Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Aug 1.
Published in final edited form as: J Adolesc. 2020 Jul 22;83:42–51. doi: 10.1016/j.adolescence.2020.07.006

Multigroup Ethnic Identity Measurement Invariance Across Adolescence and Diverse Ethnic Groups

Sycarah Fisher 1,a, Tamika B Zapolski 2,b, Lorey Wheeler 3,c, Prerna G Arora 4,d, Jessica Barnes-Najor 5,e
PMCID: PMC7906292  NIHMSID: NIHMS1615978  PMID: 32711160

Abstract

Introduction

Ethnic identity is an important protective factor for various ethnic groups and developmental periods. Although existing measures assessing ethnic identity are well known, less is known about the measurement invariance of the Multigroup Ethnic Identity Measure (MEIM) across adolescent ethnic groups. The present study evaluates the factor structure of MEIM (Roberts et al., 1999) and tests the measurement invariance across early and middle adolescence and ethnic background (N = 4,940).

Methods

Data from an ethnic minority sample of youth (54% girls; Mage = 13.88, grades 6th – 12th; 60% African American, 22% multi-ethnic, 8% Latinx, 5% Asian, 5% American Indian) in the United States of America were examined using confirmatory factor analysis (CFA) and multi-group measurement invariance via a structural equation modeling (SEM) framework. Models for invariance were tested using full information maximum likelihood-robust (FIML-R) estimation in Mplus.

Results

CFA supported a three-factor solution (i.e., cognitive clarity, behavioral engagement, and affective pride). The model indicated scalar invariance across early and middle adolescence and partial scalar invariance across the five self-identified racial/ethnic minority groups. There were no grade differences on the ethnic identity factors. Among the racial/ethnic groups, multi-ethnic youth reported the lowest levels on all three ethnic identity factors compared to the other groups.

Conclusions

Results of this study point to the validity of using the MEIM for meaningful comparisons of ethnic identity across ethnic groups and across early and middle adolescence. Implications for the interpretation and use of this measure with diverse adolescents are discussed.

Keywords: Ethnic Identity, Measurement, Diversity, Development, Adolescence

Ethnic identity development is a major task during adolescence (Phinney & Chavira, 1995; Quintana et al., 2006; Yip et al., 2006). Researchers conceptualize ethnic identity as an individual’s self-concept derived from their knowledge of their membership within an ethnic group and the emotional significance attached to that membership (Tajfel, 1981). Conceptualizations of ethnic identity development originate from social identity theory (Tajfel & Turner, 1986) and Erikson’s theory of development (1968). Specifically, social identity theory postulates that having a group identity enhances one’s self-concept and leads to a sense of belonging. Thus, the “strength and valence” of one’s connection to their ethnic group is a salient aspect of ethnic identity (Roberts et al., 1999, p. 303). Through the exploration of one’s ethnic group, adolescents may choose to align themselves with a cultural group.

One of the most widely used measures of ethnic identity is the Multigroup Ethnic Identity Measure (MEIM), developed by Phinney (1992) to measure levels of ethnic identity development. The original measure consisted of 14 items and three subscales: ethnic affirmation/belonging, ethnic behaviors, and ethnic identity achievement. The ethnic affirmation subscale captures ethnic ‘pride’ and one’s positive sense of belonging to an ethnic group, as well as feelings of attachment. The ethnic behaviors subscale captures involvement in social activities with individuals from similar ethnic backgrounds or cultural practices. Finally, the ethnic identity achievement subscale assesses an individual’s security in their sense of self and is the culmination of the ethnic identity development process.

Studies examining ethnic identity using the MEIM within and across ethnic groups, as well as across the early adolescent to late adolescent developmental period, find ethnic identity plays a major role in several adolescent health-related outcomes including self-esteem, depression, and substance use (e.g., Fisher et al., 2017; Rivas-Drake et al., 2014, Smith & Silva, 2011; Zapolski et al., 2017). Despite the widespread use of the MEIM across various racial/ethnic groups and developmental periods, limited research has examined if the measurement structure is invariant (i.e., works in the same way) across racial/ethnic groups during varying adolescent periods.

Factor Structure of the MEIM

The original MEIM development study (Phinney, 1992) conducted reliability and exploratory factor analyses with a diverse sample of high school and college students, with mean scores compared across ethnic groups to determine whether differences were statistically significant. The results of this study indicated the MEIM could be thought to have a 1-factor structure, a finding supported by several subsequent studies (Avery et al., 2007; Ponterotto et al., 2003; Worrell, 2000; Worrell et al., 2006). However, others have found contradictory factor structures with some support for 2-factor, 3-factor, and bifactor models.

For instance, findings from a study by Roberts and colleagues (1999), provided evidence that a 2-factor structure fit better than a single factor structure based upon exploratory (EFA) and confirmatory (CFA) factor analyses. Their findings also suggested the removal of two negatively worded items that did not fit within the 2-factor structure. The result was a 12-item MEIM scale consisting of two factors: exploration and commitment. This two-factor structure has also been supported in subsequent research studies (Feitosa et al., 2017; Phinney & Ong, 2007; Spencer et al., 2000)

Lee and Yoo (2004) examined the factor structure of the MEIM with 323 Asian American late adolescents/adults (Mage = 19.72) using EFA. Results of their examination found ‏evidence for three factors: ethnic identity cognitive clarity, ethnic identity ethnic pride, and ethnic identity behavioral engagement. Phinney and Ong (2007) postulated that these three factors roughly map onto a 2-factor structure with behavior engagement mapping onto the exploration factor and clarity and pride together mapping onto the commitment factor in the 2-factor model.

More recently, research conducted by Yap and colleagues (2014; 2016) with a sample of college students found evidence for a bifactor model. Instead of having items load on one factor or a general overall factor, the bifactor model allowed items to load on a general ethnic identity factor as well as one of the two factors supported by Roberts and colleagues (1999): exploration and commitment. Although the authors ultimately identified the bifactor model as the best model for their data, the bifactor model and 3-factor model fit the data similarly. Given the similarity in fit, they endorsed the bifactor model as it fit better conceptually with the literature.

The majority of the research on the MEIM factor structure was conducted using EFA (Lee & Yoo, 2004; Phinney, 1992; Reese et al., 1998; Worrell, 2000; Worrell et al., 2006). The use of EFA, as opposed to CFA, allowed for the proliferation of a multitude of factor structures, contributing to the confusion of the factor structure of the MEIM. As such, Phinney and Ong (2007) stated confirmatory factor analyses need to be conducted to test competing models for the MEIM while examining relative model fit. Thus, the first aim of the current paper is to use CFA to test competing factor structures and identify the factor structure with the best fit for the current sample.

