Toeing the Party Line: The Asymmetric Influence of Feminism on Partisans' Participation

Abstract What is the relationship between feminism and political participation? How does partisanship moderate this relationship? Prior research shows that gender attitudes, particularly sexism, rather than gender identity per se, increasingly shape vote choice and participation in US elections. However, the role played by feminism in voter behaviour remains scarcely understood. As feminist identification crosses partisanship, we argue that its impact on engagement with campaigns and turnout depends on party ID. Therefore, we expect feminist identity and how it intersects with either aligned or conflicting partisan identity to impact partisans' participation asymmetrically. Using data from the 2016 and 2020 American National Election Studies, our results support these expectations. Holding the mutually reinforcing identities of Democrat and feminist has a significant mobilizing impact, while holding the cross-cutting identities of Republican and feminist tends to lead to a decline in political participation.

The 2016 US presidential election featured the first woman nominee on a major party ticket.Hillary Clinton's loss and lackluster performance among women reinforced a long-standing finding in political science, which is that gender identity, in and of itself, is a relatively small predictor of vote choice (Huddy, Cassese, and Lizotte 2008;Sapiro 2003), and no longer a major predictor of turnout and campaign participation (Burns et al. 2018).Nonetheless, a growing body of research highlights that gender considerations, such as sexist attitudes, have become increasingly salient and consequential factors influencing vote choice, especially since the 2016 race (Schaffner 2022;Utych 2021;Valentino, Wayne, and Oceno 2018).However, considerably less attention has been devoted to the impact of sexism on political participation, with a few exceptions (Banda and Cassese 2022;Kam and Archer 2021).
We build on the literature on cross-cutting identities and gender attitudes to investigate the role of a particular type of gender-related identityfeminismin shaping who decides to engage with campaigns and vote.We expect that feminist identity influences political participation and argue that its impact depends on party ID.Indeed, gender and women's issues have increasingly become polarized along partisan lines and 'Democratic owned' (Wolbrecht 2000).Furthermore, in the last two presidential elections, the major parties significantly differed in their stances on and commitment to gender equality and equity.As feminism is a cross-cutting identity that tends to be reinforcing among Democrats while cross-pressuring Republicans, we argue that its impact on participation will be asymmetrical by party.We test this conditional relationship using the 2016 and 2020 American National Election Studies (ANES).Our results demonstrate that, in both 2016 and 2020, Democratic feminists were significantly more likely to participate relative to co-partisan non-feminists.By contrast, Republican feminists displayed a lower likelihood of participating than Republican non-feminists.In sum, the impact of feminist identity is asymmetric: mobilizing among Democrats but demobilizing among Republicans.
Aligned v. Cross-Cutting Identities, Partisanship, and Mobilization As a result of social sorting, individuals' group identities often correlate highly with their partisanship (Mangum 2013).Racial and ethnic minorities, secular individuals, and liberals increasingly identify as Democrats, while whites, religious individuals, and conservatives increasingly identify as Republicans (Mason 2018).Furthermore, partisans have become more aligned with their party, by adjusting their ideological or policy positions to the ones held by polarized party elites (Levendusky 2009).However, when conflicts between voter partisanship and political beliefs emerge, they can create cross-pressures that influence decision-making (Hillygus and Shields 2008).
On the one hand, holding multiple aligned identities can boost political engagement.Mason (2016) shows that partisans with high levels of social sorting are more likely to experience heightened emotions: anger at the opposing party and enthusiasm toward their own.In particular, anger has been found to increase political mobilization (Valentino et al. 2011).Moreover, citizens whose issue attitudes and partisanship more closely align vote more consistently for their party (Levendusky 2009).
On the other hand, when a person's other identities run in conflict with their partisanship, these cross-pressures likely depress voter participation.Looking at sexual orientation, for instance, McCabe (2017) finds that lesbian, gay, and bisexual (LGB) Republicans vote at considerably lower rates than LGB Democrats and non-LGB Republicans.More generally, Americans with cross-cutting identities are less likely to participate in campaigns and elections than those with more highly sorted and mutually reinforcing identities (Mason 2018).
Overall, identities that align or conflict with one's partisanship impact whether and to what extent individuals are politically mobilized or demobilized.Identities that are consistent with one's partisanship are likely to increase political engagement, whereas identities that crosspressure party ID are likely to reduce engagement.In this paper, we focus on feminist identity, which we argue significantly but asymmetrically influences partisans' likelihood of participating or, in other words, their willingness to help their party through active campaign involvement.This is especially the case in the context of elections where gender considerations are salient, such as in 2016 and 2020.

