The socialization of perceived discrimination in ethnic minority groups.

Contact and Perceived Discrimination

To paint a complete picture of how social relationships influence disadvantaged-group members’ perceptions of discrimination, we must consider contact with members of both the disadvantaged ingroup and the advantaged outgroup. Contact can take many forms, ranging from short-term interactions to long-term relationships such as friendships ( MacInnis & Page-Gould, 2015 ), and can be positive, negative, or relatively neutral in valence ( Schäfer et al., 2021 ). We focus specifically on direct, qualitatively positive contact and friendship, as these forms of contact have been found most consistently associated with perceived discrimination ( Reimer & Sengupta, 2023 ).

Guided by intergroup contact theory ( Allport, 1954 ; Brown & Hewstone, 2005 ), most research to date has focused on contact with outgroup members and its well-documented promise for improving outgroup attitudes ( Pettigrew & Tropp, 2006 ; Tropp & Pettigrew, 2005 ). While outgroup contact may improve outgroup attitudes, outgroup contact is also associated with other constructs including reduced ingroup identification, collective action, and support for social change in disadvantaged groups ( Cakal et al., 2011 ; Hässler et al., 2020 ; Tausch et al., 2015 ). Outgroup contact is accordingly considered to have both positive and negative consequences for various aspects of intergroup relations.

With regard to perceived discrimination, Dixon et al. (2010) proposed that, if disadvantaged-group members’ interactions with advantaged-group members were generally positive, outgroup contact experiences would reduce perceived personal discrimination and, in turn, perceived group discrimination. Additionally, if outgroup contact improved attitudes toward the advantaged, it could render discrimination less plausible. Positive attitudes toward the advantaged outgroup may therefore be linked to perceiving less discrimination from the advantaged outgroup. Consistent with this expectation, meta-analytic evidence indicates an association, albeit weak, between positive outgroup attitudes and lower perceived discrimination among members of disadvantaged groups ( r = −.10, [−.02, −.18]; Reimer & Sengupta, 2023 ). Meta-analytic evidence also indicates an association between perceived discrimination and collective action ( r = .31, [.23, .38]; Reimer & Sengupta, 2023 ; r = .35, [.30, .39]; van Zomeren et al., 2008 ), such that disadvantaged-group members who perceive greater discrimination are more likely to engage in collective action. Meta-analytic evidence thus suggests that perceived discrimination, outgroup attitudes, and collective action are related, but distinct, constructs.

Prior research has tested the hypothesis that outgroup contact reduces perceived discrimination in disadvantaged groups. A recent meta-analysis found that, across 203,637 participants in 86 studies, contact with the advantaged was, on average, associated with less perceived discrimination among the disadvantaged ( Reimer & Sengupta, 2023 ). This association, however, was small (between −.10 < r < −.04) and variable (31% of studies found a positive association). Furthermore, the available evidence consisted predominantly of cross-sectional survey studies (e.g., Dixon et al., 2010 ), which cannot establish the direction of the association: Outgroup contact could reduce perceived discrimination, or perceiving greater discrimination could lead disadvantaged-group members to engage in less outgroup contact ( Tropp et al., 2012 ). Longitudinal, experimental, and intervention studies can test the direction of the association. Yet, the available evidence includes few longitudinal studies ( Koschate et al., 2012 ; Reimer et al., 2017 ; Tropp et al., 2012 ), all of which conflated within-person change and between-person stability ( Hamaker et al., 2015 ), and only one intervention study ( Reimer et al., 2021 ) and one experimental study ( Saguy et al., 2009 ), both of which omitted a control group. Of these longitudinal, experimental, and intervention studies, only two ( Saguy et al., 2009 ; Tropp et al., 2012 ) found that outgroup contact predicted less perceived discrimination. Therefore, the available evidence generally suggests outgroup contact is associated with less perceived discrimination in disadvantaged groups, but evidence for causality is scarce.

Fewer studies have examined whether ingroup contact affects perceived discrimination. Levin et al. (2006) proposed that, for members of disadvantaged groups, ingroup peers provide an important reference group and that, to the extent that perceiving discrimination is normative within the group, ingroup contact can reinforce perceptions of discrimination. Consistent with this argument, the researchers found that, among Asian, Latino, and African American students, having more ingroup friends in their first year of college predicted perceiving more discrimination against their group in their second and third years. Similarly, Sengupta, Milojev, et al. (2015) drew on evidence for the relational nature of political attitudes (see Jost et al., 2008 , for a review) to propose that ingroup contact shapes ideological beliefs by fostering shared understandings of the social world. In a large, nationally representative sample of Māori (i.e., the indigenous peoples and the largest ethnic minority group of New Zealand), ingroup contact correlated negatively with the belief that all ethnic groups are treated fairly in New Zealand. Together, these studies provide tentative evidence that ingroup contact might increase perceived discrimination in disadvantaged groups.

