Us, them and we: How national and human identifications influence adolescents' ethnic prejudice
Abstract
Although there have been numerous studies on the relations between group identification and ethnic prejudice, it is less clear whether their associations reflect stable individual tendencies or rather situational or temporal fluctuations. This longitudinal multilevel study aimed to fill this gap by examining the between- and within-person associations of identification with the national and superordinate human groups and levels of prejudice against multiple ethnic minorities. A total of 883 Italian majority adolescents (Mage = 15.66, SD = 1.15 at T1, 49.7% females) completed questionnaires at four time points over the course of 1 year. Results showed that national identification was related to more prejudice at the between-person level but to decreases in prejudice at the within-person level. Additionally, human identification contributed to lower levels of and steeper decreases in prejudice at both the between- and within-person levels. Common and unique associations also emerged across different ethnic minority targets, but only for between-person effects. Overall, this study highlights the importance of distinguishing stable individual levels and momentary fluctuations of both ingroup identifications and ethnic prejudice in order to orient future interventions aimed at improving the quality of intergroup relationships.
BACKGROUND
We must conclude that prejudice cannot be explained entirely by approaching it at the level of generalized personality structure and dynamics. Situational, historical, and cultural factors are also important. (Allport, 1954, p. 73)
As indicated by this quote from Allport's seminal work The Nature of Prejudice, prejudice can depend on both personal and contextual factors. Still, the social psychological literature on this topic has been characterized by competing explanations whereby each set of factors is considered the main, although not unique, determinant of prejudice in spite of the other (Akrami et al., 2009; Hodson, 2009). Within this literature, the social identity approach (Abrams & Hogg, 2010; Brown, 2020; Reicher et al., 2011) is sometimes presented as a situational account, as it focuses on momentary processes of ingroup identification that guide intergroup attitudes and behaviours in a given context (Reynolds et al., 2017). However, this approach acknowledges the importance of individual differences as well (Reimer et al., 2022), which suggests that ingroup identification could explain why some people are generally more or less prejudiced than others and why they are more or less prejudiced than usual. A proper way to examine this possibility is by measuring identification and prejudice repeatedly over time and studying the relations between them at two different levels: a between-person level that represents their stable components across time, and a within-person level that represents their time-specific fluctuations. Although consistent with the social identity theorization, the empirical distinction between those levels has not been systematically addressed in the extant literature, which mainly relied on single-time assessments of identification and prejudice (Curtis, 2014; Pehrson, Vignoles, & Brown, 2009).
To fill this gap, the present research used a longitudinal multilevel design to uncover how ethnic prejudice is structurally (between-person level) and temporally (within-person level) related to national and human identifications in a large sample of Italian adolescents. Additionally, it examined whether the structural and temporal effects of identifications were outgroup dependent, by considering the most represented ethnic minority groups in Italian society (i.e. Romanians, Albanians, Moroccans, Chinese and Ukrainians; ISTAT, 2020). As a result, we could provide more nuance to the person–context debate on prejudice, and the role of ingroup identification in particular.
Ethnic prejudice: stability and fluctuations
Ethnic prejudice can be defined as a negative orientation or attitude towards others because of their different ethnic and cultural background (Allport, 1954). It implies both negative emotions (i.e. affective component) and stereotypes (i.e. cognitive component) expressed towards ethnic individuals and groups. Together, the affective and cognitive facets can lead to behavioural expressions of prejudice, which range from avoidance to aggression and discrimination (Brown, 2011; Cuddy et al., 2007).
Much of the social psychological work on prejudice has examined interpersonal differences and has relied on the assumption that these individual differences are relatively stable over time. By contrast, developmental psychologists are not only interested in stability but also in fluctuations and systematic change. Along this line, prior developmental work has examined how changes in ethnic prejudice throughout childhood and adolescence are intertwined with the progressive advancements of individual cognitive, social and moral competences and the socio-contextual influences to which youth are exposed (for a review, see Rutland & Killen, 2015). These processes can be especially impactful in adolescence, when youth face multiple developmental tasks, such as defining their personal and social identity (Crocetti et al., 2023), acquiring meaningful social and political stances (Rekker et al., 2015) and becoming engaged members of their community (Jahromi et al., 2012). Moreover, prior research has highlighted that attitudes formed in adolescence function as important organizing principles of their adult political orientations (Rekker, 2016). Relatedly, shedding light on individual and intergroup processes occurring in adolescence is relevant not only in light of the important developmental changes that characterize this life phase but also for their long-lasting impact on future generations' social and political views.
Meta-analyses (Crocetti et al., 2021; Raabe & Beelmann, 2011) have found that ethnic prejudice emerges early on, reaches a peak in middle childhood followed by a decrease and progressively consolidates from adolescence onwards. At this life stage, mean levels of ethnic prejudice as well as interindividual differences in both affective and cognitive facets remain relatively stable. This general stability trend does not necessarily imply an absence of change or a lack of temporal fluctuations (Crocetti et al., 2021). For instance, within the general population, subgroups of youth might follow multiple and divergent developmental trajectories (Bobba et al., 2023). Additionally, adolescents as well as adults might also display fluctuations around their own personal mean, as a consequence of momentary individual (e.g. social dominance orientation; Osborne et al., 2021) or macro-contextual changes (Allport, 1954). For instance, longitudinal research among adults has highlighted how temporal fluctuation in contextual features (e.g. media salience of terrorist attacks and increases in the share of immigrant population and unemployment rate) can contribute to temporal increases in prejudice against ethnic minorities (Finseraas & Listhaug, 2013; Legewie, 2013; Mitchell, 2019). Therefore, adopting a longitudinal, person-oriented approach (Bergman et al., 2003) that separates between- and within-person levels is fundamental to gaining a more nuanced understanding of the stability and fluctuations in prejudice (see Molenaar, 2004; Von Eye & Bogat, 2006).
Furthermore, socio-contextual and historical factors can impact ethnic prejudice differently, depending on the target considered. In this regard, prior studies highlighted how socio-contextual factors can differently influence changes in prejudice against some but not other ethnic minorities, rather than exerting a generalized effect across groups (Czymara & Dochow, 2018; de Rooij et al., 2015). These findings support the notion that beyond common variance among different prejudices – the so-called generalized prejudice – it is important to consider the feelings and emotions towards different ethnic groups (Bergh & Akrami, 2018). Building upon these premises, the current research took a group-specific approach by focusing on the Romanian, Albanian, Moroccan, Chinese and Ukrainian groups, which, due to immigration, are the most represented ethnic minorities in the Italian context (ISTAT, 2020). These outgroups differ not only by region of origin (i.e. Eastern Europe, Northern Africa and Asia) and history of migration to Italy (for reviews, see Abbondanza, 2017; Zincone & Caponio, 2006) but also in terms of their religious background (i.e. Catholic vs. non-Catholic) and appearance (e.g. skin colour) that could make their minority status more or less apparent. Additionally, one of the ethnic groups (Ukrainians) was involved in an international active conflict for most of the time of the current study (i.e. from early 2022 to early 2023). Together, these group-specific characteristics and socio-contextual conditions might contribute to differential evaluations of these outgroups and make group identifications more or less relevant for them.
