Introduction

Gender and ethnicity/race are salient social identities in many people’s lives that have garnered much attention in the research literature (e.g., see Miville and Ferguson 2014). However, work on each type of identity has largely occurred separately from the other. Consistent with recent calls for intersectional approaches (e.g., Cole 2009), we see room for potential theory bridging. We integrate Cameron’s (2004) multidimensional model of social identity and Egan and Perry’s (2001) gender identity model, which propose complementary sets of dimensions underlying social identities based on gender, ethnicity/race, and other groups. We tested our multidimensional model in a sample of Asian American, Latino/a, and White European American undergraduate women and men to explore how identity dimensions for both ethnicity/race and gender relate to outcomes among these diverse groups. Emerging adulthood is recognized as an important period when many persons are exploring their identities in multiple domains (Arnett 2000). Explorations of ethnicity/race and gender may be especially likely among college students who may be learning about issues of gender and ethnicity/race in their courses, dormitories, and other student life contexts (Syed and Azmitia 2008).

Ethnic–Racial Identity and Gender Identity

We begin with a review of the constructs of ethnic–racial identity and gender identity as applied in the present research.

Ethnic–Racial Identity

Ethnicity is often used to describe group members who share a common set of cultural traditions, values, and attitudes (e.g., Phinney 1990), whereas race often refers to biological and physical traits that unite a group (e.g., skin color; Quintana 1998). Relatedly, certain groups are more likely to be described in terms of their ethnic group (e.g., Latinos, Asian Americans), whereas others are more commonly referred to as a racial group (e.g., African Americans). Measures of ethnic identity and racial identity often reflect these patterns. For example, Phinney’s (1990, 1992) Multigroup Ethnic Identity Measure (MEIM) focuses on connections to the cultural traditions associated with one’s ethnic group. This measure often has been used with Latinos, Asian Americans, and European Americans (e.g., Juang and Syed 2010; Umaña-Taylor 2004). Another widely used measure is Sellers et al.’s (1997) Multidimensional Inventory of Black Identity (MIBI), which many researchers have used to assess racial identity among African Americans (e.g., Rogers et al. 2015; Sellers et al. 1998a).

Because there is often overlap in how individuals think about their ethnic and racial group (Cokley 2005), some scholars have argued for the use of a meta-construct of ethnic–racial identity instead of using the separate terms of ethnic identity or racial identity (Altschul et al. 2006; Umaña-Taylor et al. 2014). According to Umaña-Taylor et al. (2014), ethnic–racial identity is defined as “a multidimensional, psychological construct that reflects the beliefs and attitudes that individuals have about their ethnic–racial group memberships, as well as the processes by which these beliefs and attitudes develop over time” (p. 3). In accordance with this conceptualization of identity, we will use the term ethnic–racial identity in the current study. (However, we acknowledge some research questions may warrant differentiating between racial identity and ethnic identity.)

Gender Identity

Until recently, there was a general difference in how social/personality psychologists and developmental psychologists studied gender identity in adults and children, respectively. In social and personality psychology, the main emphasis from the 1970s into the 1990s was on the extent that individuals viewed themselves as expressing feminine-stereotyped (i.e., communal) and masculine-stereotyped (i.e., agentic) personality traits (e.g., Bem 1974; Spence et al. 1975). In developmental psychology, much of the work during the same period focused on whether or not children self-categorized as girls or boys (see Halim and Ruble 2010). However, there were also studies in developmental psychology examining children’s self-perceived communal and agentic traits and interests (see Leaper 2013).

Perry and his colleagues (Egan and Perry 2001; Tobin et al. 2010) noted some limitations of inferring gender identity from self-perceptions of gender-typed traits and interests. First, individuals who strongly identify with their gender may not recognize that particular traits and interests are associated with their gender group. Hence, those traits or activities may not reflect their gender self-concept. A second limitation is that individuals vary in the types of gender-typed traits and interests that they favor. For example, interest in sports and cars are two areas traditionally associated with men. One man may like sports but not cars, whereas another man may have the reverse pattern. Yet, both men might strongly identify with their gender. Additionally, individuals may exhibit psychological gender typicality (e.g., a woman who is family-oriented and nurturing), but have traditionally gender counter-stereotypical interests (e.g., pursuing a career in aeronautics). Accordingly, Perry and his collaborators argued for a model of gender identity that focuses more on various dimensions underlying people’s evaluation of their gender in-group membership. We review some components of the model later and note how they can be extended to the study of social identities based on ethnicity/race as well as gender.

Multidimensional Models of Identity

Social identity theory, as advanced by Tajfel and Turner (Tajfel 1981, 1982; Tajfel and Turner 1979), articulated ways that group identities can affect people’s self-concepts as well as their attitudes and behaviors toward ingroup and outgroup members. For example, this includes favoring characteristics tied to one’s ingroup over those associated with outgroups. Individuals commonly have multiple social identities (e.g., based on gender, ethnicity/race, religion, team affiliation, club membership). Also, individuals within a particular group vary in how they view the group. Researchers investigating social identities have noted multiple evaluative dimensions that underlie a given individuals’ identification with a particular group (e.g., see Ashmore et al. 2004; Cameron 2004, for reviews). We review two models below that guided the present research.

Cameron’s Three-Factor Model of Social Identity

Cameron (2004) proposed a multidimensional model of social identity that can be applied to ethnicity/race, gender, and other group identities (also see Ashmore et al. 2004, for a similar model). Each of the three dimensions are described below. Afterward, Cameron’s (2004) evidence for the validity of these constructs is briefly summarized.

The three factors in Cameron’s (2004) model are centrality, ingroup ties, and ingroup affect. First, centrality refers to the importance of a given group identity to the individual. As highlighted in earlier work (e.g., Gurin and Markus 1989; Luhtanen and Crocker 1992; also see Ashmore et al. 2004, for a review), individuals vary in the salience and psychological importance that particular group memberships may hold for them. For example, studies suggest that ethnic–racial identities may be more salient and important for minority group members than for majority group members (e.g., Hutnik and Sapru 1996; Phinney 1992). Also, ethnic–racial centrality can be a protective factor related to greater adjustment and academic success for ethnic/racial-minority students (e.g., Maxwell Johnson et al. 2005; Phinney et al. 1997; Sellers et al. 1998a; Smith and Silva 2011). Recently, researchers have begun to consider centrality as a moderator of gender identity (see Tobin et al. 2010). For example, Turner and Brown (2007) observed that gender was more central as a social identity for ethnic-majority children than for ethnic-minority children. Also, Bem (1981) proposed that gender schematicity partly reflected the degree that gender was an important part of adults’ self-concept.

Second, ingroup affect is the emotional valence that an individual associates with a particular group. Group members may vary in the degree that they have positive or negative feelings toward their ingroup. As Cameron (2004) acknowledges, this dimension is similar to constructs such as collective self-esteem (e.g., Luhtanen and Crocker 1992), private regard (Sellers et al. 1998b), and affective identification (Deaux 1996; also see Ashmore et al. 2004, for other similar constructs). For example, some group members may feel positive toward their ethnic–racial ingroup, whereas others feel relatively negative toward their ingroup (e.g., Demo and Hughes 1990; Rivas-Drake et al. 2014).

Finally, ingroup ties reflect the degree of emotional closeness that one experiences toward the ingroup and its members. It is one’s sense of belonging, attachment, and connection to the group. Other researchers similarly have observed how members of groups can vary in how much allegiance they feel toward the group (e.g., Bollen and Hoyle 1990; Phinney 1992; also see Ashmore et al. 2004, for a review). For example, in her model of ethnic identity in adolescents and adults, Phinney (1992) included ethnic affirmation/belonging. Various factors may affect group belonging including the degree to which one’s behavior or appearance seems congruent with group expectations (Oyserman et al. 2006) as well as opportunities to connect with in-group members (Postmes and Branscombe 2002).