Researchers have argued that people have a limited capacity for social relationships and showed that, in line with this argument, individuals who have more ingroup friendships tend to have fewer outgroup friendships and vice versa ( Pfister et al., 2020 ). Accordingly, studies need to consider both ingroup and outgroup contact in order to avoid confounding the two. While sever

Socialization and Perceived Discrimination

The research discussed thus far concerns the effects of having more or less contact with ingroup and outgroup members on disadvantaged-group members’ perceptions of discrimination. There are, however, other ways in which contact, especially friendships, may shape perceived discrimination. Individuals may adjust their attitudes, beliefs, and behaviors to become more similar to those of valued groups and individuals ( Hardin & Higgins, 1996 ; Turner, 1991 ). Kandel (1978) termed this process socialization , defined as the increase in similarity between friends over time through mutual influence. Through the process of socialization, disadvantaged-group members’ perceptions of discrimination might be influenced not merely by how many ingroup and outgroup friends they have, but also by their friends’ perceptions of discrimination. Accordingly, friends’ perceptions of discrimination should become more similar over time.

Past research suggests that socialization may shape sociopolitical beliefs, attitudes, and ideologies. Oosterhoff et al. (2022) found that friends tended to share similar levels of patriotism, right-wing authoritarianism, and anti-immigration attitudes. Similarly, Stangor and Leary (2006) identified similarities between friends in social dominance orientation, right-wing authoritarianism, and prejudice toward various outgroups. Such cross-sectional studies provide evidence of shared sociopolitical attitudes and ideologies between friends but cannot determine whether the similarity results from friends influencing each other over time (i.e., socialization) or individuals choosing to befriend similar others (i.e., selection).

Critically, both processes are plausible. Individuals can be influenced by their friends and also select as friends those who are similar with regard to various sociodemographic, behavioral, and intrapersonal characteristics ( McPherson et al., 2001 ). Longitudinal social network analysis allows one to separate and simultaneously examine socialization and selection processes. While this method remains underutilized in social psychology ( Wölfer & Hewstone, 2017 ), it has been applied to study the socialization of attitudes toward outgroups. Longitudinal social network studies have found that friends influence each other’s attitudes toward outgroups over time while controlling for individuals’ tendencies to select friends who already hold similar attitudes ( Bracegirdle et al., 2022 ; van Zalk et al., 2013 ; Zingora et al., 2020 ). These findings suggest that socialization might also shape other sociopolitical attitudes, beliefs, and ideologies, including perceived discrimination.

Several researchers have proposed that socialization explains why ingroup contact might increase perceived discrimination ( Levin et al., 2006 ; Sengupta, Milojev, et al., 2015 ; Sengupta & Sibley, 2013 ). During ingroup contact, disadvantaged-group members might exert reciprocal social influence on each other, resulting in their perceptions of discrimination becoming more similar and shifting toward the shared norm within the disadvantaged ingroup. This may not only increase the alignment of perceptions but also foster polarization in the direction of the group norm ( Haslam et al., 1999) . Reflecting disadvantaged-group members’ collective experiences and shared opposition to group-based hierarchies ( Lee et al., 2011 ), this process could explain why ingroup contact increases perceived discrimination. During ingroup contact, disadvantaged-group members might further discuss personal experiences of discrimination, which, in turn, could make group discrimination seem more prevalent.

In theory, socialization could also explain the effect of outgroup contact. Due to either a lack of direct experience or ideological reasons, advantaged-group members often perceive less discrimination against the disadvantaged outgroup than do disadvantaged-group members ( Norton & Sommers, 2011 ). Accordingly, contact with the advantaged could, through social influence, reduce perceived discrimination among the disadvantaged. Socialization is thus a plausible explanation for the opposite effects of ingroup and outgroup contact on perceived discrimination.