Measurement Invariance across Racial/Ethnic Groups

To determine the ability of a particular scale to be used across different groups, additional analyses to determine the invariance of the measure, or measurement invariance, are necessary. Scholars have identified four levels of measurement invariance that can be utilized to assess differences across groups (Chen et al., 2005; Yap et al., 2014). The most commonly used and least stringent, is configural invariance, which examines the factor loadings of the measure items to determine whether the measure is best characterized by a one, two, or three-factor structure across various groups. The second is metric invariance, which determines whether the factor loadings for the items are equivalent across groups. As noted by Yap et al. (2014), this form of measurement invariance is necessary for comparing scale correlations across groups but is not adequate for comparing scale means across groups. To examine whether groups can be justifiably compared on the same measure, an investigation of scalar invariance, which involves testing whether item intercepts are equivalent across groups, is also needed. Lastly, strict invariance involves testing the residual variances across groups. Evidence of strict invariance allows for group comparisons on observed variances and covariances (Gregorich, 2006).

Many studies examining the psychometric properties of the MEIM among adolescent samples have not examined or have been unable to find evidence of scalar invariance of the MEIM across racial/ethnic groups. This step is necessary for the comparison of MEIM scores across groups. Several studies, including the seminal study conducted by Phinney (1992) found significant differences in the ethnic identity scores across ethnic groups in their high school sample. All minority groups (i.e., Asian, Black, and Multi-ethnic) had significantly higher ethnic identity scores than White youth, with no difference in scores among themselves. This original study, however, did not examine whether the factor structure was invariant across ethnic groups.

The first study to examine measurement invariance of the MEIM across racial/ethnic groups was conducted by Roberts et al. (1999) with a sample of 5,423 students in 6th to 8th grade. They were unable to confirm equivalence of the factor loadings across racial/ethnic groups, calling for the need to examine measurement invariance of the MEIM across racial/ethnic groups among adolescent populations. Subsequent studies evaluating the measurement invariance of the MEIM have focused on adults or college students (Avery et al., 2007; Yap et al., 2014, Yap et al., 2016). For instance, Avery and colleagues (2007) examined measurement invariance in a sample of 1,349 adults (Mage = 22.85), using CFA to confirm a single factor structure. After conducting measurement invariance analyses, they found evidence of configural, but not scalar invariance.

Another study by Yap and colleagues (2014) involved 9,756 college-aged participants (Mage = 20.30). CFA was used to identify the baseline structure upon which to base the tests of invariance. Results indicated that a bi-factor model was the best fit; however, little evidence for overall scalar or partial scalar invariance was found. The authors concluded that although the MEIM may be used for studying correlates of ethnic identity in diverse groups, it should not be used for making mean-level comparisons across groups.

Yap and colleagues (2016) also conducted a follow-up study examining the measurement invariance of the MEIM across foreign-born, second-generation, and later-generation college students in the United States. The results of this study found configural and metric invariance, similar to Yap et al. (2014). However, there was no evidence of scalar invariance across generational status. These findings suggested the MEIM should not be used to compare ethnic groups across generational status with adult populations. Even though these findings may give pause to the use of the MEIM with adult populations, given the developmental nature of ethnic identity (Umana-Taylor et al., 2014), additional studies are necessary to determine if these same relationships exist during adolescence.

While studies finding scalar invariance with U.S. adolescent populations are limited, international studies of the MEIM are promising. For example, Mastrothodoros and colleagues (2012) examined measurement invariance in a sample of Bulgarian, Dutch, and Greek adolescents. The results of the study found that the MEIM demonstrated scalar invariance with these populations. Further, Musso et al. (2017) examined the measurement invariance of the MEIM-R with a sample of Italian, East European, and North African adolescents. They found full measurement invariance for the commitment subscale but partial scalar invariance for the exploration subscale. Given the limited work in this area with U.S adolescents, the second aim of the current study is to examine measurement invariance of the MEIM among a large and diverse sample of U.S. racial/ethnic minority adolescents.

Measurement Invariance across Early and Middle Adolescence

While limited research exists regarding the measurement invariance of the MEIM across racial/ethnic groups, to our knowledge, no research has been conducted on the measurement invariance of the MEIM across differing ages or developmental periods. Longitudinal research on ethnic identity development using the MEIM finds differences in trajectories of ethnic identity over time. For example, Pahl and Way (2006) modeled developmental trajectories of ethnic identity in a sample of 135 Black and Latino adolescents (Mage = 15.1). The results of their growth curve modeling following students from 10th grade to one year after 12th grade found differences in exploration and affirmation over time. Specifically, ethnic identity exploration peaked in middle adolescence, but no growth pattern was found in affirmation. Another study conducted by French and colleagues (2006) with a sample of 420 adolescents (early adolescents, Mage = 11.28; middle adolescents, Mage = 14.01) found group-esteem rose for both early and middle adolescents. Early adolescents were the only group that saw a rise in exploration.

Given that ethnic identity develops over time, it is plausible the MEIM may tap into different parts of ethnic identity across adolescence. Phinney (1992) alluded to potential measurement differences in her article, highlighting stronger reliability scores for older participants in the sample compared to those for younger participants. Phinney and Ong (2007) reiterated this point in a subsequent study stating, “it is also possible that certain MEIM items capture aspects of ethnic identity that are likely to be differentially endorsed by early, middle, or late adolescents.” (p. 279). Furthermore, in their article on methodological issues in ethnic identity research, Swartz and colleagues (2014) joined other researchers (i.e., Avery et al., 2007; Phinney & Ong, 2007) in calling for the evaluation of the measurement of MEIM across different developmental periods, citing the potential necessity for different measurement approaches given the developmental changes that occur across adolescence. Accordingly, the third aim of the present study is to examine measurement invariance of the MEIM across early and middle adolescence.

Present Study

Taken together, there has been limited work conducted that examines the invariance of the MEIM across ethnic minority groups and developmental periods (i.e., early adolescence versus middle adolescence) among U.S. adolescent populations. Further, among available studies, less stringent tests of invariance have been utilized (Ponterrotto et al., 2003; Roberts et al., 1999, Yancey et al., 2001). Based on findings from young adult studies that have utilized more stringent tests of invariance, there is evidence to suggest comparisons across ethnic groups based on MEIM scores may not be valid (Yap et al., 2014; Yap et al., 2016). Such findings call into question the validity of the wide-spread use of the MEIM to compare ethnic groups (Carter et al., 2018; Dandy et al., 2008; Gummadam et al., 2016; Rahim-Williams et al., 2007; Townsend et al., 2020), and provide the impetus to further examine the measurement structure and invariance in an adolescent population. The identification of the measurement invariance of the MEIM across ethnic groups and developmental periods is critical to ensure the appropriate use of the measure as well as the validity of the results of studies that use the measure in this way. Thus, the overall goal of the present study is to fill an important gap in the assessment literature by examining the factor structure and measurement invariance of the MEIM using stringent tests of invariance among a large and diverse sample of ethnic minority adolescents in the U.S. We first examine the factor structure comparing alternative models based on prior work (i.e., 1-factor structure, Phinney, 1992; 2-factor structure, Roberts et al., 1999; 3-factor structure, Lee & Yoo, 2004; bifactor model, Yap et al., 2014). Second, we examine measurement invariance by ethnic group, and third, we examine measurement invariance by developmental period. Based on the limited work in this area with U.S. samples, no a priori hypothesis of the invariance of the MEIM scores across groups was examined. If measurement invariance is found, given the large sample of youth from diverse ethnic minority backgrounds, an additional aim is to compare the factor means across the ethnic/racial and developmental groups.

