed in early June 2022.

2.2 Measures

All items of the scales (in English and in German) measuring the main variables of this study are reported in the supplementary materials on the OSF project page.

2.2.1 School-based ethnic discrimination

Nine items measured school-based ethnic discrimination (three items for overt discrimination by peers: e.g., ‘Today in school another student called me names or teased me because of my ethnic origin’; three items for subtle discrimination by peers: e.g., ‘Today in school another student did not interact with me in a conversation because of my ethnic origin; and three items for teacher-based discrimination: e.g., ‘Today in school the teacher picked me less often because of my ethnic origin’). The response scale for all discrimination-related items ranged from: 0 (did not happen), 1 (happened once), to 2 (happened two or more times). All items have been used in daily diary studies or slightly adapted from well-established scales developed in the United States and adapted in Germany (i.e., English et al., 2020; Titzmann et al., 2021; Wong et al., 2003). Reliability for the three subscales was estimated separately for the within-person level and the between-person level using multilevel confirmatory factor analysis (Gelfdof et al., 2014). Internal consistency for overt discrimination by peer were estimated as ωBetween = 0.95, ωWithin = 0.53, for subtle discrimination by peer as ωBetween = 0.94, ωWithin = 0.52, and for teacher-based discrimination as ωBetween = 0.99, ωWithin = 0.46.

2.2.2 School engagement

Twelve items measured daily school engagement. In line with school engagement conceptualization (e.g., Fredricks et al., 2004), each dimension contains positively and negatively formulated items. Four behavioral engagement items were adapted from a scale by Amemiya and colleagues (2020, e.g., ‘Today I participated in classroom activities’; ‘Today in school I did other things when I was supposed to paying attention in class’ [reverse coded]). For emotional engagement, four items from a scale by Patall and colleagues (2018) were adapted (e.g., ‘Today in school I was happy’; ‘Today in school I felt bad’ [reverse coded]). Cognitive engagement was measured with four items taken from Patall et al. (2018): ‘e.g., Today I put effort into learning’; ‘Today I did not stay focus’ [reverse coded]). The response scale ranged from 0 (not true), 1 (somewhat true), to 2 (completely true), with high scores indicating strong engagement. Reliability indices for behavioral engagement were estimated as ωBetween = .76, ωWithin = .44, for emotional engagement as ωBetween = .86, ωWithin = .48, and for cognitive engagement as ωBetween = .73, ωWithin = .44. The items for each dimension were averaged to create a composite score per day and per participant.

2.2.3 Gender

We assessed participants’ gender with the response options (0) male, (1) female, and (2) non-binaryFootnote 2.

2.2.4 Family heritage

Based on participants’, their parents’, and grandparents’ birth countries, participants were grouped into the following family heritage categories: (1) Turkey, and (2) predominantly Arab countries.

2.2.5 Ethnic self-identification

Next to family heritage, we assessed participants’ ethnic self-identification with one open item (Leszczensky & Santiago, 2015), ‘Some people consider themselves to be German, for example, others Turkish, and others German-Turkish. What about you? How do you view yourself?’. We then created three groups (1) German, (2) mixed identification (e.g., German-Turkish), and (3) only non-German self-identification (e.g., Turkish). More than half (55.2%) identified predominantly only with the ethnic group of their family, while a smaller group (32.2%) reported a bicultural identification (e.g., German-Turkish), only four (4.6%) identified as German only, and seven (8%) did not answer.

2.2.6 Family cultural capital

As an indicator of family cultural capital, we used one item adapted from Sieben and Lechner (2019), addressing the number of printed books at home (1  = 10 books or less to 5 = more than 200 books; M = 1.9, SD = 1.1).

2.3 Plan of data analysis

To account for the nested nature of data — Level 1 daily observations in Level 2 participants — multilevel manifest models (MLMs) were conducted (Bolger & Laurenceau, 2013) using restricted maximum likelihood estimation to handle missing observations. Analyses were performed in R using multilevelTools (Wiley, 2020; version 0.1.1), esmpack (Viechtbauer, 2021; version 1.17), and nlme packages (Pinheiro et al., 2017; version 3.1–153). To separate between- and within-person associations of school-based ethnic discrimination, we first computed person-means of school-based ethnic discrimination predictors (i.e., overt discrimination by peers, subtle discrimination by peers, and teacher-based discrimination) for each participant across all available days of assessment. Then, we calculated deviations around the person-means for each participant and each measurement occasion (i.e., centered at the person mean). To truly examine the within-person associations, we also included the between-person component in our models (Wang & Maxwell, 2015).

To test hypotheses 1 to 3, we estimated separate models for the same-day and the next-day associations of school-based ethnic discrimination on school engagement dimensions, entering all three predictors (i.e., overt discrimination by peers, subtle discrimination by peers, and teacher-based discrimination) simultaneously. The effects of school-based ethnic discrimination predictors were allowed to vary across persons, resulting in random intercept and random slopes models. For the same-day associations, to adjust for the autocorrelation of the outcome variables that were adjacent in time, we used a first-order autoregressive AR(1) error structure (Bolger & Laurenceau, 2013). To study the association of prior day discrimination on next-day engagement, we included prior day engagement as a covariate. At the student level (i.e., Level 2), we controlled for gender, family heritage, ethnic self-identification, number of books at home, and days of assessment.

We conducted additional sensitivity analyses, using dichotomous indicators of school-based ethnic discrimination. More specifically, all nine variables measuring school-based ethnic discrimination were first recoded into a dichotomous indicator (i.e., 0 = no daily discrimination, 1 = at least one daily discrimination). Subsequently, we repeated the exact same analyses to check whether the direction or the significance of the results change.

3 Results

3.1 Descriptive statistics and preliminary analyses

Compliance varied, with 59 (67.8%) participants completing eight or more daily surveys, 18 (20.7%) completing between four and seven, and 10 (11.5%) completing between one and three surveys. Out of 18,270 potential data points (87 participants x 21 daily items x 10 days), 4711 were missing (25.8%). Twenty-seven students (31%) did not report any discriminatory events for the duration of the study.

Table 1 shows descriptive statistics, within- and between-person correlations among study variables. Figure 1 depicts the sample average for the baseline levels of discrimination (previous 30 days) and daily levels of discrimination across the study period with vertical bars representing standard errors of the means. On average for the duration of study, teacher-based discrimination (M = 0.17, SD = 0.33) was significantly more frequent than overt (M = 0.12, SD = 0.20), t(869) = 6.461, p < .001, and more frequent than subtle discrimination by peers (M = 0.13, SD = 0.21), t(869) = 5.689, p <  .001. Overt discrimination was less frequent than subtle discrimination by peers t(869) = -2.247, p < .05, although the difference was small. For descriptive purposes, we visualize the patterns of experiences of school-based ethnic discrimination and the relationship with school engagement in Fig. 2. More specifically, Fig. 2 shows the day-level sample average of behavioral, emotional, and cognitive school engagement for three groups of participants (students who did not experience discrimination, students who reported low levels of discrimination (below the median = 0.13), and students who reported high levels of discrimination of all sources and behaviors combined (above the median = 0.13)). According to Fig. 2, students who did not experience discrimination report higher levels of behavioral, emotional, and cognitive school engagement than students who reported low and high levels of discrimination. However, this analysis does not model the within-person associations.

Table 1 Descriptive statistics and bivariate correlations among study variables
Fig. 1
figure 1

Baseline (previous 30 days) and daily average levels of school-based ethnic discrimination across the study period with vertical bars representing standard errors of the means (N = 87)

Fig. 2
figure 2

Baseline (previous 30 days) and daily average levels of behavioral, emotional, and cognitive school engagement for three groups of participants (students who did not experience discrimination), students who reported low levels of discrimination (below the median), and students who reported high levels of discrimination of all sources and bahaviors combined

For subtle and overt discrimination by peers as well as behavioral and cognitive school engagement, ICCs were high (0.63, 0.54, 0.87, and 0.45, respectively), indicating that more than half of the variance can be attributed to stable between-person differences, and the remaining portion to daily fluctuations. Conversely, teacher-based discrimination (0.11) and emotional school engagement (0.15) were characterized by much stronger situational variability.

3.2 Multilevel models

3.2.1 Behavioral engagement as an outcome

For behavioral engagement, there was no significant within-person association of overt discrimination by peers on same-day or next-day level (for all multi-level results see Table 2). Similarly, teacher-based discrimination was not associated with a decline in behavioural engagement on the same and next day. This indicates that on days when students reported higher levels of discrimination than usual, they did not report lower levels of school engagement. Thus, Hypothesis 1 and Hypothesis 3a were not supported. However, on the between-person level, overt discrimination by peers across the study period was significantly related to lower behavioral engagement (β = -0.32, p < .001). This suggests that students generally reporting higher levels of discrimination did show lower levels of behavioral school engagement.

Table 2 Multi-level models predicting same-day and next-day behavioral, emotional, and cognitive school engagement

3.2.2 Emotional engagement as an outcome

For emotional engagement, we observed no within-person association between subtle discrimination from peers and same-day emotional school engagement. But results indicated a statistically significant at between-person level. Hence, students generally reporting higher levels of subtle discrimination by peers did show lower levels of emotional engagement across the study period. There was no lagged association of yesterday’s subtle discrimination on today’s emotional engagement. Hence, Hypothesis 2 was not supported. However, although not hypothesized, we found a 1-day lagged association of teacher-based discrimination on lower next-day emotional engagement (β = -0.09, p < .01). This means that on days when students reported higher levels of teacher-based discrimination than usual, they did report lower levels of emotional school engagement the following day.

3.2.3 Cognitive engagement as an outcome

Discrimination by teachers was negatively related to cognitive engagement on the same-day level (β = -0.09, p = .03). Thus, partially in support of Hypothesis 3b, on days students perceived unfair treatment by teachers due to their ethnicity, they reported a decrease in cognitive engagement. On the between-person level, results echoed within-level associations (β = -0.19, p < .01). This indicates that students generally reporting higher levels of discrimination by teachers did show lower levels of cognitive engagement across the study period. There were no associations between previous-day teacher discrimination and next-day cognitive engagement.

3.2.4 Sensitivity analyses

We conducted sensitivity analyses by using dichotomous indicators of school-based ethnic discrimination (Table S2 in the Supplementary Materials). For behavioral and cognitive school engagement, sensitivity analyses yielded similar results to those obtained using a 3-point answer scale. For emotional school engagement, we replicated the findings but also found a within-person effect of teacher based-discrimination on lower same-day emotional engagement (β = -0.06, p < .001).

4 Discussion

The present study benefited from the intensive longitudinal dataset that allowed us to examine both between- and within-person associations between school-based ethnic discrimination and school engagement across a two-week period during a critical phase in adolescence. Our data suggest that mainly discrimination by teachers was negatively related to school engagement both concurrently and from day-to-day. In addition, our analyses elucidated that the associations vary across dimensions of school engagement. Specifically, cognitive engagement was only affected by teachers while emotional engagement was influenced by both sources and there were no within-effects on behavioral engagement. Overall, our multidimensional approach points out that differentiating the type of behavior (overt- vs. subtle discrimination by peers) and source (peer- vs. teacher-based discrimination) of ethnic discrimination in the school context does matter for studying daily fluctuations in specific dimensions of school engagement.

Our descriptive results indicated that nearly 70% of the sample reported at least one discriminatory episode at school during the study period. This proportion is similar to other daily diary studies conducted among adolescents in the US (Zeiders, 2017). Although the occurrence of all types of school-based ethnic discrimination was on average rather low, with less than one episode per day, our results show that there were significant variations across sources in the reported frequency of discrimination. Consistent with previous daily diary investigations that compared peer vs. adult sources (e.g., Huynh & Fuligni, 2010), our participants reported experiencing more discrimination from teachers than from peers. It is plausible that some teachers may hold biased views towards ethnic minority students (Civitillo et al., 2023), contributing to be the perception of discrimination by some students. Furthermore, the participants in our study reported experiencing more subtle forms of discrimination from their peers than overt discrimination. This finding supports theoretical assumptions (Sue et al., 2007) and empirical work (English et al., 2020) that subtle discrimination is more pervasive than overt discrimination and extends this evidence to the school environment.

Our pattern of results showed mixed support for our hypotheses. In particular, we found no significant within-person associations in behavioral engagement for any of the discrimination forms and sources. At the between-person level, however, associations were significant for overt discrimination by peers. Thus, what does the presence of between-person associations combined with the lack of within-person associations for behavioral engagement mean? One explanation is that between-person associations might be driven by confounding with some third variables (Rohrer, 2018). For example, Fuller-Rowell et al. (2021) found that college students who reported particularly high levels of internalized racism experienced more pronounced adverse consequences of ethnic discrimination. Internalized racism occurs when individuals from stigmatized ethnic groups accept the negative messages about their abilities and self-worth (Jones, 2000). As a result, interindividual differences might be responsible for both higher levels of overt discrimination and lower levels of behavioral school engagement.

On average subtle peer discrimination (between-person associations) predicted decreases in emotional school engagement. Unexpectedly, we also found a significant negative association for teacher-based discrimination at within-person level. Findings support the transactional model that emphasizes the dynamic interplay between individual and environment, leading to intra- and inter-individual variability in school engagement (Sameroff, 2009). Importantly, our measure of teacher-based discrimination resembled subtle forms of discrimination because we assumed that teachers would not resort to name calling or teasing. The complexity of subtle discrimination and the fact that it could be initiated by an authority figure such as teachers may require additional time for students to process the emotions associated with these experiences. Taken together, these findings complement previous daily diary studies (Douglass et al., 2016; English et al., 2020), by suggesting that perceiving subtle discrimination can lead to delayed associations of school-based ethnic discrimination on students’ emotional school engagement due to the complexity and meaning-make processes of these experiences.

For cognitive engagement, we found that discrimination by teachers (within-person associations) was negatively related to cognitive engagement on the same-day level. This indicates that cognitive engagement is dependent on the type of daily interactions with teachers. This finding is particularly concerning because previous research has shown that teachers are instrumental for promoting cognitive engagement (Lauermann & Berger, 2021; Wang & Eccles, 2013). Cognitive engagement is important because it involves students recognizing the value of learning. Based on social identity threat theory (Verkuyten et al., 2019), this finding suggests that teacher discrimination could signal to students that they have little control over their educational success, and that the school system does not support them, which in turn, may result in disengagement. It should be noted, however, that there were no lagged within-person effects. This raises the question of what assets enable students to bounce back. Given that secondary school students interact with multiple teachers throughout the week, positive interactions with other teachers could potentially aid in a quicker return to baseline levels of cognitive school engagement.

4.1 Limitations and future research directions

This study has several limitations that have to be considered when interpreting its results. First, one constraint is the relatively short duration of the investigation, which was limited to 10 days. As such, it is possible that we did not capture the full range of consequences that school-based ethnic discrimination can have on students’ engagement over longer time periods. Future studies could combine the examination of everyday experiences using daily diaries with more traditional longitudinal design (e.g., by-yearly assessments) that are more suitable to capture long-term changes (Nesselroade, 1991). Second, the sample size of our study was relatively small. It is important to note that the null results for behavioral engagement may also reflect a limited power to detect significant associations. Simulation studies have shown that study power is more strongly influenced by the number of participants rather than the number of time points (Bolger & Laurenceau, 2013). Thus, future studies could benefit from including a larger number of participants, while maintaining the same number of assessments for reducing the burden for students on reporting sensitive issues. Third, the internal consistency of the scales for the within-person level was rather low. There might be several explanations for this issue. Daily diaries are designed to study constructs that vary, the inherent variability should contribute to lower internal consistency across days. In addition, internal consistency was developed and defined in classical test theory for trait-like concepts and not for measures of daily fluctuations. Furthermore, this could be a consequence of the low occurrence of specific phenomena (items) on specific days. Future studies should incorporate more items in the daily scales so that more facets of experiences can be assessed while improving the reliability of the construct.

An additional important direction for future direction will be to explicate the indirect pathways underlying associations between school-based ethnic discrimination and school engagement. Although we built on insights stating that engagement may fluctuate depending on the extent to which interactions in school satisfy or dissatisfy motivational needs (i.e., control and belongingness) on a daily basis, we did not measure those needs. Following the SDT and the social identity perspective (Verkuyten et al., 2019), experiencing discrimination can thwart students’ needs, undermining school engagement. To date, there are no existing studies that assess their mediation effects with the help of daily diaries.

4.2 Implications for educational practice

Efforts to reduce school-based ethnic discrimination are likely to be of substantial importance in the movement to reduce educational inequalities. School engagement has been consistently shown to be a key predictor of academic achievement (Wang et al., 2019). Our findings show that interventions should address different manifestations of school-based ethnic discrimination as well as its differential sources. Approaches that target solely overt forms of school-based ethnic discrimination or focus only on peer-to-peer discrimination are unlikely to work because they fail to account for the complexity of this phenomenon. Moreover, it is important for educators to acknowledge that school-based ethnic discrimination can manifest as a form of violence in everyday teaching and learning situations. This acknowledgement is different from a broad stance against racism and discrimination in education.

Further implications of our study stress the role of teachers. Teachers play an important role for promoting school engagement among students of all ethnic backgrounds (Motti-Stefanidi & Masten, 2013). Teachers hold a crucial position as an authority figure and are able to observe how peer interactions shape student’s academic, behavioral, and social development (Farmer et al., 2019). This puts them in a unique position to leverage the dynamics of the classroom to promote a supportive and adaptive community of diversity learners. In addition to cultivating anti-racist ideologies, teaching education institutions should reinforce teachers’ ability to identify and address issues of racism and discrimination in the everyday school life, what Shah and Coles (2020) refer to as racial noticing. Racial noticing concerns perceiving, making sense of, and reacting to moments where race and racism are salient in classroom (Shah & Coles, 2020). While teachers can prevent discriminatory behaviors among students (Schwarzenthal et al., 2023), they should also be mindful of their own potential to perpetuate discriminatory practices.

In conclusion, the differences identified between within-person and between-person associations between school-based ethnic discrimination and school engagement underscore the scientific need to pay more attention on inter- and intra-individual processes, because the conclusions drawn from each of these approaches may differ substantially. Whereas the identification of third variables may be a key interest in between-person associations (behavioral engagement), interventions that change individual experiences may be beneficial for within-person associations (emotional and cognitive engagement). Both in combination may be particularly powerful in creating non-discriminatory learning environments.