Psychometric properties of the Multi-group Ethnic Identity Measure (MEIM) in a sample of Iranian young adults

Ethnic Identity and Subjective Well-being

Subjective well-being involves a positive feeling and a general sense of satisfaction with different areas of life (Myers & Diener, 1995). According to Pavot & Diener (2004), subjective well-being consists of two distinct, but related components: an emotional component (i.e., positive and negative affect) and a cognitive component (i.e., satisfaction with life). Positive affect reflects the occurrence of positive emotional states such as joy, interest and happiness. In contrast, negative affect involves negative emotional states such as anger, fear, sadness and guilt (Fredrickson, 2001). The cognitive component of subjective well-being refers to an assessment of the quality of different aspects of one’s life and a general feeling of satisfaction with life (Gilman, Huebner & Laughlin, 2000). Life satisfaction, positive affect, and negative affect are considered as independent phenomena (Lucas, Diener & Suh, 1996) and are therefore often evaluated individually rather than as an integrated entity (for other perspectives see Nima et al., 2020a; Nima et al., 2020b; Garcia et al., in press).

Several studies have demonstrated a significant positive correlation between ethnic identity and subjective well-being (e.g., Ali, 2006; Diaz & Bui, 2017; Oliveira, Pankalla & Cabecinhas, 2012; Syed et al., 2013; Yoo & Lee, 2005). There have also been studies indicating that strong ethnic identity can reduce negative emotions (Williams et al., 2012; Yip & Fuligni, 2002; Roberts et al., 1999) and that successful ethnic identity formation can predict an overall sense of well-being (Vleioras & Bosma, 2005; Waterman, 2007; Hejazi et al., 2009). In addition, a significant relationship between identity development and well-being has been demonstrated by few studies conducted in Iran (Alijani, 2006; Ghazanfari, 2003). Nevertheless, as most studies in this field have been conducted in western cultural contexts, the generalizability of these findings to Iran is limited, since Iran is an Islamic Middle Eastern society with a multiethnic culture. Such culture, has for example, a different approach to gender issues.

Ethnic Identity and Gender

Research related to ethnic identity and gender has shown mixed results (e.g., Kazarian & Boyadjian, 2008; Ontai-Grzebik & Raffaelli, 2004; Phinney, Cantu & Kurtz, 1997; Swenson & Prelow, 2005; Lum, 2009). For instance, in a study by Dandy and colleagues (2008), no significant gender differences were observed in identity affirmation, but females were more inclined to interact and socialize with other ethnic groups. Other studies, however, indicate higher levels of ethnic identity among males compared to females (e.g., Vo, 2008; Go & Le, 2005; Mossakowski, 2018). Yet other studies show that females score higher than males in ethnic identity (e.g., Cislo, 2008; Creagh-kaiser, 2003; Sobansky et al., 2010).

That being said, at least theoretically, the association between ethnic identity and psychological outcomes, such as life satisfaction and affect, can be mediated by gender. For example, Cross & Madson (1997) suggested that since women are more relationship-oriented, their life satisfaction may be positively affected by the sense of belongingness to their ethnic group. Other studies (e.g., Navarrete et al., 2010) have suggested that since men are more commonly the target of racial discrimination, ethnic identity has a larger impact on them compared to its impact on women (Yap, Settles & Pratt-Hyatt, 2011). Hence, despite the mixed results regarding gender, besides the relationship between ethnic identity and subjective well-being, we also investigate gender differences in ethnic identity.