Methods

Participants and Procedures

The current study involves participants drawn from a larger 5-year study examining school and health behavior outcomes among students between 4th and 12th grade. As approved by the University IRB, participants were sampled from 159 schools (21 school districts) in a large Midwestern county. Informed consent forms were sent home to parents of potential participants. Signed consent forms were obtained from 50% of parents (approximately 12,000) each year of the parent study. Given the purpose of the original study was to examine school and health behaviors on an annual basis rather than measure changes across time, retention rates across each data collection were modest with less than one-third of the participants (27.7%) completing two waves of data. For the current study, all participants who provided data at least once on the study variables of interest were retained for a final sample size of 4,940 youth. Among this sample, most of the participants self-identified as African American (60.2%), followed by multi-ethnic (22.2%), Latinx (8.0%), American Indian (4.8%), and Asian (4.8%). Participants were equally divided based on gender (53.7% female). Youth were on average 13.88 years of age (SD = 1.65). For purposes of the invariance analyses, we grouped youth based on school/developmental status, middle school (grades 6–8; early adolescence; 68.6%; Mage = 13.00, SD = 1.02) versus high school (grades 9–12; middle adolescence; 31.4%; Mage = 15.79, SD = 1.04). All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee, the 1964 Helsinki declaration, and its later amendments or comparable ethical standards.

Measures

Demographic Information

Participants were asked to indicate their gender, grade, birthdate, and racial/ethnic background (i.e., African American, American Indian, Asian, Latinx, and multi-ethnic). Participants were not allowed to select multiple categories for race/ethnicity although they had the option to identify as multi-ethnic.

Ethnic Identity

The Multigroup Ethnic Identity Measure (MEIM; Roberts et al., 1999) was used as a measure of ethnic identity that could be utilized across racial/ethnic groups. The MEIM is a 12-item scale designed to measure two components of ethnic identity: exploration and affirmation. For the exploration subscale, items included “In order to learn more about my ethnic background, I have often talked to other people about my ethnic group.” For the affirmation subscale, items included “I feel good about my cultural or ethnic background.” Items were rated on a 4-point Likert scale, with responses ranging from 1 (strongly disagree) to 4 (strongly agree). The total MEIM scale was used for the current study and had high internal consistency across all ethnic groups (α = 0.89–0.91), which is consistent with previous literature with reliability ranging from 0.81–0.92 (Ponterotto et al., 2003).

Data Analytic Approach

No data were missing for racial/ethnic group or developmental period. Data missingness of the 12-item MEIM ranged from 2.9% to 4.2%, and covariance coverage values ranged from 98% to 99%. Given the low amounts of missing data (less than the standard threshold of 5% missing) and high covariance coverage (greater than the recommended coverages of 80–90%), it was concluded that the estimates of the data analysis are trustworthy (Muthén, Muthén & Asparouhov, 2016). Therefore, Mplus full information maximum likelihood (FIML) estimation with a robust approach was used. To adjust for the nested nature of the data (i.e., adolescents nested within schools), we used the Mplus CLUSTER option to adjust for non-independence of observations by schools when computing standard errors and chi-squares (Muthén & Muthén, 1998–2017).

Given debates in the literature over the structure of the MEIM, our first aim was to use CFA to empirically identify the best baseline model for the overall sample. This included testing four potential MEIM factor structures in the overall combined sample with responses on the MEIM (N = 4,727) using Mplus 8.1 (Muthén & Muthén, 1998–2017). The four alternative factor structures were based on the literature (1-factor, Phinney, 1992; 2-factor structure, Roberts et al., 1999; 3-factor structure, Lee & Yoo, 2004; bifactor model, Yap et al., 2014). We started with the one-factor model, loading all items onto an overall ethnic identity factor. We then estimated a two-factor model following the specifications in the work by Roberts et al. (1999) reflecting either exploration (Items 1, 2, 4, 8, and 10) or commitment (Items 3, 5, 6, 7, 9, 11, and 12). In the three-factor model, based on work by Lee and Yoo (2004) in an ethnic minority sample, items were specified as indicators of correlated factors, cognitive clarity (Items 3, 6, and 7), affective pride (Items 5, 9, 11, and 12), and behavioral engagement (Items 1, 2, 4, 8, 10). Lastly, we estimated the bifactor model that isolates exploration and commitment from a general identity factor proposed by Roberts et al. (1999) and supported by Yap and colleagues (2014; 2016) with samples of college students. The bifactor model differs from a 2-factor model in that it allows items to load on both individual factors as well as an overall general factor.

To address our second and third aims, we used multi-group CFA to evaluate the measurement invariance of the MEIM scores across the five ethnic groups and two developmental periods evaluated in this study. This multi-group approach involves a series of steps to evaluate increasingly restrictive levels of measurement invariance (Millsap, 2012). These steps evaluate evidence for configural, metric, and scalar invariance. Starting with configural, this step involves specifying a single baseline factor structure across the various groups. The requirement is that the factor structure (number of factors, items loading onto factor) is similar across groups. Once this requirement is met, subsequent tests are of more rigorous forms of invariance. Metric invariance involves comparing the fit of the less restrictive configural invariance model against the fit of a more constrained nested model in which the factor loadings are constrained to be equal across groups. The absence of a meaningful decrement in model fit provides support for metric invariance (more parsimonious model). Scalar invariance is evaluated by adding constraints on manifest item intercepts and comparing model fit to the metric model. Finding the metric and scalar models do not differ meaningfully in model fit is taken as evidence in favor of scalar invariance. Though not required, we also tested for strict invariance (item error variances restricted), when evidence of scalar invariance was found.

For our last aim, we compared the factor means for the MEIM constructs to determine if there were significant differences across groups. We did this by comparing the final models from the invariance tests to models in which we constrained the means to be equal across groups (one by one for each construct). A meaningful decrease in model fit provided evidence of mean differences between groups.

