Abstract
Recent studies have shown that racial identification varies across context and time. Although sociologists recognize many contextual factors associated with racial group membership, relatively little attention has been given to understanding the specific factors—such as self-perceptions, socioeconomic incentives, and family pressures—that relate to changes in racial self-identification, especially among children of immigrants individuals who may have a relatively high propensity for inconsistency in racial identification. Using two waves of data from the Children of Immigrants Longitudinal Study and guided by social identity theory, the authors seek to (1) explore the prevalence of changes in racial self-identification over time within this sample and (2) understand the mechanisms that may contribute to changes in self-identification. The results indicate that self-esteem, self-worth, and family cohesion are related to an individual’s reporting a change in racial identification between waves. Socioeconomic status and depression are not related to changes in racial identification.
Introduction
Recent sociological inquiry into the fluid nature of racial self-identification has revealed that some individuals change their racial categories over time (Doyle and Kao 2007; Saperstein and Penner 2010; Wilkinson 2010). For example, a recent study using internal Census Bureau data from the 2000 and 2010 censuses revealed that individuals who change their racial and ethnic survey responses over time are doing so in a variety of ways, for example, reporting multiple races in 2000 but a single race in 2010, changing from a single race to multiple races, or adding or dropping Hispanic group membership altogether (Liebler et al. 2014). Consistent with prior research, the analysis of census data linking individuals over time revealed that fluidity in identification is particularly common for multiracial individuals and individuals with mixed ancestries, including Latinos and some Asian groups (Doyle and Kao 2007; Saperstein and Penner 2012; Wilkinson 2010). Yet the incorporation of children of immigrants into the study of change in racial self-identification has been slower coming, despite the rapidly growing size of this population and numerous contextual influences that uniquely influence the racial identification process for children of immigrants.
How children of immigrants self-identify racially often reflects how they view themselves within the United States’ racial hierarchy. Identification may be highly variable among children of immigrants in part because of their age, as the transition from adolescence to early adulthood is often characterized by a plethora of changes related to psychosocial development, identity formation, and interaction with the external environment (Carter 2006). Susceptibility to peer pressure and the stresses of social and academic performance further increase pressures on children seeking to carve out identities (Erikson 1968). The self-identification of children of immigrants in the United States is also shaped by larger contexts, including a national racial structure quite distinct from central and southern America; transnational experiences that influence knowledge of, and experience in, another society; and shared experiences with co-ethnics in the United States (Tovar and Feliciano 2009). Finally, it has been suggested that children of Hispanic and Asian immigrant families have relatively high propensities for racial identification change, given that their physical characteristics are often unclear indicators of their race and ethnicity. (Saperstein and Penner 2012).
Although racial fluidity continues to garner significant interest in the literature, and some recent scholarship has explored the extent of ethnic identification fluidity among children of immigrants (Feliciano 2009; Tovar and Feliciano 2009), factors that might drive racial self-identification inconsistency are relatively understudied (Kramer, Burke, and Charles 2015; Telles and Paschel 2014), especially among children of immigrants. Indeed, academic debate continues as to whether change in racial self-identification is a result of real changes in people’s life experiences or simply an effect of changes in survey measurement.
Although we note that these debates are important, scholars have long recognized that racial identities are an “acquired and used feature of human identity . . . and subject to presentation, inhibition, manipulation, and exploitation” (Lyman and Douglass 1973:351). 1 Cooley’s (1902) “looking-glass self” has been the dominant theoretical perspective from which scholars have suggested that racial category switching might occur. From this perspective, racial switching—or inconsistent racial identification—may be motivated by a need to conform to others’ expectations; thus racial fluidity occurs in accordance with how people are perceived. Yet a recent study by Saperstein and Penner (2014) found that a substantial proportion of changes in racial identification could not be explained by this perspective alone. Saperstein and Penner (2012, 2014) suggested, as one possibility, that a more complete understanding of racial category switching should incorporate theories of self-directed racial selection, whereby racial or ethnic self-identification changes when individuals alter their presentation of self and conform to the behavior of other, often more advantaged groups, in order to claim membership in a higher-status population. Although this process has often been thought to be motivated by potential economic or political gains (Espiritu 1992), racial fluidity could also be driven by self-protective strategies and the management of self-esteem (Crandall et al. 2000) and, potentially, avoidance of stigma (Goffman 1963).
Our study offers an in-depth examination of multiple factors that could explain racial fluidity among children of immigrants, adding to existing studies of this topic, which have largely focused on the identification of black or black-white multiracial individuals (see Davis 1991; Harris and Sim 2002). Historically high levels of immigration from Asia and central and southern America have challenged existing racial boundaries in the United States (Bonilla-Silva and Dietrich 2009), necessitating an exploration of factors that might drive changes in racial identification among more recently arrived groups. To date, we are aware of a very limited body of literature that has explicitly investigated how self-perceptions, depression, and other factors such as socioeconomic status and family cohesion might drive children of immigrants to move between racial categories over extended periods of time. As a result, we investigate the nature and extent of changes within individual-level racial group identification using a multi-racial sample from the Children of Immigrants Longitudinal Study (CILS).
Theoretical Orientation: Social Identity Theory
Recent sociological research indicates that racial identification inconsistency is not uncommon (Telles and Paschel 2014; see also Korgen 1998). Racial category change has been particularly well documented among individuals with multiple known racial or ethnic origins, including members of some Hispanic and Asian groups in America (Liebler et al. 2014). The propensity for racial self-identification fluidity is relatively high among individuals who identify as Hispanic and Asian (Saperstein and Penner 2012), especially as their understanding of race and racial boundaries is not derived from the legacy of slavery and institutionalized patterns of discrimination that have often shaped the black-white color divide (Lee and Bean 2004). As a result, children of immigrants may be more predisposed to changes in racial identification over time, with multiple contextual factors influencing their identification (Tovar and Feliciano 2009; Vasquez 2011).
Psychologists and sociologists alike have noted that, in general, there are multiple selves, and the adoption of a specific identity is often dependent on specific contextual factors individuals encounter (Burke 2003; Winnick and Bodkin 2009). Social identities emerge from identification with cultural and social categories, including ethnicity, nationality, and religion (Tajfel 1982). As individuals come to see characteristics of such groups inherent in themselves, a social identity forms. When an individual is able to positively identify with others, a sense of belonging and likeness is developed. Not surprisingly, racial group membership has been shown to have protective effects for an individual’s sense of self (Crandall et al. 2000).
The theory of identity development has been used to help explain the formation, and reformation, of a racial identity. This includes research on the process of identity formation among individuals with one black parent and one white parent (Doyle and Kao 2007; Khanna and Johnson 2010; Waters 1990, 1994). These studies have highlighted how blackness and attendant stereotypes are ingrained within the American collective consciousness and provide dimensions of social identity. As race creates a set of expectations by which people are judged through everyday interactions (Saperstein and Penner 2012), individuals who can plausibly claim membership in multiple groups may emphasize a particular identity over another on the basis of their lived experience. As an example, Doyle and Kao (2007) analyzed longitudinal data from the National Longitudinal Study of Adolescent Health data set to ascertain the determinants and directions of change in racial self-identification among multiracial and monoracial adolescents. The authors found strong evidence that socioeconomic status and physical appearance shape the direction of change. Similar findings have also been mirrored in other work assessing racial identity in multiracial individuals (for a review, see Brown, Hitlin, and Elder 2006; Harris and Sim 2002).
From this perspective, racial switching may be motivated by a need to conform to others’ expectations, so racial fluidity occurs in accordance with how people are perceived by others. Yet racial identity cannot simply be relegated to the internalization of institutional messages and meanings. Storrs (1999:208) concluded, “While institutional forces constrain and shape individual identities . . . [it is] misleading to characterize identity formation as simply a process of external categorization and constraint.” A recent study by Saperstein and Penner (2014) showed that a substantial proportion of changes in self-reported race could not be explained by this perspective alone. Furthermore, it remains unclear whether or not this perspective could explain trends in racial self-identification among children of Hispanic immigrants. In this vein, Hannon and DeFina (2015) analyzed data from the General Social Survey and found that more than 40 percent of Hispanics changed their racial categories (among white, black, and other) at least once within just a four-year period. Subsequently they argued that because this level of change occurred within such a short time period, meaningful identity shifts were less likely to be a result of changes in social status.
Self-perceptual Causes of Racial Identity Changes
Although social identity theory provides the broader framework in which to understand how racial identities are developed, adopted, and changed, it does not provide the means with which to understand the motivations by which an individual might choose to do so. To understand the mechanism by which change may be triggered, we draw from stigma theory, part of the broader social identity theory framework. Stigma refers to “an attribute that is deeply discrediting within a particular social interaction” (Goffman 1963:3). Prior work has suggested that racial categorization can present a stigma given particular social circumstances (Harris 1976; Storrs 1999; Tapia 2010). Howard (2006) argued that racial categorization is inherently dehumanizing, tied to unequal access to resources and structural inequality, and, in general, externally imposed on the individual, all of which circumstances qualify race as a stigmatizing condition under Goffman’s (1963) outline of stigma.
Prior work has demonstrated how race can be a stigmatizing condition, whereby one’s social identity, or the way one is viewed by others, is in some way spoiled (Goffman 1963). For example, black youth are often stigmatized as criminals, which some research has shown often results in harsher penalties for black offenders by law enforcement (Tapia 2010). In addition, prior work has shown that Mexican Americans have been considered a stigmatized racial group, whereby greater contact with whites has led to greater experiences with discrimination and stereotyping (Ortiz and Telles 2012). Perhaps unsurprisingly, racial stigma can thus result in a significant negative impact on the individual, including lower self-esteem, increased rates of depression, and a lower sense of self-worth (Gilbert and Walker 2010; Lichtenstein, Laska, and Clair 2002; Murphy, Austin, and Greenwell 2007). Although identity research has explored how race may present stigma to some individuals, particularly mixed-race individuals, and how this may relate to changes in racial identification (e.g., Storrs 1999), to date, no research has explored how stigma theory may be related to changes in racial identification among children of immigrants. This omission is surprising in light of prior work demonstrating the high propensity for racial identification switching among children of immigrants (Feliciano 2009; Saperstein and Penner 2012) and experiences of race-based stigma some immigrants undergo (Ortiz and Telles 2012).
Another aspect central to identity theory and stigma is that individuals actively work to shape their identities (see Schroeder and Mowen 2014) and are not simply passive recipients of stigmas (Howard 2006; Marzano and Romano 2007). For example, individuals can engage in coping strategies, such as devaluing stereotypical comments about race or neutralizing negative comments or beliefs about race by attributing them to racism (see Crocker and Major 1989). Although some prior research has found that these and other coping strategies may aid in avoiding stigmatization on the basis of race (Crandall et al. 2000), it is possible that some individuals are simply unable to avoid stigmatization and the corresponding negative consequences, including decreased self-esteem or depression. More specifically to considering immigrants, to the extent that some children of immigrants are considered nonwhite, they may be more stigmatized and subject to greater stereotyping. Some children of immigrants, however, might carry many indicators of whiteness relative to their parents, including non-Spanish names and other cultural or social resources (Ortiz and Telles 2012). Yet in the context in which the stigma is presented by racial identity and accompanied by depression or changes in self-esteem and self-worth, we suggest that individuals may shift their racial identification (e.g., Storrs 1999) to reflect their treatment in society. Therefore, we hypothesize (hypothesis 1) that children of immigrants with lower levels of self-esteem and self-worth and higher levels of depression will be more likely to report changes in racial identification compared with individuals with higher levels of self-esteem and self-worth and lower levels of depression. In this manner, changing racial self-identification over time could represent a form of protective strategy in order to maintain a positive racial identity.
Family and Socioeconomic Causes of Racial Identity Change
Although our conceptual focus is on how self-perceptions are associated with shifts in racial identification, invalidation or pressure, such as family pressure or socioeconomic incentives, has the potential to undermine the psychological well-being and sense of self that one can gain by group identification. Using two decades of data from the National Longitudinal Survey of Youth, Saperstein and Penner (2012) found that an individual’s racial self-identification, as well as his or her classification by others, changes in response to a number of changes in social position. As an example, Americans who lost their jobs were more likely to be seen, and to self-identify, as black. People who got married were more likely to be seen as and to identify as white. Interestingly, these effects on changes in self-identification were strengthened when multiple changes in social position occurred simultaneously, such as when marriage and employment accompanied a growth in household income. One of the important implications of Saperstein and Penner’s (2012) study is that racial self-identification change is not limited to macro-level boundary changes (i.e., the whitening of southern European populations in America). As Saperstein and Penner (2012) demonstrated, factors other than an individual’s self-perceptions may drive racial identification changes. Highlighting this further, the idea that a change in racial identification can produce rewards in society, and the labor market specifically, has strong empirical grounding in prior work (Goldsmith, Hamilton, and Darity 2006). Scholars who have argued that racism is institutional have also documented how groups that successfully define themselves as white have marshaled considerably more economic power (Feagin 2010). In this way, a change in racial identification may occur because of economic incentives (see Storrs 1999 for a study on racial identification maintenance in the workplace).
In addition to economic incentives, we know that the family is a primary agent of identity development (Steinberg 2008), as early development is often when identity is first formed. Family dynamics and traditions play an important role in racial identification. A cohesive family may support, change, or create expectations for racial consistency. Although the results of prior work are mixed, Vasquez (2011), for example, showed that there may be pressures to change or maintain particular racial self-identification even within the same family. Likewise, in exploring identity formation among young adults, Mullis, Brailsford, and Mullis (2003) showed that greater levels of family cohesion were significantly related to expectations of identity commitment. Yet in understanding the moderating effect of race, Mullis et al. found that black youth were more likely to explore interpersonal identity than white respondents when family cohesion was low (see also Soliz, Thorson, and Rittenour 2009). Although some of these findings are mixed, what is clear is that families, and family pressures or expectations, often coincide with racial inconsistency or identity maintenance.
Evidence from the CILS also demonstrates the important role of family in racial identification. CILS data have shown that the children of some immigrants report that they feel high levels of discrimination, irrespective of economic attainment (Portes and Rumbaut 2001). As a result, on the basis of the reported experiences of discrimination by teenage youth and being treated as a member of a particular racial or ethnic group, a potential group of Hispanics who originally identified as white or Hispanic-white multiracial may shift toward Hispanic identification over time. Overall, these studies offer empirical support to the idea that identification with a particular racial group may be selected for economic gain or to strengthen a sense of belonging (Espiritu 1992). In line with these findings, we hypothesize (hypothesis 2) that family cohesion and economic incentives will positively relate to a change in racial self-identification, given the broader findings on the importance of economic incentives and family on individual identification. In short, we know that some individuals switch racial identification across contexts (Saperstein and Penner 2010, 2012) and time (Liebler et al. 2014), often to maintain a prosocial identity (Crandall et al. 2000). Yet we know very little about the underlying mechanisms that contribute to chances in racial identification over time in sample of children of immigrants. To better understand this, we now turn to our current study.
The Current Project
We have outlined the literature on changes in racial identification above (e.g., Doyle and Kao 2007; Kramer et al. 2015; Saperstein and Penner 2010), but the mechanisms that cause change are still debated. Drawing on identity theory suggesting that racial stigma may relate to changes in self-esteem, lowered sense of self-worth, and depression, in addition to previous work on the importance of economic and familial pressures in identity formation and change, the goal of the current research project is to explore the underlying mechanisms by which racial identification changes occur within a sample of children of immigrants.
Methods
Data
Data used for this project come from the CILS, an investigation conducted by Alejandro Portes and Rubén Rumbaut into the educational performance and social, cultural, and psychological adaptation of children of immigrants (Portes and Rumbaut 2001). The CILS data are longitudinal and were collected in three waves between 1991 and 2003. Wave 1 data were collected from second-generation immigrant children in eighth and ninth grade schools in 1991 and 1992. The sample comprised a total of 2,503 individuals from Miami, Florida, 339 from Fort Lauderdale, Florida, and 2,420 from San Diego, California, for a total sample size of 5,262 individuals. Three years later (1994 and 1995), 82 percent of the original respondents were reinterviewed to ascertain changes over time in their family situations, school achievement, educational and occupational aspirations, language use, ethnic identification, experiences of discrimination, and psychosocial adjustment. Slightly more than 10 years after the initial sample (2002 and 2003), a follow-up survey was conducted with a total of 3,613, or 68.9 percent, of the original sample. As in all large-scale surveys, the sample sizes decreased over time. Attrition in the CILS data may have been caused by respondents’ moving without updating their contact information. However, the CILS research team suggested that it is more likely that attrition was due simply to non-response (Portes and Rumbaut 2012).
In addition to collecting data from youth, CILS researchers also collected data from parents in the first two waves. Overall, the survey questionnaires contained items measuring racial and ethnic identification, self-esteem, a number of family dimensions, academic and career aspirations, and language use. The sample included teenage youth representing 77 nationalities in areas of persistently high immigrant settlement: San Diego and Miami/Fort Lauderdale. Although some respondents had mothers and fathers from different countries of origin, previous research has relied on mothers’ countries of origin to identify the origins of children because of the absence of fathers at wave 1 in approximately 30 percent of all cases (Feliciano 2009). Following this technique, descriptive statistics show that 36.7 percent of the respondents are of Cuban origin, 20.3 percent from Mexico, 10.5 percent from Nicaragua, 6.8 percent from Colombia, 4.3 percent from Haiti, 3.8 percent from Jamaica, 2.9 percent from Dominican Republic, and 14.7 from some other country of origin (see Feliciano 2009). It is also important to note that almost 50 percent of the respondents were not born in the United States; rather, they moved to the United States as young children with their parents.
Dependent variable
To be sure that shifts represent meaningful and sociologically significant change, a question regarding racial self-identification must be asked at multiple and independent points in time (Liebler et al. 2014). CILS researchers assessed racial self-identification through questions at wave 2 and again, independently, at wave 3. Thus, we measured transitions in racial self-identification that occurred between those two waves. At wave 2, individuals were asked how they racially identified: white, black, Asian, other, multiracial, or Hispanic. During this wave, individuals could choose only one racial category. At wave 3, individuals were first asked to choose from one of the following categories: white, black, Asian, other, and multiracial. They were then asked to write in how they identified if it was different from the categories offered. We coded the former question to represent the same racial categories present at wave 2: white, black, Asian, other, and multiracial, and we used the latter question to code individuals who reported themselves as Hispanic. To create a variable representing change in racial identification, we first created dummy variables for each racial category at both times. We then multiplied each dummy variable at wave 2 by each dummy variable at wave 3.
Given the separate theoretical reasons that would explain why an individual would shift from a majority group to a minority group and vice versa, the most revealing analysis would be one that explains both pathways of shifts. However, because of the relatively small number of individuals experiencing specific shifts (e.g., white to black, Asian to Hispanic), we focus on any individual who experienced a shift in racial self-identification. Although prior literature would seem to suggest that majority- and minority-race individuals may have different reasons for moving between racial categories, it is possible that the same underlying explanations may explain shifts in either a majority or a minority individual (for an overview of how external forces influence racial identification changes for all individuals regardless of racial or ethnic group membership, see Saperstein and Penner 2012). In this vein, we posit that both white and nonwhite individuals who are presented with a stigma, or negative self-image due to their race, will attempt to deal with that stigma and may attempt to shift racial categories (see Saperstein and Penner 2010). For a complete description of this measure, the shifts and constants between each racial category are shown in Table 1.
Racial Changes between Waves.
Note: W = wave.
Data were collected in 1994 and 1995.
Data were collected in 2002 and 2003.
Table 1 reports racial self-identification at wave 2 in the rows and racial self-identification at wave 3 in the columns. Overall, 1,696 individual reported the same racial category as adults (wave 3) as they did as adolescents (wave 2). Yet 1,251 individuals reported shifts in their racial self-identification between waves 2 and 3. Overall, this shows that about 42 percent of all respondents in our sample reported inconsistent racial identification, suggesting that racial self-identification may be very fluid among children of immigrants. By way of comparison, Liebler et al. (2014) estimated that 8.3 percent of all people in the United States could have changed their race and/or Hispanic origin responses between 2000 and 2010. Although the level of change in racial identification we report is higher than that reported in prior studies of racial identification shifts, it is in line with previous studies examining ethnic identification shifts using the CILS data (Tovar and Feliciano 2009). As such, change in identification may be greater in these data given the unique histories of immigration and political contexts of San Diego and Miami, in addition to the ages of the respondents. Indeed, previous studies have found that change over time in racial self-identification is more common among individuals who were children at the first time point (Liebler et al. 2014). Involvement in new social networks and situations, such as going to college or moving away from the family as young adults, may increase the likelihood of self-conception change (Root 2003), which, though distinct from a change in identification, could certainly contribute to changes in self-identification over time. Overall, the variability among this population also serves to highlight the importance of exploring the reasons behind such changes, adding to our understanding of the experiences of children of immigrants more generally.
Of all the changes in racial identification between waves, 53 individuals reported moving into the white racial category (91 remained constant in the white racial category), 28 into the black racial category (149 constant), 114 into Asian (615 constant), 817 into Hispanic (655 constant), 60 into multiracial (105 constant), and 179 into the other racial category (81 constant). Interestingly, the number moving from multiracial as adolescents to Hispanic as young adults is much larger than the number changing from Hispanic to multiracial. This result is contrary to recent findings from a nationally representative sample, in which Hispanic-to-multiracial changes were far more numerous (Liebler et al. 2014). The contexts of the study locations, as well as the unique characteristics of the sample, may explain this observation. As an example, almost 50 percent of the respondents in the CILS data were not born in the United States; rather they moved to the United States as young children with their parents. The respondents’ experience in, and knowledge of, their home countries, in addition to greater transnational practices among the 1.5 generation, may have contributed to stronger maintenance of the Hispanic racial self-concept. In addition, greater presence of co-ethnics in Miami and San Diego and shared history may further maintain Hispanic self-concepts, whereas higher rates of marital, residential, and economic integration in other areas of the country may facilitate change toward multiracial identification.
Independent variables
As highlighted by a review of the literature, individuals engage in specific behaviors to avoid stigmatization and maintain positive levels of self-esteem, self-worth, and mitigate depression (see Schroeder and Mowen 2014). We suggest these behaviors also apply to racial identification. The current project relies on three measures of self-perception measured at wave 2, prior to the shift in racial identification. First, we performed confirmatory factor analysis to delineate empirically different measures; the results indicate three distinct factors dealing with self-esteem, self-worth, and depression. It is worth noting, too, that bootstrapping (results not shown but available on request) was applied to the confirmatory factor analysis, and on the basis of the size of the bootstrap standard error, we conclude this confirmatory factor analysis fits our data well because of very low standard errors, as all observed coefficients fit within a relatively small 95 percent confidence interval (Kline 1994).
First, the latent measure of depression is composed of three variables. Respondents were asked how often they felt sad, how often they felt depressed, and how often they felt that they could not get going (rarely, some of the time, occasionally, and most of the time), where higher scores indicate lower levels of depression. This measure has a mean of 9.45 and a standard deviation of 2.15 and ranges from 3 to 12. Table 2 shows the factor loadings for each confirmatory factor.
Confirmatory Factor Analysis.
Note: Standardized coefficients are reported; χ2 = 395, df = 45, root mean square error of approximation = .043.
Fit indices: comparative fit index = .961, Tucker-Lewis index = .952.
The second latent factor revealed by the confirmatory factor analysis captures self-esteem. Respondents were asked how much they agreed with the following statements: “I have a number of good qualities,” “I take a positive attitude to myself,” “I am satisfied with myself,” “I wish I had more respect for myself,” “I am satisfied with how I look,” and “I have good looks.” Respondents were asked if they agreed a lot, agreed a little, disagreed a little, or disagreed a lot. This measure has a mean of 20.32 and a standard deviation of 3.24 and ranges from 6 to 24.
The third and final latent variable measuring personal beliefs about oneself has to do with self-worth. Respondents were asked to respond to the following items: “I am inclined to feel I am a failure,” “I feel useless at times,” “At times I think I am no good at all”; higher scores indicate higher levels of self-worth. This measure has a mean of 9.72 and a standard deviation of 2.26 and ranges from 3 to 12. We include each factor within the model as a latent factor.
To assess how socioeconomic status may relate to changes in racial self-identification, we rely on five measures at wave 3 in the CILS data. First, respondents were asked about their levels of education; 27.5 percent reported having college degrees or higher (1 = college degree or higher, 0 = less than a college degree). Second, respondents were asked if they owned a home; 38.4 percent indicated that they owned their homes (compared with renting or living with parents). Third, CILS questionnaires assessed marital status; 18.2 percent of respondents indicated that they were married (compared with single, divorced, or other status). Fourth, respondents were asked if they had full-time employment; 62.2 percent indicated that they were employed full-time at wave 3 (compared with part-time or unemployed). Finally, to measure income, respondents were asked to report their annual incomes (1 = less than $5,000, 2 = $5,000 to $9,999, 3 = $10,000 to $14,999, 4 = $15,000 to $19,999, 5 = $20,000 to $24,999, 6 = 2$5,000 to $29,999, 7 = $30,000 to $34,999, 8 = $35,000 to $49,999, 9 = $50,000 to $74,999, 10 = $75,000 to $99,999, 11 = $100,000 to $199,999, and 12 = $200,000 or more). To use this measure of socioeconomic status as a binary measure, and thus have a consistent weight within the factor, we recoded this variable around the mean response of 7.28 (1 = above the mean, 0 = below the mean). The recoded variable represents individuals with higher incomes than the sample mean (compared with lower incomes than the sample mean). We use each of these five measures to capture socioeconomic status as a latent construct within the model.
Finally, because the family is one of the primary points of socialization and the development of identity (Steinberg 2008), we include variables to measure family cohesion in the current project. The CILS data include an index of family cohesion measured via three variables assessed at wave 2. Respondents were asked how much the family likes to spend time together, how often it is important that they feel close to family members, and how often feelings of family togetherness are important to them (1 = never, 2 = once in a while, 3 = sometimes, 4 = often, and 5 = always). To create this scale, we use the mean of all three responses, where higher scores indicate higher levels of family cohesion. This scale has a mean of 10.98 and a standard deviation of 3.01 and ranges from 3 to 15. This scale has an α of .846 (see Cortina 1993) and is included as a latent variable in the analysis.
Analytic Strategy
To assist in exploring specific pathways from the independent variable to change in racial identification, we use a longitudinal structural equation model (see Kline 1994) in Stata 12.1SE. Because of the skew (see Table 3) in some of our measures, as well as the binary outcome in the dependent variable, we use asymptotic distribution free (ADF) estimation in our structural equation model. ADF estimation, also referred to as arbitrary generalized least squares, is an approach used with binary variables in structural equation modeling because categorical variables often have underlying continuous unobserved variables, and ADF corrects for loss of information in correlations due to the categorizations of the binary outcome (see MacCallum et al. 2002).
Descriptive Statistics of Independent Variables.
As shown by the model in Figure 1, we expect higher levels of both self-esteem and self-worth to have negative effects on the probability of an individual’s experiencing a transition in racial self-identification, and higher levels of depression to have a positive direct effect on the probability of an individual’s experiencing a transition in racial self-identification. We also expect family cohesion to have a positive direct effect on a shift in racial identification. Finally, we expect socioeconomic status to have a positive direct effect on a change in racial identification. The fully identified model is shown in Figure 1.
Structural Equation Model.
Results
The results of the structural equation model are shown in Table 4. First, it is worth noting that our model does a reasonably good job of fitting the data. The χ2 value is large, but once it is divided by the degrees of freedom, we arrive at an adjusted value of 3.94, indicating a reasonably good fit in our structural equation model (Browne and Cudeck 1993). Additionally, both the root mean square error of approximation (.034) and the fit indices suggest that this model is generally a good fit to the data (see Bentler 1990).
Results of the Structural Equation Model on Any Shift (n = 2,519).
Note: Standardized coefficients are reported; χ2 = 925.74, df = 235; adjusted χ2 = 3.94. Fit indices: comparative fit index = .824, Tucker-Lewis index = .843. Root mean square error of approximation = .034 (lower bound = .036, upper bound = .041). Residual error = .9.
As shown by Table 4, self-esteem, self-worth, and family cohesion are all significantly related to an individual’s experiencing a shift in racial self-identification. In exploring the magnitude of these effects, we see that with every increase of 1 standard deviation in self-esteem, the probability of an individual’s experiencing a shift in racial identification increases by .002 standard deviations. With every increase of 1 standard deviation in self-worth, the probability of an individual’s experiencing a shift in racial identification increases by .082 standard deviations. Finally, with every increase of 1 standard deviation in family cohesion, an individual is .076 standard deviations more likely to experience a shift in racial self-identification. On the other hand, the latent constructs of socioeconomic status and depression are not significant in the model.
It is important to note that the magnitudes of the statistically significant effects are small, suggesting that other factors not included in the current model may have a strong effect on transitions in racial identification. For example, the residual variance in the outcome variable—experiencing a change in racial identification—is approximately .9, meaning that slightly less than 10 percent of the variance in changes in racial identification is explained by the fully identified model. An increase of 1 standard deviation in family cohesion, for example, would predict a 2.3 percent increase in the odds of experiencing a change in racial identification. Therefore, although the model fits the data, and some measures reach statistical significance, we are capturing a relatively small amount of the variance in racial identification shifts.
Discussion and Conclusion
Recent quantitative work has documented that racial and ethnic identification is not as consistent over time as previously thought (Doyle and Kao 2007; Harris and Sim 2002; Liebler et al. 2014). Studies have subsequently urged researchers to consider how context affects racial identification. Although contributing factors such as education, income, and other changes in social status have often been linked to the desire or pressure to shift racial categories, few studies have considered how both self-perceptions and other factors such as family cohesion and socioeconomic status may contribute to racial shifts. Furthermore, no study that we are aware of has examined how these factors are related to racial identification change among children of immigrants. In this study, we analyzed data from two waves of the CILS to offer insight into this question.
Bivariate descriptions of the data suggested that children of immigrants do report changes in racial self-identification over time. These particular shifts were consistent with the shifts in self-definitions of second-generation youth during adolescence reported by Portes and Rumbaut (2001). In that instance, a shift from ethnic abandonment to a return to an embracing of their parents’ origin reflected an awareness of their place in America’s social hierarchy. In this study, using structural equation modeling, we found that self-esteem, self-worth, and family cohesion were all significantly related to an individual’s reporting a change in self-racial categorization. On the other hand, socioeconomic status and depression were not related to a change in racial self-identification.
The first hypothesis was not supported. Within the framework of identity theory, and drawing on concepts of stigma, we suggested that individuals who reported lower levels of self-esteem and self-worth and higher levels of depression would be likely to report changes in self-identification. Furthermore, our theoretical orientation suggested that this result could be due partially to the need to maintain a socially positive identity. To the contrary, the current analysis revealed that overall, perceptions of self-esteem and self-worth, but not depression, were significantly related to shifts in racial identification, although individuals with higher levels of self-esteem and higher levels of self-worth were more likely to report changes in racial group membership, all else being equal. A recent study found that inconsistent racial identities across different settings are not related to negative emotional well-being, though this was measured as depression only (Kramer et al. 2015). These conclusions, though, are based on assessing inconsistencies in race first and emotional outcomes second, the opposite of our causal analysis. Yet the findings of Kramer et al. (2015) suggest that some measures of self-perceptions, such as emotional well-being, may not be associated with racial inconsistency. Although the causal directions of our analysis are opposite to that of Kramer et al.’s study, our findings partially support the null findings, as depression was not significantly linked to a transition in racial identification. Our findings do suggest, however, that an additional explanation for Kramer et al.’s null finding may be that changes in self-worth or self-esteem could have occurred prior to a change in racial self-identification. From this standpoint, high levels of self-worth and self-esteem may contribute to changes in racial identification because of the need to continue to maintain a socially positive identity.
Second, it is important to note that participants in our study were still in high school at the time racial self-identification was first measured. Previous studies have found that children of immigrants less engaged in school and those who experienced poorer academic outcomes are more likely to identify as Hispanic or Black; thus respondents may have associated their academic outcomes with cultural assumptions about racial groups at this time (Ortiz and Telles 2012; Wilkinson 2010). By wave 3 of the CILS, 10 years from the original survey for most respondents, academic outcomes may have become less important than family identification, which could explain some of the inconsistent racial identification. Indeed, this theory would also explain our mixed findings regarding the second hypothesis, namely, that family cohesion, but not socioeconomic status, was related to a change in racial identification. That is, individuals who reported higher levels of family cohesion were more likely to report changes in racial identification compared with individuals with lower levels of family cohesion. As we discuss below, this may be interpreted as evidence that family cohesion is associated with social status change, an important component of identity theory, thereby increasing the chances of racial self-identification change. Alternatively, we posit that family cohesion may increase the propensity to change for children of immigrants in the United States.
Although there is a limited amount of literature on the relationship between family cohesion and racial identity changes, a cohesive family could conceivably either support or restrict racial identification change (Vazquez 2011). Some prior work has noted differences in identity development across racial groups due to levels of family cohesion (e.g., Mullis et al. 2003). Although not speaking directly to inconsistent racial identities, Butler-Sweet (2011) explored identity development through qualitative interviews in transracial, biracial, and monoracial families. Findings from that study showed that irrespective of family racial structure, respondents reported rejecting their ascribed racial status and instead embraced the achieved aspects of their social identities. The importance of embracing the achieved social aspects of identity is important because it suggests that individuals may be likely to experience changes in identification to minimize negative aspects of ascribed characteristics about racial group membership (e.g., a stigma) and instead focus on minimizing, or neutralizing, stigmatization through concentration on achieved characteristics of identification. In addition, higher levels of family cohesion could coincide with a greater likelihood of achieving social status changes, for example marriage, which have been linked to increased odds of racial self-identification shifts (Saperstein and Penner 2012).
Outside of stigma and identity theory, studies of adolescent children of immigrants have also found that the closer children feel to their parents, the closer they feel to their ethnic identities and ancestry (Kiang et al. 2010). As adolescents transition to adulthood, the closeness to family and ethnic origin offers another potential source of identification. Young adults who remain close to their ethnic origins while experiencing changes in social status and social networks may have a higher propensity for racial self-identification change. Conversely, individuals who are not closely connected with their families may feel fewer options for identification.
In addition to some of the limitations noted above, there are several notable limitations to the current project. First, the sample is nonrandom and thus lacks generalizability. Using data from a localized study such as the CILS may have ensured a highly fluid sample on the basis of the unique demographics and history of San Diego and Miami/Fort Lauderdale. The larger political climates of California and Florida were also likely to shape respondents’ self-identities. In their analysis of 21 Latinos aged 20 to 26 years living in California, Tovar and Feliciano (2009) noted that bills targeting the treatment of illegal immigrants, and attendant protests, had profound effects on self-identification. Although Tovar and Feliciano’s study dealt with how ethnic identification, local context, lived experiences, and prevailing racial stereotypes continue to inform how children of immigrants understand their positions in the racial hierarchy (Romero, Hondageu-Sotelo, and Ortiz 1997).
Second, the sample is composed of children of immigrants, predominantly from central and southern America. People who report Hispanic or multiracial backgrounds are considered to have relatively high baseline propensities to change or be categorized in multiple ways (Saperstein and Penner 2012). The literature suggests that children of immigrants, specifically, also navigate multiple racial identities (Waters 1994) and are also more likely to experience changes in self-identification to assist in the assimilation process. For example, Portes and Zhou (1993) showed that immigrants will engage in specific behaviors to assimilate into the cultures they have relocated to, including adopting cultural and social norms. Although the authors did not that suggest individuals will change their racial identities, there is little reason to suspect that individuals would be willing to change their culture—including religion, gender roles, clothing, and language, among other things—and not their race. Therefore, the extent to which the present findings may relate to individuals further removed from the immigration process is unclear. Over additional generations, further socioeconomic integration, residential integration, and marital assimilation may eventually lead to the loss of distinct identities among the third and fourth generations (Gordon 1964; Waters 1990, 1994). Through interviews with three generations of Mexican American families, Vasquez (2011) revealed how racial identities change with each generation and greater socioeconomic mobility.
Finally, our study was unable to consider the important physical attribute of color (Bonilla-Silva and Dietrich 2009). As Telles and Paschel (2014) discussed, phenotype is one of the strongest predictors of racial self-identity. One would expect that the likelihood of social and environmental factors leading to a change in racial self-identity may be lower for darker skinned individuals compared with those who are phenotypically closer to white. Indeed, some groups may have more “options” to change over time than others (Waters 1990). In addition, though we assess how measures such as depression, self-worth, self-esteem, and family cohesion at wave 2 influence changes in racial identification at wave 3, it is likely that these very measures change over time as well. Unfortunately, the CILS data do not include measures of family cohesion and depression, self-esteem, and self-worth at wave 3. It is likely that these measures are also fluid over time, and we are unable to measure how changes in these constructs may also relate to changes in racial identification. Finally, because race was measured at only two time points, it is possible that individuals within this sample continued to experience further changes in racial identification, something future research projects should seek to assess.
Overall, the findings of this project have important methodological and theoretical implications. Our findings echo other sociological evidence that racial transitions are not uncommon. Therefore, it is important to move beyond the debate about whether racial identity is fixed or static and explore further the causes and the consequences of identity inconsistency (e.g., Kramer et al. 2015). Although studies have documented the fluid and dynamic nature of race and investigated the various contexts and conditions under which change in self-identification may occur, a greater appreciation should also be given to the specific self-perceptual motivations and potential benefits of fluidity (i.e., improving self-esteem and self-worth). By incorporating these individually based factors, as well as factors outside of self-perceptions, such as the role and importance of family and socioeconomic status involved in racial identity change, sociological research can offer greater evidence regarding the question as to whether racial identity inconsistency is healthy and beneficial (Kramer et al. 2015; Root 2003).