Cameron (2004) advanced the three-factor model of social identity following earlier exploratory factor analyses of social identities based on gender or ethnic identity. In his 2004 paper, he reported results from five separate samples testing for the three-factor model of social identities based on gender (two samples: Australian university students, mean ages = 19 or 20 years), university affiliation (two samples: Australian university students, mean ages = 19 years), or nationality (one sample: Australian residents, mean age = 48 years).

Results indicated satisfactory internal consistency, predictive validity, and independence of the dimensions for the three-factor model among Australian undergraduates and residents. Cameron (2004) encouraged researchers to utilize his scales to assess a variety of social identities in diverse populations. In the present research, we applied the three-factor model to examining both ethnicity/race and gender in the same individuals. Cameron also recommended further testing of possible ways that the different factors differentially predict outcomes. Accordingly, as described later, we examined how aspects of ethnic–racial identity and gender identity might be related to self-esteem. Furthermore, as reviewed next, we considered two additional social identity dimensions that we propose as potentially complementing Cameron’s (2004) model.

Perry and Colleagues’ Multidimensional Model of Gender Identity

During the last 15 years, Perry and his colleagues (e.g., Corby et al. 2007; Egan and Perry 2001; Tobin et al. 2010; Yunger et al. 2004) have advanced a multidimensional model of gender identity. It sought to move beyond earlier models of children’s and adolescents’ gender identity based on self-ratings of gender-typed traits and interests, which the respondent may or may not associate with their gender identity. Instead, the new model focuses on people’s evaluations of their gender identity. Since Perry and colleagues’ introduction of the multidimensional gender identity model, felt gender typicality and felt gender conformity pressure are two factors that have received particular attention among various researchers. Felt gender typicality refers to how similar or different one perceives oneself compared to other group members, whereas felt conformity pressure reflects the degree to which one perceives strong expectations to adhere to gender-role norms. Egan and Perry (2001) established good evidence for the independence, reliability, and predictive validity of scales to measure felt gender typicality and felt conformity pressure in a sample of American children from grades 4 through 8. Perry and his colleagues as well as other investigators have subsequently used both constructs to examine gender identity in studies of children and adolescents (e.g., Drury et al. 2013; Jewell and Brown 2014; Leaper and Brown 2008; Patterson 2012; Smith and Leaper 2006; Yu and Xie 2010; also see Tobin et al. 2010). In addition, a few studies have started to consider these facets of gender identity in adults (DiDonato and Berenbaum 2011; Dinella et al. 2014; Leaper and Van 2008; Tate et al. 2015). Researchers are also beginning to apply the construct of felt typicality to other social identities. A recent study of Mexican-origin adolescents found that having a typical ethnic physical appearance moderated the positive relationship between ethnic typicality and feelings toward their ethnic group (Santos and Updegraff 2014).

In advancing their construct of gender typicality, Perry and colleagues (e.g., Egan and Perry 2001; Tobin et al. 2010) noted that Spence (1985; Spence and Buckner 1995) had previously postulated that individuals tend to make summary evaluations of the relative congruency between their gender-related attributes and gender-role expectations. Some individuals may see themselves as possessing many attributes associated with their ingroup, whereas others may see themselves as not very typical of their ingroup. People who view themselves as atypical might be expected to experience high conformity pressure. However, Egan and Perry (2001) found these two constructs were independent and uncorrelated when establishing the validity of their gender identity model in a sample of American elementary-school children. Thus, among children who view themselves as gender typical, some may experience conformity pressure whereas others may not–perhaps partly related to whether the child has internalized gender norms (e.g., see Smith and Leaper 2006). In the present research, we sought to extend these two constructs to studying social identities more generally in adults—including both ethnic–racial identity and gender identity.

Associations Between Ethnic–Racial and Gender Identities

Prior research suggests that youth tend to follow parallel trajectories in identity exploration and commitment across different domains (Cooper and Grotevant 1987; Grotevant et al. 1982; Marcia 1980). These approaches, however, have generally not considered multiple dimensions within particular identity domains. Ethnic identity and gender identity are salient, social identities and share many similar characteristics (e.g., self-categorization, prototypicality, attachment to in-group members; see Ashmore et al. 2004). Accordingly, we hypothesized that many of the dimensions of ethnic–racial identity and gender identity would be positively and moderately correlated. For example, many individuals may feel typical of their ethnic–racial ingroup as well as their gender ingroup. But others may feel more typical of their gender ingroup than their ethnic–racial group (or vice versa). Also, some dimensions may be more strongly associated across identity domains than others. For example, people may be more likely to feel typical of both their gender and their ethnicity/race because they are generally accepted among their peers; but they may be more likely to differ in the relative centrality they attach to one identity domain over another.

Summary

The model of social identity that we investigated included 5 dimensions. Three factors were based on Cameron’s (2004) review: centrality, ingroup ties, and ingroup affect. Another two components were based on Perry and colleagues’ (e.g., Egan and Perry 2001) work: felt ingroup typicality and felt ingroup conformity pressure. We consider these dimensions of social identity to complement and build on each other. Together they cover cognitive-affective dimensions that are central aspects of identity (Tajfel 1981). Our first set of analyses tested the independence of these five dimensions separately for gender identity and ethnic–racial identity. At the same time, consistent with observations from Cameron (2004) as well as Egan and Perry (2001), we expected there would be moderate associations among the 5 dimensions within each identity domain. Further, we hypothesized moderate correlations between identity domains (gender and ethnicity/race) for each identity dimension. For example, many (but not all) people who feel typical of their gender may be likely to feel typical for the ethnic–racial group; that is, some people may generally feel a sense of normativity about themselves (e.g., DiDonato and Berenbaum 2013).

Dimensions of Ethnic–Racial and Gender Identity and Self-Esteem

Upon establishing support for the coherence of the 5-factor model of social identity for ethnicity/race and gender, our next goal was to test its predictive validity. We chose self-esteem as a candidate outcome measure. As summarized in several reviews, strong social identities are often associated with more positive self-esteem and well-being (e.g., Aberson et al. 2000; Ashmore et al. 2004; Greenwald et al. 2002). This pattern has been indicated in studies that specifically documented positive associations (a) between ethnic–racial identities and self-esteem in adolescents and adults (see Smith and Silva 2011, for a meta-analysis; average r = .21) as well as (b) between gender identities and self-esteem in children (e.g., Egan and Perry 2001) and adults (e.g., DiDonato and Berenbaum 2011).

Relations of Each Identity Dimension to Self-Esteem

Centrality

Centrality is one of the facets of identity that has been often explored. Prior studies suggest ethnic–racial centrality can be protective against negative outcomes (Sellers et al. 1998a). Research suggests that incorporating one’s ethnic–racial group into one’s self-definition can have a positive impact on psychological well-being in adolescents and adults (Maxwell Johnson et al. 2005; Phinney et al. 1997). Centrality is similarly viewed as relevant for one’s gender identity (e.g., Tobin et al. 2010). Hence, we hypothesized that individuals who consider their ethnic–racial identity or gender identity as central to their self-concept would be more likely to demonstrate higher self-esteem.

In-Group Affect

In-group affect reflects the positive or negative emotions one feels about being a group member. Prior studies of adolescents and undergraduates have found high private regard (i.e., ingroup affect) toward one’s ethnic–racial identity was associated with higher feelings of self-worth for some ethnic–racial groups (e.g., Maxwell Johnson et al. 2005; Rivas-Drake et al. 2008). In addition, researchers examining gender identity similarly noted an association between adolescents’ feelings of contentment toward their gender in-group and self-esteem (e.g., Menon 2011). Thus, we hypothesized that emerging adults’ in-group affect regarding their ethnic–racial or gender groups would be positively related to their self-esteem.