We used the Maximum Likelihood-Robust (MLR) estimator based on simulation literature on factor analyses with ordered categorical indicators, suggesting the robustness related to parameter estimates, standard errors, and chi-square statistics (Yang-Wallentin et al., 2010). Model fit was assessed using the following indices, indicating “good fit” and “acceptable fit,” respectively in the parentheses: comparative fit index (CFI > .95, .90); root-mean-square error of approximation (RMSEA < .06, .08); standardized root-mean-square residual (SRMR < .08, .10), Akaike information criterion (AIC; smaller is better), and Bayesian information criterion (BIC; smaller is better; Hu & Bentler, 1999). For tests to determine the comparative fit of nested models within the context of measurement invariance, we followed recommendations that suggest moving beyond the chi-square statistic, given the well-known dependence on sample size, to use an alternative index, change in CFI in combination with RMSEA (Chen, 2007). Following these recommendations (Chen, 2007) and prior invariance work on MEIM (e.g., Yap et al., 2014), we used a change in CFI (i.e., ΔCFI) greater than .01 supplemented with a change in RMSEA greater than .015 (i.e., ΔRMSEA) to indicate a meaningful decrement in fit between invariance models (i.e., to indicate noninvariance); though Chen does recommend CFI be used as the main criterion because the RMSEA is affected by sample size and model complexity.

Results

Identifying the Factor Structure

Correlations among items and descriptive statistics for each manifest item are presented in Table 1. Examination of the response patterns for the items ruled out the presence of floor and ceiling effects; most response being 1 or 4, and zero cells. The items met skew and kurtosis normality cutoffs; skew ranged from −1.036 to .136 while kurtosis ranged from −1.052 to .525 (non-normality: skew > 2; kurtosis > 7; Curran et al., 1996). Goodness-of-fit indicators for each baseline measurement model are shown in Table 2. These data indicate there is less support for a one-factor model as compared to the two- and three-factor models compared here. The two-factor model that is most prevalent in the literature as presented by Roberts et al. (1999) does not fit as well as the three-factor model presented originally by Lee and Yoo (2004). The bifactor model as supported by work by Yap et al. (2014; 2016) had a worse fit than the other three models. Model fit was best for the three-factor model, suggesting there is empirical support with this diverse sample of ethnic minority adolescents. Thus, we used this factor structure in the invariance analyses.

Table 1.

Bivariate Correlations and Descriptive Statistics for Items on the Multigroup Ethnic Identity Measure across Full Sample

Item 1 2 3 4 5 6 7 8 9 10 11 12
1. Spent time finding out about my ethnic group (EG) -
2. Active in organizations or social groups .41 -
3. Clear sense of ethnic background .43 .37 -
4. How life affected by EG membership .37 .42 .40 -
5. Happy that I am a member of EG .29 .25 .46 .27 -
6. Strong sense of belonging to my own EG .34 .32 .46 .34 .58 -
7. Understand what EG membership means to me .36 .34 .57 .34 .58 .62 -
8. Talked to other people about EG .50 .39 .43 .42 .34 .42 .50 -
9. Pride in EG .32 .27 .50 .28 .62 .54 .60 .41 -
10. Participate in cultural practices .40 .41 .40 .37 .34 .41 .43 .46 .41 -
11. Strong attachment toward EG .37 .36 .49 .36 .51 .56 .60 .47 .63 .54 -
12. Feel good about cultural/ethnic background .33 .26 .47 .27 .57 .51 .56 .39 .64 .41 .62 -
M 2.46 2.32 2.86 2.50 3.19 2.95 2.94 2.56 3.15 2.66 2.91 3.20
SD .98 .99 .90 .98 .89 .93 .90 .97 .87 1.02 .92 .86
Open in a new tab

Note. N = 4,727. Mplus model estimated correlations and descriptive statistics. Items are paraphrased. See Roberts et al. (1999) for complete wording. All statistics as significant p < .001.

Table 2.

Fit Indices across Baseline Measurement Models of MEIM

Model χ2 (df) AIC BIC RMSEA [90% CI] SRMR CFI
One-factor 1916.24 (54) 128473.61 128706.21 .085 [.082, .089] .060 .883
Two-factor 992.66 (53) 127006.47 127245.53 .061 [.058, .065] .040 .941
Three-factor 646.42 (50) 126444.06 126702.50 .050 [.047, .054] .030 .963
Bifactor 2243.16 (45) 129016.47 129307.22 .102 [.098, .105] .254 .862
Open in a new tab

Note. N = 4,727. All chi-square values were statistically significant (p < .001). CFI = comparative fit index (“good fit” > .95; “adequate” > .90); RMSEA = root-mean-square error of approximation (“good fit” < .06; “adequate” < .08); CI = confidence interval; SRMR = standardized root-mean-square residual (“good fit” < .08; “adequate” < .10); AIC = Akaike information criterion; BIC = Bayesian information criterion. Bold indicates the model that had the best fit.

Invariance of the Three-Factor Model across Racial/Ethnic Groups

The configural model was evaluated and found to have good fit based on all but the RMSEA and CFI, which showed acceptable fit (Table 3). Comparisons of the configural versus metric invariance models suggested there was not a meaningful decrement in fit among these models, suggesting evidence for metric invariance. The comparison of the metric versus scalar invariance models, however, did show a meaningful decrement in fit in CFI, but not in RMSEA. Thus, relying on CFI as our primary indicator of fit, we systematically unconstrained each individual intercept to determine where the misfit between groups occurred. Results suggested that freeing the intercepts for Items 2, 4, 10, and 11 led to no meaningful decrement in fit between the metric and the partial scalar invariance models, suggesting these four items were responsible for the lack of scalar invariance. Thus, this partial set of constrained items were invariant at the scalar level across groups. This model resulted in factor loadings that were all significant and greater than .40 (Table 4).

Table 3.

Model Fit Statistics and Invariance Testing across Racial/Ethnic Groups and Developmental Stage

Model χ2 (df) AIC BIC RMSEA [90% CI] SRMR CFI ACFI ∆RMSEA
Racial/Ethnic Group Models
 Configural 1181.45 (255) 126277.18 127537.09 .062 [.058, .066] .041 .949
 Metric 1299.43 (303) 126262.68 127212.45 .059 [.056, .062] .057 .945 .004 .003
 Scalar 1762.61 (351) 126687.22 127326.86 .065 [.062, .068] .080 .922 .023 −.006
 Partial scalar1 1438.64 (335) 126339.01 127082.03 .059 [.056, .062] .071 .939 .006 .000
Developmental Group Models
 Configural 841.16 (102) 126647.49 127151.45 .055 [.052, .059] .037 .956
 Metric 876.03 (114) 126632.07 127058.50 .053 [.050, .056] .038 .954 .002 .002
 Scalar 896.06 (126) 126644.87 126993.76 .051 [.048, .054] .038 .954 .000 .002
 Strict 909.27 (138) 126696.16 126967.52 .049 [.046, .052] .044 .954 .000 .002
Open in a new tab

Note. N = 4,727. All chi-square values were statistically significant (p < .001). CFI = comparative fit index (“good fit” > .95; “adequate” > .90); RMSEA = root-mean-square error of approximation (“good fit” < .06; “adequate” < .08); CI = confidence interval; SRMR = standardized root-mean-square residual (“good fit” < .08; “adequate” < .10); AIC = Akaike information criterion; BIC = Bayesian information criterion. Criterion for ∆CFI ≥ 0.010 and ∆RMSEA ≥ 0.015 would indicate noninvariance (Chen, 2007).

