The 2020 Democratic primary started with one of the most diverse presidential fields in history and ended with a White male candidate. As a means of tempering the disappointment felt by many, and in hopes of spurring voter enthusiasm, Joe Biden, the presumed Democratic nominee, vowed to select a woman as his running mate. His choice of Senator Kamala Harris―the daughter of Jamaican and Indian immigrants―came in the midst of the largest mass mobilization challenging anti-Black racism and state violence, as well as ongoing mobilization seeking to end one of the most restrictive immigration regimes in recent history. Thus, the potential significance of this nomination went well beyond that of gender. In November, when Harris took the stage as the vice president–elect, the moment was felt by many of the key constituencies that had made the victory possible. This included Asian American, Latina, Native American, and, perhaps most especially, Black women.
While there is still much debate about whether the identity of a candidate should matter in elections, there is evidence that it does, in ways that are significant for democracy. Research on political representation, as it pertains to gender and race, demonstrates that the presence of historically underrepresented groups in political office (descriptive representation) can have not only a substantive impact on policies and procedures but also a symbolic impact that changes the attitudes and even the behavior of those groups. The “minority empowerment” theory of politics holds that descriptive representation signals greater political opportunities and inclusion for these groups, something that can promote democratic engagement and participation (Bobo and Gilliam Reference Bobo and Gilliam1990). On the flip side, the absence of descriptive representation may contribute to a sense of political alienation and inefficacy, something that may suppress democratic engagement and voter turnout (Pantoja and Segura Reference Pantoja and Segura2003). Thus, perceptions of representation are not only a significant component of democratic inclusion (Mansbridge Reference Mansbridge1999) but an important consideration for political parties in mobilizing the electorate.
The dynamics of identity and its significance for representation, however, are complicated. Individuals often hold multiple identities, and the salience and meanings attached to those identities may vary in relation to each other and to the particular political context (Citrin and Sears Reference Citrin, Sears, Abdelal, Herrera, Johnston and McDermott2009; García Bedolla Reference García Bedolla2007; Valenzuela and Michelson Reference Valenzuela and Michelson2016). In an election with candidates representing different (and sometimes multiple) identities that have been historically underrepresented in politics and in a political environment in which many of these identities are particularly salient (made so by an administration perceived as actively dismantling the rights of these groups), engaging with this complexity is all the more necessary.
Traditionally, much of the research examining the meaning of representation has focused on one type of identity, usually gender or race. While scholars are increasingly taking a more intersectional approach, there is still much work to be done in understanding how different types of identities might interact. This holds true for understanding perceptions of representation. While voters may share an identity with a candidate along one dimension (such as gender), there may be a salient difference on another (such as race), or vice versa. Furthermore, there may be overlapping identities that take on a more coalitional nature, such as the cross-racial identifications of “people of color” or “women of color” that may emerge from perceived shared experiences as racial minorities. Finally, variations in group consciousness may also factor into perceptions of representation, such that the meaning attached to gender and/or racial identities may vary across and within groups. In this article, we propose a more nuanced intersectional approach to examine how multiple, intersecting, and overlapping identities shape perceptions of representation. How and to what extent do shared identities matter for perceptions of representation, and for whom?
The high salience of both gender and racial issues in the 2016 election provides an ideal opportunity for observing the complex dynamics of descriptive representation that vary across race and gender. To the extent that scholars have explored the race-gendered dynamics of the election, it has been primarily to focus on White women compared to women of color (see Cassese and Barnes Reference Cassese and Barnes2019; Frasure-Yokley Reference Frasure-Yokley2018; Junn Reference Junn2017). In this article, we are interested in focusing on potential race-gendered differences in perceptions of two minoritized ethno-racial groups, African Americans and Latinos/as. We use data from the 2016 Collaborative Multiracial Post-Election Survey (CMPS), a unique data set that allows for a robust cross-racial and intersectional analysis.Footnote 1 Respondents (Black men, Black women, Latino men, and Latinas) were asked how well they felt candidates would represent their interests. The respondents were asked about male and female candidates from multiple racial groups (Black, Latino/a, Asian American, and White) to examine different ways in which race and gender might intersect. Our findings demonstrate that while shared identity does influence perceptions of representation, the impact varies in complicated ways that are simultaneously raced and gendered. While shared race-gendered and shared ethno-racial identity both had a consistent impact on perceptions of representation, shared gender mattered only within certain race-gendered contexts. Furthermore, our cross-racial analysis found a stronger association between Latinos/as and African Americans when examining their perceptions of representation from each other compared to other out-group candidates.
IDENTITY AND PERCEPTION OF REPRESENTATION
The foundation of a representative democracy is the notion that an elected official can serve to represent the polity, if not descriptively—in which a citizen shares demographic characteristics such as race, ethnicity, gender, and so on—then substantively—on policy issues and positions. While the emphasis on descriptive representation is sometimes derided as “identity politics,” scholars have long argued that people from different groups bring a unique set of perspectives on social and economic issues to politics (Conover Reference Conover1984). Furthermore, as argued by the Combahee River Collective ([1977] 1986), “identity politics” arises from the need for groups to focus on and bring attention to their own oppression in order to achieve liberation. While the relationship between descriptive and substantive representation is by no means a perfect one, research on the political representation of historically marginalized groups has provided contingent evidence supporting the various substantive impacts made by having women and people of color in office (Brown and Gershon Reference Brown and Gershon2016; Canon Reference Canon1999; Hardy-Fanta et al. Reference Hardy-Fanta, Lien, Pinderhughes and Sierra2016; Minta and Brown Reference Minta and Brown2014; Swers Reference Swers2005; Tyson Reference Tyson2016). Additionally, descriptive representation can have symbolic meaning that shapes political behavior. The “minority empowerment” theory of politics holds that descriptive representation signals greater political opportunities and inclusion for these groups, something that can promote democratic engagement and participation (Bobo and Gilliam Reference Bobo and Gilliam1990). Research has shown that descriptive representation increases political trust, knowledge, efficacy, and engagement among women (Atkeson Reference Atkeson2003; Atkeson and Carillo Reference Atkeson and Carillo2007; Reingold and Harrell Reference Reingold and Harrell2010), Latinos/as (Barreto Reference Barreto2010; Barreto, Villarreal, and Wood Reference Barreto, Villarreal and Wood2005; Pantoja and Segura Reference Pantoja and Segura2003; Sanchez and Morin Reference Sanchez and Morin2011), and African Americans (Banducci, Donovan, and Karp Reference Banducci, Donovan and Karp2004; Bobo and Gilliam Reference Bobo and Gilliam1990; Gay Reference Gay2002; Merolla, Sellers, and Fowler Reference Merolla, Sellers and Fowler2013).
Underlying the significance of descriptive, and, in turn, symbolic representation, is the development of group consciousness. Social identity theory argues that individuals, at a minimum, must be aware of common membership with a social category to develop particular group affinities or preferences (Tajfel Reference Tajfel1981; Tajfel and Turner Reference Tajfel, Turner, Worchel and Austin1986; Turner Reference Turner and Tajfel1982). There must be a strong sense of group consciousness in order for there to be any type of in-group political cueing (Adida, Davenport, and McClendon Reference Adida, Davenport and McClendon2016; McConnaughy et al. Reference McConnaughy, White, Leal and Casellas2010). The research on race and gender demonstrates that there are important variations in the development of group consciousness within and across groups that impact the ways in which identity works as a political cue.
Racial Identity as a Political Cue
Scholars have consistently shown that racial identity serves as a powerful political cue for ethno-racial minorities. When given the choice, ethno-racial minorities often choose a candidate with a shared ethno-racial identity, unless they have a serious reason to do otherwise (Barreto Reference Barreto2010; Bejarano Reference Bejarano2013; Hero Reference Hero1992; Philpot and Walton Reference Philpot and Walton2007). For members of marginalized groups, ethnic cues “may evoke a positive sense of co-ethnic solidarity, one that symbolizes the possibility of surmounting economic and social discrimination” (Adida, Davenport, and McClendon Reference Adida, Davenport and McClendon2016, 818; see also McConnaughy et al. Reference McConnaughy, White, Leal and Casellas2010).
African Americans consistently exhibit a high degree of in-group loyalty in voting for Black candidates (Philpot and Walton Reference Philpot and Walton2007; Reese and Brown Reference Reese and Brown1995; Sigelman and Welch Reference Sigelman and Welch1984). Studies suggest that these racial preferences are so strong that they hold up when controlling for voter ideology and support for racial policies and even in the presence of contradictory cues (Kinder and Winter Reference Kinder and Winter2001; Kuklinski and Hurley Reference Kuklinski and Hurley1994; Tate Reference Tate2003). While economic heterogeneity exists among African Americans, they continue to retain high levels of group consciousness and, as a result, support for Black candidates (Dawson Reference Dawson1994; McClerking Reference McClerking2001; White and Laird Reference White and Laird2020).
Research on Latinos/as shows a similar preference for candidates with shared ethno-racial identity (Barreto Reference Barreto2010; Barreto, Villarreal, and Wood Reference Barreto, Villarreal and Wood2005; Manzano and Vega Reference Manzano, Vega, Navarro and Rosales2006). Latino/a in-group loyalty, however, is not found to the same degree as among African Americans, and it has some important caveats. Latino/a experiences with marginalization vary by national origin and other factors―such as skin color and citizenship status―something that limits the degree to which they develop a sense of linked fate (Beltrán Reference Beltrán2010). Casellas and Wallace (Reference Casellas and Wallace2015) find that support for co-ethnic representation is influenced by perceptions of shared partisanship, with Latino/a Republicans less likely to support descriptive representation in the belief that it will help Democrats.
The research on minority political behavior emphasizes the role of shared experiences with discrimination as creating a racial consciousness. Group consciousness for Blacks in the United States is argued to be particularly strong because of historically rooted experiences with racial discrimination (Dawson Reference Dawson1994; Gay Reference Gay2002; McClain et al. Reference McClain, Carew, Walton and Watts2009; Tate Reference Tate2009). Encounters with discrimination have been shown to be a central component of ethno-racial group consciousness for Latinos/as (Wallace Reference Wallace2014), but the larger historical heterogeneity between Latinos/as and Asian Americans diversifies these experiences and perceptions of them. Contemporary experiences with racism, however, have served to form newer pan-ethnic identities that can and have been mobilized, as seen in the wave of immigration protests in 2006 (Silber Mohamed Reference Silber Mohamed2017; Zepeda-Millán Reference Zepeda-Millán2017).
While there are important historical differences between ethno-racial groups that merit a particular focus, organizing coalitionally as “people of color” is an important means of expanding mobilizing power and confronting contemporary shared experiences with oppression. Research looking at the relationship between one marginalized ethno-racial group and the possibility of preference for candidates from another marginalized ethno-racial group, however, is limited and somewhat mixed. Studies of Black–Latino/a politics show the possibility of animosity between the groups, particularly when the success of one group is perceived as threatening the well-being of the other (Dyer, Vedlitz, and Worchel Reference Dyer, Vedlitz and Worchel1989; Marrow Reference Marrow2011; McClain et al. Reference McClain, Carter, DeFrancesco Soto, Lyle, Grynaviski, Nunnally, Scotto, Alan Kendrick and Cotton2006; Morin, Sanchez, and Barreto Reference Morin, Sanchez, Barreto, Telles, Sawyer and Rivera-Salgado2011). However, other studies have shown the possibility of mutually supportive relationship when the groups perceive shared common interests (Hero and Preuhs Reference Hero and Preuhs2009; Jones-Correa Reference Jones-Correa, Telles, Sawyer and Rivera-Salgado2011; Jones-Correa, Wallace, and Zepeda-Millán Reference Jones-Correa, Wallace and Zepeda-Milán2016; Masuoka and Sanchez Reference Masuoka and Sanchez2010; Stokes Reference Stokes2003).
Moving beyond Black–Latino/a relations, there is even less work to draw from exploring how either group perceives commonality with Asian Americans as another community of color subject to racial discrimination. Segura and Rodrigues (Reference Segura and Rodrigues2006) note that while Latinos/as and African Americans are often seen as the more likely coalition partners because they share a number of politically relevant characteristics (e.g., education, income level, and a number of related correlates), Asian Americans and Latinos/as share proximity in regard to immigration experience and a more moderated position in the White–Black racial hierarchy. Betina Wilkinson (Reference Wilkinson2015) finds that the complex racial dynamics are not easily reducible to simple formulae and that contextual factors matters in the formulation of interracial coalitions. While the repressive racial politics of the Donald Trump administration may enhance perceptions of commonality, such perceptions may be in varied raced, gendered, or race-gendered patterns.
Gender as an Intersectional Political Cue
Research on gender also shows a preference for candidates with shared gender identity, but with more substantial qualifications (Bejarano Reference Bejarano2013). Scholars have demonstrated an asymmetrical gender effect, with women being “far more likely than men to prefer to be represented by someone of the same gender” (Rosenthal Reference Rosenthal1995, 609; see also Paolino Reference Paolino1995; Plutzer and Zipp Reference Plutzer and Zipp1996; Sanbonmatsu Reference Sanbonmatsu2002). Much of the research emphasizes the conditional nature of these findings (Dolan Reference Dolan2004, Reference Dolan2010; Hansen Reference Hansen1997; Huddy, Cassese, and Lizotte Reference Huddy, Cassese, Lizotte and Whittaker2008; Huddy, Neely and Lafay Reference Huddy, Neely and Lafay2000). For example, studies have shown that women's preference for shared gender identity is higher when issues of gender are more salient and more likely among Democrats than Republicans (Dolan Reference Dolan2010; Paolino Reference Paolino1995). Studies of the 2016 election emphasize the role that differing gender ideologies play in partisan and candidate preferences. While women, in general, have higher egalitarian attitudes toward sex roles than men, women who voted for Trump reported higher levels of benevolent sexism, hostile sexism, and traditional attitudes toward women and significantly lower egalitarian sex role attitudes (Bock, Byrd-Craven, and Burkley Reference Bock, Byrd-Craven and Burkley2017; Cassese and Holman Reference Cassese and Holman2019).
While most scholars agree that, as argued by Huddy and Carey (Reference Huddy and Carey2009, 81), “Race has a more powerful and less ambiguous effect than does gender on voting for an in-group and against an out-group candidate,” that does not mean gender is insignificant. Rather, when and how it matters is an open empirical question that merits intersectional exploration (Hancock Reference Hancock2007a, Reference Hancock2007b). Lorrie Frasure-Yokley (Reference Frasure-Yokley2018) argues that failing to incorporate an intersectional analysis can lead us to overestimate or underestimate gender effects. For example, in her analysis of the 2016 election, she demonstrates that while hostile sexism increased the probability of voting for the GOP candidate among White women by 17 percentage points, for women of color, it was negatively associated with voting for Trump (and posed no statistically significant effects, controlling for all other factors). She argues gender-only analysis often elicits findings that are―misleadingly―attributed to all women voters.
Going one step further, Lisa García Bedolla (Reference García Bedolla2007) argues that an individual's understanding of all social categories is relational and situated within the larger political context. This dynamic understanding of intersectionality moves us away from more static questions about race, gender, and their intersection, instead leaving it a more open empirical question (as advocated by Hancock Reference Hancock2007a, Reference Hancock2007b). When and how might race, gender, and their intersection matter, and for whom? This requires engaging in the complex and often “messy” process of identity formation arising from intersectional locations and experiences (Smooth Reference Smooth2006).
One way of disentangling the “messiness” of intersectionality is to distinguish between different types of intersectionality complexity. Here, McCall's (Reference McCall2005) distinction between intra- and intercategorical complexity is instructive. Intracategorical complexity looks for differences within a particular group to highlight differences that otherwise would be obscured. For example, Black feminists have argued the importance of recognizing how gender (along with other structural positionalities) shapes Black experiences with racial oppression (Brown Reference Brown2014; Capers and Smith Reference Capers, Smith, Brown and Gershon2016; Cohen Reference Cohen1999; Nash Reference Nash2019; Simien Reference Simien2005; Simien and Clawson Reference Simien and Clawson2004). Race may still be the most salient structural position, but it does not render other positions irrelevant (Higginbotham Reference Higginbotham1992). An intracategorical intersectional approach to representation might focus on looking for gender differences between Black men and women (or between Latino men and Latinas) in their perceptions of representation by various candidates. Alternatively, it might look at racial differences within a particular gender category.
Intercategorical intersectionality looks at relationships along multiple and conflicting dimensions in a more simultaneous way. It looks at the complexity of relationships among social groups within and across analytical categories rather than the complexities within single groups or categories. Rather than looking only at gender within ethno-racial groups or race within gender groups, this approach looks across distinctive race-gendered groups. In our study, there are four race-gendered respondents (Black men, Black women, Latino men, and Latina women) and eight race-gendered candidates (Black men, Black women, Latino men, Latina women, Asian men, Asian women, White men, and White women).
Examining intersectional identity characteristics introduces important dynamics for understanding the meaning of descriptive representation that have not yet been fully explored in the extant literature. To the extent that it has been studied, the focus has been on perceptions of the candidates. For example, scholars have demonstrated that women of color candidates face unique stereotypes because of their race-gendered identities that lead voters to make assessments of their perceived traits and leadership abilities (Carew Reference Carew, Brown and Gershon2016; Cargile Reference Cargile, Brown and Gershon2016; Hicks Reference Hicks2019; Lemi and Brown Reference Lemi and Brown2019). Studies demonstrate that women of color are viewed differently, but not necessarily for the worse. For example, Bejarano (Reference Bejarano2013, 32) argues that such candidates may experience a positive interaction of their gender and racial/ethnic identity. In this article, we are interested in how perceptions of representation might be impacted by the race-gendered positioning of individual respondents in relation to the race-gendered position of a hypothetical candidate.
The 2016 Election and Expectations for Representation
The 2016 election drew much attention to race-gendered patterns in electoral behavior. In particular, attention was given to the role that White women played in electing Donald Trump, despite predictions that he might alienate them. Recent research highlights the impact that Whiteness may play in shaping gender consciousness and political behavior (Cassese and Barnes Reference Cassese and Barnes2019; Frasure-Yokley Reference Frasure-Yokley2018; Junn Reference Junn2017). While White middle-class women may be subordinate to men, they are often accorded racial privilege. White women's interdependence with White men and their desire to maintain their relative privilege may lead them to choose solidarity with men within their own ethno-racial group rather than with women across ethno-racial groups. This focus on White women is an important intervention in the traditional work on gender, but the intersectional analysis of the 2016 election might be extended further to men and women of color. What race, gendered, and race-gendered differences exist for Black and Latino/a respondents in the evaluation of Black, Latino/a, Asian, and White candidates? For women of color, who are subordinated by gender and race in a political climate where both are salient, to what extent might each be relevant? And what might an intersectional analysis find regarding men of color, who are subordinated along dimensions of race but perhaps privileged by gender (although not necessarily in every context)? What differences exist across ethno-racial group?
To explore these questions, we propose the following hypotheses:
H1: Sharing both the gender and ethno-racial identity of the candidate described will be associated with a greater likelihood that respondents will express a belief that the candidate would represent their interests, compared with candidates sharing one or neither respondent identity.
This hypothesis emphasizes the simultaneous importance of race and gender, such that gender is important within racial categories and race is important within gender categories. For example, a Black woman would be more likely to express a belief that a Black woman candidate would represent their interests, compared with a Black man or a woman of any other ethno-racial group. Approaching this as a more open empirical question, we ask, for which race-gendered groups does this hypothesis hold?
H2: Sharing the ethno-racial (but not gender) identity of the candidate described will be associated with a greater likelihood that respondents will express a belief that the candidate would represent their interests, compared with candidates sharing neither respondent identity.
This hypothesis starts from a fairly well-established premise, in that it emphasizes the strength of shared ethno-racial identity over intra-group gender differences. In and of itself, it is not an intersectional hypothesis; however, here we look at how it holds across race-gendered groups, allowing for an intra- and intercategorical intersectional analysis.
H3: Sharing the gender (but not ethno-racial) identity of the candidate described will be associated with a greater likelihood that respondents will express a belief that the candidate would represent their interests, compared with candidates sharing neither respondent identity.
While race may trump gender, this hypothesis posits that shared gender still matters. How it matters, however, may vary by group. Here we focus on the evaluation of out-group (rather than in-group) ethno-racial candidates. Another way to think of this is to consider whether gender works as an alternative point of connection when no ethno-racial one exists. Approaching this hypothesis in a more open way engages the broader race-gender complexity of both respondents and the candidates they are evaluating. For which race-gender respondents does gender matter and how? To what extent might the ethno-racial identity of the candidate also matter?
METHODS, DATA, AND RESULTS
We employ data collected by the 2016 Collaborative Multiracial Post-Election Survey from December 2, 2016, to February 15, 2017. The sample includes 3,003 Latino/a respondents and 3,102 Black respondents. Within these groups, there are 2,183 Black women, 2,037 Latinas, 971 Black men, and 966 Latino men. The CMPS is a unique data set in that it contains a large and generalizable sample of Blacks and Latinos/as.Footnote 2 In addition to using the random-recruit-to-web approach, this survey employed both listed and density-sampling methodologies (Barreto et al. Reference Barreto, Frasure-Yokley, Vargas and Wong2018). This allows for analytical comparison across and within groups.
To test our hypotheses―that voters sharing gender, ethnic, and racial identities with the candidate will more often believe that candidate would represent their interests, compared with other respondents―we rely on a series of questions asking how well respondents felt different candidates (including a HispanicFootnote 3 man, a Hispanic woman, a Black man, a Black woman, a White man, a White woman, an Asian American man, and an Asian American woman) would represent their interests. Their responses to these questions were coded dichotomously, where 1 = yes definitely or yes somewhat, 0 = maybe or no (see Appendix I in the Supplementary Material online for a complete list of measures employed in the article).
The proportion of respondents believing each candidate would represent their interests across all respondent groups are reported in Table 1. The descriptive data show support for our first and second hypotheses. Latino male, Latina, Black male, and Black female respondents more frequently and significantly (relative to the other groups) express a belief that candidates sharing their racial and gender identities would represent their interests. Furthermore, respondents sharing the racial or ethnic (but not gender) identity of the candidate significantly more often indicate that the candidate would represent their interests (compared with those not sharing the racial identity of the candidate).
Table 1. Percentage of Black and Latino/a respondents believing candidates would represent interests (N = 6,157)
Note: The table is shaded to highlight the different types of shared identity, with the darkest shade used for our first hypothesis on shared race and gender, followed by slightly later shading for shared race, slightly lighter shading for shared gender, and finally the lightest shading for when there is no shared identity.
* Mean differences are statistically significant (p < .05, two-tailed).Footnote 4
The descriptive differences between the respondents are more mixed with regard to shared gender (but not racial) identities. The data with regard to male respondents are consistent with our expectations. A higher proportion of Black men than Black women believe that Black, Latino, Asian, or White male candidates will be able to represent their interests (the differences are significant for Latino and Asian male candidate evaluations). A similar pattern is found among Latino/a respondents for the Black, Latino, Asian, and White male candidate (the differences are significant for Latino, Asian, and White male candidate evaluations).
For the female candidates, the hypothesis holds for Latina respondents’ beliefs about the Black female and Latina candidates, as well as for Black female respondents’ beliefs about the Black female candidate (these differences are significant among both Latino and Black respondents). This might be interpreted as gender mattering more for men than for women; however, it might actually be more complicated than that. In The Latina Advantage, Bejarano (Reference Bejarano2013) argues that Latina candidates (and potentially other women candidates of color) not only might draw support from those who share their gender and/or ethno-racial identity, but in ethno-racial out-groups, they might be perceived as less threatening than their ethno-racial male counterparts. The descriptive data show some support for this type of race-gendered dynamic. When Latino/a and Black respondents are asked about ethno-racial out-group candidates (those who do not share their race and ethnicity), Latinas and Black males and females generally report slightly higher support for the potential of female candidates representing their interests compared with their assessment of the male candidates. The gender responses from the Latino male respondents are less consistent, with slightly more Latino males believing a Black male candidate will be able to represent their interests, no difference in perceptions of representation for the Asian candidates, and slightly more Latino males believing a White female candidate will be able to represent their interests.
The descriptive differences in Table 1 indicate that beliefs about shared identity and representation vary among the respondents. Next, we employ multivariate models to examine these attitudes while controlling for other variables that likely shape attitudes toward political candidates and representation. To test our expectation that shared identity will be associated with beliefs about candidates, we include binary variables in our next analyses that control for whether the respondents share the candidates’ race, gender, or both of these identities. The reference (excluded) category is the respondent group whose gender and racial identities are not shared by the candidate described in each analysis (see Table 2).
Table 2. Likelihood that congressional candidate would represent interests
Note: Cell entries are logistic coefficients and standard errors.
* p ≤ .05; ** p ≤ .01, two-tailed.
We examine the association between respondent identity and attitudes toward the candidates, controlling for other variables. We control for several demographic characteristics that may be associated with variance in support for these candidates. For example, we control for whether the respondent would call himself or herself a Democrat. With little information about the candidates (as in this survey), voters are likely to use candidate identity as a cue to make inferences about their characteristics (McDermott Reference McDermott1998). Given that the overwhelming majority of Black and Latino/a members of Congress are Democrats, respondents may assume these candidates are Democrats, which may, in turn, be related to their beliefs about how well the candidate would represent their interests.
Additionally, we control for the respondent's age and education, as well as whether the respondent is unemployed. Given that over 10% of the respondents were not born in the United States, we control for nativity with the variable native born, representing whether the respondent was born in the United States. The nativity variable aims to capture additional geographical variance in attitudes toward diverse candidates, which in the case of Latinos/as may be related to whether they have more experience with diverse racial groups living in the United States. Finally, we control for the proportion of the respondents’ district sharing the racial/ethnic identity of the candidate in the Black, Latino/a, Asian American, and White candidate models. The data are from the 2015 American Community Survey and provide local context for the diversity of the population. See Appendix I for a complete list of the measures employed and Appendix II for summary statistics for each variable.
To test our hypotheses, we rely on logistic regression predicting the likelihood that Black and Latino/a respondents would agree that each candidate would “definitely” or “somewhat” represent their interests (results reported in Table 2).Footnote 5 In this intersectional analysis, we examine the association between sharing both the gender and racial identities, sharing only the gender identity, and sharing only the racial identity of the candidate described, excluding the respondent group whose race and gender identities are not shared by the candidate in each model.Footnote 6
The data reported in Table 2 illustrate the complex relationship between shared identity and beliefs about descriptive representation. Recall that our first hypothesis predicts that sharing both the gender and ethno-racial identity of the candidate described will be associated with a greater likelihood that respondents will express a belief that the candidate would represent their interests, compared with candidates sharing one or neither respondent identity. The results indicate that sharing both racial/ethnic and gender identity is associated with a significant and positive increase in the likelihood that respondents would expect the candidate described to represent their interests. Specifically, when the Black and Latino/a respondents share both the gender and racial/ethnic identity of the candidate they are being asked about, the likelihood of the respondent believing the candidate would represent their interests is positive and significant, compared with respondents who do not share either of the candidates’ identities (the excluded category).
We further explore the implications of shared identity by calculating the predicted probabilities of expressing a belief that each candidate would represent their interests (displayed in Figures 1–10). As the data in Figures 1–4 indicate, respondents sharing both gender and ethno-racial identities with the candidates have a substantively higher likelihood of expressing a belief that the candidates would represent their interests. For example, the probability that Latino male respondents will express a belief that a Latino male candidate would represent their interests is .723, compared with .668 among Latinas (who share ethno-racial identity but not gender), .496 among Black men (who share gender but not race), and .379 among Black women (who share neither gender nor racial identity).Footnote 7 Black men have the highest probability (.77) of believing the Black male candidate would represent their interests, compared with Black women (.721), Latino men (.617), and Latinas (.556). We see the same pattern with regard to Black women and Latinas, with respondent shared gender and racial identity associated with the highest probability of believing the candidates would represent their interests (see Figures 2 and 4). The overlapping confidence intervals displayed in Figures 1–4, however, indicate that men and women sharing the candidates’ ethno-racial identities do not vary substantively in their support for their co-ethnic candidates.Footnote 8
Figure 1. Belief that Black male candidate would represent interests.
Figure 2. Belief that Black female candidate would represent interests.
Figure 3. Belief that Latino male candidate would represent interests.
Figure 4. Belief that Latina candidate would represent interests.
Next, we explore our second hypothesis, which predicts that sharing the ethno-racial (but not gender) identity of the candidate described will be associated with a greater likelihood that respondents will express a belief that the candidate would represent their interests, compared with candidates sharing neither respondent identity. The data in Table 2 and Figures 1–4 offer support for our expectations, clearly indicating that shared ethnicity or racial identity is associated with a substantive and significant increase in the likelihood of believing the candidate would represent their interests, compared with the excluded respondent group. Figures 1–4 display these differences, for example, the probability that Black respondents will express a belief that their interests are represented by the Black candidates described are over 70% in both Figures 1 and 2 (compared with 56% to 63% among Latino/a respondents in these models). Similarly, there is a gap between co-ethnic beliefs about the Latino/a candidates, with the probability of Latino/a respondents indicating that their co-ethnic candidates would represent their interests 15% to 38% higher than among Black respondents (see Appendix III).
Next, we examine our third expectation, that sharing the gender (but not ethno-racial) identity of the candidate described will be associated with a greater likelihood that respondents will express a belief that the candidate would represent their interests, compared with candidates sharing neither respondent identity. Both Latino and Black male candidates show significant support for the Latino/Black male candidates that share their gender but not their racial or ethnic identity. In contrast, shared gender, with no shared race or ethnicity, does not significantly impact Latina and Black female respondent evaluations of Latina/Black female candidates.
Our next set of analyses examines our hypotheses for the Latino/Black respondent evaluations of Asian and White candidates. The results are displayed in Table 3 and Figures 5–8. In contrast with shared race or ethnicity, shared gender identity has much more inconsistent results with regard to beliefs about White and Asian candidates. For example, of all the racial out-group female candidates Latina and Black women respondents were asked about, they were not more likely (compared with the excluded category) to express a belief that the candidates would represent their interests. Additionally, the coefficients for Black women respondents are negative and significant in the White and Asian American female candidate models, indicating that they were less likely than Latino male respondents (the excluded category in this model) to believe these female candidates would represent their interests. This multivariate analysis introduces an important caveat to acknowledge, since when we control for other variables including party identification, having a shared gender identity with a White or Asian congressional candidate is not associated with a greater belief that they would represent Latina and Black female respondents’ interests.
Figure 5. Belief that White male candidate would represent interests.
Figure 6. Belief that White female candidate would represent interests.
Figure 7. Belief that Asian American male candidate would represent interests.
Figure 8. Belief that Asian American female candidate would represent interests.
Table 3. Likelihood that congressional candidate would represent interests
Note: Cell entries are logistic coefficients and standard errors.
* p ≤ .05; ** p ≤ .01, two-tailed.
Returning to the descriptive data may be instructive here. As discussed earlier in the descriptive data, the findings illustrate that Black women's views may be more a function of increased perceptions of representation by male respondents of out-group female candidates. Given that Black women are the most consistent supporters of Democratic candidates, it is imperative to further explore in future work the additional factors that impact their sense of representation across diverse candidates.
In regard to out-group male candidates, our findings are more in line with expectations. In contrast with the female respondents, the male respondents in our study are―on average―more likely to believe that out-racial group male candidates would represent their interests, as indicated by the positive and significant coefficients for the Latino male candidate among Black male respondents and for all male candidates among Latino male respondents. Latino and Black male respondents have the highest probability of believing the out-group male candidates would represent their interests, with Latino men displaying the highest probability for the White male and Asian American male candidates. However, the substantive effects of gender may not be as sizable for out-racial group candidates. As the data in Figures 5–8 indicate, the probability that Black or Latino respondents will express a belief that out-racial group male candidates would represent their interests is substantively lower than in Figures 1–4, with the probability that these Latino and Black male respondents will express a belief that White and Asian male candidates would represent their interest ranging between .33 and .46. Thus, there are important differences in how the respondents perceive different ethno-racial out-groups. In the theoretical section of this article we discuss the possibility of overlapping coalitional identities that may emerge across ethno-racial groups. There does appear to be a “Brown-Black” connection when comparing perceptions of representation by Latino/a and Black respondents with the respective ethno-racial out-group as opposed to the White and Asian American candidates. Scholars of Asian Americans have long noted that positive stereotyping of Asian Americans by Whites as the “model minority” have been an impediment to cross-racial coalitions with other minoritized groups (Kim and Lee Reference Kim and Lee2001). Future work can delve deeper into disentangling the impact of diverse identities on the contrasting evaluations or levels of support for diverse candidates. In another study, we find that a coalitional sense of minority linked fate among Blacks and Latinos/as is positively associated with evaluations of both in-group and out-group candidates (Bejarano et al. Reference Bejarano, Brown, Gershon and Montoya2020; Gershon et al. Reference Gershon, Montoya, Bejarano and Brown2019).
In addition to our variables of interest, there are several consistent results with regard to our control variables. As expected, Black and Latino/a respondents who are Democrats are more likely to believe all but one of the candidates (the White male) would represent their interests. Education levels are significantly and positively associated with the Black and Latino/a respondents’ belief that Latina, Black, and Asian American candidates would represent their interests. Unemployment is significantly and negatively associated with Black and Latino/a respondents’ belief that Black and Asian candidates would represent their interests, which may point to a disillusionment with their elected officials. Nativity in the United States is significantly and negatively associated with Black and Latino/a respondents’ belief that Latino male candidate would represent their interests (while this association is positive for the Black female candidate).
CONCLUSION
As the political environment changes, so does the meaning attached to descriptive representation. While Democratic candidate pools and voters continue to diversify in a multitude of ways, a deeper and more complex understanding of representation is required. The dynamics of identity and its significance for representation are complicated and “messy”; but, like Smooth (Reference Smooth2006), we think that it is a mess worth making. Intersectional analysis provides a more nuanced understanding of the influence different types of shared identities may have on support for diverse political candidates. It bridges and moves forward previous work that focused on shared racial identity or gender identity.
The results highlight that there is some raced, gendered, and race-gendered variation in the ways that Latinos/as and Blacks perceive other ethno-racial groups potentially representing them in politics. Sharing both gender and ethno-racial identities with the candidates provided the consistently highest levels of perceived representation. Shared ethno-racial identity was consistently important for perceptions of representation. The impact of gender, however, was much more complicated. Our results indicate that shared gender identity―particularly among racial out-groups―is not consistently associated with a belief that candidates would represent voters’ interests. This underscores the need for more careful, intersectional, and cross-racial analysis that asks not whether gender matters, but how much and in what ways does it matter.
Future research might even go one step further. In recent years, important critiques and challenges have been made to the ways in which social science uses rigid categories of gender and race that are not fully reflected in the population. Scholars have argued the malleability of race and problematized static group categories (Davenport Reference Davenport2020; Masuoka Reference Masuoka2017; Sen and Wasow Reference Sen and Wasow2016). Danielle Casarez Lemi (Reference Lemi2018) has drawn attention to an increasingly multiracial population and the ways in which this might impact our understandings of racial representation. Additionally, scholars have found that voters perceive differences in how descriptive representatives will represent the substantive needs of their group based on the legislator's skin tone (Burge, Wamble, and Cuomo Reference Burge, Wamble and Cuomo2020; Lemi and Brown Reference Lemi and Brown2019; Orey and Zhang Reference Orey and Zhang2019). Similarly, social scientists are just starting to move beyond binary approaches to gender in surveys (Medeiros, Forest, and Öhberg Reference Medeiros, Forest and Öhberg2020). In sum, intersectional analysis may need to incorporate a more fluid, dynamic, and comprehensive set of characteristics.
Ultimately, our findings demonstrate that identity still matters, but there is still much work to be done to disentangle how it matters in contemporary American politics. The increased diversity in our governing bodies, at least on the Democratic side of the aisle, is likely to mean something in regard to collective perceptions of representation. Nevertheless, the findings point to the complexity in the association between identity and representation among voters. At the same time, candidates have a lot of work to do in reaching out and connecting to a diversifying electorate. Intersectionality shows that there are more avenues for connection than most single-axis approaches assume, but that candidates will need to be able to convincingly convey their ability to represent a diverse array of constituents in order to create the electoral coalitions necessary to win elections.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S1743923X20000744.
