Masculine Republicans and Feminine Democrats: Gender and Americans’ Explicit and Implicit Images of the Political Parties

Abstract

During the past three decades Americans have come to view the parties increasingly in gendered terms of masculinity and femininity. Utilizing three decades of American National Election Studies data and the results of a cognitive reaction-time experiment, this paper demonstrates empirically that these connections between party images and gender stereotypes have been forged at the explicit level of the traits that Americans associate with each party, and also at the implicit level of unconscious cognitive connections between gender and party stereotypes. These connections between the parties and masculinity and femininity have important implications for citizens’ political cognition and for the study of American political behavior.

Over the past 40 years American society has experienced huge and controversial shifts in women’s rights and in men’s and women’s roles. Over this period the two major political parties have presented substantially different gender images to the public: they have polarized on women’s rights and abortion, female candidates have become both more common and more likely to be Democrats, a gender gap has become a regular feature of electoral politics, and Republican presidential candidates from Ronald Reagan to George W. Bush have had considerable success framing themselves as the stronger, more manly candidate. This paper demonstrates empirically that amid these developments Americans’ images of the political parties have taken on gendered characteristics, so that Democrats are understood as the more feminine party and Republicans as the more masculine.

These gendered elements of citizen’s images of the parties have been forged, I show, both at an explicit level, in the traits that Americans associate with each party, and also at an implicit level, in a set of unconscious cognitive connections between gender stereotypes and the parties. This suggests that party images and gender images are not simply parallel stereotypes with shared content but rather reflect both conscious associations and unconscious cognitive connections between the two domains. Party is, of course, a centrally-important frame of reference that people use to make sense of politics; my findings suggest that even when gender is not explicitly in play, citizens’ ideas about masculinity and femininity may nevertheless shape political evaluations more broadly than we might otherwise expect.

While the elite side of these developments has received considerable scholarly attention, we know relatively little about how individual citizens have reacted to these shifts in the parties’ public images. A partial exception is work that asks whether gender issues have precipitated a partisan realignment. Several scholars have shown that since 1980 the parties have polarized at the elite level on abortion, the Equal Rights Amendment, and other issues of women’s equality and roles (Wolbrecht 2000; Freeman 1987; Costain 1991), and Greg Adams demonstrates that this elite polarization on abortion, coupled with clear signals from the parties on the issue, has spawned mass-level partisan realignment (1997). However, beyond abortion, neither party has placed great public emphasis on gender issues, particularly those surrounding changes to gender roles, rather than formal equality for women (Sanbonmatsu 2002). Coupled with mass-level ambivalence about gender-role change, this lack of clarity has prevented a more comprehensive gender-based realignment, leading Sanbonmatsu to conclude that abortion is the exception and that “dramatic changes that have occurred in gender roles have not been absorbed into the party system” (2002, p. 220).

None of this work directly addresses the public’s broader party images; nor does it explore ways that ideas about gender not captured by issue positions might shape those images without necessarily inducing people to switch parties.¹ Most work on gender and political behavior has instead focused on the gender gap in partisan identification, vote, and public opinion; or on differences in how people react to male and female candidates.² These literatures demonstrate that gender stereotypes can shape issue opinion and candidate evaluation in subtle ways. An important recent line of work on female candidates explores the interactions between citizens’ gender and party stereotypes; often party stereotypes override gender stereotypes, although in some cases the two interact in more complex ways (Dolan 2004; Huddy and Terkildsen 1993; Koch 2002; Sanbonmatsu and Dolan 2009; Huddy and Capelos 2002; Matland and King 2002; Hayes 2009a).

The gendered facets of citizen’s party images hold the potential to shape political cognition in important ways for two reasons. First, people are quite adept at applying gender stereotypes, and in particular the attributes “masculine” and “feminine,” to a wide range of objects that are not literally male or female. Even young children reliably classify colors, types of plants and animals, shapes, and much more as masculine or feminine, leading Bem to conclude that “there appears to be no other dichotomy in human experience with as many entities assimilated to it as the distinction between male and female” (1981, p. 354; see also Leinbach et al. 1997).³ We might expect, therefore, that citizens will be likely to draw on partisan-based gender associations to make inferences about political candidates, issues, and groups. Second, precisely because gender-related issues have not been fully assimilated to the existing partisan alignment, these gendered trait associations may be particularly likely to create or reinforce cross-pressures for a significant number of citizens.

