Peer Discrimination, Deviant Peer Affiliation, and Latino/a Adolescent Internalizing and Externalizing Symptoms: A Prospective Study

Abstract

Objective:

U.S. Latino/a adolescents experience high levels of ethnic discrimination, particularly in new immigrant destinations. Due to the salience of peers during adolescence, this study examined how peer discrimination related directly and indirectly, through deviant peer affiliation, to changes in Latino/a adolescents’ internalizing and externalizing symptoms. Culture-specific moderators hypothesized to buffer discrimination impacts on adolescent symptomology included Spanish language enculturation and adolescents’ social ties to relatives in the family’s country-of-origin.

Method:

The sample of 547 Latino/a adolescent participants from the Caminos al Bienestar study (55.4% female; age M = 12.8, range = 11 – 16) was selected at random from middle schools in a large, suburban school district in Atlanta, Georgia. Three time points of survey data spaced roughly 6 months apart were collected during 2018 and 2019.

Results:

Results from longitudinal structural equation models revealed that peer discrimination was associated indirectly with increased externalizing symptoms, through increases in affiliation with deviant peers (β = 0.05; SE = 0.02; B = 0.02; 95% CI = 0.01, 0.09). We did not observe direct or indirect effects of peer discrimination on changes in internalizing symptoms, and we found no significant protective effects of either Spanish language enculturation or social ties with the country-of-origin.

Conclusions:

Ethnic discrimination by peers may lead to deviant peer affiliation and, in turn, increased externalizing behaviors. Future research identifying protective factors that buffer discrimination impacts on deviant peer affiliation is needed to inform the development of interventions that can prevent Latino/a adolescents’ externalizing symptoms.

Introduction

Anti-immigrant sentiment in the U.S. threatens the emotional and behavioral adjustment of Latino/a youth across diverse national origins and family residency statuses (Roche, Vaquera, et al., 2020). While not new in the U.S., experiences of ethnic discrimination may be increasing for U.S. Latino/as (Ee & Gándara, 2020; Natanson et al., 2020). Due to the increased salience of peer and school contexts during the adolescent years, discrimination perpetrated by peers at school may be especially harmful for Latino/a adolescents. Peer discrimination, a form of peer rejection, may lead adolescents to affiliate with deviant peer groups (Benner et al., 2018; McLoyd et al., 2009). Maintaining one’s ethnic cultural orientation – or enculturation – may protect Latino/a adolescents from the negative effects of ethnic discrimination. Using longitudinal data for a probability sample of Latino/a adolescents from diverse national origins and living in a new immigrant destination, this research examined (1) how ethnic discrimination from peers related to changes in internalizing and externalizing symptoms directly and indirectly by way of affiliation with deviant peers, and (2) how enculturation moderated the association between peer discrimination and increased adolescent symptomology.

Discrimination and Latino/a Adolescent Adjustment

Culturally informed developmental theory highlights the salience of ethnic discrimination to Latino/a adolescent adjustment. According to García Coll and colleagues’ (1996) integrative model, social stratification – social position, discrimination, and segregation – filter youth of color into various environments, including peer groups, which, in turn, promote or inhibit healthy development. Adolescent advances in cognitive skills allow young people to recognize certain behaviors as discriminatory and become aware of differential treatment rooted in structural racism (Benner et al., 2018; Gee et al., 2012), defined as “an organized system in which public policies, institutional practices, cultural representations and others’ norms work together to reinforce and perpetuate racial group inequity” (Neblett Jr, 2019). Young people may face challenges in effectively coping with the stress of discrimination as protective coping skills are still developing during early and middle adolescence (Adam et al., 2015; Gee et al., 2012).

While anti-immigrant policies and sentiment have existed in the U.S. for decades, recent changes in the political environment appear to have led to a recent spike in ethnic discrimination against Latino/as. Anti-immigrant policy actions since 2017 include the increased deportation of undocumented individuals in the country’s interior (U.S. Immigration and Customs Enforcement, 2017), the announced end of the Deferred Action for Childhood Arrivals program (U.S. Department of Homeland Security, 2017), and the separation of migrant children from parents at the U.S.-Mexico border (U.S. Department of Homeland Security, 2018). Accompanying these policy shifts, Latino/as nationwide have been exposed to media coverage of discriminatory statements (Rojas-Flores et al., 2019), and Latino/a adolescents appear to have experienced greater discrimination (Ee & Gándara, 2020; Natanson et al., 2020; Roche, Vaquera, et al., 2020). Capps and colleagues, for example, found that 24% of Latino/a high school students in 2018–19 perceived that others presumed their English was poor (Capps et al., 2020). Latino/a adolescents living in new immigrant areas, such as those in the Southeastern U.S., may experience especially high levels of discrimination (Stein et al., 2018) due to limited cultural, structural, and linguistic support for Latino/a residents (Ebert & Ovink, 2014).

From a developmental perspective, there are compelling reasons to expect that experiences of ethnic discrimination perpetrated by peers is especially salient during the adolescent years. Adolescents increasingly spend time outside the home and come to rely more on their peers as sources of support and socialization (DiClemente et al., 2009) – processes that facilitate opportunities for peer influence (Simons-Morton et al., 1999). The adolescent years are also a critical time for identity formation, a process that typically occurs within the context of social groups (DiClemente et al., 2009). Experiences of discrimination can threaten adolescent perceptions of themselves and lead to low self-esteem, thereby harming adolescent mental health and well-being (Cano et al., 2016; Dulin-Keita et al., 2011). Discrimination within the context of peer networks may be particularly harmful as adolescents navigate their identity formation (Del Toro & Hughes, 2019; Delgado et al., 2019).

Ethnic discrimination poses risks to mental health, commonly indicated by internalizing symptoms and externalizing behavioral problems (Benner et al., 2018; Umaña-Taylor, 2016). In research conducted with Latino/a adolescents, the evidence supporting links between peer discrimination and greater internalizing symptoms appears stronger than is the case for externalizing symptoms (Delgado et al., 2019; Hughes et al., 2016; Stein et al., 2019). For example, in longitudinal studies of Mexican-origin adolescents living in the southwestern U.S., peer discrimination was related to increases in internalizing, but not externalizing, symptoms (Delgado et al., 2019; White et al., 2018). As some research indicates that discrimination relates to increased externalizing symptoms for Mexican adolescents (Brittian et al., 2013), there is a need for research clarifying the relationship between discrimination and externalizing symptoms. These inquiries are especially important for Latino/a adolescents in new immigrant areas, where many Latino/a parents are foreign-born and face a more hostile context of reception than is the case in longstanding immigrant areas (Gonzalez et al., 2015; Potochnick et al., 2012). New immigrant areas also are home to many Latino/as who have escaped extreme violence and poverty in El Salvador, Guatemala, and Honduras (Cohn, Passel, & Gonzales-Barrera, 2017) Finally, today’s political climate and the sharp uptick in anti-immigrant rhetoric may exacerbate problems of ethnic discrimination for contemporary youth (Ee & Gándara, 2020; Roche, Vaquera, et al., 2020; Roche, White, et al., 2020).