1

Intercepts freed across groups for Items 2, 4, 10, 11.

Table 4.

Standardized Factor Loadings for Final Developmental and Racial/Ethnic Group Models

Racial/Ethnic Groups Developmental Groups
Item AA AI AS LX MU MS HS
Cognitive Clarity
 Item 3 .68 .69 .74 .70 .69 .69 .69
 Item 6 .73 .78 .79 .75 .74 .74 .74
 Item 7 .81 .81 .81 .83 .82 .81 .81
Affective Pride
 Item 5 .74 .72 .74 .76 .73 .74 .74
 Item 9 .82 .82 .85 .82 .81 .82 .82
 Item 11 .76 .80 .79 .81 .80 .77 .77
 Item 12 .79 .75 .77 .76 .76 .78 .78
Behavioral Engagement
 Item 1 .65 .65 .68 .65 .65 .64 .64
 Item 2 .59 .59 .56 .61 .59 .60 .60
 Item 4 .58 .59 .56 .60 .59 .59 .59
 Item 8 .72 .77 .73 .71 .74 .72 .72
 Item 10 .66 .68 .65 .65 .66 .66 .66
Open in a new tab

Note. AA = African American. AI = American Indian. AS = Asian. LX = Latinx. MU = mult-ethnic. MS = middle school (grades 6–8). HS = high school (grades 9–12).

Invariance of the Three-Factor Model across Early and Middle Adolescence

The configural model was evaluated and found to have good fit based on all but the RMSEA, which showed acceptable fit (Table 3). Comparisons of the configural versus metric invariance models suggested there was not a meaningful decrement in fit among these models, suggesting evidence for metric invariance. These constraints yielded a small increase in model fit based on the RMSEA. The comparison of the metric versus scalar invariance models also did not show a meaningful decrement in fit, suggesting evidence for scalar invariance. Given this evidence, we also tested for strict invariance. Again, model comparisons between the scalar and strict invariance models did not show a meaningful decrement in fit, suggesting strict invariance for developmental groups. The factor loadings were all significant and greater than .40 (Table 4).

Mean Differences by Group

In addressing the last aim of examining mean differences across groups, we found the multi-ethnic group means were not equal to the other four groups. The model with the multi-ethnic means unconstrained resulted in a model ΔCFI = .002 and ΔRMSEA = .000 as compared to the final partial scalar invariance model, suggesting equivalence. The constrained factor means for the four ethnic groups (i.e., African American, American Indian, Asian, and Latinx) were cognitive clarity M = 4.34 (SE = .09), affective pride M = 4.58 (SE = .11), and behavioral engagement M = 4.05 (SE = .07). The multi-ethnic group means were lower across all indicators: cognitive clarity M = 4.08 (SE = .09), affective pride M = 4.38 (SE = .10), and behavioral engagement M = 3.66 (SE = .08). We found no mean differences across developmental groups, ΔCFI = .000. Constrained factor means were cognitive clarity M = 4.26 (SE = .09), affective pride M = 4.53 (SE = .10), and behavioral engagement M = 3.91 (SE = .08).

Discussion

The present study sought to examine the factor structure and measurement invariance of the MEIM across ethnic minority groups and early and middle adolescence. The MEIM is one of the most widely used ethnic identity measures among both adolescent and adult populations (Rivas-Drake et al., 2014; Smith & Silva, 2011). Moreover, researchers have utilized the MEIM to compare ethnic identity across racial/ethnic (e.g., Anglin et al., 2012; Else-Quest & Morse, 2015; Zapolski et al., 2017) and developmental groups (French et al., 2006). However, given evidence of measurement non-invariance of the MEIM across racial/ethnic groups among young adult samples (Yap et al., 2014, 2016) and lack of examination of measurement invariance across adolescent development, further research to examine measurement invariance of the MEIM within adolescent populations was warranted. The present study aimed to fill this important gap in the literature by examining both the factor structure and ethnic and adolescent developmental period (early and middle adolescence) invariance of the MEIM among a large sample of youth who self-identified as either African American/Black, Multi-ethnic, Latinx, American Indian, or Asian.

First, the present study sought to examine the factor structure of the MEIM. Prior studies have examined the factor structure of the MEIM (Lee & Yoo, 2004; Phinney, 1992; Roberts et al., 1999; Yap et al., 2014, 2016), yet there is still disagreement in the literature regarding which factor structure fits best. Previous research supports 1-factor (Phinney, 1992), 2-factor (Roberts et al., 1999), 3-factor (Lee & Yoo, 2004), and bifactor models (Yap et al., 2014). The present study used CFA to test these four competing models and found that a 3-factor structure consisting of cognitive clarity, affective pride, and behavioral engagement, fits best for the current sample of early and middle adolescents. This factor structure is consistent with the one found by Lee and Yoo (2004) and similarly roughly maps onto the two-factor model proposed by Roberts and colleagues (1999) with cognitive clarity and affective pride paralleling affirmation and behavioral engagement paralleling exploration (Phinney & Ong, 2007). Although a 3-factor model is less parsimonious than a 2-factor model, the results of the present study found that it is a better representation of the factor structure in a large diverse sample of ethnic minority early and middle adolescents. The finding of a 3-factor structure, similar to Lee and Yoo (2004), may be attributed to both studies’ use of samples comprised of minorities only. All of the other studies conducted on the factor structure of the MEIM have included White participants (the dominant/majority group in the United States) in the sample as well. Even though the MEIM was developed for use across all racial/ethnic backgrounds (Phinney, 1992), recent research on ethnic identity questions whether White individuals interpret and respond to questions regarding racial/ethnic identity in similar ways as racial/ethnic minorities (Feitosa, et al., 2017). This influence has also been cited in MEIM measurement studies. For example, Ponterotto and colleagues (2003) state the inclusion of White participants may have led to the poor fit of their 1-factor model, and Feitosa and colleagues (2017) found their two-factor structure exhibited weak measurement equivalence with White participants.