In this paper I employ a multi-method, multi-tool approach to explore the gendered elements of citizen’s party images. First, I use nationally-representative survey data to document the explicit, public face of the gendering of the parties. Using over three decades of data from the American National Election Studies (ANES 2005), I demonstrate that citizens associate the parties with gendered—i.e., masculine and feminine—traits. These associations developed over the course of the 1980s and are most firmly established among politically knowledgeable Americans. These over-time and cross-sectional patterns suggest that these gender-party connections were shaped by the public actions and images of the parties and their candidates during this period.

Second, to explore the implicit cognitive underpinnings of these party images, I present experimental evidence from a reaction-time study conducted in a virtual psychological laboratory with a college sample. This evidence suggests that people do not merely ascribe to the Democrats and Republicans a series of traits that happen to be feminine and masculine. Rather, ideas about the parties are linked cognitively with ideas about gender. This means that when people think about the Democratic Party, they are likely to draw unconsciously on their concepts of femininity, and when people think about the Republican Party, they are likely to draw on ideas about masculinity. While each of these two sources of evidence has limitations, they also possess important counterbalancing strengths: the survey data provide a nationally-representative picture of explicit party images over a long time period, while the experimental data allow us a glimpse of the much harder to observe cognitive underpinnings of those images.

Masculinity and Femininity in American Culture and Politics

Modern American ideas about masculinity and femininity are “fuzzy sets” (Deaux 1987) made up of clusters of attributes that define the characteristics thought to be characteristic of men and women, respectively. At their core are a set of instrumental personality traits for men and expressive personality traits for women. Thus, masculine men are thought to be active, independent, and decisive; feminine women are thought to be compassionate, devoted to others, emotional, and kind. These core traits are linked with a range of other features, including other traits (masculine men are aggressive, practical, tough, hardworking, and hierarchical; feminine women are gentle, submissive, soft, ladylike, and egalitarian); physical characteristics (masculine men are big, strong, and muscular; feminine women are small, weak, and soft spoken); social roles and occupations; interests; and sexuality (masculine men and feminine women are both expected to be attracted to the other sex).⁴ Moreover, the cultural constructions of masculinity and femininity treat each as a coherent package that is defined in opposition to the other: “feminine” is thus understood as “not masculine” and vice versa (Foushee et al. 1979).⁵

It should be noted that this configuration works to associate masculinity with politics and leadership. The very idea of a political or public realm is constructed in contrast with the private, and the public/private duality is deeply gendered, with the public sphere traditionally associated explicitly with men (e.g. Phillips 1991).⁶ While formal gender segregation is now gone, both the political realm and leadership—in and out of politics—continue to have symbolically masculine connotations (Ridgeway 2001). Interestingly, Carlson and Boring present experimental evidence that male and female candidates are rated as more masculine and less feminine when described as winning, rather than losing (1981).

Finally, it should be noted that stereotypes of masculinity and femininity also include negative attributes. Thus, for example, stereotypes of men include characteristics such as greedy, hostile, and self-interested, and stereotypes of women include negative traits like spineless and gullible. In addition, some aspects of masculinity, such as aggressiveness and violence, can take on negative connotations when they appear to be excessive or when applied to an undeserving target (Spence et al. 1978, 1979).