Discrimination and Deviant Peer Affiliation

Exposure to ethnic discrimination increases Latino/a adolescents’ risk of deviant peer affiliation (Benner et al., 2018). Among a cross-sectional sample of Mexican-origin adolescents in the southwestern U.S., Delgado and colleagues (2011) found that peer ethnic discrimination was associated with greater affiliation with deviant peers (Delgado et al., 2011). Similarly, longitudinal studies of African American youth have suggested that discrimination may lead to increases in affiliation with deviant peers over time (Roberts et al., 2012; Unnever et al., 2017).

The process through which ethnic discrimination may impact peer affiliation can be explained by the individual characteristics/selection model of peer affiliation, whereby youth self-select into peer groups that hold beliefs and attitudes resembling their own (Chen et al., 2015; Deater-Deckard, 2001). Additionally, youth may seek affiliation with deviant peers following rejection due to the lack of other peer group options (Lansford et al., 2014). In an attempt to prevent isolation and loneliness (Brendgen et al., 2000), the young person experiences a selection process referred to as “default selection” (Sijtsema et al., 2010). Some scholars speculate that discriminatory experiences diminish racial/ethnic minority youth’s self-worth, which, in turn, can lead young people to endorse values that run contrary to conventional systems and institutions and to seek out friendships with peers who hold these same values (Brody et al., 2012). Youth experiencing discrimination may also seek out peers who offer some protection from future discriminatory experiences, as can occur through gang membership (Bacallao & Smokowski, 2013). Peer discrimination may represent a form of peer rejection (Martin-Storey & Benner, 2019) that leads young people to restore their self-esteem and self-worth through affiliation with deviant peer groups (Kaplan, 1980; Patterson et al., 1989). The experience of peer rejection can foster negative traits or emotions, such as hostility, aggression, or anger, that lead youth to affiliate with peer groups where these characteristics are more normative (Brody et al., 2006; Chen et al., 2015).

Although much of what has been learned about peer rejection and affiliation with deviant peers stems from studies of non-Hispanic White and African American adolescents, Latino/a adolescents in new immigrant areas may face a unique set of challenges that elevate risks for deviant peer affiliation. Not only are new immigrant areas characterized by high levels of socioeconomic disadvantage, an important correlate of adolescent deviant behavior (McLoyd et al., 2009), but new immigrant areas tend to lack the infrastructure and cultural resources important to the healthy development of Latino/a adolescents (Stein et al., 2019; White et al., 2018). Longitudinal research can help advance knowledge about the degree to which deviant peer affiliation acts as a mechanism linking ethnic discrimination from peers to increases in internalizing and externalizing symptomology for Latino/a adolescents.

Deviant Peer Affiliation and Latino/a Adolescent Adjustment

The integrative model posits that inhibiting peer environments are costly for adolescent adjustment (García Coll et al., 1996). Research using diverse samples, including Latino/a youth (Germán et al., 2009; Gonzales et al., 2017; Loukas et al., 2008), has shown that affiliating with deviant peers is associated with greater externalizing and related risk behaviors (Danzo et al., 2017; Gillies et al., 2017; Wang & Dishion, 2012). Developmental and health behavior theories and models, such as the problem behavior theory (Jessor & Jessor, 1977), social learning theory (Bandura & Walters, 1977), deviancy-training model (Dishion et al., 1996), and the social enhancement/interaction model (Chen et al., 2015) also have helped explain why deviant peer affiliation relates to youth behavior. These theories and models suggest that deviant peers provide (1) a source of behavioral modeling during a developmental stage of heightened susceptibility to social influences (Dishion & Tipsord, 2011) and (2) opportunities and positive reinforcement for participation in deviant behavior due to the peers’ own engagement in such behavior (Dishion & Patterson, 2016).

While extensive research examines how adolescents’ externalizing behaviors are shaped by deviant peer affiliation, only a few studies examine links from deviant peer affiliation to adolescent internalizing symptoms. Some research has shown, however, that deviant peer affiliation is associated with higher levels of adolescent depressive symptoms (Connell & Dishion, 2006; Hong et al., 2019; Klostermann et al., 2016). Among a sample of Mexican American 5^th graders, deviant peer affiliation was associated with greater internalizing symptoms (Roosa et al., 2010). Limited support and care typifying deviant peer relations may compromise youth’s self-esteem and emotional well-being, outcomes closely aligned with internalizing symptoms (Mrug et al., 2004; Waldrip et al., 2008). Further, deviant peer relationships often are characterized by conflict and hostility (Dishion et al., 1995), which may elevate risks for depressive symptoms (Zhang et al., 2018). There remains a need for more research examining the role of deviant peer affiliation in Latino/a adolescents’ development of internalizing symptoms.

The Mediating Role of Deviant Peers

Consistent with the integrative model, whereby social stratification can filter youth into inhibiting peer environments that may compromise their development (García Coll et al., 1996), deviant peer affiliations may play an important mediating role linking discrimination to adolescent adjustment. Using three time points of data, Buchanan and Smokowski (2009) found affiliation with deviant peers mediated the relationship between experiences of general discrimination and greater substance use among Latino/a adolescents in North Carolina and Arizona (Buchanan & Smokowski, 2009). In line with these findings, other researchers have found that deviant peer affiliation may mediate the relationship between experiences of stress and youth’s poor adjustment (e.g., Acosta et al., 2015; Roosa et al., 2010). A critical limitation of research on this topic concerns the reliance on data measuring deviant peer affiliation and adolescent outcomes at the same time point, limiting the understanding of these causal processes that occur over time (Maxwell & Cole, 2007).

