Popular accounts of the 2016 presidential contest emphasize that Donald Trump was elected primarily because he mobilized previously disinterested and disillusioned voters in key states such as Pennsylvania, Ohio, and Michigan. His core supporters shared many demographic characteristics, including identifying as working-class, evangelical Christians alienated by a perceived decline in America’s status and traditional values. Trump did especially well with white men without a college degree, winning nearly 75% of these voters (Malone 2016). Pundits largely framed Trump’s support as a knee-jerk reaction to lost manufacturing jobs in the Rustbelt and a self-serving opposition to an increasingly global and technology-centered economy. Trump’s rallying cry—including, most prominently, calls to “make America great again”—also had more than an undertone of patriarchy, racism, and American exceptionalism.
Bringing together this previously amorphous group of former union Democrats, apathetic NASCAR dads, and evangelical Christian conservatives is no small accomplishment, and scholars considering the trajectory of the Republican Party should not ignore their contributions to Trump’s victory. Nevertheless, researchers and political professionals also should not dismiss the critical role that female voters played in the election outcome. Indeed, exit polls indicated that more than half of white women voted for Trump (Malone 2016), despite his well-documented history of sexual harassment and demeaning comments about female opponents in both the primary and general elections. Who were the women who supported Trump? Were they motivated mostly by their party affiliation, as in recent presidential elections? Were they married women dutifully taking cues from their husbands? Or, did other factors—perhaps attitudes and beliefs hostile to equality—motivate these women to support his candidacy?
This analysis explores the motivations of female Trump voters in the 2016 presidential election. Contrary to pundits’ expectations, we found that the women who supported Trump were not particularly stalwart partisans, and neither were they any more Southern, working-class, evangelical, or white than male Trump voters. Instead, female voters, like their male counterparts, were most powerfully influenced by the degree to which they held racially resentful and sexist attitudes. Thus, the women who voted for Trump did so largely because they were not the equality-minded individuals emphasized in the gender-gap literature.
THE GENDER GAP AND PARTY LOYALTY IN AMERICAN POLITICS
Gender gaps in political participation and vote choice have long been evident in American politics. In fact, the gap predates the Nineteenth Amendment, which granted women the right to vote in 1920. Some commentators posit that it began as early as the Revolutionary War, when women disproportionately supported General George Washington. More likely, however, the gap’s roots lie in nineteenth-century social and political activism. The suffrage movement and its associated causes—urban reform, temperance, and peace, for example—brought women together in the names of “social justice feminism” and “good government.” The latter of these causes was particularly important in distinguishing female voters from male voters in states and localities that enfranchised women during this period (Jabour 2016).
Was there something about Trump’s campaign promises—his appeals to strong leadership and nativism—that caused many women voters to overlook this behavior? Do twenty-first-century women passively accept, and perhaps even embrace, anti-egalitarian views?
From these roots, women gained a reputation for liberal-issue advocacy and disproportionate resistance to inequality. During the second wave of the women’s movement, activists worked predominantly with the Democratic Party to advance legislation on reproductive rights and educational and employment discrimination. However, until the 1980s, the gender gap in voting behavior was less apparent. Bolstered by support from conservative Southerners, the Democratic Party enjoyed a partisan advantage among both men and women. Since the 1980 presidential election, however, men have increasingly gravitated toward the Republican Party (Kaufmann and Petrocik 1999). A consequence of this shift has been an increasing overlap among partisanship, ideology, and gender stereotypes in contemporary politics. Modern commentary treats “Democrat,” “female,” and “liberal” as essentially synonymous (Elder Reference Elder2008; Winter Reference Winter2010).
Today, the gender gap is so wide that female Republicans are a minority in their own party, with their political behavior understudied and largely underappreciated (Schreiber Reference Schreiber2012). Trump’s candidacy provides a unique opportunity to examine the source and extent of these women’s loyalty to their party and its nominees. Why would women support a presidential candidate who was caught on tape bragging about how his celebrity status enabled him to sexually assault women without repercussions? Was there something about Trump’s campaign promises—his appeals to strong leadership and nativism—that caused many women voters to overlook this behavior? Do twenty-first-century women passively accept, and perhaps even embrace, anti-egalitarian views?
HYPOTHESES, DATA, AND MEASUREMENT
Recent gender and politics scholarship (Dolan Reference Dolan2014; Hayes Reference Hayes2011) led us to anticipate that partisanship would largely explain women’s support for Trump, despite questions about his character and his rejection of many of the party’s long-standing issue positions. We also expected that female support for Trump would reflect many of the demographic characteristics emphasized in popular accounts of the election. Specifically, religiosity, region, class, education, age, and marital status are all significant predictors of Republican support generally. Trump made an explicit effort to appeal to both traditional Republicans and others who have become increasingly frustrated with modern, multicultural America. Finally, we anticipated that Trump’s female supporters might be more likely than other women to hold several anti-egalitarian beliefs identified in preliminary research on the determinants of the 2016 election outcome: authoritarian predispositions, sexism, and racial resentment (Confessore and Cohn Reference Confessore and Cohn2016; MacWilliams 2016a; Reference MacWilliams2016b; Schaffner, MacWilliams, and Nteta Reference Schaffner, MacWilliams and Nteta2017).
We examined and compared these competing accounts using data from the 2016 American National Election Study (ANES).Footnote 1 The ANES was administered in pre- and post-election waves, using a combination of nationally representative face-to-face and web samples.Footnote 2 Our analytical methods considered the survey’s complex design, and appropriate post-election sampling weights were used to calculate all of our estimates, unless otherwise noted (ANES 2016).
KEY VARIABLES
Our dichotomous dependent variable distinguishes people who said they voted for Trump from those voting for other candidates. Our independent variables fall into three categories: partisanship, demographics, and attitudes. We constructed dichotomous (1 = yes; 0 = no) indicators for each of the partisan and demographic indicators. The indicators include Republican, white, married, working-class, no (four-year) college degree, evangelical Protestant, Southern, aged 30 to 44, aged 45 and older (i.e., under 30 is the reference category), and male (overall model only).Footnote 3
We also analyzed three attitudes: authoritarianism, racial resentment, and sexism. To measure a voter’s authoritarian disposition, we followed the lead of other researchers, operationalizing the concept with a four-item battery focusing on child-rearing values (Feldman and Stenner Reference Feldman and Stenner1997; Hetherington and Suhay Reference Hetherington and Suhay2011; Hetherington and Weiler Reference Hetherington and Weiler2009).Footnote 4 Scholars previously identified this measure as an exceptionally strong predictor of supporting Trump (MacWilliams 2016). For each question, respondents indicated which of two traits is more important for children to possess: “respect for elders” or “independence,” “obedience” or “self-reliance,” “curiosity” or “good manners,” and “being considerate” or “being well-behaved.” Collectively, these items captured well authoritarians’ predilection for obedience, control, and conformity (α = 0.64). Replicating the methodology of other scholars (Hetherington and Suhay Reference Hetherington and Suhay2011; Hetherington and Weiler Reference Hetherington and Weiler2009), we coded the authoritarian response for each item as a 1 and the other as a zero. In the small minority of cases in which a respondent volunteered that both responses were equally important, the variable was coded 0.5. Each respondent’s authoritarianism-index score equals the mean response value across the items.
Racial attitudes were measured using Kinder and Sanders’s (1996) racial resentment scale (Filindra and Kaplan Reference Filindra and Kaplan2016; Knuckey and Kim Reference Knuckey and Kim2015; Tesler Reference Tesler2012). It is based on respondents’ level of agreement with four items: (1) “Irish, Italian, Jewish, and many other minorities overcame prejudice and worked their way up. Blacks should do the same without any special favors”; (2) “Generations of slavery and discrimination have created conditions that make it difficult for blacks to work their way out of the lower class”; (3) “It’s really just a matter of some people trying hard enough; if blacks would only try harder, they could be just as well off as whites”; and (4) “Over the past few years, blacks have gotten less than they deserve.” When necessary, we reverse-coded the response categories so that higher values corresponded to greater resentment. The additive index (α = 0.85) was rescaled to range from 0 to 1.
Finally, we constructed a measure of modern sexism from four ANES items frequently used for this purpose (Cassese and Holman Reference Cassese and Holman2016; McThomas and Tesler Reference McThomas and Tesler2016). Similar to racial resentment, this measure aggregates four questions tapping the extent to which individuals acknowledge the need to address ongoing gender inequality. Specifically, the items ask: (1) “How much discrimination is there in the United States today against women?”; (2) “When women demand equality these days, how often are they actually seeking special favors?”; (3) “When women complain about discrimination, how often do they cause more problems than they solve?”; and (4) “How important is it to get more women elected?” The five response options for each question were coded so that higher values corresponded to greater sexism. Combined, the items make a reliable additive index (α = 0.64), which we rescaled to range from 0 to 1.
Summary statistics for all variables are listed in the online appendix. Our sample includes 516 women and 546 men who voted for Trump and 761 women and 588 men who voted for other candidates.
FINDINGS
Bivariate statistics reveal stark differences between Trump voters and individuals who supported other candidates. As expected, Trump voters were much more Republican, white, and evangelical; married women also were modestly more likely to have supported Trump. Other than partisan affiliation, the largest differences appear on the attitudinal measures. Men who voted for Trump had significantly higher mean authoritarianism, racism, and sexism scores than men who voted for other candidates: 0.625, 0.722, and 0.526, respectively, compared to 0.420, 0.375, and 0.322. Similar gaps existed among women: 0.616, 0.735, and 0.492 for Trump supporters, respectively, compared to 0.444, 0.375, and 0.281 for other female voters.
The attitudinal differences between Trump voters and non-Trump voters are markedly larger than the distinctions between male and female Trump supporters. The gender gap among Trump voters on racial resentment, for example, was only 0.013 point, whereas the difference between Trump and non-Trump voters was 0.355. Similarly, female Trump supporters were almost as sexist as his male supporters (i.e., ∆ = 0.034); however, Trump voters held significantly more anti-egalitarian attitudes than non-Trump voters (i.e., ∆ = 0.213). (Details on these bivariate statistics are in the online appendix.)