In-Group Ties

Ingroup ties reflect one’s attachment and sense of belongingness to a group. Decades ago, Maslow (1968) proposed our sense of belongingness as a fundamental basis for self-esteem and self-actualization. This contention has been supported in several empirical studies testing for associations between adults’ feelings of ingroup ties and self-esteem (see Baumeister and Leary 1995; Cameron 2004). For example, prior studies observed positive correlations between ingroup ties (or belongingness) and self-esteem when evaluating ethnic identity among adolescents and adults (e.g., Phinney 1992; Romero and Roberts 2003; Seaton et al. 2006). In the present study, we hypothesized that stronger ingroup ties to one’s ethnic–racial or gender groups would be associated with higher self-esteem.

Felt Typicality

The potential relation between felt typicality and self-esteem is explicated in Greenwald et al.’s (2002) balanced identity theory. In their model, they distinguish between self-concepts (associations between self and attributes; e.g., “I like romance novels”) and stereotypes or attitudes (associations between group identities and attributes; e.g., “women, but not men, like romance novels; men should not like romance novels”). Balance ensues when individuals’ self-concepts are congruent (i.e., felt typicality) with the stereotypes and attitudes tied to their social identities (e.g., “I like romance novels,” “I am a woman,” and “Women like romance novels”), and positive self-esteem becomes more likely. Conversely, imbalance occurs when there is a mismatch (i.e., feeling atypical) between self-concepts, social identities, and stereotypes/attitudes (e.g., “I like romance novels,” “I am a man,” “Men do not like romance novels”); consequently, self-esteem may suffer (Greenwald et al. 2002; also see Tobin et al. 2010).

Studies provide support for a link between felt typicality and self-esteem. For example, Patterson and Bigler (2007) found that children who felt typical of their peer in-group were happier with their group identity than those who felt atypical of their peer group. In addition, researchers have observed positive associations between self-perceived gender typicality and self-esteem in studies of children, adolescents, and adults (e.g., DiDonato and Berenbaum 2011; Egan and Perry 2001; Jewell and Brown 2014; Leaper and Brown 2008; Smith and Leaper 2006). In an analogous manner, individuals who see themselves as typical of their ethnic–racial group may also tend to have stronger feelings of self-worth. Thus, we hypothesized that self-perceived ethnic–racial typicality and gender typicality would be positively associated with self-esteem.

Felt Conformity Pressure

When individuals experience conformity pressure, it implies they may not see themselves as fitting into the norms for a particular social identity. Not surprisingly, therefore, in a sample of preadolescents, Yunger et al. (2004) found that those who felt less gender typical and experienced higher pressure to conform were more likely to exhibit internalizing problems. Furthermore, felt gender-conformity pressure was negatively associated with adolescents’ self-esteem (Egan and Perry 2001). Individuals’ self-esteem during adolescence and adulthood may also suffer when they experience intra-group conformity pressures regarding their ethnic–racial group (Carter 2006; Contrada et al. 2000; Murray et al. 2012). Moreover, research on adults suggests that members of ethnic–racial minority groups may feel greater pressure to conform to ethnic–racial roles (see Contrada et al. 2000). In sum, we hypothesized that felt conformity pressure regarding one’s ethnic–racial ingroup or one’s gender would each be negatively related to self-esteem.

Relative Relations of Identity Dimensions to Self-Esteem

In some prior studies, the associations between social identity and self-esteem were not always consistent (e.g., Sellers and Shelton 2003). Accordingly, researchers have argued for the need to consider possible moderators of the links between social identities and self-esteem (see Rubin and Hewstone 1998). Along these lines, a multidimensional approach to studying social identities may prove useful; that is, some dimensions of social identity may better predict self-esteem than others. Additionally, taking into account multiple identities at once (e.g., both ethnicity/race and gender) may better account for individuals’ self-esteem (Brook et al. 2008).

Most of the research on ethnic–racial identity has focused on African Americans, Latinos, and Asian Americans (Rivas-Drake et al. 2014). In their recent meta-analysis on ethnic–racial identity and adjustment, Rivas-Drake and colleagues found that most studies find a positive correlation between ethnic–racial identity and self-esteem among ethnic–racial minority youth. Few studies, however, have investigated the relationship between ethnic–racial identity and self-esteem among White European Americans. One exception found that ethnic identity and American identity positively predicted self-esteem among European American adolescents (Phinney et al. 1997). However, these students were in the ethnic–racial minority of their school, which may have made their ethnicity/race particularly salient. Thus, the generalizability of this finding is unclear.

Few investigators have previously examined the relative contributions of different identity dimensions to self-esteem. This approach was taken, however, in Cameron’s (2004) work testing the predictive validity of the three-factor model. In one of his studies, he examined undergraduates’ social identities regarding their university affiliation. In a multiple regression, ingroup ties and ingroup affect, but not centrality, were related to self-esteem. (Centrality was associated with other personality measures.) Cameron’s (2004) findings illustrate how some identity dimensions may be better than others in predicting particular outcomes such as self-esteem. However, it is unclear if the patterns seen in Cameron’s (2004) research on social identities based on university affiliation would extend to social identities based on ethnicity/race or gender. One pertinent study did compare multiple dimensions of racial identity in relation to self-esteem. Rowley et al. (1998) considered different subscales on the MIBI (centrality, private regard, and public regard) in relation to self-esteem among African American high school and college students. Results indicated positive associations with self-esteem for private regard but not for centrality or public regard.

To the extent that the identity dimensions are independent constructs, we expected there might be variation in how well some dimensions underlying ethnic–racial or gender identity might predict self-esteem. Furthermore, the relations between particular dimensions of ethnic–racial identity or gender identity to self-esteem may vary depending on the individuals’ own gender or ethnicity/race (see Maxwell Johnson et al. 2005). For example, gender centrality may be tied to self-esteem in White European American women more than for ethnic-minority women (see Turner and Brown 2007). Little research has previously examined multiple dimensions of both ethnic–racial identity and gender identity in relation to self-esteem in samples of women and men from different ethnic–racial groups. Therefore, we conducted this set of analyses for exploratory purposes and did not advance specific hypotheses.

Combined Relations of Ethnic–Racial and Gender Identities to Self-Esteem

The final goal of our research was to examine the combined relations of ethnic–racial and gender identities to self-esteem. We have noted how the intersectional approach underscores the importance of considering the influence of multiple identities simultaneously in persons rather than focusing on only one social identity (Brook et al. 2008; Cole 2009; Kiang et al. 2008; Stirratt et al. 2008). Whereas several studies have considered either gender identity or ethnic–racial identity in relation to self-esteem, relatively few investigations have considered both identity domains together. However, there are some pertinent studies conducted with adolescents that found both ethnic identity and gender identity independently and additively contributed to adolescents’ self-esteem (DuBois et al. 2002; Rogers et al. 2015).

Building on this prior work, we hypothesized that a combination of having a positive ethnic–racial identity and a positive gender identity would additively predict emerging adults’ self-esteem. To explore this hypothesis, we contrasted emerging adults that were high or low (above or below the sample median) in each identity dimension (i.e., high–high, low–low, high–low, low–high). We predicted that, with the exception of felt pressure, those who were high in a particular dimension (centrality, ingroup ties, ingroup affect, or typicality) for both ethnic–racial identity and gender identity would report the highest average self-esteem, whereas those who were low in a dimension for both identity domains would report the lowest average self-esteem. Conversely, for felt pressure we expected the relationship to be reversed (i.e., low–low group would have highest esteem).