The Moderating Role of Enculturation

García Coll’s integrated model stresses the value of adaptive culture – roles, values, and behaviors established in response to contextual challenges such as racism and discrimination. For U.S. Latino/a adolescents, maintaining or endorsing traditional cultural orientations may occur through maintaining Spanish language use and social ties to relatives in the country of origin. Germán and colleagues found that deviant peer affiliation was less strongly associated with externalizing behaviors for Latino/a adolescents who strongly endorsed familism (Germán et al., 2009), and Cavanaugh and colleagues found that the positive associations between peer discrimination and externalizing symptoms were weaker in a context of adolescents’ greater enculturation, racial-ethnic identity, and familism values (Cavanaugh et al., 2018). By contrast, Stein and colleagues found no protective effect of familism moderating associations between peer discrimination and depression for primarily Mexican-origin youth in North Carolina (Stein et al., 2015). There remains a need to explore these processes using longitudinal data, examining adolescents in new immigrant areas with limited language and cultural supports for Latino/a immigrants (Azmitia, 2021), and focusing on adolescents growing up amidst today’s heightened anti-immigrant rhetoric (Ee & Gándara, 2020; Roche, Vaquera, et al., 2020).

Current Study

Using a probability sample of Latino/a adolescents in a new immigrant destination, the present study integrates literatures on peer ethnic discrimination, adolescent deviant peer affiliation, and enculturation to address three aims. First, we examined how peer ethnic discrimination was associated with changes in Latino/a adolescents’ internalizing and externalizing symptoms over time. Second, we examined whether deviant peer affiliation mediated associations between peer ethnic discrimination and changes in Latino/a adolescent symptomology. Third, we investigated the degree to which culture-specific protective effects buffered adolescents from the negative impacts of peer ethnic discrimination on adolescent symptomology. We hypothesized that:

H1: Peer ethnic discrimination would be associated with increases in internalizing and externalizing symptoms directly and indirectly through increases in deviant peer affiliation.

H2: Enculturation processes – adolescent Spanish language use and social ties with relatives in the country of origin – would help buffer adolescents from the harmful effects of ethnic discrimination on increases in symptomology.

Hypotheses were not pre-registered in a digital repository.

Methods

Sample and Procedure

The study sample included 547 Latino/a adolescents (ages 11–16 years) from the Pathways to Health/Caminos al Bienestar (“Caminos”) study described in detail by Roche and colleagues (2020). All students listed as “Hispanic” on 2017–18 school enrollment records were recruited from 14 middle schools in suburban Atlanta, GA. To ensure family, school, and community heterogeneity, students were selected from clusters of schools defined by the proportion of Latino/a students as “low” (<13%), “moderate” (18–25%), or “high” (>40%). Students were screened for eligibility and selected at random from grade (6^th, 7^th, 8^th) and gender strata within each of the three Latino/a student concentration clusters. Participants were ineligible for participation if: 1) they had an Individualized Education Plan, 2) were unable to read in English or Spanish, 3) self-reported as not being Latino or Hispanic, 4) had a sibling who was already a study participant, or 5) they were an age that was outside the typical range for that school grade. The response rate among eligible adolescents whose parents were contacted and provided permission was 65.2%, and the response rate among eligible adolescents contacted was 95.3%. Study participants completed their Time 1 (T1) survey online via Qualtrics. Most (77%) complete the T1 survey in the school setting (February – April 2018). As the school district unexpectedly requested an end to in-school data collection by May 2018, the remaining 23% of students completed the online survey at home from September 2018 through January 2019. Retention rates at six-month (Time 2, T2) and one-year (Time 3, T3) follow-ups were 81.5% and 76.5%, respectively. Participants received up to $65 for completing the surveys. Investigators obtained a Certificate of Confidentiality from the National Institutes of Health and Institutional Review Board (IRB) approval from The George Washington University (protocol number 121613). Parents provided oral or written consent and adolescents provided written assent. Study materials were provided in Spanish and English.

The mean age of participants at T1 was 12.8 years (SD = 1.03), and just over half of participants were female (55.4%). About two-thirds (67.3%) reported living in a household with two biological/adoptive parents; 19% reported living with a single-parent and 14% reported living with a parent and a stepparent. On average, participants’ mothers had slightly more than a high school education (M = 3.15, SD = 1.64). Roughly two-thirds (67.3%) of participants were second-generation immigrants, while 21% were third- or later-generation immigrants and 12% were first-generation immigrants. Approximately 88% of youth participants were born in the U.S., while 2.7% were born in Mexico, 2.6% in a Central American country, 3.8% in a South American country, and 2.7% in another country. Among mothers of youth participants, roughly half (51.0%) were born in Mexico, and 19.2% were born in the U.S. Remaining countries of birth for mothers included, but were not limited to: El Salvador (5.1%), Puerto Rico (4.4%), Guatemala (4.1%), Brazil (4.0%), Venezuela (2.7%), Colombia (2.7%), Honduras (2.6%), and Peru (1.8%).

Measures

Peer Ethnic Discrimination

Peer ethnic discrimination was measured at T1-T3 with the 4-item Adolescent Discrimination Distress Index (Fisher et al., 2000). Adolescents were shown the following: “The next statements are about your being treated unfairly at school because you are Latino/a,” and asked to report how often classmates at school did things such as, “say bad things about Latinos” and “threaten or harass you because you are Latino/a” (1 = almost never or never to 5 = almost always or always (T1 α = 0.77, T2 α = 0.74, T3 α = 0.85).

Affiliation with Deviant Peers

The 9-item T1-T3 measure of affiliation with deviant peers (Barrera et al., 2002; Elliott et al., 2012) assessed adolescent reports of how many of their friends engaged in risky behaviors such as cheating and stealing (1 = none to 5 = almost all; T1 α = 0.90, T2 α = 0.88, T3 α = 0.93).

Spanish Language Enculturation

T1 Spanish language use was assessed using the Brief Acculturation Rating Scale for Mexican Americans-II (Cuellar et al., 1995). Adolescents reported how often they felt 12 situations were true for them, including, “I enjoy speaking Spanish”, and “My thinking is done in Spanish” (1 = almost never/never to 5 = almost always/always; α = 0.89).

Social Ties to Family Country-of-Origin

A single item at T1 indicated how often adolescents communicated with relatives from their country of origin (0 = never to 5 = at least once a week).

