GENDER AND VOTING BEHAVIOR IN THE UNITED STATES

Like their male counterparts, the motivations for candidate choice among American women are complex and multi-dimensional. Arguably, no single measure, with the possible exception of political party identification, consistently predicts vote choice. A growing body of literature finds strong support for the role of partisanship and partisan concerns on voter decisions (Box-Steffensmeier, De Boef, and Lin Reference Box-Steffensmeier, De Boef and Lin2004; Dolan Reference Dolan1997; Reference Dolan2004; Reference Dolan2014). There is overwhelming evidence, regardless of gender; both women and men are likely to support the candidate of their own party.

While there is little previous evidence of systemic sexism on vote choice prior to 2016 election, gender biases and stereotypes have been shown to influence evaluations of candidates (Dolan and Sanbonmatsu Reference Dolan and Sanbonmatsu2009; Sanbonmatsu Reference Sanbonmatsu2002; Reference Sanbonmatsu2003; Sanbonmatsu and Dolan Reference Sanbonmatsu, Dolan, Aldrich and McGraw2012). More recently, Sharrow et al. (Reference Sharrow, Strolovitch, Heaney, Masket and Miller2016) find that candidate evaluations are influenced by the interaction of partisanship and attitudes about women's roles. Research related to black women shows that black women are “doubly bound” whereby their attitudes toward candidate and policy issues are shaped by mutually reinforcing gender and racial identities (Gay and Tate Reference Gay and Tate1998). Moreover, Philpot and Walton (Reference Philpot and Walton2007) find that black women are the strongest supporters of black female candidates.

There is a growing body of literature on the role of sexism and racial resentment on the 2016 Presidential election outcome. Using the 2016 ANES Pilot Study, Tien (Reference Tien2017) found that racial resentment, not economic class, explains support for Trump among white women. Using the 2016 ANES Times Series, McElwee and McDaniel (Reference McElwee and McDaniel2017) found support for the role of racial resentment toward blacks, and negative views toward immigration, influence vote choice for the GOP candidate in 2016, with little support for the role of economic anxiety. Bock, Byrd-Craven, and Burkley (Reference Bock, Byrd-Craven and Burkley2017) use survey data collected online immediately following the 2016 election among college students. They found that individual differences in hostile sexism and traditional attitudes toward women significantly predicted vote choice for Trump. Supporting these findings, Schaffner, MacWilliams, and Nteta (Reference Schaffner, MacWilliams and Ntetaforthcoming) find that racism and (hostile) sexism in the electorate, particularly among non-college educated whites influenced vote choice for Trump, even after accounting for partisanship, ideology, and other factors.

However, most existing studies, both prior to and following the 2016 Presidential election, were conducted using models undifferentiated by race and ethnicity. Research designs without a comparison of models separated by race and gender may obscure substantial divergence among white women voters and women of color in their preferences for both party and candidates. Again, the 2016 election revealed a fact of American politics, which was “hiding in plain sight” since 1948: white women politically behave very differently from women of color and have consistently done so, with few exceptions (Junn Reference Junn2016). Hillary Clinton's historic nomination as the first female candidate to earn a major party nomination in a Presidential election, coupled with her opponents low favorability ratings, and overall penchant for sexist, misogynist, and racist commentary, presented a rare opening for Republican women voters, raising the presupposition that this group could be pushed to vote across party lines (Brians Reference Brians2005). This was not the case for Clinton in 2016.