Republicans and Democrats Become Masculine and Feminine

There are five interrelated developments in the late 1970s and early 1980s that we might expect to have mapped masculinity and femininity—already important for politics—onto the Republican and Democratic Parties, respectively. I discuss these developments here to motivate the analyses that follow; it is beyond the scope of this paper, however, to demonstrate the causal impact that they might have on those analyses. First, as I discuss above, Wolbrecht documents the polarization of party elites over the Equal Rights Amendment and other issues of women’s equality in the late 1970s, and Adams traces the polarization of the parties on abortion over the same period. Earlier, the Republican Party was modestly more supportive of women’s rights than the Democrats, though neither party devoted much attention to the issue. By 1980, the parties had staked out the positions they hold today, and differences over abortion in particular had become an important feature of the elite-level party alignment. This polarization is reflected in stark differences in party platforms, in bill sponsorship rates, and in roll-call votes, leading Wolbrecht to suggest that “the lines have thus been drawn with considerable clarity since 1980” (2000, p. 6; Adams 1997). These partisan differences were reinforced and made more salient by the growing role within the Republican coalition of antifeminist groups and the social conservative movement and by the alliance of feminist groups with the Democratic Party (Freeman 1975, 1993; Spruill 2008).

These developments have been reinforced by the gender associations of the issues “owned” by each of the political parties. There is considerable overlap between the political issues that citizens associate with each party, on the one hand, and that they associate with men and women, on the other. Republicans are thought to handle better such issues as defense, dealing with terrorism, and controlling crime and drugs (Petrocik 1996; Petrocik et al. 2003); these are precisely the sorts of issues that Americans associate with men or with masculine traits (Kahn 1996; Alexander and Andersen 1993; Huddy and Terkildsen 1993). Conversely, Democratic-owned issues include education, health care, helping the poor, protecting the environment, and promoting peace; these are all also associated with women or with feminine traits.⁷ Rapoport et al. (1989) find that people make trait inferences about candidates based on their issue positions; we might therefore expect similar inferences about party traits based on the issues associated with each.

Third, these parallel party and gender issue competencies are reflected in and reinforced by public attention to the gender gap in vote and partisanship (Gilens 1988; for an overview of the enormous gender gap literature, see Sapiro 2003). The gender gap first achieved sustained public attention after the 1980 election as a result of efforts by women’s groups to increase their influence within the Democratic Party (Mansbridge 1985; Mueller 1988), and has been a fixture of media coverage of presidential campaigns ever since. While the size and consistency of the gender gap is often overstated in the popular media (Ladd 1997), coverage of the gap likely serves to reinforce for the public the association of the Republican Party with men and the Democratic Party with women.

Fourth, the association of women with the Democratic Party is further reinforced by the fact that substantially more women have been elected as Democrats than as Republicans over the past several decades.⁸ The Democratic nomination of Geraldine Ferraro for Vice President in 1984 was intended as a signal that the Democrats were the party of women (e.g. Wolbrecht 2000, pp. 52–53). While Mondale and Ferraro did not win, the number of Democratic women elected at the federal and state levels has increased faster than the number of Republican women, generating what Laurel Elder has called a “partisan gap” among female elected officials (2008). As depicted in Fig. 1, since the mid-1980s the number of Democratic women elected to the US Senate, the US House, and to state legislatures has increased steadily, while the number of Republican women has increased much more slowly if at all. This means that citizens are likely to observe more women in politics who are Democrats than Republicans.

Fig. 1

Number of female elected officials, by party. Senate and House figures are from CAWP (2010). State legislature data provided by Elder (2008)

Finally, as I discuss above, the concept of leadership and the political realm itself both carry symbolically male connotations. Perhaps because of this, since the early republic presidential candidates and their campaigns have often emphasized their own masculine credentials and tried to undermine those of their opponents (Kann 1998; Etcheson 1995; Duerst-Lahti 2006, 2008; Bederman 1995; Hoganson 1998; Fahey 2007; Ducat 2004). More anecdotally, journalistic and academic observers alike have suggested that from Ronald Reagan in 1980 through George W. Bush in 2004, the Republicans have frequently won the battle to appear more manly, through a combination of claims about personal character and assertions that they are strong—and their opponents weak—on issues ranging from standing up to foreign enemies to being tough on crime and drugs (e.g., Orman 1987; Mihalec 1984; Jeffords 1994; Kimmel 1987; Ducat 2004; Malin 2005; Fahey 2007; Rich 2004). These gendered differences in candidate presentation and substance dovetail with linguist George Lakoff’s argument that conservatives and liberals—and by extension the Republican and Democratic parties—operate in different, and metaphorically gendered, moral universes. In Lakoff’s account, different approaches to the appropriate role of the government metaphorically evoke different views on parenting: the Republicans are the party of the strict father, while the Democrats are the party of the nurturing mother (Lakoff 2002).