Present Study

Our study built on prior research utilizing multidimensional models of ethnic–racial identity and felt gender identity. We started with Cameron’s (2004) three-factor model of social identity that includes centrality, in-group ties, and in-group affect. We sought to expand this model by adding typicality and felt conformity pressure from Perry and colleagues’ (Egan and Perry 2001; Tobin et al. 2010) model of gender identity. The five factors were tested as predictors of self-esteem for both ethnic–racial identity and gender identity in a sample of emerging adults attending a public university in the United States. Relatively few studies have examined both ethnic–racial identity and gender identity together. Moreover, prior studies utilizing the multidimensional model of gender identity have primarily focused on children and adolescents (e.g., Egan and Perry 2001; Leaper and Brown 2008), whereas studies of ethnic–racial identity have considered mostly adolescents and young adults (e.g., Phinney 1992; Syed and Azmitia 2008). Furthermore, even less work has considered the intersection of ethnic–racial and gender identities in relation to people’s outcomes, such as self-esteem.

Our sample was comprised of young women and men who self-identified as Asian American, Latino/a, or White European American. We conducted the following analyses. First, we tested the independence and internal reliability of each of the five dimensions separately for ethnic–racial identity and gender identity. Consistent with prior investigations, we predicted moderate correlations among the dimensions within each identity domain. Second, we examined the associations between ethnic–racial identity and gender identity for each dimension. We hypothesized each dimension would be moderately associated between the two identity domains. Third, we tested the degree to which the identity dimensions for ethnicity/race and gender predicted self-esteem. We hypothesized self-esteem would be (a) positively correlated with centrality, ingroup ties, ingroup affect, and typicality; and (b) negatively correlated with felt conformity pressure. Fourth, we examined if the associations between identity dimensions and self-esteem varied depending on the intersection of participants’ ethnicity/race and gender. These were exploratory analyses and we did not advance any specific hypotheses. Finally, we tested the combined relationships of ethnic–racial identity and gender identity dimensions to self-esteem. We hypothesized those who scored high in each identity dimension for both identity domains would indicate the highest average levels of self-esteem; however, we predicted the opposite association regarding felt conformity pressure.

In our analyses, we took into account whether the participants’ parents were recent immigrants to the United States as well as the parents’ education. Traditional cultural expectations (including traditional gender roles) may be more common among recent immigrants to the United States. Conversely, traditional attitudes tend to be less likely among parents with higher education levels (e.g., Leaper and Valin 1996; Qin 2009).

Method

Participants

We recruited 848 undergraduates enrolled in psychology classes at a California public university as potential participants. The ethnic–racial demographics of undergraduates enrolled at the university include White European American (39 %), Latino (28 %), Asian (21 %), Multiracial (6 %), African American (2 %), and Other (4 %). Although the undergraduates on the campus are relatively gender balanced (52 % Women, 48 % Men), the psychology major is disproportionately female (72 %).

Because the study was examining ethnic–racial identity during emerging adulthood, we excluded 169 participants (74 % female) from the current analyses if either (1) they self-identified to an ethnic–racial group with low occurrences (n = 15 African American; n = 11 Middle Eastern); (2) they self-identified as having multiple ethnic–racial identities (n = 67), with some exceptions (described below); or (3) they were older than 23 years of age (n = 62).

The effective sample was 662 young adults (75 % female, 25 % male) ranging in age from 18 to 23 years (M = 19.86, SD = 1.28). The three ethnic–racial groups examined in the present study were White European American (45.3 %), Latino/a (28.2 %), and Asian American/Pacific Islander (26.4 %) participants. Participants were allowed to check all of the ethnic–racial groups of which they were members. Those who indicated that they were members of more than one ethnic–racial group were asked to report if they identified more strongly with one of their ethnic–racial groups. Multi-ethnic individuals who reported that they more strongly identified with one ethnic–racial group that included White European American (n = 85), Latino/a (n = 85), or Asian American (n = 13) were classified as that respective ethnic–racial group. Individuals who indicated that they were multi-ethnic and did not report that they more strongly identified with one of these three ethnic–racial groups were not included in the current sample.

The majority of Latino (66 %) and Asian American (71 %) participants were second-generation immigrants (i.e., participant was born in the United States and one or both parents were born outside of the United States). There was also socioeconomic diversity among participants as measured by parents’ education. The range of mothers’ highest education level was as follows: 17 % with graduate degree, 22.9 % with bachelor’s degree, 41.5 % with high school diploma, and 18.9 % with no high school diploma. Fathers’ highest education level ranged from: 21 % with graduate degree, 21.4 % with bachelor’s degree, 36.7 % with high school diploma, and 20.5 % with no high school diploma.

Procedure

Participants completed an online survey. As described below, it included multidimensional measures of identity for ethnicity-race and gender and a measure of self-esteem.

Measures

Parents’ Education

Participants indicated their mother’s and father’s highest education level using the following rank-ordered scale: 1 = elementary school, 2 = some high school, 3 = high school graduate, 4 = some college, 5 = college degree (bachelor’s), 6 = some graduate school, 7 = graduate degree (master’s, doctorate, medical, law).

Self-Esteem

A short version of Rosenberg’s (1979) Self-Esteem Scale was used to measure self-esteem (e.g., “On the whole, I am satisfied with myself”). Participants rated 10 items on a 4-point scale (1 = strongly disagree to 4 = strongly agree). The scale had satisfactory internal consistency (α = .90).

Multidimensional Measures of Identity

To create an integrated multidimensional model of identity, we combined elements from Cameron’s (2004) social identity scales (centrality, in-group affect, in-group ties) and Egan and Perry’s (2001) gender identity scales (felt typicality, felt conformity pressure). Centrality refers to the perceived importance of a particular social identity to one’s self-concept. In-group affect is the sentiment one has about belonging to a group. In-group ties reflect the extent to which one feels connected to group members. Felt typicality refers to how representative one perceives the self in relation to other group members. Felt conformity pressure reflects the degree that one feels pressured to adhere to the social norms of a group. Each of the five dimensions (i.e., scales) was measured separately for ethnic–racial identity and gender identity. For items measuring ethnic–racial identity, the reference group was “people within my ethnic/racial group” (e.g., “I have a lot in common with other people within my ethnic/racial group”). For items measuring gender identity, the reference group was same-gender undergraduates (e.g., “I have a lot in common with other undergraduate [women/men]”). Each item was rated on a 5-point scale (1 = disagree strongly to 5 = agree strongly). The items and alpha coefficients for each scale are presented in Table 1.

Table 1 Items in multidimensional identity measures and alpha indices of internal consistency
Full size table

Two items for felt conformity pressure from Egan and Perry’s (2001) original scale were not used. Upon inspection, the wording of the items “I don’t like people within my [identity in-group] who sometimes do things that [identity out-group] usually do” and “I think my parents would be upset if I wanted to do activities [identity out-group] usually do” did not appear to measure felt conformity pressure; rather, they assessed evaluations or affect towards atypical group behavior.

We performed two sets of exploratory factor analyses (principal axis factor extraction with direct oblimin rotation) with all items for gender identity and ethnic–racial identity. Exploratory factor analyses revealed satisfactory factor loadings for each dimension of ethnic–racial identity and suggested a 5-factor model. However, original factor analyses for gender identity revealed less clear loadings across identity dimensions. Items for gender typicality and gender in-group ties loaded on the same factor, suggesting they were not independent constructs. Closer inspection of the factor loadings for these two dimensions suggests that the typicality item “I don’t feel I fit in with other [identity in-group]” was more closely related to in-group ties. Further, the in-group ties item “I have a lot in common with other [identity in-group]” appeared more appropriate for typicality. Therefore we switched these two items for both gender identity and ethnic–racial identity to better reflect the distinctions between in-group ties and typicality. The changes are reflected in the presentation of items in Table 1.

With a few exceptions, loadings for each factor were at least .40 and in expected directions. However, factor loadings for felt gender-conformity pressure loaded on two separate factors. Questions concerning felt pressure to conform from parents loaded on to one factor, and conformity pressure from peers loaded on to a separate factor. However, prior studies have combined them together (e.g., Corby et al. 2007), and the reliability of the combined construct for our sample was satisfactory (α = .84). Therefore, we analyzed felt conformity pressure as one construct.