In the end, party loyalty trumped gender identity, again. However, several questions remain regarding the extent to which views toward sexist attitudes influenced women's vote choice in 2016, particularly between white women and women of color. Jackman (Reference Jackman1994) argues that dominant groups—and in the case of gender, males—act strategically to avoid overt hostility with subordinate groups (women) by practicing coercion within the context of an ideology of paternalism. Love and affection are given to subordinates only under the condition that they recognize and adhere to the hierarchical relationship and accept their subordinate status. Jackman argues persuasively that members of the dominant group are not alone in this arrangement; subordinate group members (women) must also comply by embracing the ideology and agreeing to choose the proffered “velvet glove”—in this case, Trump's promises of protection and being cherished—albeit knowing that it serves to cover the iron fist of patriarchy. This arrangement is justified by love and caring that supports the promise of the provision of goods for members of the subordinate group, and for protection by the dominant group. Resources are then controlled within the larger group, in this case, white Americans, as long as the unequal relationship between men and women is preserved. The preservation of the status quo is maintained, and justified through “system justifying ideologies” or a psychological orientation toward structural inequality among groups (Jost and Banaji Reference Jost and Banaji1994; Sidanius and Pratto Reference Sidanius and Pratto1999). Importantly, the intersection between Jackman's work and the longstanding research in “system justification theory” (Jost and Banaji Reference Jost and Banaji1994) or “social dominance orientation” (Sidanius and Pratto Reference Sidanius and Pratto1999; Sidanius et al. Reference Sidanius, Pratto, van Laar and Levin2004) is the interdependence between men and women in the promotion, reinforcement, and internalization of gender roles and stereotypes toward the maintenance of the status quo.

One lens to examine Jackman's theory is ambivalent sexism, which stresses the creation of ideologies about women, while also drawing attention to how some women accept these system-justifying ideologies (Glick and Fiske Reference Glick and Fiske1996; Jost and Banaji Reference Jost and Banaji1994). In gender relations, members of the two groups often have close intimate/romantic and familial relationships—within traditional gender roles there is power differentiation (dominance and sub-ordinance). For example, while men may compete with women for resources and power, many of the same men devote their resources to providing for the women in their lives (Glick and Fiske Reference Glick and Fiske1996). There are two forms of ambivalent sexism, benevolent and hostile. Benevolent sexism is a chivalrous attitude where one sees women as weak creatures in need of protection. Hostile sexism is an antagonistic attitude where one sees women as seeking to control men.

Ambivalent sexism creates a system of rewards and punishments, which encourages women to maintain conventional gender roles, mostly out of fear or anxiety, thus sustaining systemic gender inequality (Glick and Fiske Reference Glick and Fiske2001). The underlying rationale being that relationships between men and women are rooted in male dominance in high status roles (patriarchy and power), and intimate interdependence between the sexes leads some men and women to have universal and rigid gender prescriptions. Ambivalent sexism aims to capture the nuance of sexism as it was changing in the late 1980s and early 1990s. This accounts for not only hostile remarks, but also non-overtly hostile remarks. This new measure of sexism moves beyond both old fashioned as well as modern sexism (i.e. attitudes toward women scale) that failed to account for this important nuance.

The Ambivalent Sexism Inventory (ASI), originally created by Glick and Fiske (Reference Glick and Fiske1996), employs a series of 22 survey questions composed of two subscales that measure both hostile and benevolent sexism, with 11 questions measuring each subscale. Importantly, women who reject traditional gender roles or are career-oriented are the traditional targets of hostile sexism. On the other hand, women who adopt domestic norms, such as stay-at-home mom are subject to benevolent sexism. Overall women are more likely than men to reject hostile and benevolent sexism (Glick and Fiske Reference Glick and Fiske1996; Killianski and Rudman Reference Kilianski and Rudman1998); however, these studies did not disaggregate models by race and ethnicity.

Hostility toward Clinton, by both the Republican and Democratic parties, has brewed for nearly 25 years of her life in the political spotlight (Gates Reference Gates1996). Hillary Clinton's personal and professional career choices, down to her infamous pantsuits and perceived “likable enough” personality, made her a target for hostile sexism. As noted, more recently scholars have examined the ASI through the lens of American electoral politics. Bock, Byrd-Craven, and Burkley (Reference Bock, Byrd-Craven and Burkley2017, 192) argue that the backlash against Clinton symbolized the extent to which her ascend violated traditional norms underscoring persistent distain for Clinton among some women.

I hypothesize that the extent to which then candidate Trump's paternalist construction of gender relations resonated with female voters was a function of their attitudes about sexism. I test these hypotheses by treating respondents’ level of opposition to the ambivalent sexism measures as a proxy for their acceptance of patriarchal relations. I expect that white female voters who agree with ambivalent sexist statements asked about in the 2016 ANES will be more supportive of candidate Trump, than white female voters who do not support these views, while holding constant all other factors. In contrast, these attitudes should have no effect on the likelihood of supporting Trump among women of color. The reason for these differences is the intersectional position of race and gender for women voters.