In sum, these interrelated developments all conspire to associate the Republican Party with men and masculinity and the Democratic Party with women and femininity. These gendered associations have their foundation in political issues that deal explicitly with questions of gender, and are reinforced through recent political campaigns and other public discourse surrounding the parties. In the sections that follow I explore empirically the images Americans have of the two political parties, looking first at the gendered traits that Americans associate explicitly with the parties, and second at implicit cognitive connections between gender and party.

Gendered Traits are Explicitly Associated with the Contemporary Parties

This first analysis explores the gendered trait associations contained in Americans’ images of the contemporary political parties over the past three decades, drawing on the ANES open-ended questions about respondents’ likes and dislikes about the political parties. In each pre-election study, the ANES asks respondents to mention up to five things they like and an additional five things they do not like about each of the political parties, along with parallel questions about each major-party presidential candidate (in presidential years) and major-party House candidates in the respondent’s district. The analysis in this paper focuses on the political parties—up to 20 distinct mentions per respondent.⁹

This measure has several important advantages for this analysis. First, the questions are open-ended, which means that respondents can mention, in their own words, anything they consider salient about each of the parties. Respondents’ remarks can reflect their own ways of seeing and thinking about the parties; this is an important feature when measuring something as heterogeneous as party images, and especially when searching for themes—such as masculinity and femininity—that have not been the focus of sustained prior research. Second, because these questions have been included since 1972 in the ANES with comparable coding of responses, they allow for controlled comparisons over long stretches of time. Finally and most pragmatically, the these questions are the only nationally-representative source for open-ended information about citizen’s views of the party over this time period; they are the best data available for my purpose. Moreover, the likes/dislikes battery has been the measure of choice for those exploring the nature of party images (Baumer and Gold 1995; Sanders 1988; Trilling 1976; Geer 1991); the parallel questions about the candidates have also been used to explore candidate images and vote choice (e.g. Hayes 2009b).

One drawback of this measure is non-response: in each ANES study approximately a quarter respondents decline to mention any reason to like or dislike either party.¹⁰ These respondents appear to be relatively detached from politics in general and from the parties in particular—they are, for example, substantially more likely than others to classify themselves as “pure” independents, to refuse to rate the parties on the feeling thermometer scale or to rate both at the neutral, 50-degree mark, and ANES interviewers rate these respondents substantially lower in political knowledge.¹¹ It may be reasonable to expect, therefore, that these respondents do not mention things they like and dislike simply because they do not hold much of a meaningful image of the political parties. If this is the case, then their absence from the analysis does not hamper my ability to characterize the gendered features of the aggregate images of the parties among the public. Nevertheless, of course, we can not be sure of this in the absence of close-ended measures that tap relevant aspects of party images; as it is my findings are conditioned to apply to that part of the public that is able and willing to articulate some reasons to like or dislike the parties.

A second concern is that ANES does not report respondents’ actual remarks; rather, each remark is coded into one of 699 “party-candidate master codes” or categories. For my analysis, therefore, these party-candidate master codes were recoded into one of three categories: stereotypically masculine traits, stereotypically feminine traits, and a residual category for everything else, including non-gendered traits and all mentions of groups, issues, specific individuals, and miscellaneous other things.¹² Thus, ANES categories were coded as masculine if they refer to traits or personality characteristics that are associated in contemporary American gender stereotypes with men or masculinity, and classified as feminine if they refer to traits or personality characteristics that are associated with women or femininity. References to political issues and to social groups were not coded as masculine or feminine. The ANES codes were classified independently by the author and two graduate student research assistants, both of whom were familiar with the gender stereotypes literature but were blind to the hypotheses of this study.¹³ After classifying the codes independently, the three coders met together to discuss differences and ambiguous cases and came to agreement on final classification of each code.¹⁴