Internalizing and Externalizing Symptoms

Internalizing and externalizing symptoms were measured at T1-T3 using the Youth Self-Report/11–18 (Achenbach & Edelbrock, 1991). Adolescents indicated the degree to which symptoms were true for them within the past six months (0 = not true, 1 = somewhat or sometimes true, 2 = very true or often true). Internalizing symptomology subscales included anxious/depressed (12 items; e.g., “I am nervous or tense”), withdrawn/depressed (8 items; e.g., “I would rather be alone than with others”), and somatic complaints (9 items; e.g., “I feel dizzy or lightheaded”; T1 α = 0.85, T2 α = 0.86, T3 α = 0.87). Externalizing symptomology subscales included aggressive behavior (17 items; e.g., “I destroy my own things”) and rule-breaking behavior (13 items; e.g., “I lie or cheat”; T1 α = 0.82, T2 α = 0.84, T3 α = 0.85).

Demographic Variables

Demographic characteristics assessed at T1 included adolescent age in years, gender (0 = female, 1 = male), and immigration generational status, indicated by dummy coded variables for first-generation (i.e., parent and adolescent were both foreign born), second-generation (i.e., adolescent was U.S. born but has at least one parent who was foreign born), and third-/later-generation (i.e., parent and adolescent were both U.S. born; the omitted reference group). Maternal educational attainment was treated as a continuous variable (1 = 8^th grade or less, 2 = some high school, 3 = completed high school, 4 = some college, 5 = completed college, and 6 = graduate or profession school after college). Dummy-coded variables for household structure assessed living with a single parent, a parent and stepparent, or two biological/adoptive parents, the omitted reference group. Finally, dummy-coded variables indicated a high, moderate, of low (omitted reference group) Latino/a student concentration in the school.

Analytic Strategy

We estimated structural equation models (SEM) using Mplus 8.2 to examine how peer ethnic discrimination was associated with internalizing and externalizing symptoms and to test the mediating role of deviant peer affiliation and the moderating role of enculturation factors.

Measurement Model.

Measurement models run using confirmatory factor analysis (CFA) estimated loadings of each manifest indicator on the respective latent construct. Following Little and colleagues (2013), we used parceling techniques, whereby the average score of at least two items was calculated and included as indicators predicted by a latent variable. Parceling reduces the number of indicator variables to no more than three for each construct, thereby improving reliability of the indicator variables, decreasing the likelihood of violations to assumptions of normality, and decreasing the likelihood of correlated residuals and dual factor loading (Little et al., 2013). Facet representative parceling was used for externalizing and internalizing symptoms; each parcel consisted of items comprising a symptomology subscale. A balance of high and low loadings was used to create parcels for discrimination, deviant peer affiliation, and Spanish language enculturation. To ensure a meaningful metric, we identified scales using effects coding, which sets the scale based on the average of all indicator variables (Little et al., 2006). We correlated parcel residuals for each construct across time and allowed all constructs to correlate within time. Measurement invariance across time and gender was establishing using multi-group CFA models with increasing constraints and comparing model fit. Model fit was determined using a comparative fix index (CFI) change of less than 0.01 and a root mean square error of approximation (RMSEA) value within the 95% confidence interval of the reference model indicated a non-significant change in model fit, and invariance in loadings and intercepts across groups (Putnick & Bornstein, 2016).

Structural Model, H1:

Figure 1 details our hypothesized structural model for H1. We used autoregressive structural models to test for an indirect effect from T1 peer discrimination to T3 internalizing and externalizing symptoms, through T2 affiliation with deviant peers. A strength of autoregressive models is that they account for dependency of variables measured across time and assess for the stability of that measure over time, an important assumption in testing longitudinal mediation (Cole & Maxwell, 2003; MacKinnon et al., 2013). We followed the process for testing longitudinal mediation put forth by Cole and Maxwell (2003) by testing 1) equilibrium in variances and covariances of the latent variables across time; 2) other possible pathways needed to explain the relationships between variables by estimating all cross-lagged and autoregressive pathways; 3) estimating only the structural paths included in the hypothesized mediation pathway; and 4) the full mediation model (Cole & Maxwell, 2003). The full mediation model included direct pathways from T1 peer discrimination to T2 deviant peer affiliation; direct pathways from T2 deviant peer affiliation to T3 internalizing and externalizing symptoms; and, indirect pathways from T1 peer discrimination to T3 outcomes through T2 deviant peer affiliation. We also estimated pathways from T1 peer discrimination to T2 symptomology. The model included autoregressive paths between latent variables across time and accounted for within-time correlations of latent variables. T3 outcomes were regressed on all background demographic variables. Acceptable model fit for the measurement and structural models based on the following criteria: CFI value of 0.95 or greater, standardized root mean residual (SRMR) value less than 0.08, and RMSEA value less than 0.06 (Schumacker & Lomax, 2016). Following Little’s (2013) recommended tests of mediation, we estimated the total indirect effect using the MODEL INDIRECT command in Mplus and used bootstrapping to determine statistical significance (Little, 2013). Structural invariance by gender was assessed by sequentially adding constraints to each structural path, one at a time, to be equal across males and females and comparing model fit to the freely estimated model (Sass & Schmitt, 2013). Changes in the CFI and RMSEA values were used to determine structural invariance. Finally, we ran a series of alternate models as a validity check for final structural model results.

Figure 1.

Hypothesized structural model for H1.

Structural Model, H2:

Two separate models tested the protective effect of Spanish language enculturation and communication with relatives in the country of origin on the relationship between peer ethnic discrimination and later internalizing and externalizing symptoms. Latent interaction terms for each protective factor X peer discrimination were included in interaction-effects structural models predicting each outcome. Model fit was determined using the same criteria listed above.

Missing Data

Retention rates at T2 and T3 were 81.5% and 76.5%, respectively. Greater attrition at Time 2 was spurred by changes in school leadership resulting in a subsequent request to shift survey administration out of the school. Data missing due to item non-response ranged from 1.5% to 4.5%, except for mothers’ educational attainment at T1, which had 16.1% missing. Those lost to attrition at T2 included a higher proportion of males (28.7%) versus females (16.8%) and reported greater T1 externalizing symptoms (M = 10.97, SD = 8.36) than those retained (M = 8.71, SD = 6.92). There were no differences in study variables missing due to item non-response or to attrition from T2 to T3.