Unlike white women, women are color, are rarely able to extricate the intersection of their racialized, gendered identity and positionality (Collins Reference Collins1990; Gay and Tate Reference Gay and Tate1998; Hooks Reference Hooks1981; Reference Hooks1984; Jordan-Zachery Reference Jordan-Zachery2007; King Reference King1988; Prestage Reference Prestage1977). I hypothesize that white women, given their position as “second in sex, but first in race” will be the only group of female voters to choose the velvet glove over the extended hand of a white woman running for U.S. President. In contrast, women of color are second in both sex and race. For women of color, I hypothesize that descriptive attitudes in agreement with the ambivalent sexism measures will not influence a positive choice for the Republican candidate who exemplifies patriarchal values. Moreover, white women who reject unequal patriarchal arrangements would reject the candidacy of Trump in the same way as the vast majority of minority women voters. Basing this analysis of these outcomes on empirical tests of Jackman's (Reference Jackman1994) theory of intergroup oppression, through a measure of ambivalent sexism, will allow us to explore how gender and race simultaneously help to explain white women's support for Trump.

While ambivalent sexism has been internationally studied for over 20 years, in nearly 20 nations and across six continents, as McMahon and Kahn (Reference McMahon and Barsamian Kahn2015) contend, there is a lack of systematic investigation into how hostile or benevolent sexist attitudes may be shaped by the target women's race and the race/ethnicity of the respondent (170).

To further examine these factors and their potential effect, I turn to the role of racial resentment. Racial resentment is synonymous with symbolic racism, where one sees racial prejudice as no longer an obstacle for minority economic advancement. Some central tenets of the theory argue that continuing disadvantage for African-Americans is their own fault because they simply do not work hard enough, and this group “takes what they have not earned.” Thus, claims of continued discrimination and persistent calls for racial equality are unjustified (Kinder and Kam Reference Kinder and Kam2009; Kinder and Sanders Reference Kinder and Sanders1996; Kinder and Sears Reference Kinder and Sears1981; Tesler and Sears Reference Tesler and Sears2010). To measure racial resentment, most researchers use a version of the original racial resentment scale, which debuted in the 1986 ANES (Kinder and Sanders Reference Kinder and Sanders1996; Tesler and Sears Reference Tesler and Sears2010; Tuch and Hughes Reference Tuch and Hughes2011). The scale was initially development with six survey items, and later shortened to a four-item scale. As detailed in the next section, I account for racial resentment in order to isolate the effects of ambivalent sexism, while separating out the effects of potential racial animus. I further examine some of the limitations in using the racial resentment scale, particularly for women of color. I hypothesize that women with high levels of racial resentment on a scale of 0–1, are more likely to favor Trump than women with lower levels of racial resentment.

RESEARCH DESIGN AND METHODS

This study uses data from the 2016 ANES, which is part of a time-series collection of national surveys fielded continuously since 1948. The ANES are designed to present data on Americans’ social backgrounds, enduring political predispositions, social and political values, perceptions and evaluations of groups and candidates, opinions on questions of public policy, and participation in political life. As with all Time Series studies conducted during years of presidential elections, respondents were interviewed during the 2 months preceding the November election (pre-election interview), and then re-interviewed during the 2 months following the election (post-election interview). Like its predecessors, the 2016 ANES was divided between questions necessary for tracking long-term trends and questions necessary to understand the particular political moment of 2016 (ANES 2017).

This 2016 ANES study features a dual-mode design with both traditional face-to-face interviewing (n = 1,181) and surveys conducted on the Internet (n = 3,090), and a total sample size of 4,271 (ANES 2017 release). There are 3,649 (85%) respondents who completed both the pre- and post-interviews. Of those respondents, 2,691 reported voting in the 2016 Presidential election. In this study, I examine all female, self-reported voters in the 2016 presidential election (n = 1,424). I also create a separate model to examine non-Hispanic white women (n = 1,062) and another to examine women of color (n = 362).