Both positive and negative traits were classified; for example, masculine traits include both references to being statesmanlike, energetic, or efficient, and also references to such negative traits as being cold or being selfish as well as references to sex scandals. Feminine traits included kind, gentle, and compassionate as well as weak and indecisive. A complete listing of the master codes classified as masculine and feminine appears in the Appendix (Tables 6, 7, 8, and 9). Analyses are based on tallies of all mentions, separately for each of the four types (Democratic Party likes, Democratic Party dislikes, Republican Party likes, and Republican Party dislikes).¹⁵

My analysis explores how often stereotypically masculine or feminine traits are among the reasons that respondents like or dislike each party. The unit of analysis is the mention, meaning that I exclude respondents who gave no mentions at all, and also respondents who gave no mentions of a particular type. That is, when examining Democratic Party likes, I analyze the universe of mentions in that category, and therefore exclude respondents who had nothing positive to say—gendered or not—about the Democratic Party. This has the effect, of course, of yielding a more informed and knowledgeable sample than the nation as a whole, and of weighting more heavily the views of those respondents who gave more mentions in a particular category. This is appropriate for the purpose of this paper, which is to examine the parties’ aggregate images among the public.¹⁶

Because the likes and dislikes battery was excluded in a number of non-presidential years, my analysis focuses on presidential years between 1972 and 2004.¹⁷ Over that period there were a total of 55,127 things mentioned as likes or dislikes for the parties, and 72% of respondents mentioned at least one “like” or “dislike” about at least one of the parties.

Americans Associate Gendered Traits with the Parties

Table 1 shows masculine traits as a percentage of each party’s total likes and dislikes, as well as the ratio of masculine percentages between the two parties. This table indicates that Americans mention masculine traits much more often when thinking about the Republican Party, as I expect. Likes are tallied in the first column of the table, which shows that masculine traits make up 9.3% of the things that people mentioned as a reason to like the Republican Party, and make up 1.3% of the things that people mention as reasons they like the Democrats (all of the partisan differences are statistically significant, p < 0.001). Thus, masculine traits are about seven times more likely to be mentioned as a reason to like the Republicans than as a reason to like the Democrats. The second column tallies dislikes, and depicts a rather more muted version of the same pattern: masculine traits are about 60% more likely to be mentioned as reasons to dislike the Republican Party (4.2% of all dislikes) than to dislike the Democratic Party (2.6% of all dislikes).

Table 1 Masculine party trait impressions, 1972–2004

Table 2 presents the analogous analysis of feminine traits; here we observe the mirror image of the partisan patterns in Table 1. When thinking about reasons to like the Democratic Party, Americans are almost six times more likely to mention feminine traits as they are when thinking about the Republican Party—5.9% of all Democratic likes are feminine traits, compared with 1.0% of Republican likes. Similarly, feminine traits are over four times as likely to be mentioned as reasons to dislike the Democrats as they are as reasons to dislike the Republican Party (4.4% vs. 1.0%, respectively).

Table 2 Feminine party trait impressions, 1972–2004

These results confirm my expectations about gendered trait associations for the political parties: masculine traits appear much more frequently as reasons to like or dislike the Republicans, and feminine traits appear much more frequently as reasons to like or dislike the Democrats. Gendered traits represent a modest, but I would argue important, component of the overall images of the two major political parties, especially in light of the fact that the denominators for the percentages reported in Tables 1 and 2 include all likes and dislikes, including references to issues, to groups, and to individuals associated with the parties. Not surprisingly, these latter categories, which by definition could not be coded as masculine or feminine traits, make up a very large part of all like and dislike about the parties—about 72%.¹⁸ Thus, while gendered traits certainly do not dominate party impressions, insofar as people associate traits with the parties, they tend to associate masculine ones with the Republicans and feminine ones with the Democrats.¹⁹

Current Gendered Party Associations Solidified in the 1980s

Having documented the basic gendered patterns of party trait associations in the modern era, I turn now to the emergence of these associations over time. As I discuss above, by 1980 party elites had polarized on gendered issues and in subsequent years the various other gendered messages about the parties—from increasing numbers of female Democratic elected officials to the gender gap—emerged or increased in frequency or intensity. Insofar as those public messages affected the public views on the parties, we should expect the patterns of gendered trait associations to strengthen through the 1980s.