Socialization may, however, be more likely to occur through relationships with ingroup members than with outgroup members. Theoretical work contends that individuals conform to the norms of groups with which they identify ( Sherif & Sherif, 1953 ; Turner et al., 1987 ) and adjust their attitudes in accordance with similar others to achieve uniformity within a social group ( Festinger, 1954 ). Furthermore, disadvantaged-group members may be influenced only by ingroup members’ perceptions of discrimination because outgroup members’ perceptions could be considered biased, untrue, or based on indirect experience. Bracegirdle et al. (2022) conducted the first social network study to distinguish ingroup and outgroup socialization and found that ingroup, not outgroup, friends’ attitudes influenced individuals’ attitudes toward ethnic outgroups. Consistent with these findings, socialization through ingroup friends, rather than outgroup friends, may shape perceived discrimination in disadvantaged groups. Research to date could not test whether socialization shapes perceived discrimination because it measured only how much contact with ingroup and outgroup members someone has—and not what those ingroup and outgroup members’ perceptions of discrimination are. In the present research, we simultaneously examine both contact and socialization, in order to consider the multiple ways in which social relationships may influence perceived discrimination.

The Present Research

The reviewed research gives reason to expect that perceived discrimination may be shaped through the processes of contact and socialization. On the basis of this literature, we propose three hypotheses specifying how social relationships with ingroup and outgroup members might influence perceptions of discrimination among disadvantaged-group members. First, we hypothesize that greater outgroup contact will predict decreased perceived discrimination (termed an “outgroup contact” effect). Second, we hypothesize that greater ingroup contact will predict increased perceived discrimination (termed an “ingroup contact” effect). Third, we hypothesize that ingroup friends’ perceptions of discrimination will influence individuals’ perceptions of discrimination, such that their perceptions will become more similar over time (termed an “ingroup socialization” effect). Finally, we explore the extent to which outgroup friends’ perceptions of discrimination influence individuals’ perceptions of discrimination (termed an “outgroup socialization” effect). As no study to date has investigated the socialization of perceived discrimination, we seek to determine whether socialization occurs only through ingroup friends, or also through outgroup friends.

We combine different state-of-the-art methods to test our hypotheses across three studies. 2 The first study examines the longitudinal associations between ingroup and outgroup contact and perceived discrimination among ethnic minority groups in New Zealand. This study does not investigate socialization but rather seeks to determine whether the associations identified in prior research will replicate when separating within-person change and between-person stability ( Hamaker et al., 2015 ). The second study uses Bayesian multilevel network autocorrelation modeling to examine ingroup and outgroup socialization, in addition to ingroup and outgroup contact, among ethnic minority students in the United Kingdom. This study investigates the associations between individuals’ perceptions of discrimination, their numbers of ingroup and outgroup friends, and their friends’ perceptions of discrimination. The third study uses longitudinal social network analysis to explore the direction of the associations tested in Study 2 among a sample of Asian British students. This study examines whether friends have similar perceptions of discrimination because individuals are influenced by their friends’ perceptions (i.e., socialization) or because individuals choose friends with similar perceptions (i.e., selection). The intergroup contexts and ethnic groups examined therefore differ across the three studies.

We consider two forms of contact, consisting of direct positive contact in Study 1 and friendship in Studies 2 and 3. Research suggests that direct positive contact and friendship are more consistently associated with perceived discrimination than indirect, neutral, and negative contact ( Reimer & Sengupta, 2023 ). Friendship may be especially important for shaping individuals’ perceptions of discrimination because friendship occurs over an extended period of time and entails active engagement, self-disclosure, empathy, intimacy, and trust ( MacInnis & Page-Gould, 2015 ; Swart et al., 2011 ). Furthermore, although individuals may be influenced by unfamiliar others in short interactions ( Blanchard et al., 1991 ), socialization may be more likely to occur through long-term intimate interactions with those who are liked, especially one’s friends ( Paluck, 2011 ; Sinclair et al., 2005 ). In the present research, we investigate the extent to which both friendship and generic positive contact shape perceived discrimination.

Study 1

Prior research testing the effects of ingroup and outgroup contact on perceived discrimination has relied almost exclusively on cross-sectional surveys (e.g., Dixon et al., 2010 ; see Reimer & Sengupta, 2023 ). For this reason, it is unclear whether contact leads to a change in perceived discrimination, perceived discrimination leads to a change in contact, both, or neither. Studies that assess contact and perceived discrimination over time can provide evidence for the temporal order of effects by identifying whether changes in ingroup or outgroup contact precede changes in perceived discrimination or vice versa. Temporal precedence is a necessary (but not sufficient) condition for causal inference.