While individual models fully separated by racial and ethnic group would be ideal, the 2016 ANES sample sizes for non-white female voters are too sparse, and do not allow researchers to fully disaggregate by gender, race, and ethnicity. Lumping the responses of women of color together is problematic given the vast differences in backgrounds, cultures, experiences with systems of oppression, and avenues toward political incorporation. However, in this study as it relates to the dependent measure, Presidential vote choice, this mode of investigation is appropriate for the following reason. Women of color remain consistent supporters of the Democratic Party with black women being its most loyal stalwarts. On the other hand, the majority of white women are consistent Republican Party loyalist, voting in line with the GOP. Weighted 2016 ANES data reveal that 42% of American women voters reported casting a ballot for Trump in 2016. However, once this figure is disaggregated by race, on average, 52% of white women voted for Trump, while on average, only 15% of all women of color.

The ANES remains problematic for in-depth intersectional analysis along race/ethnicity, given its historical exclusion of large samples of people of color. Notably, the first ANES oversample of African-Americans occurred in 1964. However, the sample sizes remained relatively low, with some exceptions, including the oversamples of African-Americans in 2008 and 2012. Post-election questionnaires were produced in both English and Spanish versions for the first time in 1992. However, 2008 marked the first full Spanish-language instrumentation, programmed together with the English-language instruments. The first ever Latino oversample was conducted in 2008 and again in 2012. Unfortunately, there were no oversamples of any racial or ethnic groups in the 2016 ANES. The practice of oversampling is costly; however, it allows research to select respondents so that some groups make up a larger share of the survey sample, than they do in the population. Sampling weights are then employed to bring the population back in line with their actual share of the population, while allowing researchers a larger sample to conduct more in-depth analysis of the group (Mercer Reference Mercer2016).

The 2016 ANES surveyed 1,591 white women (1,062 voters), 237 black women (163 voters), 66 Asian women (30 voters), 14 Native American (seven voters), 216 Hispanic (103 voters), and 94 women (52 voters) who self-identified as other (including multiple races). Combining these groups, through imperfect, provide an opportunity to begin an empirical investigation into differences in political behavior between white women and women of color. Despite the data limitations, the 2016 ANES is useful for this analysis because it features the following four (only) measures drawn specifically from the Hostile Sexism Scale of the ASI—developed by Glick and Fiske (Reference Glick and Fiske1996).

• • “Women fail to appreciate what men do for them.”

• • “Many women interpret innocent remarks or acts as being sexist.”

• • “Women seek to gain power by getting control over men.”

• • “Once a woman gets a man to commit to her, she tries to put him on a tight leash.”

These four measures were added to the ANES for the first time in 2016. It is the only known (publically available) data source to include measures from the Ambivalent Sexism Index, along with the measures of political attitudes, predispositions, socio-demographics, and vote choice.Footnote ¹ The hostile sexism scale is constructed from these four statements by coding the five potential responses to each assertion from 0 to 1 in intervals of .25, with 0 being the least supportive of these statements and 1 being the most supportive of these statements. These answers are summed and divided by the number of items to provide an easily interpretable 0–1 scale. A score of .50 marks the 50% mark, so that people below 50 hold less hostile sexist views and people above the 50% are hold more hostile sexist views. The high scale reliability of these items (Cronbach’ α score = .788) reflects an internally consistent construct, rather than four multiple and diverse themes in which it is operationalized.

Table 1 describes the distribution of agreement with the hostile sexism measures, this study's main explanatory variable, by gender and race.Footnote ² Women of color are, on average, slightly more likely than white women to agree or strongly agree that “women fail to appreciate what men do for them,” “women seeks to gain power by getting control over men,” and more likely, on average, to agree with the statement “once a woman gets a man to commit to her, she tries to put him on a tight leash.” On the other hand, white women, on average, are more likely to report agreement with the statement that “women interpret innocent remarks or acts as being sexist,” than women of color. Despite these differences, once scaled together, there is little variation in mean hostile sexism scores between these groups (36% for all women, 37% for white women, and 38% for women of color).