The second aim of the study was to examine the measurement invariance of the MEIM across racial/ethnic groups. When examining measurement invariance across the included ethnic groups we found configural and metric invariance. However, for scalar invariance, findings provided support for partial invariance as there were four items (three from the behavioral engagement subscale and one from the affective pride subscale) that were noninvariant across racial/ethnic groups. The three items from the behavioral engagement subscale that contributed to the lack of full scalar invariance were “I have a clear sense of my ethnic background,” “I think a lot about how my life will be affected by my ethnic group membership,” and “I participate in cultural practices.” “I have a strong attachment toward my ethnic group” from the affective pride subscale was the last item that contributed to the lack of full scalar invariance. This finding suggests that although there are items within the MEIM that can be justifiably compared across racial/ethnic groups, particularly within the cognitive clarity and affective pride subscale, a majority of the items within the behavioral engagement subscale are not comparable given differences in item intercepts across groups.

This finding of noninvariance across groups for the behavioral engagement subscale (also referred to as the exploration subscale within the 2-factor model) is consistent with findings by Musso et al (2018) who also found noninvariance among their sample of adolescents residing in Italy. However, unlike the Musso et al. (2018) study which found the noninvariance effect between ethnic majority versus minority youth, the current study’s findings of noninvariance was found among racial/ethnic minority youth, This finding suggests that there may be qualitatively different cultural experiences among racial/ethnic minority youth within the United States that impact their responses on items assessing behavioral engagement in social organizations or cultural practices of their group, as well as the extent to which youth contemplate how their group membership impacts their life.

Based on the unstandardized coefficients within the scalar invariance model, findings suggest that such experiences may be more pronounced among African American youth than those of other racial/ethnic groups. Specifically, the intercept scores for African American youth were the highest among the racial/ethnic groups for items 2 and 4 and were similar to those of Asian and Latinx youth for item 10. There is evidence in the literature to support these observed differences in behavioral engagement, with studies documenting higher levels of ethnic identity exploration among African American youth compared to youth of other racial/ethnic groups (Else-Quest & Morse, 2015; Hughes et al., 2009). Studies have also documented a high degree of cultural socialization regarding African American culture by African American caregivers towards their youth (Else-Quest & Morse, 2015; Hughes et al., 2006; Hughes et al., 2009), likely resulting in more active engagement of African American youth in their culture (e.g., Grills et al., 2016).

However, further work is needed in this area. For example, although scores were relatively high for African American youth on behavioral engagement items, Latinx youth reported the lowest intercept score for item 2 regarding activity in organizations or social groups that include mostly members of their ethnic group. Moreover, American Indian youth reported the lowest score on item 10 regarding participation in the cultural practices of their own group. Additional research needs to be conducted to better understand why behavioral engagement in cultural practices differs across racial/ethnic minority youth populations in the United States. Given the heterogeneity of the Latinx (e.g., Valezquez & Avila, 2017), it is plausible that some groups within the Latinx community may engage in greater exploration than other groups due to immigrant status and acculturation in the United States. As for American Indian youth, there is evidence to suggest that American Indian youth tend to be heavily involved in cultural activities to help foster their cultural identity regardless of whether they live on or off a reservation (Kulis et al., 2013; Schweigman et al., 2011). Thus, it is surprising that scores on item 10 were lowest for American Indians in comparison to the youth of other racial/ethnic groups. For the current study, participants were recruited from communities in the Midwest, which may not be representative of practices among American Indians across the country. As a result, more research is needed in this area to better understand the cultural factors that may impact responses across racial/ethnic groups on ethnic identity measures of behavioral engagement.

The third aim was to examine measurement invariance across developmental periods in our sample of youth. This aim was significant in that to our knowledge, the current study was the first study to look at the scalar invariance of the MEIM across early and middle adolescence. Results indicated the MEIM demonstrated configural, metric, scalar, and strict invariance across the two age groups (6th–8th and 9th–12th). This suggests that the MEIM measures the same construct across these developmental periods during adolescence. This finding contrasts literature that has found differences in exploration between early and middle adolescents (French et al., 2006), as well as assumptions that the MEIM items may be differentially understood and endorsed across adolescent development due to the developmental nature of ethnic identity (Phinney & Ong, 2007). Findings from the present study, however, indicate it is not the case.

There are also several methodological differences across studies that may explain the contradictory findings between the current study and French et al. (2006) study. Specifically, the French et al. (2006) study was based on data collected among 5th, 6th, 8th, and 9th graders, with 5th and 6th grade representing early adolescence and 8th–9th grade representing middle adolescence. This grade stratification is important, as the current study operationalized early adolescence as 6th–8th grade and middle adolescence as 9th–12th grade. This difference in grade stratification suggests that when examining item endorsement of the MEIM based on distinct developmental periods (i.e., middle school versus high school), reporting on the MEIM is comparable although variability may exist among youth within each grade. There were also important distinctions between the studies based on the demographic composition of the study participants and sample size. In particular, the French et al. (2006) study included a total of 420 adolescents self-identifying as African American, Latinx, and White, which is much smaller than the current study which was comprised of 4,940 youth self-identifying as African American, multi-ethnic, Latinx, American Indian, and Asian. Since the French et al. (2006) study included White youth and was smaller, these findings may not be generalizable to a larger sample of racial/ethnic minority youth, such as the current study. Given that the current study utilized a comprehensive set of measurement invariance metrics among a large sample of diverse youth across distinct developmental periods, it is posited that these findings are a more accurate assessment of the MEIM for this population of youth.

Lastly, the present study sought to examine mean differences in MEIM scores across the included ethnic minority groups and developmental periods. Multi-ethnic youth group means were found to be different than the other four groups (i.e., African American, American Indian, Asian, and Latinx). Specifically, Multi-ethnic youth scored significantly lower across all indicators including cognitive clarity, affective pride, and behavioral engagement. This is consistent with previous research that included multi-ethnic youth in the sample (Phinney, 1992). These findings suggest that special attention is needed particularly for multi-ethnic youth to better understand what factors may hinder, as well as foster, ethnic identity development among this population of youth. Finally, regarding developmental periods, no mean differences were found across early and middle adolescence in cognitive clarity, affective pride, and behavioral engagement. This suggests that the mean endorsement of items was similar between early and middle adolescent youth.

Limitations and Future Directions

The present study contributes significantly to the field by examining scalar variance in a widely used tool with an understudied population, though, limitations should be noted. First, the sample used for the present study was drawn from a limited geographical location, and results may not be generalizable to the broader population. Despite this fact, results provide preliminary evidence for the use of the MEIM across ethnic groups and early/middle adolescence. Second, although this measurement study includes ethnic groups not commonly included in previous studies of the MEIM, sample sizes of some of the included underrepresented groups are small. Future research should examine the utility of the MEIM with larger nationally representative samples. Lastly, while Robert’s and colleagues (1999) version of the MEIM is still a widely used tool, Phinney and Ong (2007) developed a shorter revised version of the MEIM. Future research should also examine the measurement invariance of the revised version of the MEIM (MEIM-R) across racial/ethnic groups and developmental periods to determine whether findings from the current study are consistent with the use of the revised measure.