The Multigroup Ethnic Identity Measure (MEIM)

Several models and measures for the assessment of ethnic identity have been proposed. Most of which are based on the works of Erikson, Tajfel, and Phinney (see Umana-Taylor, 2004). Many of the earlier measures were designed for specific ethnic groups, thus, making cross-ethnic comparisons impossible. In order to measure ethnic identity across cultures, Phinney (1992) introduced a 20-item ethnic identity measure (i.e., MEIM-O) that included five items to measure Affirmation and Belonging, seven items to measure Ethnic Identity Achievement, two items for Ethnic Behaviors, and six additional items to assess Other Group Orientation (OGO). Phinney (1992) considered the two factors, Ethnic Identity and OGO, for MEIM-O based on an exploratory factor analysis done on data collected among American students from different ethnic groups. Later on, in 1999, the original MEIM was revised by Roberts, Phinney, Masse, Chen, Roberts, and Romero, who carried out an exploratory factor analysis on 14 MEIM items (not including the six OGO items). They administered the test to 5,423 young adolescents (sixth to eighth graders) from a variety of ethnicities including African American (n = 1, 237); Central American (n = 253); Chinese American (n = 177); European American (n = 755); Indian American (n = 188); Mexican American (n = 755); Pakistani American (n = 155); Vietnamese American (n = 304); Pacific Islander (n = 101); and mixed ancestry (n = 342). The results showed that the two items assessing Ethnic Behaviors (i.e., being active in ethnic organizations and participating in cultural practices) were misunderstood among younger adolescents and were therefore removed. Thus, arriving to the version of the MEIM used in the present study; which consists of 12 items and two factors, namely exploration and commitment. More specifically, of the 12 remaining items, five items represent exploration and seven items represent commitment (Herrington et al., 2016).¹

Prior Research on Ethnic Identity in Young Iranian Adults

For instance, Moghaddas Jafari, Sheikhavandi & Sharif Pour (2008) studied 347 Kurdish university students of Saqqez Payam-e-Noor University, a Kurdish university located in the west of Iran. They found that Kurdish students were to a large extent committed to all components of ethnic and national identity. They also showed that perceived social inequality was the most important factor explaining and predicting commitment to ethnic and national identity. It is worth mentioning that Iranian ethnic groups reside mostly in the borders of Iran, but also beyond the borders. For example, Kurds partly dwell in Iran and partly in neighboring countries including Turkey and Iraq. Therefore, it has long been a concern of the Iranian central government to maintain the unity of the country in order to strengthen the central power and national security. In this regard, over different eras, political leaders have tried to promote the common aspects of the Iranian culture as opposed to different cultures of ethnic groups within the country. Quite often this has led to the negligence or even suppression of ethnic communities (Amanolahi, 2005; Saleh, 2013). The national media in Iran, for instance, mostly promulgates the Persian and Shi’i values and culture while the global media emphasizes cultural integration and globalization (Gulyas, 2017; Saleh, 2013). Indeed, modernization and globalization tendencies have, in certain ways, notably blurred the distinctive characteristics of different ethnic groups. Studies show that, although in recent years there have been several measures taken in Iran to appreciate ethnic differences, for example by teaching Kurdish and Turkish language courses at certain universities, there is still a long way to go before satisfactory conditions are created, where all Iranian ethnicities and sub-cultures are respected and appreciated (Gulyas, 2017; Saleh, 2013).

Nevertheless, Hajiyani (2008) demonstrated that both ethnic and national identity are strong and prominent among the six main Iranian minorities including Turks, Kurds, Lurs, Balochs, Arabs and Turkmen. They also found a significant positive relationship between the sociocultural dimensions of ethnic identity and the cultural and somewhat social dimensions of national identity. Despite the importance of these studies, a major problem is that a majority, if not all, have not used standard models and measures when assessing ethnic identity. This makes comparisons across and within cultures impossible. Standard measures for ethnic identity assessment, such as the MEIM, are under attention in other countries, making research findings comparable.

The present study

It is important to notice that Iran is a very diverse country with seven main distinctive ethnicities: Farses with 61% of the population are the largest group, Azerbaijanis (16%), Kurds (10%), Lurs (6%), Arabs (2%), Balochs (2%), and Turkmens and Turkic tribes (2%). In addition, there is very small number of Armenians, Assyrians, and Georgians (World Population Review & Iran population, 2020). All these ethnicities are referred to as “Iranian”, since they are basically from Iran and over the centuries, they have inhabited the Iranian land. At the same time, due to modernization, industrialization, and other economic and political issues, the country has experienced an increase in internal migration patterns, which have in turn resulted in cities becoming more and more multi-ethnic over the last few decades.