Table 7 in Appendix 1 summarizes factor loadings for gender identity items, and Table 8 in Appendix 2 includes loadings for ethnic–racial identity items.

Results

Gender Identity and Ethnic–Racial Identity Correlations

Two sets of bivariate correlations were conducted to test the associations among the five investigated identity dimensions (centrality, in-group affect, in-group ties, self-perceived typicality, and felt conformity pressure). In the first set of analyses, we tested the degree to which the five dimensions correlated with one another within each identity domain (ethnic–racial identity and gender identity). These results are presented in Table 2.

Table 2 Intercorrelations between identity dimensions within ethnic–racial and gender identity domains
Full size table

Many of the dimensions were significantly correlated. The magnitudes of most coefficients were moderate, with the strongest associations seen between typicality and ingroup ties for both identity domains. The dimensions therefore appeared to be related—yet independent—facets of identity, as we expected.

The second set of analyses tested for correlations between ethnic–racial identity and gender identity domains for each dimension. These results are summarized in Table 3. For each of the five identity dimensions, scores for ethnicity/race and gender were significantly and positively correlated. When examining the entire sample, the largest association was seen between felt gender conformity pressure and felt ethnic–racial conformity pressure (r = .60). Otherwise, the other correlations were moderate in magnitude. Thus, as expected, people’s ethnic–racial identity and gender identity were somewhat but not strongly related.

Table 3 Bivariate correlations between gender identity and ethnic–racial identity dimensions for overall sample and by participant gender and ethnicity
Full size table

When each identity dimension was broken down by participants’ gender and ethnicity/race, different patterns of association were implicated (see Table 3). For example, the association between felt gender pressure and felt ethnic–racial pressure was relatively strong for Asian American women (r = .66) but comparatively modest for Asian American men (r = .24). Thus, the relationship between gender identity and ethnic–racial identity appeared to vary depending on both the gender and the ethnicity/race of the participants.

Tests for Average Group Differences in Identity Variables

The following set of exploratory analyses tested for group differences based on participants’ gender and ethnic–racial backgrounds. Although we did not advance any hypotheses, these analyses were conducted to probe into possible ways that the identity dimensions varied by gender and ethnicity/race. Table 4 includes means and standard deviations for the gender identity and the ethnic–racial identity dimensions. Means are presented for the combined sample and also broken down by participants’ gender and ethnicity/race (Latino/a, Asian American, or White European American). To test for average group differences in these measures, we conducted 2 (gender) × 3 (ethnic/racial group) MANCOVAs separately with the ethnic–racial identity dimensions and the gender identity dimensions. As previously described, the identity variables had modest to moderate correlations, suggesting that MANCOVA tests were appropriate.

Table 4 Comparisons of ethnic–racial identity and gender identity dimensions for overall sample and by participants’ ethnic–racial group
Full size table

In the MANCOVA tests, we included parents’ highest education level and generational status as covariates in the model. We averaged the highest level of education for both parents to analyze parents’ education. To control for parents’ generational status, we included dummy coded variables separately for mothers and fathers (1 = parent born in the United States, 0 = parent not born in the United States). The covariates were not significant with either identity domain.

A significant multivariate effect for participant ethnicity/race occurred in both MANCOVAs. In addition, a significant multivariate effect for participant gender occurred in the MANCOVA with gender identity but not with ethnic–racial identity. The Gender × Ethnicity/Race interaction was significant for both ethnic–racial identity and gender identity. The corresponding univariate effects are summarized below.

Group Comparisons of Ethnic–Racial Identity

There was a significant multivariate effect for participant ethnicity/race regarding the ethnic–racial identity dimensions, Wilks’ λ = .87, F(10, 1100) = 7.76, p < .001, partial η2 = .07. There were not significant multivariate effects for participant gender, Wilks’ λ = .99, F(5, 550) = 1.57, p = .167, partial η2 = .01. The Gender × Ethnicity/Race interaction was significant, Wilks’ λ = .96, F(10, 1100) = 2.23, p = .014, partial η2 = .02.

While controlling for parent education and immigration status, univariate tests revealed significant main effects for participant ethnicity/race for each dimension except in-group affect. These results are presented in Table 4. On average, White European Americans scored higher on ethnic in-group ties than did Asian Americans (with Latino/a participants not differing from either group). Also, White European American undergraduates scored higher in felt typicality and lower in felt conformity pressure and centrality than did Latino/a and Asian American undergraduates. Latino/a undergraduates were higher on in-group affect than either Asian American or White European American undergraduates.

The only dimension associated with a significant Participant Gender × Participant Ethnicity/Race interaction was ethnic–racial in-group affect. To interpret the interaction effect, we conducted three follow-up univariate ANCOVAs to compare gender differences in ethnic–racial in-group affect for Asian Americans, Latinos/as, and White European Americans. Results suggested that gender moderated the relationship between ethnic–racial in-group affect and participants’ ethnicity/race. Women scored higher than men on ethnic–racial in-group affect only among Asian Americans and Latinas. This gender effect was not found among European American women and men.

Group Comparisons of Gender Identity

Significant multivariate effects in the MANCOVA testing the gender identity measures were indicated for participant gender, Wilks’ λ = .83, F(5, 548) = 22.54, p < .001, partial η2 = .17; and for participants’ ethnicity/race, Wilks’ λ = .95, F(10, 1096) = 2.66, p < .001, partial η2 = .02. Also, the Gender × Ethnicity/Race interaction was significant, Wilks’ λ = .97, F(10, 1096) = 1.87, p = .046, partial η2 = .02.

The univariate tests revealed significant gender effects with two gender identity measures. On average, women scored significantly higher on gender centrality (M = 3.22, SD = .80) than did men (M = 2.91, SD = .54), F(1, 552) = 11.64, p = .001, partial η2 = .02. Men reported more average felt conformity pressure (M = 2.72, SD = .76) than did women (M = 2.10, SD = .69), F(1, 552) = 76.56, p < .001, partial η2 = .12. There were no significant gender differences for gender in-group ties, gender in-group affect, or gender typicality.

The univariate tests of participants’ ethnicity/race in relation to the gender identity dimensions are summarized in Table 4. As seen in the table, significant group differences occurred with two of the measures. In-group affect was weaker among Asian American participants than among White European American and Latino/a participants. Also, felt gender conformity pressure was stronger and in-group ties were weaker among Latino/a and Asian American participants than European American participants. There were no significant ethnic–racial group differences associated with gender centrality, gender typicality, or gender in-group affect.

The effect of participants’ ethnicity/race on gender in-group ties depended on one’s gender. That is, White European American men scored higher on gender in-group ties than did Latino men and Asian American men. However, there was not a significant difference for gender in-group ties among women between ethnic–racial groups.

Relationship Between Identity Dimensions and Self-Esteem

Bivariate Correlations

Table 5 summarizes the Pearson bivariate correlations between identity dimensions and self-esteem. These correlations were performed separately for ethnic–racial identity and for gender identity. In addition to running these correlations for the entire sample, we performed them separately for women and men within each of the three ethnic groups. We hypothesized that individuals scoring higher on centrality, typicality, in-group affect, and in-group ties for ethnic–racial identity and gender identity would be more likely to score higher on self-esteem. In contrast, those scoring higher on felt conformity pressure in both measures would score lower on self-esteem.

Table 5 Bivariate correlations between identity dimensions and self-esteem for overall sample and by participants’ ethnic–racial group and gender
Full size table

When the combined sample was analyzed, four of the five ethnic–racial identity dimensions were significantly correlated with self-esteem in expected directions (see Table 5). Ethnic–racial centrality was not correlated with self-esteem. A similar pattern was seen when examining the gender identity dimensions. Except for gender centrality, each dimension was significantly associated with self-esteem. An inspection of Table 5 reveals some variations in the relations of particular identity dimensions for both ethnicity/race and gender to self-esteem that depended on the participants’ gender and ethnic–racial group. As described next, these patterns were further explored in regression analyses.