To address missing data, descriptive analyses were run using a multiple imputed (MI) grand mean dataset representing averaged estimates for imputed values of missing data across 200 MI data sets (Howard et al., 2015). Data imputation was performed using the PcAux package in R which optimizes recovery of auxiliary information under missing at random (Enders, 2010). For analytic models using SEM, the non-imputed dataset was used, and missing data were handled with full information maximum likelihood (FIML) estimation which produces unbiased estimates under missing at random (Enders & Bandalos, 2001). FIML estimation results in more robust standard errors compared to analyses with the grand mean dataset.

Results

Descriptive Statistics

Table 1 presents means, standard deviations, and bivariate correlations for study variables. Mean scores for peer ethnic discrimination were 1.59 (SD = 0.76), 1.49 (SD = 0.67), and 1.53 (SD = 0.72) at T1 – T3, respectively. The T1 to T2 decrease in ethnic discrimination was statistically significant (p < 0.01). Mean scores for deviant peer affiliation were 1.45 (SD = 0.58), 1.44 (SD = 0.50), and 1.52 (SD = 0.62) at T1 – T3, respectively. The increased score from T2 to T3 was significant (p < 0.01). Mean scores for internalizing symptoms were 0.48 (SD = 0.36), 0.52 (SD = 0.39), and 0.51 (SD = 0.39) at T1 – T3, respectively. None of the changes in internalizing symptom scores were statistically significant. Finally, mean scores for externalizing symptoms were 0.30 (SD = 0.24), 0.34 (SD = 0.28), and 0.33 (SD = 0.28) at T1 – T3, respectively. Mean externalizing symptoms scores significantly increased from T1 to T2 (p < 0.01). Finally, there were moderate to strong positive correlations (based on standard thresholds for correlation coefficients; (Taylor, 1990) between the primary independent variables and adolescent internalizing and externalizing symptomology (p < 0.01).

Table 1.

Descriptive Statistics and Correlations for Study Variables

Variable	1	2	3	4	5	6	7
1. Peer Discrimination (T1)	--
2. Peer Discrimination (T2)	0.56***	--
3. Peer Discrimination (T3)	0.55***	0.53***	--
4. Deviant Peers (T1)	0.37***	0.31***	0.37***	--
5. Deviant Peers (T2)	0.35***	0.50***	037***	0.57***	--
6. Deviant Peers (T3)	0.27***	0.32***	0.46***	0.40***	0.64***	--
7. Internalizing Symptoms (T1)	0.47***	0.37***	0.31**	0.33***	0.25***	0.21**	--
8. Internalizing Symptoms (T2)	0.39***	0.48***	0.34***	0.26***	0.36***	0.32***	0.69***
9. Internalizing Symptoms (T3)	0.40***	0.40***	0.43***	0.24***	0.32***	0.43***	0.65***
10. Externalizing Symptoms (T1)	0.44***	0.41***	0.41***	0.63***	0.47***	0.42***	0.57***
11. Externalizing Symptoms (T2)	0.35***	0.51***	0.42***	0.45***	0.56***	0.58***	0.42***
12. Externalizing Symptoms (T3)	0.37***	0.43***	0.52***	0.33***	0.51***	0.64***	0.37***
13. Adolescent Age	0.13*	0.13**	0.14*	0.16**	0.13**	0.12**	0.14**
14. Adolescent Gender: Male	−0.08	−0.05	−0.02	0.00	−0.03	0.06	−0.25***
15. Moderate Latino/a Concentration	0.07	0.02	0.06	0.02	−0.02	0.02	0.04
16. High Latino/a Concentration	−0.15**	−0.06	−0.10*	0.06	0.07	0.08	−0.03
17. HH Structure: Single Parent	−0.03	0.03	0.01	−0.03	−0.01	−0.01	0.04
18. HH Structure: Stepparent	0.06	0.15**	0.05	0.03	0.03	0.06	0.07
19. Mother’s Education	0.00	0.02	−0.01	−0.16**	−0.13*	−0.07	−0.06
20. First Generation Immigrant	−0.02	−0.02	−0.05	−0.09*	−0.09*	−0.02	−0.01
21. Second Generation Immigrant	−0.06	−0.05	−0.05	0.04	0.02	−0.01	0.03
22. Spanish Language Enculturation	0.07	0.03	0.05	0.03	0.00	0.02	−0.01
23. Social Ties w/Relatives	−0.05	−0.02	0.00	−0.01	−0.02	−0.04	−0.10*
Mean	1.59	1.49	1.53	1.45	1.44	1.52	0.48
Standard Deviation	0.76	0.67	0.72	0.58	0.50	0.62	0.36

Variable	8	9	10	11	12	13	14	15
8. Internalizing Symptoms (T2)	--
9. Internalizing Symptoms (T3)	0.70***	--
10. Externalizing Symptoms (T1)	0.49***	0.41***	--
11. Externalizing Symptoms (T2)	0.70***	0.56***	0.70***	--
12. Externalizing Symptoms (T3)	0.45***	0.71***	0.56***	0.68***	--
13. Adolescent Age	0.10*	0.08	0.15**	0.11**	0.09	--
14. Adolescent Gender: Male	−0.22***	−0.22***	0.00	0.01	0.04	−0.01	--
15. Moderate Latino/a Concentration	0.02	0.05	0.07	0.06	0.09*	0.02	−0.02	--
Mean	0.52	0.51	0.30	0.34	0.33	12.80	--	--
Standard Deviation	0.39	0.39	0.24	0.28	0.28	1.03	--	--

Variables	16	17	18	19	20	21	22	23
16. High Latino/a Concentration	--
17. HH Structure: Single Parent	0.01	--
18. HH Structure: Stepparent	0.03	−0.19***	--
19. Mother’s Education	−0.40***	0.03	−0.03	--
20. First Generation Immigrant	−0.02	0.00	0.12**	0.05	--
21. Second Generation Immigrant	0.26***	−0.06	−0.08	−0.30***	−0.53***	--
22. Spanish Language Enculturation	0.21***	−0.04	0.01	−0.28***	0.10*	0.22***	--
23. Social Ties w/Relatives	0.05	−0.00	−0.00	−0.03	0.12**	−0.04	0.21***	--
Mean	--	--	--	--	3.15	--	3.47	3.01
Standard Deviation	--	--	--	--	1.64	--	0.91	1.56

Note. Unless otherwise specified, all variables were measured at T1

Latino/a Concentration = Concentration of Latino/a student in youth’s school; HH structure = Household structure

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Examining raw scores for internalizing symptoms at T3, we found that 66.9% of the sample fell into the normal range, 21.9% were within the clinical range, and 11.2% were borderline. For externalizing symptoms at T3, 83% were within the normal range, 10.7% were within the clinical range, and 6.1% were borderline.