Table 1. Distribution of Mean Views on Ambivalent Sexism Statements, by Gender and Race (Voters Only)

Source: 2016 ANES.

Notes: All estimates are adjusted using survey weights (*Cronbach's α score .788).

The racial resentment scale used in this study follows Tesler and Sears (Reference Tesler and Sears2010) and is constructed from how strongly respondents agreed or disagreed with the following assertions:

• • “Irish, Italian, Jewish, and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors”;

• • “Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class”;

• • “Over the past few years, blacks have gotten less than they deserve”; and

• • “It is really a matter of some people not trying hard enough; if blacks would only try harder they could be just as well off as whites.”

The scale was created by coding the five potential responses to each assertion from 0 to 1 in intervals of .25, with 0 being the most racially liberal response and 1 being the most racially conservative. These answers are summed and divided by the number of items to provide an easily interpretable 0–1 scale. A score of .50 marks the 50% mark, so that people below 50 are considered more racially liberal, while people above the 50% are considered more racially conservative in their views toward blacks. The Cronbach’ α score for these for measures is .844.

Racial resentment was originally conceptualized as a measure of white prejudice toward blacks. To date, exactly what racial resentment represents among non-whites toward blacks remains unclear (Tesler and Sears Reference Tesler and Sears2010). Kam and Burge (Reference Kam and Burge2017) examine the open-ended reactions of both white and black Americans to these four racial resentment measures, which as they note, a vast majority of published articles have largely ignored black responses, despite consistently surveying this group (but see Orey Reference Orey2012; Tesler and Sears Reference Tesler and Sears2010). They find variations for both groups in how much they agree and disagree with the four measures as well as what goes through respondent's minds when answering the questions. They note that both groups “consider negative traits, individualism, and discrimination as they reflect on these questions” (2). These scholars call for a recasting of the scale as “Structural versus Individual Attributions for Black Americans’ Economic and Social Status (SIA), where low values emphasize structural inequalities and high values individual failings (6).” Further, some scholars argue for an explicit measure of racial resentment that modernizes and clarifies its operationalization, meaning of its conceptual framework, and refines its measurement properties (Wilson and Davis Reference Wilson and Davis2011).Footnote ³

Table 2 presents the distribution of women voters’ mean views on the racial resentment statements, by race and gender. In the descriptive statistics, I provide a separate category for black women in order to examine average views of black women voters toward racial resentment, in comparison to white women and other women of color voters. There is great variation in the mean racial resentment scores between these groups (52% of all women, 56% for white women, and 49% for non-black women of color and 31% for black women).

Table 2. Distribution of Mean Views on Racial Resentment Statements, by Gender and Race (Voters Only)

Source: 2016 ANES.

Notes: All estimates are adjusted using survey weights (*Cronbach's α score .844).

As Jackman (Reference Jackman1994) contends, women and men of all backgrounds can hold both gender and racial based prejudices about one another and internally. The empirical question for this study is the extent to which such views influence political behaviors and how these behaviors vary between white women and women of color.

The 2016 election season also underscored candidate Trump's persistently race-baiting, dog-whistle, overtly racist remarks concerning and toward immigrants. To test the role of views toward immigrants, I include the measure of responses toward the question: “Now I would like to ask you about immigration in recent years. How likely is it that recent immigration levels will take jobs away from people already here?” This measure is scaled from 0 to 1 (0 = not likely at all; 1 = extremely likely). I hypothesize that women who respond that immigrants are extremely likely to take their jobs away are more likely to favor Trump, than those who do not hold these views.

The role of white economic anxiety took center stage as scholars and pundits scrambled to disentangle the election outcome (Parker Reference Parker2016; Schaffner, MacWilliams, and Nteta, Reference Schaffner, MacWilliams and Ntetaforthcoming; Vance Reference Vance2016). Economic anxiety can be described as a feeling that one's personal economic status is threatened. For example, economic anxiety has been shown to alter public opinion toward immigration where individuals higher in economic anxiety exhibit less support of immigrant-friendly policies and support the Tea party in 2012 (Parker and Barreto Reference Parker and Barreto2013). To examine the influence of economic anxiety, I include a measure of personal economic well-being: “We are interested in how people are getting along financially these days. Would you say that [you/you and your family living here] are [much better off financially, somewhat better off, about the same, somewhat worse off, or much worse off financially] than you were a year ago?” (1 = somewhat or much worse off/0 = otherwise).