Hierarchical Regressions

To test for the independent contributions of each of the five identity dimensions to variations in self-esteem, two sets of hierarchical multiple regressions were conducted. One analysis examined the relation of the ethnic–racial identity dimensions to self-esteem, whereas the other analysis tested the relation of the gender identity dimensions to self-esteem.

In step 1, we entered participant gender, participant ethnic–racial group, and the Participant Gender × Participant Ethnicity/Race interaction. Participant gender was dummy coded (1 = women, 0 = men). Participant ethnic–racial group were entered as dummy coded variables: Latino (1 = Latino/a, 0 = not Latino/a), Asian American (1 = Asian American, 0 = not Asian American). Step 2 included the five identity variables (centrality, typicality, pressure, in-group ties, in-group affect). Step 3 included interaction terms for Participant Gender × Identity Dimension and Participant Ethnic–Racial Group × Identity Dimension. Finally, step 4 included the 3-way interaction terms for Participant Gender × Participant Ethnic–Racial Group × Identity Dimension. Tests revealed that all variables had a VIF below 6, which suggests that multicollinearity was not indicated and therefore multiple regression tests were appropriate.

Ethnic–Racial Identity and Self-Esteem

Only the first two steps significantly added to the model in the regression; therefore, step 2 was used as the final model, F(10, 624) = 7.83, p < .001, adjusted R 2 = .10. Two of the five dimensions were significant: in-group affect (β = .14, p = .005) and in-group ties (β = .15, p = .007). The other factors in the model were not significant: centrality (β = −.03, p = .459), felt pressure (β = −.07, p = .088), typicality (β = −.03, p = .597), Gender × Latino (β = −.029, p = .745), and Gender × Asian American (β = −.060, p = .458). We additionally conducted follow-up analyses to test whether there was a difference between Asian Americans and Latinos/as; none was indicated.

Gender Identity and Self-Esteem

As seen in the prior regression, only the first two steps significantly added to the model, and we focused on step 2 as the final model, F(10, 621) = 18.23, p < .001, adjusted R 2 = .21. Four of the five identity dimensions were significant in the model: typicality (β = .09, p = .047), centrality (β = −.08, p = .030), in-group affect (β = .27, p < .001), and in-group ties (β = .18, p < .001). The following factors were not significant: felt pressure (β = −.03, p = .510), Participant Gender × Latino (β = −.08, p = .320), and Participant Gender × Asian American (β = −.07, p = .349). We also ran follow-up analyses testing whether there was a difference between Latino/as and Asian Americans; none was indicated.

Intersection of Ethnic–Racial Identity and Gender Identity in Relation to Self-Esteem

In our last set of analyses, we considered whether ethnic–racial identity and gender identity might additively contribute to self-esteem. Using a median-split we classified individuals as high or low in each dimensions separately for ethnic–racial identity and gender identity. We subsequently contrasted those who were (1) high in both identity dimensions (e.g., high in ethnic–racial typicality and high in gender typicality), (2) high in the ethnic–racial identity and low in gender identity, (3) high in gender identity and low in ethnic–racial identity, or (4) low in both identity dimensions. We hypothesized individuals who were high in the identity dimensions (except for felt pressure) for both ethnicity/race and gender would score highest in self-esteem, whereas individuals who were low in both dimensions would score lowest in self-esteem. (The opposite pattern was predicted for felt pressure.)

For each identity dimension, we performed 4 (Identity intersection: high–high, high-low, low–high, low–low) × 3 (Participant ethnicity/race) × 2 Gender ANOVAs in relation to self-esteem. As summarized in Table 6, a significant main effect for the identity intersection factor was indicated with all of the identity dimensions except for centrality. In all of the significant effects, there was a consistent pattern for persons who were high in an identity dimension for both ethnicity/race and gender to indicate higher self-esteem than those scoring low in both identity domains. In contrast, those scoring high in a dimension for one identity domain and low in the other domain generally fell between these two groups in self-esteem.

Table 6 Self-esteem in relation to intersection of ethnic–racial (E–R) identity and gender identity in each dimension for overall sample
Full size table

The only significant interaction effect was with Participant Gender × In-group Affect, F(3, 620) = 4.65, p = .003, partial η2 = .022. The identity intersection factor for in-group affect was significant for both male participants, F(3, 149) = 4.79, p = .003, partial η2 = .088; and for female participants, F(3, 471) = 22.80, p < .001, partial η2 = .127. Among men, self-esteem was significantly stronger among those expressing high in-group affect in both identity domains than any of the other three groups. Among women, self-esteem was strongest among those who either indicated high in-group affect for both ethnicity/race and gender or who scored high in-group affect for gender and low in-group affect for ethnicity/race; and self-esteem was lowest among women who either (a) scored low in-group affect regarding both ethnicity/race and gender or (b) scored low in-group affect for gender and high in-group affect for ethnicity/race. Thus, for men, it would appear that in-group affect for ethnicity/race and gender had an additive impact on their self-esteem; for women, it appeared that in-group affect for gender was important while in-group affect for ethnicity/race was negligible.

Discussion

Existing measures of gender identity (Egan and Perry 2001) and social identity (Cameron 2004) share complementary constructs and distinctions that lend themselves to integration. Bridging aspects of these measures of identity potentially offer a better understanding of how different identity dimensions relate to outcomes among diverse emerging adults. Additionally, examining how ethnic–racial identity and gender identity relate to outcomes across ethnic–racial groups and genders is important. For example, previous research reports conflicting findings concerning the effect of ethnic–racial identity on self-esteem (Sellers and Shelton 2003; Phinney et al. 1997). However, certain aspects of identity may be more important for self-esteem than others. Furthermore, these associations may also vary based on one’s ethnicity/race or gender. Finally, strong identities in two domains, such as ethnicity/race and gender, may contribute additively to one’s self-esteem. Our study sought to address these issues and questions.

In our study, we adapted and validated multidimensional measures of ethnic–racial identity and gender identity by integrating constructs from prior work on social identity (Cameron 2004) and gender identity (Egan and Perry 2001). We included three dimensions of Cameron’s (2004) social identity measure (centrality, in-group affect, in-group ties) and two dimensions of Egan and Perry’s (2001) gender identity measure (felt typicality, felt-pressure). We adapted each of these dimensions for ethnic–racial identity and gender identity. We view the dimensions that we investigated as complementary (rather than contradictory) with those advanced in other models (e.g., Sellers et al. 1998b). As discussed later, we believe additional identity dimensions are worth exploring in future research.

Reliability and Correlations Among Dimensions

Our first goal was to test the independence and reliability of five dimensions of gender identity and ethnic–racial identity separately. As expected, our analyses indicated centrality, in-group affect, in-group ties, typicality, and felt pressure were interrelated yet independent dimensions within each identity domain (gender and ethnicity/race). Most of the correlations among dimensions were significant but moderate in magnitude. This suggests that these dimensions are appropriate adaptations for ethnic–racial identity and gender identity.

Within both identity domains, the two dimensions that were most strongly associated were ingroup ties and felt typicality. These constructs differ in emphasis. Whereas ingroup ties reflect people’s sense of connection to the in-group, felt typicality refers to how similar they consider themselves to other in-group members. Understandably, the two constructs are likely related (e.g., people who feel similar to group members are apt to feel a connection to them). Nonetheless, the regression analyses indicated these two dimensions independently contributed to self-esteem (discussed later).

The second goal was to investigate the relationship between ethnic–racial identity and gender identity domains. In support of our hypothesis, we observed moderate and positive correlations between the two identity domains for each dimension. These findings suggest that gender identity and ethnic–racial identity are related yet independent constructs. For example, the moderate correlation (r = .35) between gender typicality and ethnic–racial typicality for the overall sample suggests that participants had separate understandings of what it means to be typical of their gender and their ethnic–racial group. Because many people likely view themselves as typical for both groups (or atypical for both groups), the two dimensions are positively correlated. However, the magnitude of the correlation implies that there were also many individuals who might view themselves as more typical for one group than the other. Rogers et al. (2015) similarly found moderate, positive correlations between gender identity and racial identity dimensions (centrality and private regard) among African American male adolescents.

The strongest correlation between ethnic–racial identity and gender identity domains was for felt conformity pressure (r = .60). Individuals who encounter many expectations for group behavior may feel a global sense of conformity pressure spanning across domains. As discussed below, cultural expectations of behavior may explain group differences in the correlation for felt conformity pressure between ethnic–racial identity and gender identity.

It is important to note that associations between gender identity and ethnic–racial identity varied depending on participants’ gender and ethnicity/race. For example, we found small differences for the association between gender identity dimensions and ethnic identity dimensions among Asian American men, but significant associations among White European American men. This may reflect cultural differences in what it means to be a man. For instance, in a study among Asian American men and European American men in early adulthood, Chua and Fujino (1999) found that, among European American men, views of masculinity were associated with power (e.g., independence, dominance) and in opposition to femininity. Asian American men were more likely than European American men to associate communal traits with masculinity (e.g., caring, politeness, reliability). Thus, ideas of gender and ethnicity/race might be more likely to overlap for White European American men than they are for Asian American men.

We also observed meaningful within-group differences. For example, there was a stronger correlation between ethnic–racial in-group ties and gender in-group ties among Latino men than among Latina women. There was also a significant correlation between felt conformity pressures for gender identity and ethnic–racial identity among Asian American women–but not among Asian American men. For Asian American women, pressures to conform to gender roles and ethnic–racial roles may have been more intertwined than they were for Asian American men. In Qin’s (2009) study among Chinese immigrant adolescents, both boys and girls negotiated between conflicting cultures of home and school when constructing their identity. However, girls were more likely to adhere to Chinese traditions in school, whereas boys were more likely to actively resist academic stereotypes of Chinese students by participating in non-academic activities to fit in with peers. The results from Qin’s research and our own study suggest how cultural expectations associated with one’s ethnic group might have different outcomes for young women and men. Further consideration of the intersection of gender and ethnicity/race in future studies will help us better understand these potential conflicts across identity domains.

Group Differences in Identity Variables

Our findings further pointed to group differences based on participants’ gender andethnicity/race in relation to particular identity dimensions. These analyses were exploratory, and we did not advance hypotheses for these group differences. However, we observed some interesting differences that we discuss below.

Ethnic–Racial Identity

We found that ethnic–racial identity varied by participants’ ethnic–racial group. Two patterns are notable. First, White European Americans scored higher in ethnic–racial typicality than did Latino/a or Asian American participants. There may be more intragroup variability in ethnic–racial identity for members of ethnic–racial minority groups than for ethnic–racial majority group (Celious and Oyserman 2001; Phinney 1996). If so, there may be a greater likelihood of not fitting into certain in-group norms associated with ethnic–racial minorities. Studies suggest there are often more specific norms associated with group typicality among ethnic–racial minorities. For instance, in a study among Latina/o and African American high school students, participants indicated that in-group peers are often judged based on adherence to typical speech styles, dress, and music tastes (Carter 2006). Many of these behaviors that denote typicality are often based on stereotypical or culturally normative ideas for what it means to be a member of one’s ethnic–racial group. Although there are stereotypes related to what it means to be European American (Ghavami and Peplau 2012), there is little knowledge concerning whether these stereotypes serve as a basis for perceiving oneself as typical or a basis for in-group conformity pressure.

In addition, Asian Americans and Latino/as expressed higher felt ethnic–racial conformity pressure than did White European Americans. Felt conformity pressure is common in many groups. Individuals are more likely to favor in-group members who exhibit normative group behavior (Marques et al. 1998), and those who do not conform may encounter in-group discrimination or feel disconnected from the group (Oyserman et al. 2006). Felt pressure to conform may also intersect with generation status. Latino/a youth who are first-generation college students may feel disconnected from their communities because they are in college (e.g., Cooper et al. 2002). Thus, they may feel greater pressure to prove they are still connected to their friends and family.

Gender Identity

Our analyses of gender identity similarly indicated some variations based on the intersection of participants’ gender and ethnic–racial group. On average, men scored lower on gender centrality than did women. This is expected, as those in majority groups may be less likely to think about their identity because it is usually less salient in their environment (Bigler and Liben 2007). Also, men tended to score higher on felt gender conformity pressure than did women. Prior research (e.g., Dinella et al. 2014; Egan and Perry 2001; Leaper 2015) has also found stronger felt conformity pressures among boys and men than girls and women. Additionally, men and boys may feel pressure to be accepted by in-group peers because group membership is often important for self-definition. Our finding supports previous research that found Latino adolescent boys who did not feel typical of their ethnic–racial group felt pressure to behave in more masculine ways to increase their peer acceptance among other boys (Oyserman et al. 2006). Future studies might include measures of the specific ways in which some women and men feel pressure to conform to group expectations.

We also observed ethnic–racial group variations in gender identity. Gender in-group affect and gender in-group ties were weaker among Asian Americans than the other ethnic–racial groups. The ethnic–racial composition of the university may be a factor in this finding. Participants were asked to compare themselves to other undergraduate peers. Because Asian Americans in our sample were in the ethnic–racial minority on their campus, they might not have felt connections to other undergraduate women and men who did not share their ethnic–racial group membership. Furthermore, White European American men are often represented as the norm in society because they reflect the ethnic–racial majority group as well as the dominant gender group; hence, these men may solely consider other European American men as their gender in-group (Ghavami and Peplau 2012). However, when ethnic–racial minority men are considering their gender in-group, they may think either about all other men or about other men within their ethnic–racial group. Because of this reference group variation, there may be a greater likelihood that ethnic-minority men may feel less cohesion and in-group ties to other men.

In addition, we found that Latino/a and Asian American participants reported higher felt gender-conformity pressure than did White European Americans. (As discussed earlier, an analogous pattern was also found regarding ethnic–racial identity.) In a sample of American fifth graders, Corby et al. (2007) found a similar pattern of higher felt gender-conformity pressure among ethnic-minority children (Latinos and African Americans) than White European American children. Variations in the meaning associated with what it means to be a woman or a man within particular ethnic–racial groups may help to explain this difference. For instance, studies assessing the intersection of gender and ethnic–racial identity suggest that there is a great deal of variation in the meaning associated with what it means to be a Latino man (Torres et al. 2002). Furthermore, compared to other ethnic groups in the United States, Latinos/as and Asian Americans may face stricter expectations to adhere to traditional ideas of what it means to be a woman or a man (Addis and Mahalik 2003). Thus, these individuals may feel greater pressure to conform to gender-role expectations.

Correlations Between Identity Dimensions and Self-Esteem

To explore the predictive validity of identity dimensions for ethnicity/race and gender, we tested how well they were correlated with self-esteem. We hypothesized that self-esteem would be (a) positively correlated with centrality, ingroup ties, ingroup affect, and typicality; and (b) negatively correlated with felt conformity pressure.

Ethnic–Racial Identity

We found that most dimensions of ethnic–racial identity were associated with self-esteem in the bivariate correlations (with only centrality being unrelated). Furthermore, in regression analyses, we evaluated the relative contributions of the ethnic–racial identity dimensions to self-esteem. These results indicated two factors emerged as significant predictors: in-group affect and in-group ties. These two dimensions of identity relate to holding a sense of group belonging and pride. Prior research suggests group belonging and ties to peers are important for self-esteem among adolescents (Brown and Lohr 1987) and adults (Baumeister and Leary 1995). This may explain why ethnic–racial in-group affect and ethnic–racial in-group ties were especially associated with self-esteem in our sample. Recent research related to social identity theory supports this finding as well. For example, Bratt (2015) examined multiple social identities regarding family, ethnic, school-class, and nationality among ethnically diverse Norwegian adolescents. The measure of social identity emphasized one’s sense of belongingness with the ingroup. In a multiple regression, he found family identity was the only identity domain that independently predicted self-esteem.