Feminist Identity and Political Participation
Recent scholarship has called attention to how attitudes about gender, such as modern or hostile sexism, shaped voter decision-making and behaviour in the 2016, 2018, and 2020 elections (Cassese and Barnes 2019;Oceno, Valentino, and Wayne 2023;Schaffner 2022;Schaffner, MacWilliams, and Nteta 2018;Utych 2021).Research further shows that sexism has played a larger role in these elections than in previous ones (Valentino, Wayne, and Oceno 2018).This suggests that gender considerations became more salient during the Trump era: with the first major party's female presidential nominee in 2016, the heightened prominence of #MeToo in 2017, Kavanaugh's confirmation hearings in 2018, the largest number of women in US history running for a presidential nomination in 2020, and the first woman becoming Vice-President in 2021.Importantly, the growing influence of gender in US elections and politics is not about gender identity per se, but rather about how the American public 'feel[s] about issues that affect women', such as sexual harassment and assault, gender discrimination, and gender equality (Sides, Tesler, and Vavreck 2020).Indeed, women across partisanship tend not to see themselves as part of a cohesive group (Klar 2018), and many women lack gender group consciousness (Gurin 1985).
This paper focuses on the role played by a particular type of gender considerationsfeminist identityin influencing political participation in 2016 and 2020.Thus, we address two gaps in the existing literature.On the one side, recent scholarship about gender-related attitudes and identities has primarily examined vote choice, rather than participation, as their dependent variable (Cassese and Barnes 2019;Schaffner, MacWilliams, and Nteta 2018).Conversely, the few research cases addressing the association between gender predispositions and participation have focused on sexism, while neglecting feminism (Banda and Cassese 2022;Kam and Archer 2021).
Feminism is a strong social and political identity whose influence on political values and issue preferences extends beyond gender identity alone (Conover 1988;Oceno 2020).Social psychologists have found feminist ID to be positively associated with believing in the efficacy and appropriateness of collective action and activism in the struggle for women's rights and gender equality (Duncan 2010;Yoder, Tobias, and Snell 2011).However, the political implications of identifying as a feminist remain largely understudied and insufficiently understood, particularly regarding elections.Harbin and Margolis (2022) show that feminist identification is associated with a stronger willingness to recognize racial discrimination among white Americans.As for electoral behaviour, Oceno, Valentino, and Wayne (2023) found that, in 2016, anti-feminist attitudes largely dampened, while feminist identity significantly increased support for Clinton.Our research builds on these findings by examining how the relationship between feminist identity and participation in US elections is moderated by partisan attachments.

Feminism and Partisanship as Aligned or Cross-Cutting Identities
The Democratic and Republican stances on gender-related issues have become increasingly divergent (Wolbrecht 2000).Consequently, along with ideological and social sorting, a partisan gender gap has also developed: a relatively larger proportion of women identify with the Democratic Party, while a relatively larger proportion of men identify with the Republican Party (Gillion, Ladd, and Meredith 2020;Ondercin 2017).But partisans have become even more divided when it comes to gender attitudes rather than simply gender identity (Sides, Tesler, and Vavreck 2020).Indeed, modern or hostile sexism levels are significantly higher among Republicans than Democrats (Banda and Cassese 2022;Kam and Archer 2021).Importantly, sexism differences based on partisanship are much larger than those based on gender identity (Cassese and Holman 2019).Do partisans also differ in terms of feminist identification?
Similar to McCabe's findings (2005), a look at the 2016 and 2020 ANES by feminist, gender, and party ID highlights that there are feminists among women and men and in both major parties.1 Nonetheless, there are clear partisan differences: Democrats are considerably more likely than Republicans to identify as feminist and do so more strongly, although there are nonnegligible subgroups of feminists within both parties, particularly among women.This begs the question of how feminists are differently cross-pressured by their partisan identities, and how those partisan cross-pressures moderate the relationship between feminism and participation, particularly in elections in which gender considerations are salient and prominent, such as in 2016 and 2020.