As further statistical controls, I also include measures to account for the role of partisanship (0–1 scale from strong Democrat to strong Republican)Footnote ⁴; conservative ideology (1 = conservative/0 = liberal or moderate)Footnote ⁵; race (1 = non-Hispanic white/0 = otherwise (full sample only); 1 = non-Hispanic black/0 = otherwise (women of color sample only)); educational attainment (0–1 from high school graduate or less to master's degree or more); marital status (1 = married/0 = otherwise); religious identification (1 = evangelical/0 = otherwise)Footnote ⁶; and age (18–89 years). Finally, to account for whether or not respondents heard of the Access Hollywood video, I include the measure, “In October, the media released a 2005 recording of Donald Trump having a crude conversation about women. Have you heard about this video, or not?” (1 = yes, heard about the video/0 = no, have not heard about it).

FINDINGS AND DISCUSSION

This study examines the role of ambivalent (hostile) sexism toward women in the 2016 election, controlling for racial resentment, partisanship, views toward immigration, personal economic anxiety, and socio-demographics, among white women and women of color. Hostile sexism and racial resentment were two powerful influences on women's Presidential vote choice in 2016, even while controlling for partisanship, conservative ideology, and other factors. In the sample of all women, hostile sexism toward women is a positive, and significant predictor of voting for Trump in 2016. However, this finding is quite misleading once an intersectional approach is undertaken, which disaggregates by both gender and race. Once the model is separated by race and gender, some interesting trends emerge. To aid in the interpretation of the logit regression coefficients, column II in Table 3 presents the marginal effects of each measure, while holding the other variables in the models, constant at their means. As the table indicates, once disaggregated by race and gender, the effect of hostile sexism increases the probability of voting for the GOP candidate by 17 percentage points among white women. However, for women of color, hostile sexism was negatively associated with voting for Trump and, in fact, posed no statistically significant effects, controlling for all other factors.Footnote ⁷ As hypothesized, despite the fact that women of color, on average reported they were more likely than white women to agree that “women fail to appreciate what men do for them,” “women seeks to gain power by getting control over men,” and “once a woman gets a man to commit to her, she tries to put him on a tight leash,” these measures, which make up part of the ANES hostile sexism scale, posed a negative relationship to the candidate and, after accounting or all other factors, was not a driving force in their voting decisions.

Table 3. Predictors of GOP Vote Choice in 2016 Presidential Election, by All Women, White Women, and Women of Color (Column I: logistic regression coefficients, Column II: marginal effects)

Source: 2016 American National Election Study (ANES), dependent variable 1 = Trump/0 = otherwise (self-reported voters only).

Notes: Standard errors in parentheses (***p < .001, **p < .01, *p < .05). Statistical procedures for complex sample designs were used to obtain correct estimates of sampling errors and correct indications of statistical significance. All estimates are adjusted using the appropriate survey weights.

Next, the racial resentment scale indicates that a change in probability of voting for Trump, when women's attitudes toward black people goes from racially liberal to racially conservative, increase by 32 percentage points in the full sample of women. Again, once the model is disaggregated by gender and race, the effect of racial resentment appears to be slightly larger for white women (.38) and significantly smaller for women of color (.14). While the racial resentment literature “controls” for the role of gender, it largely fails to disaggregate by gender. Therefore, although we have 30 years of racial resentment literature, very little of it provides insight into differences between white men and white women. Moreover, we have much more work to do toward understanding and disentangling the role of racial resentment for women of color, disaggregated by race, ethnicity, or national origin group.

These models confirms the longstanding findings concerning the role of partisanship on vote choice, whereas stronger leaning Republican women were 36 percentage points more likely than left leaning women to cast a ballot for Trump, this effect was similar for white women (.37) but the role of partisanship was less salient for women of color at .28. Moreover, if we view female voters as a monolith, it appears that conservative women were more likely to support Trump versus liberal and moderate women. However, when this measure is disaggregated by race and gender, white female conservatives are more likely to support Trump, while this measure is negative and has no statistically significant influence on vote choice for women of color, controlling for all other factors.

The additional control measures also present some interesting findings. As hypothesized, views toward immigrants posed a positive and statistically significant effect on vote choice in the 2017 election, among all women (.11), white women (.10), and women of color (.11), controlling for all other factors. Moreover, regarding economic anxiety, the model for all women suggests that respondents who believe their personal economic condition is worse off than 1 year ago are .05 percentage points more likely to cast a ballot for Trump than those who did not believe their economic condition worsened. However, once the model is separated by gender and race, economic anxiety posed no statistically significant effect on vote choice for white women, nor women of color, controlling for all other factors. Next, for white women, racial identity has a positive and statistically significance influence on vote choice for Trump versus non-white women. In this study, I included a control variable for black women to account for differences in racial identity among black women relative to other women of color in the sample. This measure was negative and statistically insignificant.

There are several main takeaways from this research. First, importantly, the same stimulus among different groups can have different effects or no effect at all on vote choice. Allport (Reference Allport1954) writings about the nature of prejudice still ring true today, “The same heat that melts the butter also hardens the egg.” In this case, white women behavior politically differently than women of color. Second voting behavior must be analyzed intersectionally. Scholarship in gender and politics that does not account for group differences in race and ethnicity may present misleading results that are underestimated or overestimated. Scholars of gender and politics who continue to invest in the monolithic myth of the partisan gender gap, ignore the role race/ethnicity at the peril of generating incorrect answers for which an intersectional and more nuanced analysis can reveal.

Many have pointed out the hegemony of white womanhood in the gender and politics literature (Cathy Reference Cohen and Carroll2003; Smooth Reference Smooth2006). Scholars such as Smooth (Reference Smooth2006) raise important questions concerning what we lose by an investment in the gender gap narrative as the primary lens through which to examine women and politics, writing,

This research expands the literature in women and politics as well as racial and ethnic politics. It is one step toward a working methodological framework for how to explore intersectionality empirically, which includes the need to examine models, separated race, and gender. The role of sexism and racial animus among and between women need further investigation. Future research should continue to measure and refine racial resentment (Kam and Burge Reference Kam and Burge2017; Wilson and Davis Reference Wilson and Davis2011) as well as incorporate measures of the Ambivalent Sexism Index, including both hostile and benevolent measures.

As the U.S. becomes increasing racially, ethnically, and linguistically diverse, it is time to push the boundaries of our data collection procedures to include large and generalizable samples of racial and ethnic groups and to allow for within-group comparison and analysis of an individual racial group, or comparative analysis across groups (Barreto et al. Reference Barreto, Frasure-Yokley, Vargas and Wong2018; Jackson et al. Reference Jackson, Torres, Caldwell, Neighbors, Nesse, Joseph Taylor, Trierweiler and Williams2004). This may be done through oversampling or via new innovative sampling methods to reach racial and ethnic minority populations beyond a ‘one-size-fits-all’ approach (Berry, Chouhoud, and Junn Reference Berry, Chouhoud, Junn, Atkeson and Michael2016). One approach called random-recruit-to-web has been implemented in mixed-mode phone-web public opinion research among Latino voters (Barreto and Segura Reference Barreto and Segura2014). A similar method was also used for the registered voter samples in the 2016 Collaborative Multiracial Post-election Survey (CMPS), conducted immediately following the 2016 Presidential election.Footnote ⁸

While survey data have their advantages, intersectional research calls for a diversity of methodologies, whereas qualitative modes of inquiry such as focus groups and in-depth interviews are equally important. Such methods will help researchers to examine how women from various racial and ethnic groups, native-born and naturalized, voters and non-voters, conceptualize and evaluate the role of sexism, racism and other factors when making political choices at the local, state, and federal levels.