We used multivariate analysis to further explore how gender, partisanship, and attitudinal factors influenced individuals’ propensity to support Trump. Table 1 is a series of logistic-regression models. To aid in interpretation, figure 1 displays how the predicted probability of voting for Trump changes as a function of belonging to different groups or displaying the highest rather than the lowest levels of authoritarianism, racial resentment, and sexism. The first column in table 1 summarizes results for female voters. Models for males and all voters in the second and third columns provide reference points.
Why Women and Men Voted for Trump
Notes: Cell entries are logistic-regression coefficients, with standard errors in parentheses. The estimates are weighted and adjusted for sample-design effects. * p<0.05; ** p<0.01; *** p<0.001, two-tailed. Pseudo R-square statistics are approximate and were calculated separately with models using weighted data and robust cluster options.
Change in Predicted Probability of Men and Women Voting for Trump as Predictors Shift from Minimum to Maximum Values
Notes: For authoritarianism, racial resentment, and sexism, the bars plot the differences in the probability of voting for Trump for people with the highest versus the lowest level of these attributes. For other variables, the bars represent differences in the predicted probability of voting for Trump when compared to individuals not in the reference group. The lines indicate 95% confidence intervals, which are weighted and adjusted for sample effects.
Unsurprisingly, self-identified Republicans were considerably more likely than other voters to support Trump (i.e., pr. = 0.634 versus pr. = 0.349). Despite the prevailing narrative about Trump’s electoral base composed of older, working-class, not-college-educated men, none of these demographic indicators—including gender—attained statistical significance in the full model. Southern men were only modestly more likely to support Trump, as were married women and evangelical Protestants of both genders. Although race also is significant for both men and women, after controlling for other factors including racial resentment and sexism, the 14-percentage-point difference between white and non-white women’s support for Trump shown in figure 1 is smaller than expected, given the overall race gap in American politics.
Despite the prevailing narrative about Trump’s electoral base composed of older, working-class, not-college-educated men, none of these demographic indicators—including gender—attained statistical significance in the full model.
The data plotted in figure 1 indicate that beliefs linked to the fears on which Trump preyed were as powerful in shaping women’s vote choices as they were for men. Females with the highest levels of racial resentment were more than four times as likely (i.e., 68% versus 16%) to support Trump than those with the lowest levels of racial animosity. Among women, those with the highest levels of sexism were 54 percentage points (i.e., 76% versus 22%) more likely to support Trump than those who expressed no sexist attitudes. It is interesting that after accounting for racial resentment and sexism, authoritarianism was not a significant predictor of men’s or women’s vote choice.
The results clearly show that racial bias and sexism had virtually identical influences on male and female voters. Moreover, there is no evidence to suggest that Trump’s female supporters voted for him primarily because of partisanship rather than prejudice. Controlling for the influence of other factors, possessing the levels of sexism and racism for the typical female Trump voter increased the probability that a woman would vote for him by 37 percentage points, when compared to women with sexism and racism scores typical of a non-Trump female voter. By comparison, being a female Republican increased the probability that a woman voted for Trump by 29 points.
DISCUSSION AND CONCLUSIONS
Our results challenge the popular wisdom that Republican, middle-aged, working-class, not-college-educated men and their loyal wives delivered victory to Donald Trump. Although many of Trump’s female supporters shared these characteristics, it appears that attitudes hostile to gender and racial equality were more decisive motivators of vote choice in 2016. This finding defies the popular pre-election wisdom that men and women would diverge in their reactions to Trump’s campaign messaging and especially his behavior toward women. Most analysts, in fact, assumed that Trump’s inevitable electoral defeat would be largely attributable to his inability to secure women’s votes. It is obvious that those predictions were wrong, and our analyses instead show that the women and men who supported Trump were strikingly similar with respect to the role of prejudice in determining their vote choice.
Trump’s victory provides important insights into the limits of the gender gap for explaining the attitudes and political behavior of American women. Although the gap in the 2016 presidential election was the largest in American history, it was only modestly greater than in 2012. This outcome underscores the fact that women—despite Trump’s anti-feminist rhetoric and actions—were and are a critical element of both of the major parties’ electoral coalitions. For scholars, this finding is an important reminder about the dangers of oversimplifying modern politics to gender differences. Women who supported Trump, for example, were more Republican than those who did not. However, and more important, they held sexist and racially resentful attitudes more similar to males supporting Trump than to their female counterparts supporting other candidates. These attitudes reflect trepidation toward the loss of “traditional American family values,” including the preservation of separate spheres for men and women. They also suggest that many women fear how “outsider” groups may be altering the political landscape, an attitude that observers attribute primarily to angry white men. Researchers, therefore, cannot continue to treat women as comprising a monolithic liberal voting bloc; neither can we differentiate between only Republican and Democratic women. Instead, we must more deeply consider the attitudes, beliefs, and values that motivate all voters’ choices on Election Day.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S1049096518000355