In sum, assessing measurement structure and invariance, the current study found support for the 3-factor model of the MEIM after testing four competing models. Furthermore, the study is the first to find partial scalar invariance of the MEIM across a diverse group of racial/ethnic minority adolescents who self-identify as either African American/Black, Multi-ethnic, Latinx, American Indian, or Asian. This finding is significant given the widespread use of the MEIM with adolescent populations, despite the absence of research supporting the use of the measure to compare means across ethnic groups. Lastly, the present study answered a much-needed call for the examination of the measurement invariance of the MEIM across development (Avery et al., 2007; Phinney & Ong, 2007; Swartz et al., 2014), finding evidence for strict invariance across early and middle adolescence. The results of this study thus point to the validity of using the MEIM for meaningful comparisons of ethnic identity across ethnic groups and across early and middle adolescence.

Acknowledgments

This work was supported by the National Institutes of Health 5KL2-TR001996, KL2TR001106, R25DA035163, and UL1TR001998.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Contributor Information

Sycarah Fisher, The University of Georgia.

Tamika B. Zapolski, Indiana University Purdue University Indianapolis.

Lorey Wheeler, University of Nebraska-Lincoln.

Prerna G. Arora, Columbia University.

Jessica Barnes-Najor, Michigan State University.

References

  1. Anglin AE, Johnson-Pynn JS, & Johnson LR (2012). Ethnic identity and civic attitudes in Latino and Caucasian youth. Journal of Youth Studies, 15(5), 621–643. 10.1080/13676261.2012.663887 [DOI] [Google Scholar]
  2. Avery DR, Tonidandel S, Thomas KM, Johnson CD, & Mack DA (2007). Assessing the Multigroup Ethnic Identity Measure for measurement equivalence across racial and ethnic groups. Educational and Psychological Measurement, 67(5), 877–888. 10.1177/0013164406299105 [DOI] [Google Scholar]
  3. Carter J, Peralta RL, & Xi J (2018). The association between ethnic identity and non-medical prescription drug use among a sample of college students: Does a sense of ethnic belonging matter?. Substance Use & Misuse, 54(2), 203–213. 10.1080/10826084.2018.1501065 [DOI] [PubMed] [Google Scholar]
  4. Chen FF (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling, 14, 464–504. 10.1080/10705510701301834 [DOI] [Google Scholar]
  5. Chen FF, Sousa KH, & West SG (2005). Teacher’s corner: Testing measurement invariance of second-order factor models. Structural Equation Modeling, 12(3), 471–492. 10.1207/s15328007sem1203_7 [DOI] [Google Scholar]
  6. Curran PJ, West SG, & Finch JF (1996). The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1, 16–29. 10.1037/1082-989X.1.1.16 [DOI] [Google Scholar]
  7. Dandy J, Durkin K, McEvoy P, Barber BL, & Houghton S (2008). Psychometric properties of Multigroup Ethnic Identity Measure (MEIM) scores with Australian adolescents from diverse ethnocultural groups. Journal of Adolescence, 31(3), 323–335. 10.1016/j.adolescence.2007.06.003 [DOI] [PubMed] [Google Scholar]
  8. Else-Quest NM, & Morse E (2015). Ethnic variations in parental ethnic socialization and adolescent ethnic identity: A longitudinal study. Cultural Diversity and Ethnic Minority Psychology, 21(1), 54–64. 10.1037/a0037820 [DOI] [PubMed] [Google Scholar]
  9. Erikson EH (1968). Identity: Youth and crisis. W. W. Norton & Company. [Google Scholar]
  10. Feitosa J, Lacerenza CN, Joseph DL, & Salas E (2017). Ethnic identity: Factor structure and measurement invariance across ethnic groups. Psychological Assessment, 29(9), 1129–1141. 10.1037/pas0000346 [DOI] [PubMed] [Google Scholar]
  11. Fisher S, Zapolski T, Sheehan C, & Barnes-Najor J (2017). Pathway of protection: ethnic identity, self-esteem, and substance use among multiracial youth. Addictive Behaviors, 72, 27–32. 10.1016/j.addbeh.2017.03.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. French SE, Seidman E, Allen L, & Aber JL (2006). The development of ethnic identity during adolescence. Developmental Psychology, 42(1), 1–10. 10.1037/0012-1649.42.1.1 [DOI] [PubMed] [Google Scholar]
  13. Gregorich SE (2006). Do self-report instruments allow meaningful comparisons across diverse population groups? Testing measurement invariance using the confirmatory factor analysis framework. Medical Care, 44(11 Suppl 3), S78–S94. 10.1097/01.mlr.0000245454.12228.8f [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Grills C, Cooke D, Douglas J, Subica A, Villanueva S, & Hudson B (2016). Culture, racial socialization, and positive African American youth development. Journal of Black Psychology, 42(4), 343–373. 10.1177/0095798415578004 [DOI] [Google Scholar]
  15. Gummadam P, Pittman LD, & Ioffe M (2016). School belonging, ethnic identity, and psychological adjustment among ethnic minority college students. The Journal of Experimental Education, 84(2), 289–306. 10.1080/00220973.2015.1048844 [DOI] [Google Scholar]
  16. Hughes D, Hagelskamp C, Way N, & Foust MD (2009). The role of mothers’ and adolescents’ perceptions of ethnic-racial socialization in shaping ethnic-racial identity among early adolescent boys and girls. Journal of Youth and Adolescence, 38(5), 605–626. 10.1007/s10964-009-9399-7 [DOI] [PubMed] [Google Scholar]
  17. Hughes D, Rodriguez J, Smith EP, Johnson DJ, Stevenson HC, & Spicer P (2006). Parents’ ethnic-racial socialization practices: a review of research and directions for future study. Developmental Psychology, 42(5), 747–770. 10.1037/0012-1649.42.5.747 [DOI] [PubMed] [Google Scholar]
  18. Kulis S, Wagaman MA, Tso C, & Brown EF (2013). Exploring indigenous identities of urban American Indian youth of the Southwest. Journal of Adolescent Research, 28(3), 271–298. 10.1177/0743558413477195 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Lee RM, & Yoo HC (2004). Structure and measurement of ethnic identity for Asian American college students. Journal of Counseling Psychology, 51(2), 263–269. 10.1037/0022-0167.51.2.263 [DOI] [Google Scholar]
  20. Millsap RE (2012). Statistical approaches to measurement invariance. Routledge. [Google Scholar]