Not all of the ethnic–racial identity dimensions independently contributed to self-esteem in the regression; however, different patterns may occur with other outcomes. For example, previous research has indicated that ethnic–racial centrality among adolescents was especially related to their perceptions of discrimination (Sellers and Shelton 2003) and academic achievement (Okeke et al. 2009). If identity dimensions were differentially related to various outcomes, it would further support the advantage of a multidimensional approach. Of course, additional theoretical work is needed to explain how and why specific identity dimensions would be related to particular outcomes.

Gender Identity

We also examined the dimensions of gender identity in relation to self-esteem. In the regression analysis, most of the dimensions independently contributed to self-esteem (with only felt conformity pressure being nonsignificant). The association for each of the significant factors in the model was in the expected direction except for centrality. Curiously, gender centrality was negatively related to self-esteem in the regression. These two factors were unrelated in the bivariate correlations. Hence, it may be that high gender centrality might put some individuals’ self-esteem at risk when they feel strong ties and affect with their gender in-group.

Overall, our findings expand on earlier studies linking gender identity to self-esteem (e.g., DiDonato and Berenbaum 2011; Egan and Perry 2001; Jewell and Brown 2014; Smith and Leaper 2006) through the inclusion of additional dimensions. Also, whereas most prior studies examining multiple dimensions of gender identity and self-esteem were conducted with children and adolescents (see Tobin et al. 2010), our study suggests these associations may continue into early adulthood.

Intersection of Participant Ethnicity/Race and Gender

Our fourth goal was to investigate how the relationship between identity and self-esteem differs based on the intersection of ethnicity/race and gender. We explored if and how the intersection of participants’ gender and ethnicity/race would moderate the associations between ethnic–racial identity and self-esteem. We found evidence suggesting possible ways this might occur. For example, among Latina women in our sample, multiple dimensions of gender identity were correlated with self-esteem. In contrast, ethnic–racial identity dimensions were not significantly associated with their self-esteem. In future research, including measures of caballerismo and marianismo may offer further insights into the relationship between the intersection of ethnicity/race and gender among Latino/a men and women (e.g., Arciniega et al. 2008).

Additive Contributions of Ethnic–Racial Identity and Gender Identity to Self-Esteem

In our last set of analyses, we tested the additive contributions of ethnic–racial identity and gender identity to self-esteem. As hypothesized, across most identity dimensions, those high in both ethnic–racial and gender identity domains scored higher in self-esteem than those low in both identity domains. In an analogous finding, Kiang et al. (2008) observed those scoring low across several identities (ethnic, American, family, religious) tended to have lower self-esteem than others.

We also noted an interaction effect, whereby the effect of in-group affect for both ethnicity/race and gender dimensions depended on participants’ gender. For men, there was an additive influence of in-group affect for ethnicity/race and gender on self-esteem. However, for women, levels of gender in-group affect appeared more strongly associated with self-esteem than did ethnic–racial in-group affect. Research with children found that gender identity was more salient for females than for males (e.g., Turner and Brown 2007). Perhaps for this reason, gender identity may be more strongly related to self-esteem for women than men among adolescents (e.g., see DuBois et al. 2002) and young adults. The intersection of ethnic–racial and gender identity may be especially relevant among ethnic-minority youth. In traditional ethnic-minority families, girls often face stricter parental supervision and expectations than do boys (Phinney 1990; Qin 2009; Suárez-Orozco and Qin 2006). Brook et al. (2008) have also argued that individuals’ well-being is affected by a combination of multiple identities. Furthermore, in a study of adolescent Black males, Rogers et al. (2015) observed that both racial private regard and gender private regard independently predicted self-esteem.

Directions for Future Research

The present study supports and builds upon previous multidimensional models of identity regarding ethnicity/race (e.g., Sellers et al. 1998b) and gender (e.g., Egan and Perry 2001). One potential advantage of the present model is that it allows researchers to consider analogous identity dimensions across different social identities such as ethnicity/race and gender. In future research, we recommend considering additional dimensions that might complement the ones that we examined. This might include intergroup bias (e.g., see Tobin et al. 2010). Furthermore, some dimensions or facets of identity need to be examined specifically in relation to particular types of social identities. These might include the ideologies or social norms associated with a particular ethnic–racial group (e.g., see Sellers et al. 1998b), a particular gender (e.g., Levant et al. 2007, 2010), or the intersection of gender and ethnicity/race (e.g., Arciniega et al. 2008; Liu and Iwamoto 2006).

We also recommend considering correlates of ethnic–racial identity during developmental periods and in populations that were not examined in the present study. First, we hope to see research examining these dimensions of ethnic–racial and gender identity during adolescence, as this developmental period is an important time of identity exploration (e.g., Cooper et al. 1998). We are aware of recent work exploring ethnic–racial identity among adolescents that used some of the dimensions that we employed (Santos and Updegraff 2014). Second, future work should explore the validity of our measure in other racial–ethnic groups. Of particular note, many of the issues concerning ethnic–racial identity and its effects appear especially important for many African Americans (e.g., see Byrd 2012). Also, exploring how these dimensions relate to multi-ethnic individuals is essential. Multi-ethnic individuals may experience issues with in-group ties, typicality, and conformity pressures in different ways than those who identify with one ethnic–racial group (AhnAllen et al. 2006; Root 1990). Third, we advise studying ethnic–racial and gender identities in different sociocultural contexts (see Gjerde 2014). Research suggests that identity salience often varies by context (Gjerde 2014; Phinney 1992; Steele and Aronson 1995). For example, the salience and meaning of one’s ethnic–racial identity may differ for undergraduates depending on whether their in-group constitutes a majority or minority at the college (e.g., MacDonald et al. 2007; Van Camp et al. 2009).

Our last recommendation is to consider the relations of the identity dimensions to other factors besides self-esteem. We observed some identity dimensions for ethnicity/race or gender were more strongly related to esteem than were other dimensions. We expect that different patterns would be seen depending on the kinds of outcomes, which could be one of the major advantages of multidimensional models of identity. Other relevant outcomes to consider in relation to the intersections of ethnic–racial and gender identities include academic achievement and experiences with discrimination (e.g., Byrd and Chavous 2011; Else-Quest et al. 2013; Leaper et al. 2012; Rowley et al. 2008; Syed et al. 2011; Umaña-Taylor et al. 2012).

Conclusion

Our findings indicate that an integration of constructs from measures of social identity (Cameron 2004) and gender identity (Egan and Perry 2001) is valid and useful for understanding outcomes among emerging adults. Our unified multidimensional model offers a more in-depth understanding of the relationship between ethnic–racial identity, gender identity, and self-esteem. Knowing how specific aspects of identity relate to outcomes will help researchers and practitioners understand and improve well-being and adjustment among youth and emerging adults. For example, our finding that in-group affect and in-group ties predict higher self-esteem suggests that fostering feelings of belongingness is especially important for positive adjustment. Furthermore, our research highlights how strong identities in multiple domains, such as ethnicity-race and gender, may additively contribute to positive outcomes.

There has been conflicting evidence concerning the effect of ethnic–racial identity and gender identity on adjustment across diverse groups (Phinney et al. 1997; Smith and Silva 2011). Our model suggests that examining underlying dimensions of identity will aid in understanding both similarities and differences among groups. A multidimensional approach to understand identity can illuminate important relationships between identity and development during adolescence and emerging adulthood across diverse groups that might otherwise be lost. Thus, in order to gain a more nuanced understanding of the effect of identity on developmental outcomes, it is important to investigate multiple dimensions of identity as well as how they differ across various groups.