Despite the fact that the MEIM has been extensively used in research around the world, to the best of our knowledge, no psychometric research has been done on the 12-item MEIM in the Iranian young adult population. Additionally, the 12-item MEIM has not been used for assessing ethnic identity in different Iranian ethnic groups. Therefore, this study aimed to confirm the two-factor structure of the Persian version of the 12-item MEIM in a sample of young Iranian adults using confirmatory factor analysis (CFA). In addition, we examined gender differences, test–retest validity, and the reliability of the MEIM. In other words, this research explored the cross-national generalizability of the MEIM’s factor structure and also provided an additional test of its construct validity and reliability. Parallel to these aims, we also investigated the association between ethnic identity and subjective well-being (positive affect, negative affect, and life satisfaction). In short, the present paper had three main objectives: (1) To evaluate the factorial validity and test-retest reliability of the MEIM among a sample of young Iranian students, (2) To examine ethnic identity, as measured by the MEIM, in relation to subjective well-being in order to test the concurrent validity of the MEIM (3), and to examine possible gender differences in ethnic identity. Regarding subjective well-being, we expected that (a) ethnic identity will be positively associated with positive affect, (b) negatively associated with negative affect, and (c) and positively related to life satisfaction.

Participants

Participants were considered eligible if they were within 17 to 40 years of age, had at least completed one semester of university studies, and provided written informed consent for participation in the study. Those who reported psychiatric disorders, organic brain disorder, and/or drug dependence were excluded from the study. Sample size estimation was based on the rule of thumb suggested by Bentler & Chou (1987). A total of 500 undergraduate students volunteered to participate in the normative sample, but 50 (10%) students were excluded because they had responded to less than 90% of the questions. Out of the remaining 450 participants, 24 were withdrawn from the study due to noncompliance. Hence, the final sample included 426 students (91.8% of whom reported being single and 54.7% were males). The percentages of participants from different ethnic groups were as follows: 33.1% Turks, 28.9% Farses, 21.4% Lors, and 16.7% Kurds. The average age was 21.98 years and ranged from 18 to 31 years. The students were 60.3% undergraduates and 39.7% graduates and were from three faculties: Social Sciences (45.1%), Foundational Sciences (33.3%) and Engineering (21.6%). Most of the participants lived in dormitory (70.7%) and the remaining lived with their parents or spouses. All participants were offered the option to take the MEIM on one or two occasions. Those who wanted to take the test twice were part of the test-retest subsample and took the test for the second time after a four-week interval. The test-retest sub-sample (n = 50; 25 males) had a mean age of 23 years (age range 18–30 years).

Measures

Procedures

With permission from its author (personal communication, 2013), the MEIM was translated into Persian and back-translated to English by a research team (a linguist and two mental health experts) fluent in both Persian and English languages (Guillemin, Bombardier & Beaton, 1993; Villagran & Lucke*, 2005). A primary Persian version of the MEIM was prepared and its clarity was evaluated by means of a pilot study on 30 students (15 females) who were asked to rate the fluency of items from 0 (Not Understandable) to 5 (Completely Understandable). In this pilot study, participants were asked to report any misunderstanding or lack of clarity related to concepts or wording. The most common rating by participants was completely understandable (response option number five), with this response option endorsed at 95% or higher across all items. Thus, participants’ responses indicated no need for item revision. The back-translation was done by other mental health and linguistic professionals. The back-translation followed by experts‘ judgments revealed that, apart from a few minor adjustments, regarding wording and layout, the Persian version of the MEIM was a precise reflection of its original version. Ethical approval for carrying the study was obtained from The Shahid Beheshti University’s ethics board (95-03-211-31609). To carry out the sampling process, using convenience sampling, research assistants at The Shahid Beheshti University invited students to participate in the study. The ones who accepted the invitation, were handed the questionnaires in in their classrooms in a group survey administration format. Before asking participants to sign the written consent form, they were informed that they could withdraw from participating in the study whenever they wanted to do so, they were given brief description about the aim of the study, and informed about the confidentiality of their participation. The three questionnaires were distributed in different orders, that is, in a counterbalanced design to control for order-effect. The questionnaires included the MEIM, the PANAS, the SWLS, and a demographic information sheet (i.e., gender, age, marital status, and educational status).