  21. Musso P, Moscardino U, & Inguglia C (2018). The Multigroup Ethnic Identity Measure–Revised (MEIM-R): Psychometric evaluation with adolescents from diverse ethnocultural groups in Italy. European Journal of Developmental Psychology, 15(4), 395–410. 10.1080/17405629.2016.1278363 [DOI] [Google Scholar]
  22. Muthén LK, & Muthén BO (1998–2017). Mplus user’s guide (8th ed.). Muthén & Muthén. [Google Scholar]
  23. Muthén BO, Muthén LK, & Asparouhov T (2016). Regression and mediation analysis using Mplus. Muthén & Muthén. [Google Scholar]
  24. Pahl K, & Way N (2006). Longitudinal trajectories of ethnic identity among urban Black and Latino adolescents. Child Development, 77(5), 1403–1415. 10.1111/j.1467-8624.2006.00943.x [DOI] [PubMed] [Google Scholar]
  25. Phinney JS (1992). The multigroup ethnic identity measure: A new scale for use with diverse groups. Journal of Adolescent Research, 7(2), 156–176. 10.1177/074355489272003 [DOI] [Google Scholar]
  26. Phinney JS, & Chavira V (1995). Parental ethnic socialization and adolescent coping with problems related to ethnicity. Journal of Research on Adolescence, 5(1), 31–53. 10.1207/s15327795jra0501_2 [DOI] [Google Scholar]
  27. Phinney JS, & Ong AD (2007). Conceptualization and measurement of ethnic identity: Current status and future directions. Journal of Counseling Psychology, 54(3), 271–281. 10.1037/0022-0167.54.3.271 [DOI] [Google Scholar]
  28. Ponterotto JG, Gretchen D, Utsey SO, Stracuzzi T, & Saya R Jr (2003). The Multigroup Ethnic Identity Measure (MEIM): Psychometric review and further validity testing. Educational and Psychological Measurement, 63(3), 502–515. 10.1177/0013164403063003010 [DOI] [Google Scholar]
  29. Quintana SM, Aboud FE, Chao RK, Contreras-Grau J, Cross WE, Hudley, Hughes D, Liben LS, Nelson-Le Gall S & Vietze DL (2006). Race, ethnicity, and culture in child development: Contemporary research and future directions. Child Development, 77(5), 1129–1141. 10.1111/j.1467-8624.2006.00951.x [DOI] [PubMed] [Google Scholar]
  30. Rahim-Williams FB, Riley III JL, Herrera D, Campbell CM, Hastie BA, & Fillingim RB (2007). Ethnic identity predicts experimental pain sensitivity in African Americans and Hispanics. Pain, 129(1–2), 177–184. 10.1016/j.pain.2006.12.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Reese LRE, Vera EM, & Paikoff RL (1998). Ethnic identity assessment among inner-city African American children: Evaluating the applicability of the Multigroup Ethnic Identity Measure. Journal of Black Psychology, 24(3), 289–304. 10.1177/00957984980243002 [DOI] [Google Scholar]
  32. Roberts RE, Phinney JS, Masse LC, Chen YR, Roberts CR, & Romero A (1999). The structure of ethnic identity of young adolescents from diverse ethnocultural groups. The Journal of Early Adolescence, 19(3), 301–322. 10.1177/0272431699019003001 [DOI] [Google Scholar]
  33. Schweigman K, Soto C, Wright S, & Unger J (2011). The relevance of cultural activities in ethnic identity among California Native American youth. Journal of Psychoactive Drugs, 43(4), 343–348. 10.1080/02791072.2011.629155 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Smith TB, & Silva L (2011). Ethnic identity and personal well-being of people of color: A meta-analysis. Journal of Counseling Psychology, 58(1), 42–60. 10.1037/a0021528 [DOI] [PubMed] [Google Scholar]
  35. Spencer MS, Icard LD, Harachi TW, Catalano RF, & Oxford M (2000). Ethnic identity among monoracial and multiracial early adolescents. The Journal of Early Adolescence, 20(4), 365–387. 10.1177/0272431600020004001 [DOI] [Google Scholar]
  36. Tajfel H (1981). Human groups and social categories: Studies in social psychology. Cambridge University Press. [Google Scholar]
  37. Tajfel HT, & Turner J JC (1986). The social identity theory of intergroup behavior. Psychology of intergroup relations, 7–24. 10.4324/9780203505984-16 [DOI] [Google Scholar]
  38. Townsend TG, Kaltman S, Saleem F, Coker-Appiah DS, & Green BL (2020). Ethnic disparities in trauma-related mental illness: Is ethnic identity a buffer?. Journal of Interpersonal Violence, 35(11–12), 2164–2188. 10.1177/0886260517701454 [DOI] [PubMed] [Google Scholar]
  39. Worrell FC (2000). A validity study of scores on the Multigroup Ethnic Identity Measure based on a sample of academically talented adolescents. Educational and Psychological Measurement, 60(3), 439–447. 10.1177/00131640021970646 [DOI] [Google Scholar]
  40. Worrell FC, Conyers LM, Mpofu E, & Vandiver BJ (2006). Multigroup Ethnic Identity Measure scores in a sample of adolescents from Zimbabwe. Identity, 6(1), 35–59. 10.1207/s1532706xid0601_4 [DOI] [Google Scholar]
  41. Yang-Wallentin F, Jöreskog KG, & Luo H (2010) Confirmatory factor analysis of ordinal variables with misspecified models, Structural Equation Modeling, 17, 392–423. doi: 10.1080/10705511.2010.489003 [DOI] [Google Scholar]
  42. Yap SC, Donnellan MB, Schwartz SJ, Kim SY, Castillo LG, Zamboanga BL, Weisskirch RS, Lee RM, Park IJ, Whitbourne SK, & Vazsonyi AT (2014). Investigating the structure and measurement invariance of the Multigroup Ethnic Identity Measure in a multiethnic sample of college students. Journal of Counseling Psychology, 61(3), 437–446. 10.1037/a0036253 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Yap SC, Donnellan MB, Schwartz SJ, Zamboanga BL, Kim SY, Huynh QL, Vazsonyi AT, Cano MA, Hurley EA, Whitbourne SK, & Castillo LG (2016). Evaluating the invariance of the Multigroup Ethnic Identity Measure across foreign-born, second-generation and later-generation college students in the United States. Cultural Diversity and Ethnic Minority Psychology, 22(3), 460–465. 10.1037/cdp0000068 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Yip T, Seaton EK, & Sellers RM (2006). African American racial identity across the lifespan: Identity status, identity content, and depressive symptoms. Child Development, 77(5), 1504–1517. 10.1111/j.1467-8624.2006.00950.x [DOI] [PubMed] [Google Scholar]
  45. Zapolski TCB, Fisher S, Banks D, Hensel D, & Barnes-Najor J (2017). Examining the protective effect of ethnic identity on drug attitudes and use among a diverse youth population. Journal of Youth and Adolescence, 46(8), 1702–1715. 10.1007/s10964-016-0605-0 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES