Playing the Woman Card: Ambivalent Sexism in the 2016 U.S. Presidential Race

Abstract

Late in the 2016 U.S. Presidential primary, Donald Trump attacked Hillary Clinton for playing the “woman’s card.” Theories of system justification suggest that attitudes about gender, particularly endorsement of hostile and benevolent sexism, likely shaped reactions to this campaign attack. Using a set of two studies, we find that hostile sexists exposed to the attack showed increased support for Trump and decreased support for Clinton. Benevolent sexists, however, reacted to Trump’s statements with increased support for Clinton, consistent with benevolent sexism’s focus on protecting women (Study 1). We further found that the woman card attack produced distinct emotional reactions among those with low and high levels of hostile and benevolent sexism. The attack also increased political participation among hostile sexists (Study 2). Our results offer new insights into the role of sexism in the 2016 presidential contest and further the discipline’s understanding of the gendered dimension of negative campaigning.

How does sexism affect women’s paths to political office? In this article, we use the 2016 U.S. Presidential Election as a case study to evaluate the system justifying nature of hostile and benevolent sexism in the context of overt sexist appeals in a political campaign. A robust literature on gender and political leadership demonstrates the complex and multifaceted nature of voter reactions to gender dynamics in political campaigns. While it is clear that women running for office behave in strategic ways to increase or decrease the emphasis on their gender, we know little about the effects of explicit attacks on a candidate’s gender on voters’ perceptions of candidate and their vote choice. The 2016 U.S. Presidential race offered numerous opportunities to study gender‐based campaign attacks. Here, we focus on Donald Trump’s oft‐repeated critique of Clinton’s use of gender in campaign messaging, famously captured in the statement: “the only card she has is the women’s card” (Rappeport, 2016). Trump’s “woman card” attack and related rhetoric from his campaign11 Other examples include: “I watched her the other day and all she would talk about was, ‘Women! Women! I’m a woman! I’m going to be the youngest woman in the White House!’” (Newton‐Small, 2015), and Clinton “has got nothing else going. Frankly, if Hillary Clinton were a man, I don't think she would get 5% of the vote” (Luhby & Collinson, 2013).
presents an opportunity to better understand how attitudes about gender and power can be activated by the political environment to shape perceptions of candidates (Bauer, 2015; Krupnikov & Bauer, 2014) and mobilize voters (Banaszak & Ondercin, 2016; Ondercin & Bernstein, 2007) when women run for office.

We argue that beliefs about gender hold the key to understanding a broad set of reactions to attacks such as this one. Using system justification theory, which contends that individuals hold beliefs or “legitimizing ideologies” that relate to preferences for or against the status quo (Jost, Banaji, & Nosek, 2004), we theorize that gender‐based campaign attacks resonated differently among voters based on attitudes about the social position and power women hold (Deckman & McTague, 2015; Diekman & Eagly, 2000), which is measured via levels of hostile and benevolent sexism (Jost & Kay, 2005). These two forms of sexism reflect distinct patterns of thinking about gender. Hostile sexism encompasses a set of beliefs focusing on threats to men’s power over women, whereas benevolent sexism involves endorsement of positive and negative gender stereotypes (Barnes & Beaulieu, 2014; Glick & Fiske, 2001). Both forms of sexism are stable attitudes that can be situationally triggered by campaign messaging. Thus, we expect where voters fall on the hostile and benevolent sexism continua, above and beyond their party and gender identifications, will determine their reactions to gender‐based campaign attacks.

In evaluating these relationships, we rely on two unique studies of the public’s reactions to the woman‐card attack; one conducted the week following Trump’s “women’s card” comments and another conducted several months later.22 The second study was conducted prior to both the emergence of the Access Hollywood tape of Trump and accusations of Trump’s unwanted sexual advances towards several women.
In our first study, we find evidence of system justification processes at work in the 2016 U.S. Presidential race. Individuals with high levels of hostile sexism who viewed the attack evaluated Trump more favorably and Clinton less favorably and expressed lower willingness to vote for Clinton. By contrast, benevolent sexists exposed to the attack evaluated both candidates more favorably and were more inclined to vote for Clinton, a finding we attribute to benevolent sexists’ protectionist attitudes toward women.

In our second study, we test whether the attack produces unique emotional responses based on levels of hostile and benevolent sexism. We hypothesize and find that exposure to the woman‐card attack elicits enthusiasm about the campaign among individuals high in hostile sexism and anger among individuals low in hostile sexism. At the same time, the attack increases anxiety among those scoring high on benevolent sexism. Controlling for emotional reactions to the woman‐card attack, we find that it mobilizes hostile sexists while demobilizing people low in hostile sexism. By contrast, benevolent sexism does not moderate the effect of the attack on intentions to participate in the campaign. In both studies, the effects of hostile and benevolent sexism cut across gender and party lines, pointing to the power of gendered campaign attacks to broadly shape attitudes toward candidates and the importance of understanding the activation of competing identities and attitudes among voters (Klar, 2013).

Our findings engage with a larger debate in the campaigns and elections literature as to whether negative campaigning mobilizes voters (Mattes & Redlawsk, 2015) and also to a robust body of scholarship that finds that gender plays a role in political campaigns in a complex, contextually dependent manner in both the United States (Bauer, 2015, 2017 ; Cassese & Holman, 2017; Holman, Merolla, & Zechmeister, 2016; Mo, 2014) and abroad (i.e., Johnson, 2015; Tobar, ; Williams, 2018). The results also shed new light on the scope of sexism, its system‐justifying functions and political consequences, as well as the effects of using gender as the focus of campaign attacks. While gender stereotypes are not inevitably at play when women run for office (Dolan, 2014), the use of overtly gender‐based attacks like Trump’s woman‐card attack heightens the salience of gender and activates hostile and benevolent sexism, with significant implications for how voters evaluate candidates and engage with political campaigns.

System Justification Theory and Sexism

Individuals are motivated to view the world in ways that make prevailing institutions and social systems seem fair and good (Jost & Banaji, 2015). System justification theory argues that individuals hold beliefs or “legitimizing ideologies” that relate to preferences for maintaining and protecting the status quo and that these beliefs structure political thinking and behavior (Jost et al., 2004). System justification beliefs are widespread and held both by those advantaged and disadvantaged by the system; both types of groups will often work to protect it, even when the system contains injustices or inequalities (Jost & Banaji, 2015). Indeed, people can use system justification to rationalize their own group’s disadvantaged place in society; as a result, “a group’s disadvantaged status reinforces prejudice” (Glick & Fiske, 2001, p. 110).

Views about gender represent a key set of system‐justifying beliefs that underlie support for traditional gender roles in society and politics (Glick & Fiske, 2001; Jost & Kay, 2005). Hostile and benevolent sexism, two related (but distinct) attitudes that form the ambivalent sexism inventory (Glick & Fiske, 1996), represent a key opportunity to evaluate reactions to Trump’s gender‐based campaign attacks. Hostile sexists see the relationship between men and women as a power play and a zero‐sum game, whereby if women gain power, it is at men’s expense. Hostile sexists are explicitly antagonistic toward women, who they see as seeking control over men (Glick & Fiske, 2001). They are generally less likely to support women in positions of power and will engage in direct actions to elevate men over women, such as hiring a less qualified man over a more qualified woman (Christopher & Mull, 2006). Hostile sexists are also more likely to condone aggression and violence toward women, endorse rape myths (Begany & Milburn, 2002), and show a proclivity to engage in acquaintance rape and sexual aggression (Masser, Viki, & Power, 2006). New research also suggests that hostile sexists were more supportive of Trump in the 2016 U.S. election (Bäck, Carroll, Hansen, & Bäck, 2017; Valentino, Wayne, & Oceno, 2018).

Benevolent sexists, by contrast, view “women as wonderful but fragile creatures who ought to be protected and provided for by men” (Glick et al., 2004, p. 715). Benevolent sexism is rooted in the belief that women should be cared for by men and that biological differences between men and women give women advantages in domestic duties (Swim, Aikin, Hall, & Hunter, 1995). It often involves endorsement of ostensibly more positive stereotypes about women relative to hostile sexism, but these stereotypes reinforce women’s subordinate position relative to men and maintain gender inequality. Benevolent sexism persists because these attitudes are widely held by both men and women (Glick & Fiske, 2001).

Given the distinctiveness of these two forms of sexism, we expect they will produce differing reactions to the woman‐card attack. Trump’s attacks contain elements of hostile sexism, including focusing on the advantages that Clinton claims by virtue of her gender. For instance, Trump said that men are “petrified to speak to women anymore” and that “women get it better than we do” (Khalid, 2016). Trump’s attacks may legitimize the attitudes held by hostile sexists and may be particularly powerful in attracting support among them. As such, we posit:

H1: Hostile sexists who are exposed to the woman‐card attack will hold more favorable evaluations of Trump and more unfavorable evaluations of Clinton, resulting in a reduced willingness to vote for Clinton and an increased willingness to vote for Trump.

How will benevolent sexists react to the woman‐card attack? On one hand, Trump’s treatment Clinton violates benevolent sexist norms, which are akin to protective paternalism. As such, benevolent sexists may view Trump more negatively after exposure to the attack because it violates a norm of civility toward women. On the other hand, Clinton is a woman seeking a position of power in the public sphere, which violates gender norms linked to benevolent sexism. In this way, benevolent sexists might be reluctant to support Clinton in her efforts to undermine traditional gender roles (Rudman, Moss‐Racusin, Phelan, & Nauts, 2012). Research shows that benevolent sexists’ protectionism toward women has limits; it often disappears when women are seen as “disobedient” or deviate from gender norms (Glick, Sakalli‐Ugurlu, Ferreira, & Souza, 2002). This leads us to our next hypothesis:

H2: Exposure to the attack will decrease evaluations of both Trump and Clinton among benevolent sexists; as a result, the likelihood of voting for Clinton or Trump will not change.

We test Hypotheses 1 and 2 in the first of our two studies.

System Justification, Emotion, and Political Mobilization

We theorize that the “woman card” and similar campaign attacks may influence political attitudes, and also elicit emotional and behavioral responses. System‐justifying beliefs like sexism have been positively linked to system‐enhancing behaviors (Jost & Hunyady, 2005) and negatively associated with system‐challenging behaviors like political protest (Jost et al., 2012; Osborne & Sibley, 2013). This suggests that those advantaged by a system engage to protect their advantage when given an opportunity. Trump’s candidacy presented an opportunity for hostile sexists to bolster a system characterized by male advantage.

Emotional reactions to campaign attacks like this one may help explain whether they mobilize voters. The link between sexism and emotion has yet to be explored in the context of political campaigns, but it seems particularly relevant in the context of gendered campaign attacks. We know that emotions run high in political campaigns, with consequences for voters’ learning and behavior (Petersen, 2010; Valentino, Brader, Groenendyk, Gregorowicz, & Hutchings, ). Distinct emotions are associated with hostile and benevolent sexism in nonpolitical contexts. For example, hostile sexists react with anger when their masculinity is threatened (Dahl, Vescio, & Weaver, 2015), and benevolent sexists react with anxiety when confronted with the objectification of women (Krawczyk, 2013). Research has also evaluated reactions to observing the expression of sexism directed against others; for example, scholars have found that witnessing hostile sexism can elicit anger (Barreto & Ellemers, 2005). However, it remains unclear whether endorsement of hostile and benevolent sexism interact with political events—such as campaign attacks—to produce specific emotional and behavioral reactions.

Trump’s attack on Clinton’s use of the woman card were likely perceived as threatening to some voters and affirming for others. For those endorsing hostile sexism, Trump’s comments are affirming (e.g., women are demanding special consideration). Trump’s woman‐card attack criticizes Clinton for opportunistically capitalizing on her gender: Trump said on CNN “She is a woman, she is playing the woman card left and right … Frankly, if she didn’t, she would do very poorly. If she were a man and she was the way she is, she would get virtually no votes.” This sentiment speaks directly to hostile sexist beliefs, for example, the idea that women seek special favors, such as hiring policies that favor them over men, under the guise of asking for equal treatment.

Our expectations for benevolent sexists are less straightforward. As we articulate in Hypothesis 2, Clinton’s candidacy violates benevolent sexists’ ideas about women’s appropriate role in society. At the same time, Trump’s attack on Clinton is a violation of expected behavior for men, who are supposed to protect women. We expect that this conflict will produce anxiety among those endorsing benevolent sexism. People low in hostile and benevolent sexism—those holding more progressive views on gender—will likely also have distinctive reactions to the attack, responding to the attack with less enthusiasm and more anger. Taken together, we expect that the attack will produce the following emotional reactions:

H3: Hostile sexists will react to the attack with greater enthusiasm and less anger and anxiety relative to those low in hostile sexism.

H4: Those low on hostile and benevolent sexism will react with less enthusiasm and more anger.

H5: Benevolent sexists will react to the attack with higher levels of anxiety.

Understanding this heterogeneity in response to the woman‐card attack is an important first step in uncovering its implications for political behavior. Discrete negative emotions have distinctive consequences for preferences and participation (Huddy, Feldman, & Cassese, 2007; Petersen, 2010). While anger and enthusiasm have different valences (one is negative, and one is positive), both are associated with increased approach or action tendencies, which increases political participation (e.g., Carver & Harmon‐Jones, 2009). Anxiety or fear, by contrast, is associated with an avoidance behavioral tendency and can reduce participation (Valentino et al., ). Because people may experience distinct emotional reactions to the attack, we expect that:

H6: Benevolent sexists exposed to the attack will participate less, while exposure to the attack will accelerate participation among those with high and low levels hostile sexism.

We test Hypotheses 3–6 with Study 2.

STUDY 1

Methods, Participants, and Study Design

The Study 1 sample (n = 957) was recruited on Amazon’s Mechanical Turk (MTurk) in early May of 2016—less than a week after Trump’s “woman card” attack. MTurk is an online platform on which workers completed our survey in exchange for a small payment. MTurk samples typically overrepresent Caucasians, men, liberals, and younger adults compared to random samples conducted by telephone (Paolacci, Chandler, & Ipeirotis, 2010), but they are more diverse than convenience samples like college students (Berinsky, Huber, & Lenz, 2012). This was true for our sample, which is described in the first column of Table 1. As a brief validation of the responses provided by participants in our sample, we compared agreement with the question “Do you think that Hillary Clinton gets more of an advantage in the election because she’s a woman?” in our control conditions to a representative sample obtained via telephone interview by The Washington Post less than a month after our study (Ross, 2016). Overall, 32% of those in the representative sample agreed, as compared to 31% of our control group (see Table S1 in the online supporting information for demographic comparisons). Our distribution of hostile and benevolent sexists also closely mirrors other research on sexism in the 2017 election (Bäck et al., 2017; Valentino et al., 2018). Thus, although our data source is not a replacement for a probability sample, we can be reasonably confident that our sample resembles a nationally representative sample on gendered attitudes.

Table 1. Sample Characteristics

	Study 1	Study 2
N	957	409
Gender
Male	56.8%	50.5%
Female	43.2%	49.5%
Age	Mean: 36.1	Mean: 37.2
Income	Median range: $25,001 to $50,000	Median range: $25,001 to $50,000
Education
High School	12%	11%
Some College	40%	42%
College Degree	36%	38%
Graduate Degree	12%	10%
Race
Caucasian (Baseline)	78%	74%
Black	7%	8%
Hispanic	5%	8%
Asian	9%	9%
Other	1%	2%
Party
Democrat (Baseline)	56%	58%
Independent	17%	17%
Republican	27%	25%
News Consumption
Daily	38%	33%
Several times a week	38%	39%
About once a week	14%	17%
A few times a month	5%	8%
Rarely	5%	3%
Never	1%	1%
Registered to Vote
Yes	91%	91%
No	9%	9%

Note.

• Values may not sum to 100 due to rounding. All data are self‐reported by study participants.

In the survey, we first captured whether each survey respondent was exposed to Trump’s woman‐card attack by asking: “Recently in the Presidential Primary Campaign, Donald Trump accused Hillary Clinton of ‘playing the woman’s card’ insisting that she would get fewer votes if she were a man. Have you heard anything about this?” Survey participants indicating “yes” were coded as having been exposed to the attack, and participants responding “no” were coded as not having been exposed. About three‐quarters (76%) of survey participants were exposed to the attack (n = 725), and a quarter (24%) were not (n = 231), demonstrating a high level of awareness of the attack in the general public. Participants then completed the following measures.

Candidate Evaluations and Vote Choice.

We measure candidate evaluations using a traditional feeling‐thermometer rating of both Hillary Clinton and Donald Trump. For vote choice, participants also answered: “If the 2016 presidential election were being held today and the candidates were Hillary Clinton and Donald Trump, who would you vote for?” Responses were coded 1 if respondents chose Clinton and zero otherwise.

Sexism

Hostile and benevolent sexism were measured using an abbreviated version of the Ambivalent Sexism Inventory (Glick & Fiske, 2001). Survey participants indicated their level of agreement with three statements on a 7‐point Likert scale ranging from strongly disagree to strongly agree, which were combined to form a standardized scale. Hostile sexism included: (1) Most women fail to appreciate all that men do for them; (2) Women seek to gain power by getting control over men; and (3) Most women interpret innocent remarks or acts as being sexist (α = .88). Benevolent sexism included: (1) Women should be cherished and protected by men; (2) Many women have a quality of purity that few men possess; and (3) A good woman ought to be set on a pedestal by her man (α = .81).33 A shortened version of ASI has been shown to have the same psychometric properties as the original ASI (Rollero, Glick, & Tartaglia, 2010) and has been used in political science research (Barnes, Beaulieu, & Saxton, 2018).

Control Variables

See Table 1 for descriptive and coding information.

Results

Gender and Partisan Differences in Sexist Beliefs

Beliefs about gender, particularly benevolent sexism, cut across gender and partisanship. To evaluate the distribution of sexist beliefs across survey respondents, we plotted the mean values of benevolent and hostile sexism by respondent gender and party using data from Study 1 (and Study 2 to demonstrate the consistency of differences) in Figure 1. While women consistently expressed lower levels of hostile sexism, differences based on party are far more pronounced than differences based on gender. Among Republicans, we see that women express high levels of hostile and benevolent sexism, consistent with our system justification framework. To isolate the effects of hostile and benevolent sexism on political attitudes and behavior, we control for party and gender when testing each of our hypotheses below.

Figure 1

Hostile and benevolent sexism by gender and party. Gender and party identification self‐reported by survey respondents. Study 1: Observational data collected via MTurk in the week following Trump’s comments; includes those exposed or not exposed to Trump’s comments. Abbreviated ASI Study. Study 2: Experimental data—includes responses aggregated across the conditions. Complete ASI.

Modeling the Effects of Exposure to the Woman‐Card Attack

To evaluate whether the woman‐card attack activated hostile and benevolent sexism, bringing them to bear on decision making, we estimated a series of models predicting candidate evaluations and vote choice as a function of hostile and benevolent sexism, among other key covariates. In Hypothesis 1, we speculated that hostile sexism would boost support for Trump while depressing support for Clinton among those exposed to the attack. In Hypothesis 2, we posited that benevolent sexists exposed to the attack will have lower evaluations of Trump. At the same time, we expected that Clinton might be punished by benevolent sexists for her norm‐deviating behavior; that is, a woman seeking a powerful position traditionally held by men, such that benevolent sexism would also be associated with lower evaluations of Clinton among those exposed to the woman‐card attack.

Separate models are provided for respondents based on exposure to the woman‐card attack for comparison purposes, and standardized coefficients are included to further facilitate these comparisons (see Table 2). To ensure that respondents who were exposed to the campaign attack did not differ significantly on any of our covariates from those not exposed, which would introduce selection bias, we estimated a logit model predicting exposure to the attack as a function of sexism, and political and sociodemographic controls (see Table S2 in the online supporting information). We find that only voter registration and reported frequency of news consumption predicted exposure to the attack. We control for these factors in the subsequent analysis.44 As a robustness check, we reestimated the models using a Heckman selection approach. The results are consistent for both modeling strategies, though the diagnostics suggest a selection approach is not necessary (see Table S3 in the online supporting information). These robustness checks suggest that our approach is appropriate give that our data in Study 1 is observational rather than experimental (see Study 2 for experimental results).

Table 2. Effect of the Woman‐Card Attack on Candidate Evauations (Study 1)

Exposed to “Woman Card” Attack?	Clinton thermometer		Trump thermometer		Clinton vote
	No	Yes	No	Yes	No	Yes
Hostile sexism	‐0.08	‐0.13^***	0.17^*	0.21^***	‐0.22	‐0.45^***
	(0.06)	(0.04)	(0.07)	(0.04)	(0.19)	(0.11)
Benevolent sexism	0.04	0.07^*	0.01	0.08^*	0.13	0.23^*
	(0.06)	(0.04)	(0.07)	(0.03)	(0.19)	(0.10)
Female	0.02	0.04	‐0.14^*	‐0.06^*	0.18	0.17^+
	(0.06)	(0.03)	(0.06)	(0.03)	(0.19)	(0.10)
Age	0.00	0.07^*	‐0.02	0.06^+	0.09	0.84
	(0.06)	(0.03)	(0.07)	(0.03)	(0.20)	(1.07)
Independent	‐0.35^***	‐0.31^***	0.17^*	0.23^***	‐1.38^***	‐0.79^***
	(0.06)	(0.03)	(0.07)	(0.04)	(0.23)	(0.10)
Republican	‐0.49^***	‐0.40^***	0.49^***	0.52^***	‐1.45^***	‐1.29^***
	(0.06)	(0.04)	(0.07)	(0.04)	(0.28)	(0.13)
Income	‐0.06	0.03	‐0.12^*	‐0.03	0.11	0.08
	(0.06)	(0.03)	(0.06)	(0.03)	(0.19)	(0.11)
Education	0.22^***	0.07^*	0.01	0.01	0.78^***	0.29^**
	(0.06)	(0.04)	(0.07)	(0.03)	(0.21)	(0.11)
Black	0.09	0.09^**	0.02	0.00	‐0.08	0.29^*
	(0.06)	(0.03)	(0.07)	(0.02)	(0.21)	(0.12)
Hispanic	0.15^***	0.02	‐0.04	0.01	0.00	0.00
	(0.05)	(0.04)	(0.05)	(0.02)	(0.00)	(0.00)
Other race	‐0.02	0.13^***	0.01	0.01	‐0.36^+	0.23^*
	(0.06)	(0.03)	(0.05)	(0.03)	(0.22)	(0.10)
News consumption	‐0.04	‐0.07^*	0.02	0.00	0.04	0.10
	(0.06)	(0.03)	(0.07)	(0.03)	(0.20)	(0.11)
Registered voter	‐0.04	0.03	0.04	0.00	0.28	0.09
	(0.06)	(0.03)	(0.06)	(0.03)	(0.23)	(0.09)
Constant	0.00	0.00	0.00	0.00	‐0.78^***	2.28
	(0.05)	(0.03)	(0.06)	(0.03)	(0.20)	(3.38)
Adj/Pseudo R^2	.37	.30	.33	.44	.36	.31
N	222	708	222	704	215	708

Note.

• Entries are for the thermometer models are OLS regression coefficients with robust standard errors in parentheses. Entries for the vote choice models are logit coefficients with robust standard errors in parentheses. All coefficients are standardized to facilitate effect size comparisons. Significance tests are two‐tailed: ⁺ p < .10, ^* p < .05, ^** p < .01, ^*** p < .001.

Hostile Sexism and Reactions to the Woman‐Card Attack

The results provided in Table 2 show that hostile sexism shapes candidate evaluations and vote choice among those exposed to the woman‐card attack, consistent with Hypothesis 1. Looking first at Clinton evaluations, one can see that hostile sexism has no effect among survey participants who were not exposed to the attack. Among those exposed to the attack, however, hostile sexism is associated with a significant decline in Clinton evaluations. Whether exposed to the woman‐card attack or not, hostile sexists evaluate Trump more favorably. However, the effect is substantively larger among those exposed to the woman‐card attack. Hostile sexism also shapes vote choice by significantly reducing the probability of voting for Clinton, but only among those exposed to the woman‐card attack. These results support our contention that the woman‐card attack activates hostile sexism, bringing it to bear on candidate evaluations and vote choice. In the absence of the attack, hostile sexism does not influence Clinton evaluations or the likelihood of voting for her.

Benevolent Sexism and Reactions to the Woman‐Card Attack

Our results contradict Hypothesis 2, in that exposure to the woman‐card attack results in more favorable evaluations of both candidates rather than lower evaluations as expected. Benevolent sexism was unrelated to vote choice among those not exposed to the attack but significantly increased the likelihood of voting for Clinton among those exposed to it. These results may suggest that Trump’s norm‐deviating behavior was seen as worse than Clinton’s behavior, because the net effect of the attack was an increase in the likelihood of voting for Clinton. These findings fit with extant research, which demonstrates that benevolent sexism may lead to the belief that women in political office are more honest and less corrupt (Barnes & Beaulieu, 2014; Barnes, Beaulieu, & Saxton, in press). In this case, it appears that benevolent sexists extend the benefit of the doubt to Clinton when attacked by Trump. Judging from the standardized coefficients, the effect of benevolent sexism on candidate evaluations is comparatively weaker than the effects of hostile sexism, suggesting that while this attack activated both forms of sexism, its effects on hostile sexism were likely more consequential in terms of electoral outcomes.

It is worth noting that despite the explicitly gendered nature of the attack, respondent gender plays only a modest role in shaping candidate evaluations and vote choice. Women’s thermometer ratings of Clinton do not differ significantly from those of men. Women do evaluate Trump more negatively than men, though the effect of gender is smaller than the effects of hostile and benevolent sexism among women exposed to the woman‐card attack. Women are more likely to vote for Clinton (the effect is marginally significant), though again the effects of hostile and benevolent sexism are larger than respondent gender among those who saw the woman‐card attack.55 We explored the possibility that a subgroup of women was driving these results by evaluating the interaction between respondent gender and a series of sociodemographic factors including age, race, and education, but we did not find any evidence to support this idea.
This result is consistent with extant research arguing that competing identity and value primes can cancel out the effects of an identity (Klar, 2013). Regardless, the results illustrate that beliefs about women—captured by hostile and benevolent sexism and not just gender alone—significantly shaped candidate evaluations.

Substantive Effects of Hostile and Benevolent Sexism

To illustrate the divergent effects of hostile and benevolent sexism on candidate evaluations and vote choice, we plotted predicted values from the Table 2 models among those who reported exposure to the woman‐card attack. The results are presented in Figure 2. The top row shows predicted Clinton evaluations at minimum, mean, and maximum levels of hostile sexism (left) and benevolent sexism (right) with 95% confidence intervals. As hostile sexism increases, evaluations of Clinton steadily decrease. The reverse is true for benevolent sexism. The relationships are more similar across sexism types for Trump evaluations (middle row, Figure 2) with a positive relationship observed for both, though a stronger effect is present for hostile sexism. The relationships between hostile sexism, benevolent sexism, and vote choice (bottom row) exhibit patterns similar to the Clinton evaluations, with hostile sexism decreasing the likelihood of voting for Clinton, while benevolent sexism increases the likelihood of voting for Clinton.

Figure 2

Predicted candidate evaluations and vote choice among those exposed to the “woman card” attack (Study 1). Entries are predicted Clinton Thermometer ratings (top row), predicted Trump Thermometer ratings (middle row), and the predicted probability of voting for Clinton (bottom row) with 95% confidence intervals. Predicted values were calculated at the minimum, mean, and maximum values of hostile and benevolent sexism using the model specifications presented in Table 2, holding all other independent variables to their mean values.

STUDY 2

Methods, Participants, and Study Design

We conducted Study 2 several months later to further evaluate whether the “woman card” attack’s activation of hostile and benevolent sexism shapes emotional reactions to the campaign and political participation. We employed an experimental design that randomly assigned survey participants to read an article about the woman‐card attack from The New York Times or a control article about the use of social media in the campaign, rather than relying on participant recall of the attack.66 The Times article was attributed to the Associated Press (AP) to avoid partisan attributions to the source. The control article’s content was from The Huffington Post and was edited for length and content to more closely mimic the style of the treatment Times article. Complete study materials are available in the online supporting information.
Prior to reading the article, participants completed the full Ambivalent Sexism Inventory. After reading the article, participants reported their emotional reactions to the article they read and intentions to participate in electoral politics. We recruited the Study 2 sample, (n = 409) in the same fashion as Study 1, on September 16–17, 2016. Table 1 provides sample characteristics.77 Given that there are only two conditions, power calculations (obtained via GPower) demonstrate a need for an N of 107 per condition. Our sample exceeds this requirement.

Emotions. We captured emotional reactions to the woman‐card attack by asking participants to “indicate to what extent you are feeling each emotion in reaction to the article you just read.” Participants rated their experience of eight different emotions on a scale ranging from 1 (not at all) to 5 (extremely). We combined and rescaled the emotion measures to form standardized variables. The anger scale incorporates ratings of the terms “angry,” “hostile,” and “disgusted” (α = .90). Anxiety includes ratings of the emotion terms “anxious” and “afraid” (α = .85), and enthusiasm was measured using responses to the terms “enthusiastic,” “proud,” and “hopeful” (α = .89).

Political Participation is measured using respondents’ likelihood of engaging in the following acts of political participation: attending campaign events, donating money to a political campaign, volunteering for a political campaign (i.e., going door to door, making phone calls), attending other political events, discussing the election with others in person or online (e.g., e‐mail, Facebook, etc.), and voting in the election. Responses were provided on a Likert scale ranging from 1 (definitely won’t) to 5 (definitely will). As is common in extant scholarship (Farris & Holman, 2014), items were combined to form a standard scale (α = .77).

Exposure to woman‐card attack is captured by dummy variable coded 1 if the respondent was assigned to the treatment condition and read the gender‐card attack article and coded 0 if the respondent was in the control condition and read the social‐media article.

Hostile and benevolent sexism were measured using the full Ambivalent Sexism Inventory (Glick & Fiske, 2001), which we asked prior to the treatment to ensure exogeneity. The items were combined to form two reliable subscales (α_benevolent = .89; α_hostile = .92). Factor analysis confirmed the two scales are distinct and moderately correlated (r = .43).

Control Variables

Control variables for Study 2 were measured using the same approach discussed for Study 1; see Table 1.

Manipulation Checks

To determine whether respondents responded in the expected ways to the treatment condition, they were asked to identify the tone of the article they read on a 5‐point Likert scale ranging from 1 (very negative) to 5 (very positive). Participants rated the mean tone of the treatment condition (M = 1.9) significantly lower than the control (M = 3.4), t(386) = 19.7, p < .001.

Emotion Checks

Asking emotions‐check questions following the treatment is a traditional means for assessing the emotional reaction to the treatment (Huddy et al., 2007; Valentino et al., ) and allows us to assess both the effects of the treatment on emotions and the effect of these elicited emotions on our outcome variables. We compare discrete emotional reactions to the treatment and condition conditions using t‐tests via responses to the standardized emotion scales described above. Mean anger was significantly greater is the treatment condition (M = .41) compared to the control (M = −.43), p < .001. Enthusiasm was significantly lower in the treatment (M = −.22) compared to the control (M = .16), p < .001, and anxiety was greater in the treatment (M = .12) compared to control condition (M = −.15), p < .01. Collectively, these results indicate that the treatment condition was viewed to be more negative and aroused negative emotional states while depressing positive emotional states.

Results

Emotional Reactions to the Woman‐Card Attack

We hypothesized that the attack would produce specific emotional reactions among those endorsing high and low levels of hostile and benevolent sexism. Specifically, we expected hostile sexists would react to the attack with heightened enthusiasm and reduced anger and anxiety (H3); benevolent sexists would react primarily with anger and anxiety, but reduced enthusiasm (H4), and those low on both of these measures would react with heighted anger and reduced anxiety and enthusiasm (H5). To test these hypotheses, we estimated a series of OLS regression models where the three emotion scales (anger, enthusiasm, and anxiety) are modeled as a function of exposure to the woman‐card attack (the experimental treatment condition), hostile and benevolent sexism, and the interaction between these kinds of sexism and the attack, along with a series of control variables. The results are presented in Table 3.

Table 3. Emotional Responses to the “Woman Card” Attack (Study 2)

	Enthusiasm	Anxiety	Anger	Intent to Participate
Woman‐card attack	‒0.22^***	0.16^**	0.46^***	‒0.04
	(0.05)	(0.05)	(0.04)	(0.06)
Hostile sexism	0.23^**	0.10	0.13^**	‒0.18^*
	(0.08)	(0.07)	(0.05)	(0.08)
Benevolent sexism	0.17^*	0.08	0.09^*	‒0.12
	(0.08)	(0.07)	(0.04)	(0.07)
Attack × hostile	0.00	‒0.08	‒0.28^***	0.17^*
	(0.08)	(0.08)	(0.06)	(0.08)
Attack × benevolent	‒0.06	0.16^+	0.08	0.06
	(0.08)	(0.09)	(0.07)	(0.07)
Anger				0.17^*
				(0.07)
Anxiety				0.01
				(0.06)
Enthusiasm				0.32^***
				(0.05)
Female	‒0.02	0.12^*	0.10^*	0.02
	(0.05)	(0.05)	(0.04)	(0.05)
Age	0.03	‒0.07	‒0.01	0.10^+
	(0.04)	(0.06)	(0.05)	(0.05)
Independent	‒0.07	‒0.11^*	‒0.08	‒0.14^**
	(0.05)	(0.05)	(0.05)	(0.05)
Republican	‒0.06	‒0.21^***	‒0.25^***	0.03
	(0.06)	(0.06)	(0.05)	(0.06)
Income	0.09^+	‒0.05	‒0.06	0.02
	(0.05)	(0.05)	(0.04)	(0.05)
Education	‒0.02	0.00	0.02	0.10^*
	(0.05)	(0.05)	(0.05)	(0.05)
Black	0.00	‒0.09^*	‒0.03	0.03
	(0.05)	(0.04)	(0.04)	(0.06)
Hispanic	0.07	‒0.03	0.07	0.04
	(0.05)	(0.04)	(0.04)	(0.05)
Other race	‒0.05	‒0.11^*	‒0.11^**	‒0.07^+
	(0.05)	(0.04)	(0.04)	(0.04)
Constant	0.00	0.00	0.00	0.00
	(0.05)	(0.05)	(0.04)	(0.05)
R ^2	.16	.12	.33	.18
N	392	392	392	392

Note.

• Entries are standardized OLS regression coefficients with robust standard errors in parentheses. Significance tests are two‐tailed: ⁺ p < .10, ^* p < .05, ^** p < .01, ^*** p < .001.

Contrary to our expectations in Hypothesis 3, hostile sexists did not respond with more enthusiasm to the attack, although they do indicate a higher level of overall enthusiasm about the Presidential campaign regardless of whether they are in the treatment or control condition. Hostile sexists reported more anger in the control condition relative to the woman‐card attack condition, consistent with our expectations, though feelings of anxiety about the campaign were not affected. Benevolent sexism played a role in moderating anxiety in response to the attack as we expected in Hypothesis 4. The coefficient on the woman‐card attack variable gives the effect of the attack at minimum levels of benevolent sexism and shows that it increased reported anxiety by .16 of a standard deviation. As benevolent sexism increases from its minimum to maximum value, the effect of the woman‐card attack increases by an additional .16, such that the effect size doubles among those high in benevolent sexism.88 Respondent gender was significantly related to emotional expression, with women reporting higher levels of anxiety and anger relative to men. However, these gender differences were not conditioned on exposure to the woman‐card attack; we evaluated interactive models to explore this possibility and obtained null results.

Benevolent sexists also reported higher levels of enthusiasm about the campaign regardless of whether they were exposed to the woman‐card attack. This finding runs counter to expectations, but perhaps fits with our observation from Study 1 that benevolent sexists reported more favorable thermometer ratings of both Donald Trump and Hillary Clinton. Based on these results, we find partial support for Hypotheses 3 and 4. Hostile sexists respond to the attack with reduced anger, likely due to the match between the attack and their perspective on Clinton’s campaign. Benevolent sexists responded with heightened anxiety, though this anxiety was somewhat tempered by enthusiasm.

Among survey respondents low on hostile at benevolent sexism, we find strong support for Hypothesis 5. With the sexism × attack interaction terms included in the models, the coefficients for the woman‐card‐attack variable (the first row of coefficients in the table) gives the effect of exposure to the attack among those with minimum levels of both hostile and benevolent sexism. Among this group, exposure to the woman‐card attack significantly decreased enthusiasm, while significantly increasing anxiety and anger. The effect for anger is particularly strong—amounting to about a half of a standard deviation increase in anger in terms of its standardized effect. To illustrate the substantive effects of the woman‐card attacks on reported anger and anxiety, we plotted predicted values from the Table 3 models in Figure 3A. The figure highlights divergent emotional responses to the attack based on one’s levels of hostile and benevolent sexism, highlighting angry responses to the attack among those low in hostile sexism and anxious responses to the attack among those high in benevolent sexism.

Figure 3

Reactions to the “woman card” attack (Study 2). Predicted values were calculated based on the model specifications presented in Table 3, holding all other independent variables to their mean values. Panel A: Entries are predicted anger at minimum and maximum levels of hostile sexism (left) and predicted anxiety at minimum and maximum levels of benevolent sexism (right). Panel B: Entries are predicted intentions to participate in politics at minimum and maximum levels of hostile sexism (left) and predicted anxiety at minimum and maximum levels of benevolent sexism (right).

Behavioral Reactions to the Woman‐Card Attack

We next considered how the woman‐card attack shaped intentions to participate in politics. The results are presented in the fourth column of Table 3. We include emotional reactions as predictors of political participation. Given that we ask the emotion questions posttreatment, we follow Bullock and Ha (2011) and evaluate the effects of emotions in the model with our control variables. At the same time, we recognize the limitations of suggesting a mediating relationship without probes of the mediating process (Green, Ha, & Bullock, 2010). As such, we focus on the direct effect of our treatment and emotions on our dependent variable (Bullock & Ha, 2011).99 See the Figure in the online supporting information for a graphic presentation of the effects of anger, enthusiasm, and anxiety on intentions to participate in politics.

Consistent with Hypothesis 6, we find that the woman‐card attack increased intentions to participate in politics among hostile sexists, as the coefficient on the interaction term is positive and statistically significant (p < .05). However, we find that when controlling for emotional responses to the attack, individuals low in hostile sexism are demobilized. In addition, benevolent sexism does not condition responses to the attack. To illustrate these effects, we graphed participatory intentions by experimental condition and levels of hostile and benevolent sexism in Figure 3B. The results show that when controlling for emotional reactions, the woman‐card attack is mobilizing for people high in hostile sexism and demobilizing for people low in hostile sexism. The same pattern is true for benevolent sexists, although the relationships are weaker than for hostile sexists.

What were the consequences of these patterns of mobilization and demobilization in terms of candidate outcomes? Those low in hostile sexism were much more likely to express an intention to vote for Clinton (60.7%) compared to those high in hostile sexism (31.3%), with much smaller differences in vote choice among people low (47% vote for Clinton) and high (47%) in benevolent sexism (see Table S2 in the online supporting information for additional information). This suggests the consequences of the attack were likely worse for Clinton than any backlash faced by Trump. Our findings regarding mobilization point to the dangerous potential of explicitly gendered attacks for women candidates and further illustrate the importance of hostile sexism in explaining voters’ reactions to this kind of campaign attack.

Conclusions

Our results suggest that system‐justifying beliefs about gender played a key role in shaping candidate evaluations and political behavior in the 2016 U.S. Presidential race. Clinton’s historic candidacy explicitly brought issues of gender forward in the 2016 campaign, and Trump frequently referenced Clinton’s gender in campaign attacks. As such, the 2016 campaign is an ideal setting to investigate system justification processes in campaigns and elections. We found that Trump’s “woman card” campaign attack resonated with hostile sexists, polarizing their evaluations of both candidates and reducing their intention to vote for Clinton. The consequences of evaluative and emotional reactions to the attacks also extended beyond vote choice; exposure to the “woman card” attack mobilized hostile sexists to participate in electoral politics at a higher rate, meaning their preferences may have registered more strongly in the campaign.

Hostile and benevolent sexism are beliefs endorsed by both men and women; this is consistent with broad findings that many women endorse traditional beliefs about gender roles and gendered notions of authority (Cassese & Holman, 2017; Deckman, 2016). Across both studies, party, more so than gender, accounts for endorsement of hostile and benevolent sexism. And with the inclusion of hostile and benevolent sexism in our models, we see few effects of gender. Collectively, our findings suggest that one’s belief about gender roles and the legitimacy of the power differential between men and women rather than one’s gender per se offers the most explanatory power in this context. This finding is useful for interpreting the election returns, as 53% of White women voted for Trump (CAWP, 2016), and also for understanding White women’s historic tendency to support GOP candidates (Deckman, 2016). Given the high rates of hostile sexism among Republican men and women (see Figure 1), Trump’s campaign rhetoric may have kept Republican women in the fold by activating their existing beliefs and preferences surrounding gender relations (Cassese & Barnes, 2017; Strolovitch, Wong, & Proctor, 2017).

Extant research on the use of gender in campaigns has focused on how voters react to the positive use of gender targeting (Herrnson, Lay, & Stokes, 2003; Holman, Schneider, & Pondel, 2015). At the same time, research on the negative use of identity in political ads has examined the effectiveness of implicit racial cues (over explicit cues), given American’s commitment to “norms of equality” (Mendelberg, 2001). Our findings represent a departure from these conclusions, given that we find Trump’s explicit attack on Clinton’s gender to be relatively effective. This result may be due to gender stereotyping being more acceptable than racial stereotyping “perhaps because of biological differences between the sexes and a sexual division of labor that appears to be natural” (Sanbonmatsu, 2002, p. 31). Alternatively, scholars, activists, and pundits may have simply overestimated our country’s commitments to equality. Research in other country contexts have noted the frequency of explicitly gendered attacks and media narratives, even in places that have successfully chosen women as key leaders (Johnson, 2015; Tobar, ; Williams, 2018). Future research might build on what we have presented here to understand the effects of explicit gendered attacks across campaigns and national contexts.

These findings raise several other avenues for future research and exploration, particularly around how candidate gender interacts with sexist beliefs, partisanship, and context. While the woman‐card attack represented a key event in the 2016 election, it was far from the only gendered attack that Trump leveraged against Clinton. These included explicit attacks like calling her a “nasty woman” during a debate and more implicit ones, such as questioning her stamina and ability to lead. Future research might explore how hostile and benevolent sexists respond to implicit versus explicit gendered attacks in political campaigns. And while the methodological approach employed in Study 1 allowed us to evaluate the real‐time effects of how individuals respond to explicit gendered attacks, using a convenience sample also limits the application of our findings. Future research might replicate the media exposure model using a randomized sample to increase confidence in the results.

Clinton was not the only woman running in the 2016 Presidential race; Carly Fiorina ran in the Republican primary. Her attacks on Clinton also focused on gender; for example, an e‐mail to supporters entitled “How Hillary’s Bullying Women” noted “I'm proud to be a woman. But I also know gender is not an accomplishment…. The only way she can win is by playing the gender card.” Future research might evaluate how the characteristics of the attacker plays a role in the success of a gender‐based attack, especially given past research demonstrating that candidate gender shapes reactions to campaign attacks (Krupnikov & Bauer, 2014).

In the two studies presented here, we reach the conclusion that sexism played a role in the 2016 presidential race. And yet, postelection coverage of Trump’s victory often focuses on the role of racial resentment (Ingraham, 2017), anti‐elitism (Lawler, 2017), and education (Silver, 2016) to explain the outcome of the election. These factors certainly mattered, but to consider the election without evaluating the full effect of gender is to paint an incomplete picture. Using the 2016 U.S. presidential election as a case reaffirms the importance of considering how women running for office—and the actions of their opponents—may activate gendered attitudes.

Acknowledgments

Funding for this project was provided by the Newcomb College Institute and the Tulane Council on Research. Thanks to the Gender and Political Psychology Writing Group for their feedback on this work. An early version of this article was presented at the 2016 European Political Science Association meeting in Brussels and the 2017 Resisting Women’s Political Leadership Conference at Rutgers University. Correspondence concerning this article should be addressed to Erin C. Cassese, Department of Political Science and International Relations, University of Delaware, 347 Smith Hall, 18 Amstel Ave., Newark, DE 19716. E‐mail: [e-mail]