Measurement Model

The measurement model fit the data well: RMSEA (95% CI) = 0.05 (0.04–0.05); CFI = 0.97; SRMR = 0.05; χ²(333) = 695.08, p < 0.001. All indicators had significant (p < 0.001) loadings on respective latent constructs, and all standardized loading values fell above 0.70. Multi-group CFA indicated invariance in the measurement model across both gender and time.

Structural Model: H1

In our test for equilibrium of the variances and covariances of the latent construct, the freely estimated model fit the data better than the constrained model, suggesting possible non-equilibrium in variances and covariances over time. However, the constrained model did fit the data well (RMSEA = 0.05, 95% C.I: 0.04–0.05; CFI = 0.96; SRMR = 0.06; χ²(325) = 739.31, p < 0.001), and differences in the CFI between the freely estimated and constrained models was 0.01, and the constrained model RMSEA was outside the 95% confidence interval of the freely estimated model by only 0.001. As equilibrium of variances and covariances is not considered essential for testing longitudinal mediation (Little, 2013), we continued with steps needed to test mediation. We began by estimating a model including all cross-lagged and autoregressive pathways, finding significant pathways from T1 internalizing and externalizing symptoms to T2 deviant peer affiliation. Our next model included only pathways consistent with our mediation hypothesis, with estimates of the indirect effects. We observed a significant indirect effect, providing support for our hypothesized mediation process. We then estimated our final mediation model, which included the mediation pathways, all autoregressive paths, and demographic covariates.

Results from the final mediation model testing the indirect effect of T1 peer discrimination on T3 adjustment outcomes, through T2 deviant peer affiliation are shown in Table 3 and Figure 2 (see Table S2 for full model results). Fit statistics indicated acceptable model fit: RMSEA = 0.04, 95% C.I. = 0.04–0.05; CFI = 0.94; SRMR = 0.06; χ²(616) = 1258.94, p < 0.001. Results indicated a significant positive indirect effect from T1 peer discrimination to T3 externalizing symptoms, through T2 deviant peer affiliation (β = 0.05; SE = 0.02; B = 0.02; 95% CI = 0.01, 0.09). The direct effect of T1 peer discrimination on T3 outcomes was non-significant, consistent with indirect-only mediation (Zhao et al., 2010). Specific direct effects revealed that T1 peer discrimination was associated with a significant T1 to T2 increase in affiliation with deviant peers; T2 deviant peer affiliation, in turn, was associated with a significant T2 to T3 increase in externalizing symptoms. We did not observe a significant effect of T1 peer discrimination on T3 internalizing symptoms directly, or indirectly through affiliation with deviant peers at T2. Model results also indicated that T1 internalizing symptoms were associated with significant decreases in T2 affiliation with deviant peers (β = −0.19; SE = 0.06; B = −0.27; 95% CI = −0.29, −0.08) and that T1 externalizing symptoms were associated with significant increases in T2 affiliation with deviant peers (β = 0.30; SE = 0.09; B = 0.66; 95% CI = 0.14, 0.45). Finally, the only significant demographic finding was that males reported a greater decrease in internalizing symptoms over time than did females (β = −0.07; SE = 0.03; B = −0.05; 95% CI = −0.12, −0.01). Results from the multi-group structural models indicated invariance across gender (Supplemental Table S1).

Table 3.

Standardized Path Estimates from Mediation Model: Direct and Indirect Effects

	Pathways	β (SE)	95% CI
Direct Effects	T1 Peer Discrimination → T3 Externalizing Symptoms	0.06 (0.04)	−0.01, 0.13
	T1 Peer Discrimination → T3 Internalizing Symptoms	0.03 (0.04)	−0.04, 0.10
Indirect Effects	T1 Peer Discrimination → T2 Deviant Peers → T3 Externalizing Symptoms	0.05* (0.02)	0.01, 0.09
	T1 Peer Discrimination → T2 Deviant Peers → T3 Internalizing Symptoms	0.02 (0.02)	0.00, 0.04

^*Note. p < 0.05

Figure 2.

Standardized Regression Coefficients from Mediation Model: T1 Peer Discrimination, T2 Deviant Peer Affiliation, and T3 Symptomology. Note. Non-significant paths (p > 0.05) shown with dashed lines. All variables correlated within time and parcel residuals correlated across time; arrows not shown to improve legibility. T3 internalizing and externalizing symptoms regressed on adolescent age, gender, immigration status, school Latino/a concentration, household structure, and mother’s educational attainment. Model fit statistics: RMSEA (95% CI) = 0.04 (0.04–0.05); CFI = 0.94; SRMR = 0.06; χ2 (616) = 1258.94, p < 0.001. * p < 0.05, ** p < 0.01, *** p < 0.001.

Structural Model: H2

Results from models testing interactions indicated no statistically significant moderating effects of culture-specific factors (i.e., Spanish language enculturation; social ties with the family’s country of origin) on the associations between peer ethnic discrimination and adolescent internalizing and externalizing symptomology.

Alternate Models

To further explore the lack of significant links between T1 peer discrimination and T2 symptomology, two sets of alternative models were run. The first set of models considered that the lack of this T1 to T2 effect may have stemmed from the short time lag (roughly 6 months) between surveys. Thus, we estimated a structural model with pathways from peer discrimination at T2 to outcomes at T3, also capturing a 6-month time lag. As those results indicated that T2 discrimination was positively associated with T3 externalizing symptoms at T3 (β = 0.23; SE = 0.08; B = 0.10; 95% CI = 0.10, 0.36), there was no evidence to suggest that time lag explained the lack of a significant T1 to T2 association between these variables. The second set of models considered that the lack of significant T1 discrimination to T2 symptomology paths was due to adolescents’ age. Thus, we ran multiple group structural models comparing pathways for those under age 13 to those over age 13 at baseline. Comparisons of model fit between groups, however, revealed no statistically significant differences in associations between peer discrimination, deviant peer affiliation, and symptomology. Thus, there was no evidence to suggest that either age or survey time lag explained the lack of a statistically significant association between T1 peer discrimination and T2 symptomology.