However, the few longitudinal studies in this literature have used cross-lagged panel modeling (CLPM; e.g., Tropp et al., 2012 ), which produces biased estimates because CLPMs fail to separate time-invariant, between-person stability from within-person change ( Hamaker et al., 2015 ; Orth et al., 2021 ; Osborne & Little, in press ). Applied to contact research, between-person stability reflects the degree to which those who report high (vs. low) levels of contact across all time points also report high (vs. low) perceived discrimination across all time points. Between-person stability can be considered a confound when estimating causal effects ( Granger, 1980 ) and may indicate that unobserved variables produce an artificial correlation between stable individual differences in contact and perceived discrimination (i.e., possible third-variable effects) or may reflect prior changes in a subsequently stable environment. Most importantly, between-person stability tells us little about the extent to which a change in contact leads to a change in perceived discrimination. In contrast, identifying within-person change would mean that an increase or decrease from one’s average level of contact at a given time point precedes an increase or decrease from one’s average level of perceived discrimination at the following time point . Thus, after adjusting for between-person stability, within-person change demonstrates the temporal precedence of contact relative to perceived discrimination, which provides stronger evidence of an effect of contact on perceived discrimination.

Hamaker et al. (2015) developed an extension to the CLPM, termed the random-intercept cross-lagged panel model (RI-CLPM; see Figure 1 ), to separately estimate between-person stability and within-person change. Although there are alternative methods for separating between-person and within-person effects, RI-CLPMs are the most likely to converge ( Orth et al., 2021 ) and do not require intensive longitudinal data (e.g., over 30 assessments). Thus, RI-CLPMs are quickly becoming the method of choice amongst longitudinal researchers interested in within-person effects ( Friehs et al., 2023 ; O’Donnell et al., 2021 ; Osborne & Little, in press ). Here, we apply this model to test, for the first time, whether levels of ingroup and outgroup contact predict within-person change in perceived discrimination over time. This method also allows us to investigate potential reciprocal effects, in which perceived discrimination predicts within-person change in contact over time.

Study 1 draws on data from a large-scale national longitudinal study in New Zealand: The New Zealand Attitudes and Values Study (NZAVS). New Zealanders of European descent (hereafter, Europeans) represent the advantaged ethnic group. The indigenous Māori, as well as later immigrant groups from the Pacific Islands and Asia, represent disadvantaged ethnic groups. Ethnic minority groups in New Zealand face longstanding deficits relative to Europeans across a range of socioeconomic indicators including income, unemployment, life expectancy, and life satisfaction ( Ministry of Social Development, 2016 ). Thus, our analyses focus on the contact each ethnic minority group—Māori, Pacific Islanders, and Asians—has with their own group (ingroup contact) and with Europeans (outgroup contact) and whether this affects their perceptions of discrimination against their own ingroup.

The NZAVS began in 2009 (i.e., Time 1). Invitations to participate in Time 1 were sent to 40,500 people randomly selected from the electoral roll, 6,518 (16.6%) of whom returned completed surveys. In 2010, Time 2 surveys were sent to all participants who responded at Time 1. This survey was completed by 4,423 (67.9%) participants from the Time 1 sample. From Time 3 to Time 5 (2011–2013), in addition to sending out surveys to existing participants, booster sampling was conducted to increase the overall sample size (see Sibley, 2018 ). Thus, Time 3 (2011) contained responses from 6,844 participants, Time 4 (2012) had 12,182 participants, Time 5 (2013) had 18,264 participants, and Time 6 (2014) had 15,822 participants.

The measures required for the present research (i.e., contact and perceived discrimination) were only included in the NZAVS survey at Times 3, 4, 5, and 6. Accordingly, data for the present study were drawn from the only four consecutive waves of the NZAVS that contained all the items required for our model: Time 3 (2011) to Time 6 (2014). This yielded a total sample size of 4,325 ethnic minority participants (2,929 Māori, 558 Pacific Islanders, and 839 Asians) who provided partial or complete responses to the variables of interest and who responded to at least one of the four surveys ( M age = 47.43, SD = 13.73; 65.87% women). Of these participants, 842 (19%) completed all four surveys, 1,271 (29%) completed three surveys, 1,920 (44%) completed two surveys, and 292 (7%) completed only one survey. Across surveys, 81% (Time 4), 84% (Time 5), and 84% (Time 6) of participants who participated in the preceding wave (Times 3–5) participated in the subsequent wave (Times 4–6). 3 Descriptive statistics for all variables included in the model are presented in online Appendix A. 4 No a priori power analy

Perceived Group Discrimination

Consistent with prior theory and research on the construct of perceived discrimination, we assessed perceived discrimination in terms of people’s perceptions that their ingroup is discriminated against in society (see Branscombe et al., 1999 ). We used a single-item measure that asked, “Do you think people from your ethnic group are discriminated against in New Zealand?” (1 = definitely no , 7 = definitely yes ).