Statistical strategy

A CFA using LISREL, version 8.72 (Jöreskog & Sörbom, 2005) was applied to examine the two-factor structure of the MEIM (cf. Phinney (1992) and Phinney (1992) )). This method offers a variety of statistical tests and indices designed to assess the ”goodness-of- fit” of identified models (Mulaik et al., 1989). For this purpose, in the present study, the goodness-of-fit was evaluated using the following statistics: the goodness-of-fit index (GFI >.9), the adjusted goodness-of-fit index (AGFI >.90), the non-normal fit index (NNFI >.90), the comparative fit index (CFI >.90), the root mean square residual (RMSR <.08), the normal chi-square (3 >χ²/df <2) and the root mean square error of approximation (RMSEA) and its 90% confidence interval (<.05: Breckler, 1990; Mulaik et al., 1989). In short, the extent to which present data was compatible with different models (and modifications) was examined using LISREL, version 8.72 (Joreskog & Sorbom, 2005). The first model specified a one-factor model (M₁) in which all 12 items were forced to load on a single global factor (i.e., Reese, Vera & Paikoff (1998)); the second model presented a two-factor orthogonal model (M₂); the third examined a two-factor oblique model (M₃) as reported in the EFA procedure by Phinney (1992) and Phinney (1992). The concurrent validity of the MEIM was investigated by correlations between the MEIM scores and the subjective well-being scores. To evaluate the test-retest reliability of the MEIM, Intra-class correlation coefficients, for the total scale and the two sub-scales, were calculated on two occasions over four weeks apart. Cronbach’s alpha and mean inter-item correlation coefficients were calculated for the total MEIM score and its sub-scales. Preliminary analysis of the data showed that normality was violated. The Z score for the univariate skewness values ranged from −4.40 [Item 9, “I have a lot of pride in my ethnic group”] to −.76 [Item 1, “I have spent time trying to find out more about my ethnic group, such as its history, traditions, and customs.”] (See Table 2), and Relative Multivariate Kurtosis was 1.16 and test of multivariate normality for skewness and kurtosis was χ² = 255.29, p < .001. Due to univariate and multivariate non-normality of the MEIM items, the weighted least squares (WLS) estimation method was applied in CFA (Bentler & Bonett, 1980).

Aim 1: Factorial Validity and Test-Retest Reliability of the MEIM

Table 2 presents the fit estimates for all models. The one-factor model and the two- factor orthogonal model did not meet the previously specified fit criteria and the two factors oblique model showed inadequate fit to the data (M₁ to M₃). The modification by correcting errors in the two-factor model revealed some improvement (M_3a; RMSEA = .053). The chi-square test results were significant for all models, but that is to be expected with models with large degrees of freedom and relatively large sample sizes (Hu & Bentler, 1995). An examination of the remaining fit indices (Satorra & Bentler, 2001) for nested models suggested that the oblique and correlated errors model (M_3a in Table 2) was significantly more acceptable than the two factors oblique model (M₃; χ² = 102.07; df = 4, P < .001). In a comparison of the nested models, the χ² (Jöreskog & Sörbom, 1993) showed that the two-factor oblique correlated errors model provided a better fit [S-B χ ²/df = 2.19; CFI = .90; NNFI = .88; and RMSEA = .053 ([CI] 90% = .039, .067]. The correlation found between exploration and commitment latent variables was .88 (p < .001).