Hypotheses
Our descriptive analysis supports the expectation that the effect of feminist ID on participation is likely to vary asymmetrically by party.As partisan positions have become increasingly polarized about gender starting in the 1970s (Wolbrecht 2000), issues and goals promoted by feminists have emerged as growing sources of partisan cleavage.This polarization along gender and feminist issues has accelerated since the 2016 presidential campaigns.In both the 2016 and 2020 elections, Democratic and Republican voters received highly distinct cues about their party's stance on gender issues.This is true for both candidates and issue positions.The Democratic presidential and vice-presidential nominees were female in 2016 and 2020, respectively, while the Republican presidential nominee was consistently criticized for his statements and behaviour toward women in both years.The Democratic platforms were also disproportionately more focused than the Republican platforms on 'feminist-minded' policies and goals that explicitly sought to advance gender equity.In other words, similar to Carmines and Stimson's (1986) argument on issue evolution, feminism has progressively become an additional layer of the partisan divide.
Accordingly, we argue that the interaction of feminist identity and partisanship is likely to affect whether campaign involvement and turnout are amplified or dampened.We expect that when feminist and partisan identities align, this will lead to increased participation.Since these two identities are more aligned among Democrats, this alignment will result in a greater likelihood of Democratic participation.By contrast, when these identities are in conflict, this will be politically demobilizing.This is likely to be the case among Republicans, thereby leading to weaker campaign engagement and a lower probability of voting.
Hypothesis 1: Among Democratic voters, feminist identity strength tends to be associated with higher levels of campaign participation and turnout.
Hypothesis 2: Among Republican voters, feminist identity strength tends to be associated with lower levels of campaign participation and turnout.

Data and Methods
To test our hypotheses about how feminist identity, conditional on partisanship, impacts participation, we employ data from two nationally representative surveys: the 2016 and 2020 ANES.The sample includes 3,648 adults in 2016 and 7,453 adults in 2020 who completed both survey waves.
To examine participation, our model includes 7 and 8 dependent variables in 2016 and 2020, respectively. 2Specifically, respondents were asked if they engaged in the following forms of campaign participation: (1) talking to other people and trying to show them why they should vote for or against one of the parties or candidates, (2) attending a rally, (3) attending an online rally (in 2020), (4) wearing a button or displaying a sticker or yard sign, (5) working for a party or candidate, (6) donating to a candidate, (7) donating to a party, and (8) donating to any group that supported or opposed candidates.We combine these items to create an additive scale of campaign participation ranging from 0-1 as well as evaluate each measure separately.Finally, we analyze turnout: we rely on validatedrather than self-reportedvoting measures in both 2016 and 2020. 3 Our main explanatory variable is the interaction between feminist identity strength and partisanship.To measure feminist ID strength in 2016, we combine four items into an additive scale (α = 0.87). 4The first is the traditional 0-100 degree feeling thermometer toward feminists.The other three items were added to the ANES in 2016: (1) 'Do you consider yourself a strong feminist, a feminist, or are you not a feminist?',(2) 'How well does the term "feminist" describe 2 See Appendix A for survey question wording.
3 Drawing on Verba, Schlozman, and Brady (1995), our analyses distinguish between voting and forms of campaign engagement.The decision to become involved in a campaign demonstrates a stronger commitment not only to supporting a candidate, but also to actively devoting additional time, money, and/or effort.you?' and (3) 'How important is it to you to be a feminist?'To measure feminist ID strength in 2020, we employ three of the same questions as in 2016 and combine them into an additive scale (α = 0.68). 5All the analyses also control for ideology, gender, race, ethnicity, age, education, income, religiosity, political interest, modern sexism, and racial resentment. 6

Results
We first examine participation in the context of the 2016 presidential election.In Table 1, we analyze how the interaction of feminist identity strength and partisanship as a 7-point measure impacts each form of participation. 7We employ logistic regression for all binary dependent variables and OLS for the campaign participation scale.As expected, the interaction term coefficient between feminist and party ID is significant, large, and negative across all participation items, except for attending a rally.Figure 1 further shows the marginal effect of feminist ID interacted with partisan -Democratic or Republicandummies. 8The figure reveals a clear pattern: feminist ID is consistently associated with higher levels of participation among Democrats, while it either fails to mobilize or demobilizes Republican voters.9These results provide some support for H1 and H2. 5 The question about how well the term 'feminist' describes respondents was not asked in 2020.