Analysis Strategy

Following the RI-CLPM procedure outlined by Hamaker et al. (2015) , each observed variable was modeled as a function of two latent variables. One latent variable was the time-invariant random intercept, which was modeled as loading equally on each congeneric indicator across time (e.g., the random intercept of ingroup contact loaded equally on ingroup contact at Times 3, 4, 5, and 6). The second latent variable loaded only on its corresponding indicator at each individual wave, thus representing a time-specific departure from the average response over time. The error variances of the observed variables were constrained to zero meaning that all variation in the observed scores was explained by the within-person and between-person factor structure.

The random intercepts of ingroup contact, outgroup contact, and perceived discrimination were then correlated with each other to estimate between-person effects. These correlations indicate the degree to which stable individual differences in contact are associated with stable individual differences in perceived discrimination (i.e., trait-like stability). Having thus accounted for between-person effects, the time-specific latent variables for ingroup contact, outgroup contact, and perceived discrimination were regressed on each other in the same manner as a traditional CLPM, yielding autoregressive and cross-lagged coefficients (e.g., perceived discrimination at Time 4 regressed on ingroup and outgroup contact at Time 3; ingroup and outgroup contact at Time 4 regressed, in turn, on perceived discrimination at Time 3). These coefficients represent within-person change over time—the degree to which an individual’s deviation from their expected score (i.e., their own mean across waves) on the predictor variable at a given time point explains their deviation from their expected score on the outcome variable at the following time point.

Finally, we estimated the model as a stationary process by constraining the corresponding autoregressive and cross-lagged effects to equality across time. This approach maximizes power and represents an assumption of a continuous process with an unknown starting point ( McArdle, 2009 ). We estimated the model using Mplus 8.4 ( Muthén & Muthén, 2015 ). The outcome variables, which we analyzed as continuous variables, had 7-point response scales. Analyzing ordinal data as if they were metric data can distort estimates of effect sizes and inflate rates of false-positive and false-negative findings ( Liddell & Kruschke, 2018 ). But, as the RI-CLPM has not been implemented or described for ordinal variables, we nonetheless proceeded to model all variables as continuous variables. The materials and analysis code for Study 1 are provided online, and the data are available upon request (see online Appendix D).

Between-Person Correlations

Results from the RI-CLPM showed that the random intercept for ingroup contact was positively correlated with the random intercept for perceived discrimination ( r = .30, SE = .04, p < .001; 99% CI [.21, .37]), whereas the random intercept for outgroup contact was negatively correlated with the random intercept for perceived discrimination ( r = − .22, SE = .03, p < .001; 99% CI [−.31, −.16]). The correlation coefficients indicate small to medium associations between contact and perceived discrimination. These results show that those who reported greater ingroup contact across the four-wave assessment period also reported higher perceived discrimination across the same period, but those who reported greater outgroup contact across the four-wave assessment period reported lower perceived discrimination across the same period. This indicates that stable individual differences in contact and perceived discrimination correlate in the population but does not indicate a longitudinal effect in which changes in contact precede changes in perceived discrimination. The random intercept for ingroup contact was also positively correlated with the random intercept of outgroup contact ( r = .27, SE = .02, p < .001; 99% CI [.21, .31]).

Within-Person Coefficients

When adjusting for between-person stability, the within-person effects of ingroup and outgroup contact on subsequent levels of perceived discrimination were nonsignificant, as were the within-person effects of perceived discrimination on subsequent levels of ingroup and outgroup contact (see Table 1 , for all within-person effects in the model). This indicates that an individual’s deviation from their trait level of contact in 1 year was not associated with a deviation from their trait-level perceived discrimination 1 year later. The results also showed that the reciprocal association did not emerge. That is, an individual’s deviation from their trait level of perceived discrimination was not associated with a deviation from their trait-level contact 1 year later. These findings are not consistent with an effect of contact on perceived discrimination because they do not show that changes in contact precede changes in perceived discrimination or vice versa. 5

Thus, the findings of Study 1 do not provide support for our hypotheses that ingroup contact predicts increased perceived discrimination over time or that outgroup contact predicts decreased perceived discrimination over time. Rather than a within-person effect of contact on perceived discrimination, our results showed time-invariant associations between contact and perceived discrimination that may reflect third-variable effects. We note that, in Study 1, we analyzed ordinal variables as if they were metric variables, which, as discussed in the next section, risks degrading statistical inferences ( Liddell & Kruschke, 2018 ). We did so as the RI-CLPM has yet to be implemented or described for ordinal variables.