Discussion

Ethnic discrimination from peers may be especially meaningful during the adolescent years. During this time, young people navigate the developmental task of identity formation within peer groups and increasingly rely on their peers for support and approval (Kerr et al., 2003). Consistent with García Coll’s integrative model of minority child development and with research on general discrimination (e.g., Benner et al. 2018) and on peer rejection (e.g., Waldrip et al., 2008), peer ethnic discrimination is associated with increases in externalizing symptoms among the Latino/a adolescents in this study. Advancing ideas about potential mechanisms for this association, youth’s increased affiliation with deviant peers appears to be an intervening factor explaining discrimination impacts on externalizing symptoms. Contrary to expectations and some research (e.g., Delgado et al., 2019), peer discrimination was not related to changes in internalizing symptoms. Contrary to expectations, there is no evidence that enculturation –adolescent Spanish language use and communication with relatives in the country of origin – buffers adolescents from the harmful effects of peer discrimination on symptomology.

The present study focused on discrimination experiences perpetrated by peers at school, a critical context of adolescent socialization (Kerr et al., 2003). In new immigrant areas, where schools and youth are less familiar with, and accepting of, ethnic diversity (Natanson et al., 2020), experiences of peer ethnic discrimination may be especially harmful to Latino/a adolescents. Consistent with prior research, our findings suggest that ethnic discrimination by peers may represent a form of peer rejection, precipitating deviant peer affiliation as to cope with loneliness and/or to bond with young people who share similar negative sentiments toward peers who rejected them (Brendgen et al., 2000; Kaplan, 1980). As in other studies (e.g., Germán et al., 2009) and consistent with the integrative model of child development, the present study’s results suggest that peer discrimination may filter youth into “inhibiting” peer environments – those unsupportive of positive development – and, in turn, increase adolescent risks for problem behaviors. Problem behavior theory recognizes peer engagement in deviant behavior as a critical precursor to adolescent problem behaviors (Jessor & Jessor, 1977), and social cognitive theory highlights observational learning and behavioral modeling as important to the development of new behaviors (Bandura & Walters, 1977). These theories thus help explain the mediating role of deviant peer affiliation shown in the current study.

In contrast to our hypothesis and to research on adolescent depressive symptoms (Delgado et al., 2019; Stein et al., 2019), our results do not support the idea that peer ethnic discrimination directly, or indirectly through deviant peer affiliation, impacts Latino/a adolescents’ internalizing symptoms. Our findings may not be comparable to those from prior research due to our assessment of a broad range of internalizing symptoms, as opposed to depressive symptoms only. We also speculate that this study’s 18-month time period may have been too brief to capture effects of discrimination on changes in internalizing symptoms. Prior evidence for links between discrimination and Latino/a adolescent depressive symptoms has been shown in studies following youth over a period of several years (e.g., Delgado et al., 2019). Further, null findings for internalizing symptoms may be due to the present study’s focus on a new immigrant area. The greater cultural inclusion of Latino/as in longstanding immigrant areas may confer greater harm from peers’ ethnic exclusion and mistreatment on internalizing symptoms than is the case in areas where peer discrimination is more normative and/or expected. It also is interesting that our research did not indicate a link between deviant peer affiliation and changes in adolescent internalizing symptoms, which has been shown among adolescents of non-Latino origins (Connell & Dishion, 2006; Hong et al., 2019; Klostermann et al., 2016). Given the centrality of family to Latino/a adolescent development (Lawton & Gerdes, 2014), Latino/a adolescents’ internalizing symptomology may be more responsive to experiences occurring in the family as opposed to peer context. For instance, higher levels of parental acceptance and support are strong correlates of declines in internalizing symptoms for Latino/a adolescents (Zeiders et al., 2015).

Our inclusion of a demographically diverse population of Latinos/as in a new immigrant destination during a time of heightened anti-immigrant sentiment in the U.S. suggests the importance of the larger geographic and historical context for understanding risks to outward-directed problem behaviors (i.e., externalizing) for a pan-Latino/a sample of adolescents. Unlike the findings shown for Mexican-origin adolescents surveyed from 2002 to 2013 and living in Arizona, an area with a longstanding presence of Mexican-origin residents (Delgado et al., 2019; White et al., 2018), we find that peer ethnic discrimination relates to increases in externalizing symptoms, but not to changes in internalizing symptoms. It may be that Latino/a adolescents engage with anti-social peers to cope with feelings of rejection in hostile immigrant environments. Perceptions of a community as less open and accepting of immigrant populations have been associated with greater risk taking and aggressive behavior among Latino/a youth (Forster et al., 2015). Future research should seek to identify community-level protective factors, such as co-ethnic concentration (White et al., 2018) and/or social support (Rivas‐Drake, 2011), that might buffer Latino/a adolescents from harm conferred by discrimination and/or deviant peer affiliation.

This study’s prospective study design is an important strength. Bidirectional relationships between deviant peer affiliation and externalizing symptoms were consistent with prior research (e.g., Mrug et al., 2004) and with the social interaction/enhancement model, which posits that engagement in deviant behavior increases youth’s affiliation with deviant peers, elevating youth’s risks for later conduct problems (Chen et al., 2015). Unlike prior research findings (e.g., Connell & Dishion, 2006), the present study indicates that internalizing symptoms may lead to declines in later deviant peer affiliation. Adolescents with increased symptomology may face unique challenges with regard to social skills and interpersonal relationships, leading to fewer interactions with peers, including those who are deviant (Parker et al., 2006). Future research should investigate the different ways in which adolescent symptomology affects peer selection.

Contrary to expectations, the present study did not indicate a protective role of adolescent enculturation in mitigating harm conferred by peer ethnic discrimination for increased symptomology. This may be due to the powerful role that peers play in the lives of youth (DiClemente et al., 2009). It also is possible that protective factors examined in the current study are not powerful enough to buffer young people from the harmful experience of being discriminated against by classmates. Given the centrality of family in the lives of many Latino families (Rivera et al., 2008; Stein et al., 2014), future research should investigate family-based strengths as potential sources of resilience.

There are a few study limitations to note. First, this research did not assess personal characteristics that might act as confounders and/or modifiers in the associations between discrimination, peer affiliation, and adjustment outcomes. For example, adolescent temperament and resilience are important individual-level factors relevant to discrimination experiences, peer affiliation, and adolescents’ adjustment (García Coll et al., 1996; Roosa et al., 2011). Second, all variables were assessed using youth self-report, an issue that may be especially limiting for our understanding of deviant peer affiliation since adolescents tend to perceive greater similarity between themselves and their peers than may be true in reality (Mrug et al., 2004). Our measure of deviant peer affiliation, therefore, may reflect descriptive norms, or youth’s perception of how their peers behave, as opposed to actual peer behavior. It is important to consider, however, that descriptive norms demonstrate strong associations with youth risk behavior (Lapinski & Rimal, 2005). Third, it is not known whether the effects of discrimination observed in this study differ from those experienced prior to 2017 or during times of less overtly hostile immigrant environments. Research does suggest, however, that discrimination may be a particularly salient stressor for today’s Latino/a young people (Huang & Cornell, 2019). Relatedly, we also were unable to determine if this new immigrant area is less receptive to immigrant populations than more established immigrant destinations. A fourth limitation concerns the insufficient number of individuals from the diverse country of origins to facilitate analyses of important within-group differences. Finally, despite the strengths of auto-regressive models for testing longitudinal associations, some researchers have suggested that this method limits the ability to assess for within-person differences. Thus, future research using alternate methods such as the random intercepts cross-lagged panel regressive model is warranted (Berry & Willoughby, 2017; Hamaker et al., 2015).

Despite these limitations, this study is characterized by several strengths. First, we employed a rigorous multi-step process for testing statistical mediation using three time points of data, increasing support for a causal process. Second, the signficant mediating pathway linking Time 1 ethnic discrimination to changes in deviant peer affiliation at Time 2 and, in turn, to changes in externalizing behaviors at Time 3 held even after accounting for a significant path from Time 1 externalizing symptoms to Time 2 deviant peer affiliation. Finally, the use of probability-based sampling techniques to select adolescent participants strengthens the generalizability of this study’s findings to Latino/a adolescents living in other new immigrant areas in the U.S.

Conclusions

The current political climate surrounding issues related to immigration highlights the need to better understand risk and protective factors for Latino/a adolescents, one of the fastest growing populations in the U.S. This study addresses gaps in the current literature by using longitudinal data to examine the mechanism through which peer ethnic discrimination affects the adjustment of Latino/a adolescents in a new immigrant area, where experiences of discrimination may be particularly salient. Future research is needed to assess factors that may protect Latino/a adolescents from the harms conferred through peer discrimination. It also is important to examine how these associations differ across characteristics such as skin color, country of origin, migration experience, and immigration status given the immense heterogeneity of the U.S. Latino population and variations in risk and reliance factors across Latino subgroups (Del Toro & Hughes, 2019; Fuentes et al., 2021; Umaña-Taylor & Fine, 2001). Given that Latino populations have been disproportionately impacted by the COVID-19 pandemic and recent research indicated elevated risks to Latino/a adolescent adjustment as a result of the pandemic (author cite), there also is a need to consider experiences of discrimination amid the pandemic (Podewils et al., 2020; Riley et al., 2021; Vargas & Sanchez, 2020). Findings from the current study confirm prior research suggesting pernicious effects of discrimination on the adjustment of Latino/a adolescents. Importantly, the present study also identifies deviant peer affiliation as a potentially important mechanism through which this effect occurs. This research also highlights areas for future research to better understand this mechanism to inform the development of interventions aimed at improving the health and well-being of contemporary Latino/a adolescents.

Table 2.

Standardized Factor Loadings and Standard Errors from Measurement Model

Indicator Variable Parcel	T1	T2	T3
	β (SE)
Peer Discrimination
Parcel 1: call you a name that is bad or insulting to Latino/as + say bad things about Latino/as	0.91 (0.03)	0.83 (0.03)	0.88 (0.03)
Parcel 2: threaten or harass you + exclude you from activities	0.75 (0.03)	0.76 (0.03)	0.77 (0.03)
Deviant Peer Affiliation
Parcel 1: stole something worth less than $50 + used marijuana + stole something worth more than $50	0.91 (0.01)	0.87 (0.02)	0.88 (0.01)
Parcel 2: damaged or destroyed property + used alcohol + sold hard drugs	0.91 (0.01)	0.87 (0.02)	0.95 (0.01)
Parcel 3: cheated on school tests + started a fight + broke into a vehicle or building	0.79 (0.02)	0.78 (0.02)	0.83 (0.02)
Internalizing Symptoms
Parcel 1: cry a lot + afraid to go to school + afraid I might think or do something bad + feel that I have to be perfect + feel that no one loves me + feel worthless or inferior + am nervous or tense + too fearful or anxious + feel too guilty + am self-conscious or easily embarrassed	0.93 (0.01)	0.91 (0.02)	0.91 (0.02)
Parcel 2: there is very little that I enjoy + would rather be alone than with others + refuse to talk + am secretive or keep things to myself + am too shy or timid + don’t have much energy + am unhappy, sad, or depressed + keep from getting involved with others	0.79 (0.02)	0.82 (0.02)	0.82 (0.02)
Parcel 3: have nightmares + feel dizzy or lightheaded + feel overtired without good reason + have aches and pains + have headaches + have nausea or feel sick + have rashes or other skin problems + have stomachaches + vomit or throw up	0.70 (0.02)	0.72 (0.02)	0.74 (0.02)
Externalizing Symptoms
Parcel 1: don’t feel guilty after doing something I shouldn’t + break rules + hang around kids who get in trouble + lie or cheat + would rather be with older kids + run away from home + set fires + steal at home + steal from places other than home + swear and use dirty language	0.82 (0.02)	0.84 (0.02)	0.88 (0.02)
Parcel 2: argue a lot + am mean to others + try to get a lot of attention + destroy my things + destroy things belonging to others + disobey my parents + disobey at school + get in many fights + physically attack people + scream a lot + am stubborn + my moods or feelings change quickly + am suspicious + have trouble sitting still	0.83 (0.02)	0.82 (0.02)	0.87 (0.02)
Spanish Language Enculturation
Parcel 1: I speak Spanish + I enjoy speaking Spanish	0.82 (0.02)	--	--
Parcel 2: I enjoy Spanish language TV + I enjoy Spanish language movies + my friends are Latinos	0.84 (0.02)	--	--
Parcel 3: I enjoy reading in Spanish + my thinking it done in the Spanish language	0.81 (0.02)	--	--

Note. Model Fit: RMSEA (95% CI) = 0.05 (0.04–0.05); CFI = 0.97; SRMR = 0.05; χ²(333) = 695.08, p < 0.001