Another series of CFAs were conducted in order to determine if the factorial structure of the MEIM was consistent across gender. In short, testing the factor structure of MEIM in both males and females is the main starting point in evaluating multivariate CFA across genders and building baseline models for each group separately without any specification of equality constrained across groups. For testing equivalency across genders, both baseline models were modified to include the specification of four error covariances for males, and for females. First, a CFA was conducted for males and females separately. Several steps were taken in order to obtain satisfactory model fit and these steps were guided by parsimony and meaningfulness perspectives (Byrne, Shavelson, & Muthén 1989). The comparison of the baseline model (χ²/ df = 1.55; CFI = .93; NNFI = .91; and RMSEA = .051 ([CI] 90% = .034, .066) and the modified models for males (M_m; χ²/ df = 1.72; CFI = .91; NNFI = .88; and RMSEA = .056 (90% CI [.035–.076]) and females (M_f; χ²/ df = 1.38; CFI = .96; NNFI = .94; and RMSEA = .044 ([CI] 90% = .0095, .069), indicated that the modified models for males (Δ²= 67.47, df = 49, p < .05), and for females (Δ²= 84.51.4, df = 49, p < .01) were significantly different from the diagonal error covariance baseline model. Hence, these models provided an adequate level of fit for males and females as baseline models. Table 2 (model M_b) shows that the hypothesized factor structure of MEIM revealed good fit across genders, but the pattern of factor loadings (model M_c) varied across genders (Table 2). All constrained models in Table 2 were compared with χ² statistics against a starting or baseline model (M_b) in which no constraints in estimation of parameters were specified. The results indicated that the variances and covariance of commitment and exploration were consistent across gender.

In the case of configural invariance, χ², RMSEA, CFI, NNFI, and other fit indices were used to examine whether or not the combined models had a good model fit. Furthermore, for metric, scalar, residual, and latent variances and covariance, the RMSEA values and RMSEA confidence intervals of the hierarchical (nested) models were compared. For example, in the case of comparison of the weak and strong factorial invariance models, if RMSEA values fall within one another‘s confidence intervals, this shows strong factorial invariance. Then, the changes in the CFI of hierarchical (nested) models were examined. Also, the change in CFI for the weak and strong factorial invariance models was assessed. A change of less than .01 shows strong factorial invariance (Cheung & Rensvold, 2002). That is, it shows equal form (i.e., the number of factors and the pattern of factor-indicatorrelationships are the same), equal factor loadings, equal thresholds (i.e., when observed scores are regressed on each factor, the thresholds are equal), and equal residual variances across gender.

The internal reliability coefficients and the mean inter-item correlation for the MEIM are presented for the total sample and separately for males and females in Table 3. These findings suggest that the scale has acceptable internal consistency. For test-retest reliability of the MEIM, 50 Students (15 males) completed the scale with four weeks between measurements (i.e., time 1 and time 2). Intra-class correlation coefficients between the total and subscale scores at time 1 and 2, respectively, ranged from .82 to .91 (see Table 4).

Aim 2: Concurrent Validity of the MEIM

Table 5 presents the Pearson correlation coefficients between MEIM and its subscales with positive affect, negative affect, and life satisfaction. The results indicate that, as expected, the MEIM and its subscales have a positive relationship with positive affect and Life satisfaction and a negative relationship with negative affect. The shared variance for the total score of the MEIM indicates weak relationships between ethnic identity and subjective well-being measures.

Aim 3: Gender Differences and Ethnic Identity

A noticeably different pattern of results was found for males and females (Table 3). Female students had a significantly higher score on the MEIM compared to male students (t(424) = 2.71, p = .007). A multivariate analysis of variance (MANOVA) was conducted to evaluate the effects of gender on the MEIM sub-scales, with Bonferroni adjustment for multiple comparisons (Huberty & Morris, 1989). The MANOVA results showed a significant effect: Hotelling’s Trace = .02, F(3, 1158) = 4.33, p = .014, η²= .02 (Table 3). Subsequent examination of between-subject effects showed that the female group scored significantly higher on exploration (F(1, 402) = 7.50, p = .003, η²= .018) and commitment (F(1, 402) = 4.99, p = .016, η²= .014).