Ethnic prejudice and group identification: the social identity approach
The social identity approach is one of the most widely used perspectives within the social psychological study of intergroup attitudes and behaviours (Abrams & Hogg, 2010; Brown, 2020; Reicher et al., 2011). It includes social identity theory (SIT; Tajfel & Turner, 1979) and self-categorization theory (SCT; Turner et al., 1987) and holds that outgroup prejudice depends on the extent to which individuals categorize themselves and others as group members, and the meanings they derive from these categorizations. SIT postulates that when group identities are psychologically salient, people are motivated to prefer their ingroups over the outgroups because this positive distinctiveness reflects positively on their selves (Tajfel & Turner, 1979). Ingroup preference can take the form of outgroup negativity and prejudice, but this is not inevitable and depends on other factors, including the ways that people define and understand their group (McGarty, 2001; Reicher et al., 2011). SCT explains when group identities are psychologically salient by specifying the conditions under which individuals self-categorize as group members rather than unique individuals. Because it posits that the activations and meanings of social identities are context dependent, the social identity approach is typically regarded as a situational account of prejudice (Hodson & Dhont, 2015). However, it also acknowledges that the variation in how people categorize themselves depends on individual differences in addition to contextual conditions. On the one hand, SCT claims that people are more likely to self-categorize as a member of a particular group when the differences between this group and other groups in a particular situation are perceived to be relatively large (comparative fit) and in line with expectations (normative fit). On the other hand, it also states that some individuals are more likely to use particular categorizations than others (perceiver readiness), and the degree to which they identify with the groups in question is typically regarded as an indicator of this (Turner et al., 1994).
Although it has long been acknowledged that ingroup identification is not necessarily related to outgroup negativity (Brewer, 1999; Hinkle & Brown, 1990), findings on the link between national identification and ethnic prejudice are largely in line with the social identity approach. Specifically, prior research has highlighted that individuals with higher levels of identification with their national group tend to report more negative attitudes against immigrant minorities (Luedtke, 2005; Meeus et al., 2010; Pehrson, Brown, & Zagefka, 2009; for meta-analyses, see Anderson & Ferguson, 2018; Crocetti et al., 2021). However, these associations are not inevitable and depend on other factors (see Pehrson, Brown, & Zagefka, 2009; Pehrson, Vignoles, & Brown, 2009; Smeekes et al., 2011; Spiegler et al., 2022). How national identity is represented is one of these factors, and research has shown that national identification is associated with more anti-immigrant prejudice in countries where there is a more cultural definition of nationhood (Pehrson, Vignoles, & Brown, 2009). In Italy, the context of the present study, such a definition seems to be present as well. For example, individuals of immigrant descent (i.e. born abroad or from immigrant parents) who live in Italy have limited opportunities to be involved in the country's political sphere, to vote or obtain nationality (MIPEX, 2020). Relatedly, while generally supportive of integration policies for invidual with a migrant background (Maratia et al., 2023), Italian (i.e. ethnic majority) adolescents have been found to mostly endorse a cultural definition of citizenship (Reijerse et al., 2015). Therefore, it would be reasonable to expect a positive relation between their national identification and levels of ethnic prejudice overall. However, as argued below, this relation may unfold differently at the within-person versus the between-person level.
Furthermore, SCT also posits a more inclusive level of self-categorization of humanity. Self-categorization at this level implies a focus on similarities with other humans rather than differences between ethnic groups. Therefore, consistent with the common ingroup identity model (Gaertner et al., 1993; Gaertner & Dovidio, 2000), considering oneself as a human being is assumed to facilitate a more positive attitude towards individuals belonging to different ethnic groups (see also Albarello et al., 2018; Albarello & Rubini, 2012). Along this line, research has shown that a stronger identification with humanity goes together with less ethnocentrism and less prejudice (for a review, see McFarland et al., 2019).
Stability and fluctuations in national and human identifications
In the social identity approach, group identification is typically conceived of as the degree to which group membership is incorporated into the self-concept (McGarty, 2001; Reimer et al., 2022). This conception suggests stability. However, social identity theorists have warned against ‘the idea that identification expresses some kind of fixed and stable self-structure or personality trait which is chronically salient across situations’ (Turner & Reynolds, 2001, p. 139). Instead, there is the acknowledgement that group identification ‘varies from individual to individual and from situation to situation’ (Reimer et al., 2022, p. 276, italics added). This suggests that identification can have both stable and fluctuating components. However, to the best of our knowledge, those have not been systematically differentiated in the existing literature.
The stable component of a particular group identification (e.g. national and human) can be estimated by measuring it repeatedly over time and calculating the mean across different measurement occasions. It indicates the degree to which the group membership is structurally important to the individual, and it varies between persons. By contrast, the fluctuating component varies within persons. It represents the extent to which an individual's group identification at a particular time deviates from their personal (stable) mean, and thus whether the group membership is more or less important than it usually is. Whereas the stable component could be used to examine why some individuals are generally more or less prejudiced than others (stable means) or become so over time (change rates), the fluctuating one could be used to examine why they are temporally more or less prejudiced than they normally are (see Thijs et al., 2023). Importantly, these are different questions. Thus, results at the between- and the within-person levels might not be the same, and this may have significant consequences for theory and intervention. For instance, several recent studies on intergroup contact (Friehs et al., 2023; Sengupta et al., 2023) found not only that outgroup contact was positively associated with more outgroup positivity and solidarity at the between-person level – which was in line with theory (see Allport, 1954) – but also that it had no significant effect at the within-person level. The latter indicates that personal increases in contact did not result in personal improvements in intergroup attitudes and behaviours.
Just like contact, national identification might be differentially related to ethnic prejudice at the between- and within-person levels. The social identity approach does not provide clear expectations about this because prior empirical studies within this research field have yet to uncover the implications of the distinction between both levels. However, one tentative possibility is that positive within-person level fluctuations in the strength of one's national identification reflect a more deliberate processing of the content of one's identity, which could either activate its default meaning or lead to a critical re-evaluation of it. Given their cultural definition of nationhood (see Reijerse et al., 2015), the default meaning of national identity for Italian adolescents may be one that excludes ethnic others, implying positive association of national identification with mean levels of prejudice at the between-person level. However, at the within-person level, the active processing of the content of their national identity might lead individuals to temporarily embrace either a more or less exclusive view of it. On the one hand, temporal increases in national identification might result in a momentarily heightened sense of nationalism that is based on exclusive views of group membership (Mihelj & Jiménez-Martínez, 2021; Zhuravlev & Ishchenko, 2020) and thus contribute to higher levels of prejudice against ethnic minorities (Pehrson, Brown, & Zagefka, 2009). On the other hand, temporal increases in national identification might represent moments of strengthened commitment to the national group resulting from the thorough and active exploration of its meaning and implications (Bosma & Kunnen, 2001; Luyckx et al., 2006), as typical of the identity consolidation cycle (for a review, see Branje et al., 2021). Although limited attention has been paid to the implications of exploration-based ingroup identification, some studies have found that youth with an achieved ingroup identity (i.e. high levels of commitment coupled with exploration) reported more positive attitudes towards ethnic outgroups (Phinney et al., 2007; Whitehead et al., 2009). Additionally, experimental research has highlighted that inducing participants to explore their national identity weakened the identification–prejudice link and led to more positive attitudes towards ethnic minorities (Spiegler et al., 2022). Thus, nuanced effects may be uncovered.
Whereas national identification may generally imply the exclusion of ethnic others but might nonetheless foster positive intergroup attitudes upon active and thorough exploration, the impact of human identification may be more unequivocal. Human identification implies an inclusive way of thinking about self and others as members of the same superordinate group. Therefore, both its stable levels and temporal fluctuations can lead individuals to adopt more positive views, generally and momentarily, about ethnic diversity.
The current study
The purpose of the current study is threefold. First, this research aims to study the associations between national identification and ethnic prejudice at the between- and within-person levels. At the between-person level, it examines whether and how stable differences in levels of national identification are associated with average levels of and average rates of change in affective prejudice. At the within-person level, it studies whether and how temporal fluctuations in levels of identification with the national group are linked to temporal fluctuations in prejudice. National identification is expected to be linked to significantly higher average levels of prejudice at the between-person level, whereas its association with average rates of change will be examined from an exploratory perspective. Furthermore, at the within-person level, fluctuations in levels of identification with the national group could either contribute to increases or decreases in prejudice.
Second and similarly, this study aims to examine whether and how stable differences and fluctuation in levels of human identification are associated with average levels and between-person changes on the one hand, and within-person fluctuations on the other in ethnic prejudice. At both levels, human identification is expected to be significantly associated with lower stable levels and momentary decreases in prejudice against all ethnic minorities, while the associations between stable levels of human identification and average rates of changes in prejudices will be examined from an exploratory point of view. Last, this study will explore whether the effects of national and human identifications on ethnic prejudice are group specific or rather similar across the five most represented ethnic minorities in the context of the study. Figure 1 outlines the expected associations between identifications and prejudices at the between- and within-person levels.
METHODS
Participants
Data for this research are drawn from the longitudinal project IDENTITIES ‘Managing identities in diverse societies: A developmental intergroup perspective with adolescents’, a cohort sequential study conducted in the north-east part of Italy (i.e. Emilia-Romagna region). Specifically, participants included in the current study were 883 adolescents (Mage = 15.66, SD = 1.15 at T1, 49.72% females) attending, at the beginning of the study (i.e. 2022), the 1st (48.24%) and 3rd (51.76%) year of high school. Participants completed four assessments in January/February 2022 (T1), April/May 2022 (T2), September/October 2022 (T3) and January/February 2023 (T4) respectively.
Only adolescents of Italian descent (i.e. whose parents were both born in Italy) were included in the current study. At baseline, adolescents reported that most of their fathers (48.33%) and mothers (50%) had a medium educational level (i.e. high school diploma). Among fathers, some of the remaining (26.67%) had a low (i.e. up to middle-school diploma) followed by those (25%) with a high (i.e. university degree or higher) educational level. Conversely, most of the remaining (34.88%) mothers had a high and only a few (15.12%) had a low educational level.
All adolescents included in the present study completed at least two of the four assessments, while more than half (59.68%) completed questionnaires at all time points. Within each assessment, the completion rate was high (ranging from 70.10% of items at T4 to 87.32% of items at T2) and missingness was mostly due to participants not filling out the questionnaire because they were not in school on the day of data collection. Little's (1988) missing completely at random (MCAR) test yielded a normed χ2 (χ2/df = 4405.62/3283) of 1.34, indicating that data were likely missing completely at random. Therefore, the total sample of 883 participants was included in the analyses, and missing data were handled with the full information maximum likelihood (FIML) procedure available in Mplus (Enders, 2013b).
Procedure
The present study was approved by the Ethics Committee of the Alma Mater University of Bologna (Italy) as part of the ERC-Consolidator project IDENTITIES ‘Managing identities in diverse societies: A developmental intergroup perspective with adolescents’. Schools were selected through a stratified (by track and level of urbanization) randomized method and principals were approached to present the project. Upon their approval, the study was presented to students and their parents who also received written and detailed information. Active consent from parents was obtained prior to their children's participation. Active consent was also obtained from adolescents of age, while their underage peers provided their assent to participate in the project. Participation in the study was voluntary, and students were informed that they could withdraw their consent at any time. At each wave, adolescents completed an online questionnaire during school hours. Research assistants were present in the class to answer possible questions from students. Adolescents were required to create a personal code to ensure confidentiality and pair their answers over time.
Measures
Ethnic prejudice
Ethnic prejudice was assessed using the Feeling thermometer (Haddock et al., 1993; for the Italian version, see Bobba & Crocetti, 2022), a scale that has been extensively used to examine the affective component of prejudice among adults (for a review, see Dovidio et al., 2001) and adolescents (for a review, see Crocetti et al., 2021). In this version of the measure, participants are asked to rate how much they like the five most represented ethnic minorities in the Italian context (i.e. Romanians, Albanians, Moroccans, Chinese and Ukrainians; ISTAT, 2020) on a sliding scale from 0° (not at all) to 100° (very much). The scale was reversed to simplify the interpretation of results, with higher scores indicating higher prejudice against each of the ethnic minority groups.
National Identification
Identification with the Italian national group was assessed with a shortened version of the Group Identification Scale (Thomas et al., 2017). This shortened version included three items (i.e. ‘I have a lot in common with the other members of the Italians' group’; ‘Being a member of the group of Italians is important to who I am’; and ‘I identify with the Italians' group’), which adolescents rated on a 5-point Likert type scale from 1 (completely false) to 5 (completely true). Cronbach's alphas were .74, .76, .82 and .84 at T1, T2, T3 and T4 respectively.
Human identification
Adolescents' identification with the human group was assessed with the Human Identification Scale (Albarello & Rubini, 2012). This scale included four items (i.e. ‘I identify with all human beings’; ‘I feel strong bonds with human beings of any social group’; ‘I myself am equal to all human beings, no matter ethnic, politic, religious, social, or ideologic differences’; and ‘I feel proud of being part of mankind’) which adolescents rated on a 5-point Likert type scale from 1 (completely false) to 5 (completely true). Cronbach's Alphas were .78, .78, .84 and .83 at T1, T2, T3 and T4 respectively.
RESULTS
Preliminary analyses
Descriptive analyses were performed in IBM SPSS Version 28.0 for Windows, while the remaining analyses (i.e. measurement invariance and multilevel models) were conducted in Mplus version 8.10 (Muthén & Muthén, 2017) using the maximum-likelihood (ML) estimator (Finch & Bolin, 2017). Means and standard deviations of the study variables are reported in Table S1, while correlations are reported in Table S2 (see Supporting Information). Rank-order stability levels were high for ethnic prejudices (.47 < r < .77), Italian identification (.54 < r < .58) and human identification (.55 < r < .59). Overall, national identification correlated positively and human identification correlated negatively with measure of ethnic prejudices. As a preliminary check, longitudinal measurement invariance of Italian and human identification scales was tested. Results are reported in Table S3 of the Supporting Information. Both Italian and human identification scales reached partial scalar invariance, therefore we could proceed with the main analyses.
Furthermore, the intraclass correlation coefficients (ICCs) of national and human identifications were examined by running an unconditional multilevel model in Mplus. Results indicated that slightly more than half of the variance in identification with the Italian (51.30%) and the human (52.40%) groups was at the between level, while the remaining (48.70% for national and 47.60% for human) was at the within-person level. This means that separating the stable (between person) from the fluctuating (within person) components was necessary to capture the interplay of social identity and ethnic prejudice. It is important to note, however, that part of the within-person variance could also be attributed to measurement error.
Multilevel analyses
Multilevel modelling was used to examine the associations between national and human identifications and prejudice against several ethnic minority groups at both the within-person (Level 1) and between-person (Level 2) levels. Specifically, Level 1 represents the associations between within-person (or overtime) fluctuations (i.e. deviations from an individual stable mean) in ethnic prejudice and within-person (overtime) fluctuations in both Italian and human identifications. Conversely, Level 2 examines whether adolescents' stable levels of identification with the national and human groups would be associated with stable overtime mean and change (i.e. linear slope) in levels of ethnic prejudice against the Romanian, Albanian, Moroccan, Chinese and Ukrainian groups. Group-mean centring was used for the predictors included at the within-person level, while grand-mean-centred cluster means of the Level 1 predictor were included at the between-person level (Enders, 2013a).
The final multilevel model was built through multiple steps. First, the fit of each model was evaluated relying on a combination of low deviance (−2LL) scores and small AIC and BIC values as indicative of good fit. Next, nested models were compared against each other, and a significant likelihood-ratio test (Δ−2LL) indicated a significant improvement from the simpler to the more complex model (Finch & Bolin, 2017; Hox et al., 2018). Last, once the final fully constrained model with predictors at both levels was established, the log-likelihood ratio was used to understand whether the associations between each identification and prejudice were significantly different depending on the ethnic minority group examined, at both the within- and between-person (i.e. intercepts and slopes) levels. The most parsimonious and best-fitting model was retained. In all models, the residuals for the ethnic prejudice scores were allowed to correlate at both Level 1 and Level 2. Model fit indices are reported in Table 1, while unstandardized regression coefficients of the multilevel analyses are reported in Table 2. A schematic representation of results is provided in Figure 2.
| Model | Model fit | Residual variance | Prejudice vs. | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| LL (df) | AIC | BIC | Δ−2LL | Romanians | Albanians | Moroccans | Chinese | Ukrainians | ||
| Model 1 Unconditional | −64,916.674 (35) | 129,903.349 | 130,112.876 | Level 1 | 418.946*** | 393.272*** | 408.029*** | 416.522*** | 511.868*** | |
| Level 2 | 650.495*** | 685.755*** | 762.299*** | 752.961*** | 593.853*** | |||||
| Model 2 Fixed linear slope | −64,893.807 (40) | 129,867.615 | 130,107.075 | 45.734 (5)*** | Level 1 | 418.643*** | 393.210*** | 406.697*** | 416.515*** | 508.375*** |
| Level 2 | 650.219*** | 685.755*** | 763.085*** | 752.891*** | 593.644*** | |||||
| Model 3 Random linear slope | −64,839.344 (50) | 129,778.688 | 130,078.014 | 108.926 (10)*** | Level 1 | 398.591*** | 383.180*** | 377.124*** | 392.442*** | 507.779*** |
| Level 2 | 720.708*** | 751.738*** | 841.179*** | 790.957*** | 592.470*** | |||||
| Model 4 Fully constrained predictors at Levels 1 and 2 | −64,051.216 (56) | 128,214.432 | 128,549.102 | 1576.256 (6)*** | Level 1 | 386.085*** | 376.999*** | 369.267*** | 384.316*** | 490.435*** |
| Level 2 | 618.444*** | 637.549*** | 650.647*** | 714.105*** | 506.496*** | |||||
| Model 5: Partially constraineda predictors at Levels 1 and 2 | −64,017.039 (60) | 128,154.079 | 128,512.654 | 68.354 (4)*** | Level 1 | 386.302*** | 376.892*** | 369.656*** | 384.602*** | 492.394*** |
| Level 2 | 615.135*** | 638.387*** | 639.993*** | 712.130*** | 503.459*** | |||||
- Abbreviations: AIC, Akaike information criterion; BIC, Bayesian information criterion; df, degree of freedom; LL, log likelihood; Δ, change in fit indices.
- *** p < .001.
- a In this model, the following paths were unconstrained: (a) the Level 2 association between Italian identification and overtime mean in prejudice against Moroccans; (b) the Level 2 association between Italian identification and overtime mean in prejudice against Chinese; (c) the Level 2 associations between human identification and overtime mean in prejudice against Moroccans; and (d) the Level 2 association between human identification and overtime change in prejudice against Chinese.
| Outcomes: Prejudice vs. B [95% CI] | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Romanians | Albanians | Moroccans | Chinese | Ukrainians | ||||||
| Level 1 | Level 2 | Level 1 | Level 2 | Level 1 | Level 2 | Level 1 | Level 2 | Level 1 | Level 2 | |
| Model 2: Fixed linear slope | ||||||||||
| Time | −0.43 [−1.13, 0.27] | – | −0.01 [−0.69, 0.67] | – | 0.86* [0.17, 1.56] | – | 0.01 [−0.69, 0.72] | – | −1.59*** [−2.36, −0.82] | – |
| Model 3: Random linear slope | ||||||||||
| – | – | – | – | – | – | – | – | – | – | |
| Model 4: Fully constrained predictors at Level 1 and Level 2 | ||||||||||
| Italian Identification (w) | −1.40* [−2.79, −0.01] | – | −1.40* [−2.79, −0.01] | – | −1.40* [−2.79, −0.01] | – | −1.40* [−2.79, −0.01] | – | −1.40* [−2.79, −0.01] | – |
| Human Identification (w) | −3.04*** [−4.43, −1.64] | – | −3.04*** [−4.43, −1.64] | – | −3.04*** [−4.43, −1.64] | – | −3.04*** [−4.43, −1.64] | – | −3.04*** [−4.43, −1.64] | – |
| Italian Identification (b) → Average | – | 11.10*** [8.18, 14.02] | – | 11.10*** [8.18, 14.02] | – | 11.10*** [8.18, 14.02] | – | 11.10*** [8.18, 14.02] | – | 11.10*** [8.18, 14.02] |
| Italian Identification (b) → Slope | – | −0.28 [−1.28, 0.72] | – | −0.28 [−1.28, 0.72] | – | −0.28 [−1.28, 0.72] | – | −0.28 [−1.28, 0.72] | – | −0.28 [−1.28, 0.72] |
| Human Identification (b) → Average | – | −14.97*** [−17.80, −12.15] | – | −14.97*** [−17.80, −12.15] | – | −14.97*** [−17.80, −12.15] | – | −14.97*** [−17.80, −12.15] | – | −14.97*** [−17.80, −12.15] |
| Human Identification (b) → Slope | – | −1.26* [−2.23, −0.29] | – | −1.26* [−2.23, −0.29] | – | −1.26* [−2.23, −0.29] | – | −1.26* [−2.23, −0.29] | – | −1.26* [−2.23, −0.29] |
| Model 5: Partially constrained predictors at Level 1 and Level 2 | ||||||||||
| Italian Identification (w) | −1.40* [−2.79, −0.06] | – | −1.40* [−2.79, −0.06] | – | −1.40* [−2.79, −0.06] | – | −1.40* [−2.79, −0.06] | – | −1.40* [−2.79, −0.06] | – |
| Human Identification (w) | −3.01*** [−4.4, −1.62] | – | −3.01*** [−4.41, −1.62] | – | −3.01*** [−4.41, −1.62] | – | −3.01*** [−4.41, −1.62] | – | −3.01*** [−4.41, −1.62] | – |
| Italian Identification (b) → Average | – | 9.69*** [6.75, 12.63] | – | 9.69*** [6.75, 12.63] | – | 14.58*** [11.36, 17.79] | – | 11.95*** [8.60, 15.30] | – | 9.69*** [6.75, 12.63] |
| Italian Identification (b) → Slope | – | −0.32 [−1.31, 0.68] | – | −0.32 [−1.31, 0.68] | – | −0.32 [−1.31, 0.68] | – | −0.32 [−1.31, 0.68] | – | −0.32 [−1.31, 0.68] |
| Human Identification (b) → Average | – | −13.60*** [−16.40, −10.80] | – | −13.60*** [−16.40, −10.80] | – | −19.02*** [−22.12, −15.92] | – | −13.60*** [−16.40, −10.80] | – | −13.60*** [−16.40, −10.80] |
| Human Identification (b) → Slope | – | −1.59** [−2.57, −0.60] | – | −1.59** [−2.57, −0.60] | – | −1.59** [−2.57, −0.60] | – | −0.22 [−1.36, 0.92] | – | −1.59** [−2.57, −0.60] |
- Note: B = Unstandardized regression coefficients; (w) = predictors entered at the within-person level; (b) = predictors entered at the between-person level.
- Abbreviation: CI, confidence interval.
- * p < .05.
- ** p < .01.
- *** p < .001.
Intraclass correlation
As a preliminary step, a null model was specified that partitioned the variances of the (correlated) dependent variables in their within-person (Level 1) and between-person (Level 2) components and allowed the calculation of the intra-class correlations (ICC). Results of this model (Model 1) indicated that between half and two-third of the variance in ethnic prejudice measures was at the between-person level (ranging from 53.7% for Ukrainians to 65.1% for Moroccans). Thus, a substantial portion of variance (ranging from 34.9% for prejudice against Moroccans to 46.3% for prejudice against Ukrainians) was at the within-person level (although it included measurement error), indicating that time-specific fluctuations matter and multilevel analyses are warranted to examine the correlates of ethnic prejudice at both levels.
Stability and change in ethnic prejudices
To examine within-person fluctuations, as well as overtime means and changes at the between-person level in ethnic prejudice, the effect of time was first added as a predictor of ethnic prejudices at Level 1 (Model 2). As shown in Table 1, adding linear slopes for time significantly improved model fit. However, as can be inferred from Table 2, time was not a significant predictor of ethnic prejudice against the Romanian, the Albanian and the Chinese groups. Conversely, while ethnic prejudice against Moroccans displayed an increase over time, ethnic prejudice against Ukrainians showed a significant linear decrease.
In the next step, the linear slopes for time were randomized at the between-person level, meaning they were freely estimated for all participants to acknowledge the possibility that individuals differ in their rate of change in prejudice (Model 3). This resulted in improved model fit and decreased residual variance of ethnic prejudices at both Level 1 and Level 2. Estimates of the overtime means and slopes in ethnic prejudices at Level 2 are reported in Table 3. Regarding overtime means, adolescents displayed prejudice levels that were below the scale mid-point for all the five ethnic minority groups considered. This result suggests that overall evaluations of ethnic minorities were positive. Additionally, the variance of participants' average prejudice scores was high and significant, highlighting substantial heterogeneity in how youth think of diverse others. As can be inferred, most slopes displayed a significant variance, indicating that adolescents differed from each other in the linear rate of change in ethnic prejudice against the Romanian, Moroccan and Chinese groups. Conversely, no variation emerged for the linear slope of ethnic prejudice against the Albanian and Ukrainian groups. In other words, adolescents in the sample did not display a significant change in prejudice against the Albanian group, whereas they reported significant decreases in ethnic prejudice against Ukrainians.
| Prejudice vs. | Level 2 parameter estimates | |||
|---|---|---|---|---|
| Average | Slope | |||
| M (SE) | σ2 (SE) | M (SE) | σ2 (SE) | |
| Romanians | 43.22 (1.09)*** | 720.71 (40.95)*** | −0.43 (0.37) | 10.89 (3.29)** |
| Albanians | 40.23 (1.10)*** | 751.74 (42.75)*** | 0.00 (0.35) | 5.20 (2.99) |
| Moroccans | 43.85 (1.15)*** | 841.18 (46.96)*** | 0.84 (0.37)* | 16.57 (3.77)*** |
| Chinese | 40.91 (1.13)*** | 790.96 (46.05)*** | 0.00 (0.38) | 14.70 (3.79)*** |
| Ukrainians | 41.87 (1.07)*** | 592.47 (39.15)*** | −1.60 (0.39)*** | 0.16 (4.34) |
- Note: M = Level 2 means; SE = Level 2 Standard Error; σ2 = Level 2 variance.
- * p < .05.
- ** p < .01.
- *** p < .001.
The role of Italian and human identifications
In the following model, fluctuations in Italian and human identifications were entered as predictors of fluctuations in ethnic prejudice at Level 1, and overtime means of both identifications were entered as predictors of overtime means and changes (i.e. slopes) in ethnic prejudice at Level 2 (Model 4). In this model, the path from each identification to the prejudice scores was constrained to equality across the five ethnic minority groups, separately for each predictor at each level (e.g. the within-person effect of Italian identification on prejudice was fixed to be equal across the five minorities). This fully constrained model with predictors at Levels 1 and 2 resulted in a significant improvement in model fit and a decrease in the residual variances of ethnic prejudices at both levels.
Next, to tackle the third goal of the current study, constrained paths in Model 4 were freed one by one to examine whether the effects of national and human identifications significantly differed depending on the ethnic group examined. To this end, a stepwise procedure was employed comparing different models against each other to identify the best-fitting and most parsimonious solution to represent the current data. The full procedure is detailed in the Supporting Information (see Tables S4 and S5). As can be inferred, a model with only a few unconstrained paths (Model 5) provided the most parsimonious and best-fitting representation of the data. Across the multiple steps procedure, the following between-person level paths were found to significantly differ from the others and were therefore freed: (a) from Italian identification to overtime means of prejudice against Moroccans; (b) from Italian identification to overtime means of prejudice against Chinese; (c) from human identification to overtime means of prejudice against Moroccans; and (d) from human identification to overtime change in prejudice against Chinese. The results of this final model are reported in Table 2 and outlined in Figure 2.
At the within-person level, fluctuations in identification with the Italian group and fluctuations in identification with the human group were both negatively associated with fluctuations in ethnic prejudices. This means that within-person increases in identification with these two groups were associated with within-person decreases in ethnic prejudices. Furthermore, these associations were identical across all the five ethnic minority groups.
At the between-person level, the stable means of Italian and human identifications were included as predictors of both the means and slopes of adolescents' ethnic prejudices. Higher scores on Italian and human identifications were significantly associated with, respectively, higher and lower mean levels of ethnic prejudice. Further inspection of the similarity in regression coefficients across ethnic groups revealed that the strength of the effect for Italian and human identification depended on the target group. Specifically, Italian identification was more strongly linked to higher mean levels of prejudice against the Chinese, and even more so against the Moroccan groups, compared to the other Eastern European minorities (i.e. Romanian, Albanian and Ukrainian). This finding was partially replicated for human identification, which had a stronger effect in reducing mean levels of ethnic prejudice against Moroccans compared to the other groups.
Regarding associations between identifications and slopes of prejudice, Italian identification was not significantly associated with changes in ethnic prejudice against any of the minority groups considered. Conversely, higher levels of human identification were found to contribute to significantly steeper decreases in ethnic prejudice against the Romanian, Albanians and Ukrainian groups, and to less steep increases in prejudice against the Moroccan group. No significant association emerged between stable levels of human identification and changes in prejudice against the Chinese group.
DISCUSSION
From its early beginnings, the psychological study of prejudice has attempted to unravel whether this social phenomenon depends on stable personal characteristics, or rather fluctuates and changes under specific situational conditions (Allport, 1954). The current longitudinal study aimed to contribute to this debate by examining how Italian adolescents' national and human identifications were related to their ethnic prejudice, both at the between- and within-person levels. Additionally, it took an outgroup-specific approach to unravel whether the identification–prejudice link would vary depending on the ethnic minority group considered. Our findings highlighted significant associations between the stable and fluctuating components of national and human identifications and prejudice against multiple ethnic groups, although the strength and direction of these links varied considerably at the between- and within-person levels.
Stable or fluctuating national identification: the distinction matters
The first goal of the current study was to examine whether and how stable differences and temporal changes in levels of national identification are associated, respectively, with the stable and fluctuating components of affective prejudice. According to the social identity approach, strong group identifiers are more likely to make ingroup–outgroup distinctions, which could translate to outgroup negativity depending on how the ingroup is defined (Reicher et al., 2011). As Italian adolescents seem to endorse a cultural definition of nationhood (Reijerse et al., 2015), and because such a definition implies a limited inclusion of newcomers (Pehrson, Vignoles, & Brown, 2009), we anticipated a positive relation between their national identification and prejudice towards ethnic minorities at the between-person level, while within-person associations were examined from an exploratory point of view. Overall, the strength and direction of these associations differed at the between- and within-person levels.
At the between-person level, mean levels of identification with the Italian majority were indeed linked to higher mean levels of prejudice against all ethnic minorities considered. In line with the social identity approach (Brown, 2020), this finding suggests that adolescents who usually rely more on ingroup–outgroup distinctions tend to approach the social world in dichotomous terms that are conducive to more ethnic prejudice and less inclusivity. More specifically, their stable levels of identification might be indicative of their general readiness to distinguish Italians from non-Italians, resulting in less positive evaluations of the latter (Reicher et al., 2011).
Conversely, at the within-person level, fluctuations in national identification were negatively and significantly associated with ethnic prejudice and this effect was equally strong across the five ethnic minority groups. In other words, when youth displayed a momentary increase in the salience of and attachment to their national group, they also reported lower levels of affective prejudice against ethnic minorities. A possible explanation for this finding is that fluctuations (increases) in identification with the national group might represent moments of strengthened commitment through exploration (Bosma & Kunnen, 2001; Luyckx et al., 2006), which provide youth with clarity and security over their sense of self (Becht et al., 2017; Schwartz et al., 2011). In other words, at a given moment, youth might increase their sense of identification with their national group by actively seeking information about and reflecting on the meaning of their group membership. Such exploration-based ingroup identification, accompanied by feelings of self-certainty and security, has been previously linked to more positive intergroup attitudes and behaviours (Phinney et al., 2007; Spiegler et al., 2022; Whitehead et al., 2009). Relatedly, and in line with these studies, our findings provide further insight into the nuanced role of ingroup identification for intergroup outcomes. These conclusions align with the idea that a secure sense of ingroup identification represents the psychological basis for reducing ethnic prejudice levels and fostering open and positive attitudes towards the social world (Allport, 1954). Further research is needed to understand the mechanisms through which these associations occur. However, for now, it is important to note that the negative within-person association between national identification and prejudice obtained in our study is not at odds with the social identity approach. Adolescents' national identity was still important for their intergroup relations but not in the exclusive way that was characteristic for the between-person level.
The protective role of common group identities: the case of human identification
The second goal of the current study was to investigate the associations between stable and fluctuating components in levels of identification with the common human ingroup and prejudice against multiple ethnic minorities. Given the inclusive nature of this identity, both its stable and fluctuating components were expected to be linked to lower levels of and momentary decreases in prejudice against ethnic minorities. Our findings supported the protective role of human identification at both levels. Specifically, when youth rely on a superordinate level of categorization, they are also more prone to overcoming differences, transcending dichotomous views of society and endorsing more inclusive attitudes towards others (McFarland et al., 2019).
At the between-person level, stable means of identification with the group of humanity were linked to lower stable levels and slopes of ethnic prejudice, leading to a general reduction in negative attitudes towards the minority groups considered. Youth who generally adopt this superordinate level of categorization are also less prejudiced and display significant reductions in their negative feelings against diverse others. These findings align with prior research on the protective role of global human identification for reducing prejudice and supporting inclusiveness (McFarland et al., 2012) and fostering intergroup helping and prosocial behaviour (Hamer et al., 2017; Sparkman & Hamer, 2020).
Furthermore, not only stable levels but also temporal fluctuations in identification with the human group matter for momentary reductions in prejudice. Specifically, and in line with the common ingroup identity model (Gaertner et al., 1993; Gaertner & Dovidio, 2000), the situational (increased) activation of the superordinate human categorization level is accompanied by significant temporal decreases in affective prejudice against ethnic minorities. This evidence is in line with prior experimental works highlighting the effectiveness of priming human identity for reducing prejudice (Albarello & Rubini, 2012; Wohl & Branscombe, 2005). That is, whenever individual identification with humanity increases in salience and importance, a reframing of ingroup–outgroup boundaries occurs and feelings, attitudes and behaviours align with a more inclusive view of self and others, thus improving the quality of intergroup relationships.
Minority group matters: common and differential effects across ethnic targets
The current study sought to provide a nuanced understanding of the stable and situational antecedents of prejudice by examining outgroup-specific levels of prejudice against the most salient ethnic minorities (i.e. Romanians, Albanians, Moroccans, Chinese and Ukrainians) in the Italian context. Specifically, it tackled patterns of stability and fluctuations in prejudice across these groups and tested whether associations between national and human identifications and affective prejudice differed depending on the target group considered. Our findings highlighted a few similarities and some differences in average levels, stability and change patterns, as well as differential effects of ingroup identifications depending on the target group.
Regarding similarities, adolescents displayed generally low levels of ethnic prejudice against all the five ethnic minorities examined. This is consistent with prior research on adolescents' prejudice conducted across different contexts (Bagci & Gungor, 2019; Boer & van Tubergen, 2019; Taylor & McKeown, 2021; Vezzali et al., 2020; Wölfer et al., 2016) and suggests that, on average, youth have favourable attitudes towards diverse others. Thus, especially the younger generations appear to be accepting of diversity and willing to accommodate it (Harris et al., 2023). However, significant variability also emerged in ethnic prejudice average levels, highlighting that not all adolescents approach diversity in the same way. More research is needed to identify subgroups of youth at risk of developing ethnic prejudice and to understand the individual and contextual conditions that can hinder their positive adjustment to multicultural societies (Bobba et al., 2023).
Regarding differences in patterns of stability and change, overtime changes in levels of affective ethnic prejudice emerged for two ethnic minority groups (i.e. Moroccans and Ukrainians) and showed opposite trends. Ethnic prejudice against the Moroccan group increased significantly. This might be a consequence of the concomitant increase in migration flows originating from Africa to the Italian coasts during the year of data collection, which made this topic an intensively debated issue in both the political campaign for the 2022 national elections and in the media. Specifically, this substantial growth in number of migrants arriving in Italy between January 2022 and 2023 involved mostly individuals coming from Africa (Ministero dell'Interno, 2023). The Moroccan group, although not directly involved in the migration flows, might have been regarded as representative of the African minority, thus contributing to increased negative feelings against this ethnic group. In contrast, ethnic prejudice levels against the Ukrainian group displayed a significant decrease over the course of the data collection. This finding is in line with prior research highlighting that socio-contextual events (such as the ongoing Russia–Ukraine war; Bobba et al., 2024), and how they are recounted in the media, can contribute to shifting emotions and attitudes towards ethnic minorities (Finseraas & Listhaug, 2013; Mitchell, 2019).
Regarding the associations between prejudices and identifications, consistent differences emerged at the between-person level of analysis. Specifically, stable levels of both national and human identifications were more strongly associated with (respectively, higher and lower) stable levels of prejudice against the Moroccan and Chinese (only for Italian identification) groups compared to the other minorities. This finding can be interpreted in light of physical, cultural and historical differences between the five ethnic groups considered in this study that might heighten the salience of self and other group membership. Compared to the Eastern European (i.e. Romanian, Albanian and Ukrainian) groups, the Moroccan minority substantially differs from the Italian group, both in terms of appearance and cultural and religious backgrounds. These features can contribute to enhancing perceived differences among groups (i.e. comparative fit) and therefore drive processes of marginalization of Muslim minorities (Perocco, 2018). Relatedly, compared to both the European and the Chinese minorities, the Moroccan group is quite often the target of suspicion, distrust and hostility (Kunst et al., 2012; Rizzo et al., 2020) and is perceived as culturally incompatible with the ethnic majority (Cicognani et al., 2018). Furthermore, media depictions (Cervi et al., 2021) and political discourses (Cervi, 2020) might convey and even exacerbate representations of the Moroccan minority as a highly salient and distinct (out)group compared to one's ingroup. Consequently, when youth strongly identify with the Italian group, such ingroup–outgroup demarcation can foster threat perceptions and heighten the levels of prejudice against the Moroccan group minority. On the contrary, highly marked distinctions between ingroup (i.e. Italian) and outgroup (i.e. Moroccan) members coupled with strong identification with the superordinate human group can be at the basis of social identity complexity (Roccas & Brewer, 2002). In other words, youth might still perceive themselves and others as members of two distinct groups, thus avoiding a ‘colorblind’ approach that neglects existing differences, within the context of a strong superordinate identity (Gaertner & Dovidio, 2012). Along this line, prior experimental research found that maintaining the salience of subgroups within an equally salient superordinate group led to substantial decreases in intergroup bias (Crisp et al., 2006). Similarly, in this study, the presence of salient and clearly marked ingroup–outgroup distinctions, such as the one between Italian and Moroccan groups, which are although comprehended within an overarching group – that of humanity – can still lead individuals to display generally lower prejudice against members of this ethnic minority.
Conversely, the relations between adolescents' fluctuations in Italian and human identifications on one side, and ethnic prejudice on the other, were consistent regardless of the target group. Thus, when their national and human identifications are temporarily important to them, this has common positive implications for different outgroups. The finding for national identification is also in line with the aforementioned interpretation in terms of identity exploration and (re)evaluation. Presumably, the effects of fluctuations indicate a heightened focus on the meaning of the national group vis-à-vis other groups in general, rather than its specific differences with a particular outgroup.
Theoretical and practical implications
The current study has important theoretical and practical implications. From a theoretical perspective, it advances the person versus situation debate by offering a comprehensive understanding of the associations between ingroup identification and ethnic prejudice. Specifically, by adopting a multilevel longitudinal methodology and separating the stable and fluctuating components of prejudice and its social identity antecedents, this research highlighted the role of national and human identifications at different levels. Regarding the former, youth who highly identify with the national group tend to report more negative attitudes towards ethnic minorities, while momentary increases in levels of identification are associated with steeper decreases in prejudice. Regarding the latter, relying on the superordinate category of humanity, both as a stable tendency and as a result of situational increases in salience of this group membership, appears to favour more positive feelings and behaviours in intergroup contexts.
These findings have potentially important implications for future interventions. Specifically, it seems that strengthening people's national identity can be an effective strategy for reducing ethnic prejudice, even though high national identifiers are typically more prejudiced than low national identifiers. Clearly, more research is needed to confirm this recommendation, but the results of this study suggest that stimulating people to explore and reflect upon the meanings and positive implications of this identity might decrease their prejudice against ethnic others (Spiegler et al., 2022). This intervention strategy might be especially appropriate for adolescents, who are in the process of consolidating their stable levels of identification with relevant social groups and forming coherent views of themselves and others (Crone & Fuligni, 2020; Svensson & Syed, 2023). Overall, interventions aimed at improving the quality of intergroup relationships should also strive to account for the situational conditions that can support adolescents in forming positive attitudes and endorse inclusive views of current multicultural societies (Beelmann & Lutterbach, 2021).
Limitations and suggestions for future research
Findings from the current study should be read considering some limitations. First, the current study focused exclusively on the affective component of ethnic prejudice while less is known about negative stereotypes attributed to different ethnic minorities and how they are influenced by self- and other-categorization processes. In light of the multidimensional nature of ethnic prejudice (Brown, 2011; Crocetti et al., 2021), future research should strive to assess multiple facets of this phenomenon to understand the factors underpinning affects, cognitions and behaviours against ethnic minorities. While affects might be more susceptible to socio-contextual changes, negative stereotypes and beliefs might be more enduring and therefore display lower fluctuations and be less affected by momentary changes in levels of identification with the ingroup. Furthermore, affective and cognitive components of ethnic prejudice guide discriminatory and aggressive behaviours against diverse others (Dovidio et al., 2010). Therefore, future research could further examine how processes of identification are intertwined with ingroup bias, outgroup derogation and negative intergroup behaviours (Greenwald & Pettigrew, 2014; Hodson, 2021). Additionally, the current study relied on the Feeling Thermometer Scale, which is formulated in terms of liking (or positive attitudes) more than disliking (or negative attitudes). Although the latter certainly implies the former and this instrument has been extensively adopted to evaluate (affective) prejudice among youth across diverse contexts (Bratt et al., 2016; Vezzali et al., 2020; Weber, 2019), future studies could examine whether the current results are replicated using other assessment methods.
Second, although this study focused on the role of time-specific contextual variation, it was conducted in one particular setting (i.e. the Italian context) and examined prejudice against the five most represented ethnic outgroups. More specifically, this research was conducted in the Emilia-Romagna region, an area characterized by the highest percentage of ethnic minority population in the Italian school context (Ministero della Pubblica Istruzione, 2022). This increased opportunity for intergroup contact with diverse peers at school (Karataş et al., 2023) might have influenced how youth deal with self- and other-oriented processes of categorization and ultimately their levels of prejudice. These aspects should be considered in the generalization of current findings to other contexts with different levels of ethnic diversity.
Third, although this study relied on longitudinal data collected from a large sample of youth, which allowed to distinguish between stable or trait-like levels and momentary fluctuations in both identification and prejudice, the current multilevel analyses were not suitable for establishing causality. Testing the identification–prejudice link within the same temporal situation, in line with the premises of the social identity approach, removes the temporal precedence of one variable (i.e. social identifications) over the other (i.e. prejudice) and thus cannot provide any causal evidence of their association. Future experimental studies can complement findings from the current study by testing the causal dynamics of these psychological processes. Furthermore, the present research did not disentangle the direction of influence between national and human identifications and ethnic prejudice to examine if the former predicts a relative change in the latter or vice versa. Future longitudinal studies complemented by advanced analytic approaches, such as the random intercept cross-lagged panel model (Hamaker et al., 2015), could address this novel question and provide additional insight into the identification–prejudice link.
Next, it should be noted that a significant portion of variance at the within-person level and, even more so, at the between-person level still remains unexplained. This means that additional factors and conditions could contribute to stable levels and overtime changes in ethnic prejudice. Future research should strive to address this gap by assessing other stable individual characteristics (e.g. personality traits and social dominance orientation; Albarello et al., 2020; Crocetti et al., 2021) and how they contribute to the consolidation of attitudes towards diverse others. Moreover, the present findings indicate that fluctuations in group identifications can be relevant for prejudice, but it is important to examine where those fluctuations themselves stem from.
Last, this research examined between- and within-person associations between national and human identifications and ethnic prejudice among a sample of adolescents. Current findings highlighted the importance of separating stability and fluctuations to gain a more nuanced understanding of these social phenomena, in line with increasing attention to these differences in multiple research fields (Perinelli et al., 2023; Zuffianò et al., 2023). Future studies should apply a similar approach to adult samples in order to understand whether results are replicated across age groups or are rather dependent on the developmental phase taken into account. Although social and political attitudes progressively stabilize in adulthood (Rekker et al., 2015), momentary fluctuations in levels of prejudice can still occur as a consequence of events and changes in the macro-context (e.g. terrorist attacks; Legewie, 2013). More research is needed to unravel whether group identification also contributes to fluctuations in adults' attitudes towards diversity.
CONCLUSION
The associations between prejudice and group identification have been extensively examined from the social identity approach. However, prior studies have neglected to account for the stable and fluctuating components of both social phenomena. By adopting a longitudinal multilevel design, the current research aimed to understand whether and how stable levels of and fluctuations in identification with the national and superordinate human groups were associated, respectively, with stable levels of and temporal changes in ethnic prejudice, and whether these associations varied depending on the ethnic target group. Regarding national identification, stable levels were linked to higher average levels of affective prejudice, while fluctuations were negatively associated with fluctuations in prejudice. Conversely, both stable levels of and fluctuations in human identification were related to lower average levels of and steeper decreases (or less steep increases) in ethnic prejudice. Associations across temporal fluctuations remained the same regardless of the ethnic minority target, whereas the link between stable levels of (national and human) identifications and prejudice was found to be stronger for the Moroccan group compared to all the others. These findings highlight the importance of examining associations between prejudice and social identity processes in more complex ways to gather a nuanced understanding of these phenomena as part of the person–context debate.
AUTHOR CONTRIBUTIONS
Beatrice Bobba: Conceptualization; investigation; writing – original draft; data curation; formal analysis; methodology. Jochem Thijs: Conceptualization; methodology; writing – original draft; supervision. Elisabetta Crocetti: Conceptualization; funding acquisition; writing – original draft; supervision; methodology; resources.
FUNDING INFORMATION
This work was supported by a grant from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation programme (ERC-CoG IDENTITIES grant agreement N. [101002163]; principal investigator: Elisabetta Crocetti).
CONFLICT OF INTEREST STATEMENT
The authors report no conflict of interests.
Open Research
DATA AVAILABILITY STATEMENT
Data, analyses codes and outputs supporting the results of this research can be retrieved from the Open Science Framework (OSF) page of the project at the following link: https://osf.io/9gh8d/.