6
These controls were chosen as they are important predictors of feminist ID and are likely to influence the relationship between feminism, partisanship, and participation (see Appendix D).

8
The omitted category is Independent.Next, we turn to the 2020 election.Table 2 includes regression models of each form of participation on the interaction between feminism ID and partisanship as a 7-point variable.
Similar to 2016, this interaction is a powerful, significant, and negative predictor of all campaign participation items as well as turnout.Additionally, Fig. 2 shows the marginal effect of feminist ID by party. 10The asymmetric pattern displayed for 2016 remains consistent and becomes even stronger in 2020, thus providing further clear support for both H1 and H2.It is noteworthy that, in both elections, the impact of feminism by party tends to be large and statistically significant not only for 'cheap' forms of participation, such as talking to others and displaying a button or bumper sticker, but also for 'costly' ones, such as donating money to a campaign (Valentino et al. 2011).
Finally, Fig. 3 portrays the predicted probabilities of both voting and our campaign participation scale by party as feminist ID strengthens in 2016 (left panel) and 2020 (right panel).In all four plots, as respondents identify more strongly with feminism, predicted levels of participation become higher for Democrats, while they tend to decline for Republicans.This provides additional support for our two hypotheses and illustrates how the asymmetric relationship between feminism and participation by party is directly linked to strength of feminist identification, namely the degree to which individuals embrace the feminist label.

Discussion
While gender identity has been found to be a relatively weak predictor of vote choice and political participation, our results demonstrate that feminist identification, conditional on partisanship, is important for explaining campaign involvement and turnout.In the context of the 2016 and 2020 presidential elections, feminist ID was associated with increased participation among Democrats but lower participation among cross-pressured Republicans across a wide range of activities, particularly those related to donation and persuasion.These findings shed light on the cross-cutting nature of feminist identity and its relationship to political behaviour.They emphasize that party elites' competing messages, both explicit and implicit, about their commitment (or antipathy) toward feminist ideals can influence who is mobilized and who is demobilized in the US electorate in countervailing ways.They further suggest that fostering feminist identification may be a particularly fruitful strategy among Democratic elites and candidates.By contrast, it may end up damaging Republican candidates' electoral chances by demobilizing potential Republican voters.
Our analyses focus on two US elections where gender considerations have been particularly salient.A question that future research should address is whether and to what extent gender considerations, such as feminist identification, will remain salient and powerful drivers of turnout and engagement with campaigns or whether they will gradually recede in electoral races following the Trump era.Furthermore, future work may consider the role of feminist identity in different contexts.In particular, the electoral impact of feminist identification can be expected to be substantively different in other countries, especially countries where the women's movement did not go through a period of sorting into one particular party, as in the United States.
Finally, our analyses point to the asymmetric influence of feminist identity, rather than feminist attitudes.Although we account for sexist attitudes and previous work has shown that sexism cross-pressures Democrats, sexist and feminist attitudes are not merely opposites.A low level of sexist beliefs cannot necessarily be interpreted as an endorsement of feminist goals.Future work should, therefore, study the impact of feminist attitudes and to what extent they asymmetrically influence participation and other political behaviours similarly to feminist identity.

Figure 1 .
Figure 1.Marginal Effect of Feminist ID by Partisanship on Participation in 2016.Dependent variables on the y-axis.All estimates come from logit regression models except for the campaign participation scale, which was estimated using OLS.Partisans include leaners.Controls included.83.5 per cent confidence intervals shown.

Figure 2 .
Figure 2. Marginal Effect of Feminist ID by Partisanship on Participation in 2020.Dependent variables on the y-axis.All estimates come from logit regression models except for the campaign participation scale, which was estimated using OLS.Partisans include leaners.Controls included.83.5 per cent confidence intervals shown.

Figure 3 .
Figure 3. Predicted Participation by Feminist ID Strength in 2016 and 2020.The figure shows predicted probabilities with 83.5 per cent confidence intervals.Partisans include leaners.Controls included.

Table 2 .
Predicting Participation by Feminist and Party ID in 2020