Figure 2 shows the development of these gendered associations over time. It displays the percentage of all party mentions that are masculine and feminine traits, separately for each year between 1972 and 2004. The patterns of change over time are consistent with what we would expect as the public has been exposed to gendering images through the past three decades. The top-left panel of Fig. 2 shows the proportion of masculine traits among each party’s “likes.” It indicates that the association of the Republicans with masculine traits jumped sharply from 3.9% of likes in 1972 to 11.1% in 1980, and has since varied between about 8 and 14% of all likes. Meanwhile, the Democratic Party has drawn a much lower—and essentially unchanging—proportion of masculine trait likes over the entire period. The pattern of masculine dislikes is less clear over time; as in the aggregate figures reported in Table 1, the Republican Party draws somewhat more masculine trait dislikes than the Democrats, although the differences are smaller and only clear in the mid-1980s.²⁰

Fig. 2

Gendered party mentions by year. Figure shows masculine and feminine traits as a percentage of all likes or dislikes for each party in each year

The mirror-image of this pattern holds for feminine traits, as I expect. In the bottom panels of Fig. 2, we see that the feminization of the Democratic Party—in terms of both likes and dislikes—first begins to appear in 1980, and is solidified in 1984, after which it remains fairly steady over time. There is a noticeable jump in feminine dislikes for the Democrats in 2004; much of this increase is driven by a spike in references to the party lacking a definite philosophy. This reflects, perhaps, the prominence of this theme in Republican campaigns in 2004, and in particular the “flip-flopper” attacks on John Kerry. In any case, from 1972 through 2004 the Republican Party draws a consistently tiny set of feminine trait likes and dislikes.

Politically Knowledgeable Citizens are Most Prone to Hold Gendered Impressions of the Parties

Over the course of the 1980s, then, the public appears to have absorbed the gendered themes surrounding the parties, and this is reflected in the ways they evaluate the parties. This pattern is consistent with the idea that these associations have their roots in the various gendered aspects of the images projected by the parties during this time. If this is the case, then turning from over-time to cross-sectional variation, I would expect those citizens who pay the most attention to politics to be the most likely to have absorbed these gendered images, and therefore to mention masculine traits about the Republican Party and feminine traits about the Democratic Party. Following John Zaller, I use political knowledge as a proxy for habitual attention to politics (1992). I expect, then, that those with greater political knowledge to be more likely to mention a feminine trait as a reason to like or dislike the Democratic Party, and to be more likely to mention a masculine trait as a reason to like or dislike the Republican Party. Conversely, I do not expect political knowledge to influence the likelihood of mentioning the opposite, non-dominant gendered traits.

While habitual attention to politics should moderate reception of the various gendered messages about the parties, I do not expect there to be systematic variation in who accepts or rejects gendered messages. Citizens may not even recognize messages as gendered, and even if they do few would necessarily reject them on that basis. Therefore I expect the gendered party images to be held relatively homogenously among different members of the American public, aside from the variation due to political attention. In particular, I do not expect systematic differences between men and women, nor among independents, Democrats, and Republicans. To be sure, different partisans have different images of the parties—in particular, Democratic identifiers generally have positive images of the Democratic Party and negative images of the Republican Party, and Republican identifiers have the opposite pattern. Nevertheless, I do not expect either Democrats or Republicans to be systematically different in their gendering of the parties. Insofar as Democrats have positive or negative things to say about the Republican Party, for example, I expect them to be just as likely to mention masculine traits as Republican identifiers; conversely, Republican identifiers who have positive or negative things to say about the Democratic Party should be as likely as Democrats to mention feminine traits. Finally, for similar reasons I expect men and women to hold similarly gendered images of the two parties.

I explore these cross-sectional hypotheses among respondents to ANES studies from 1984 through 2004, the period during which the gendered party images were fully in place among the public as a whole. I estimate a series of models of the individual-level antecedents of mentioning gendered traits about the parties. Specifically, I constructed a set of dichotomous variables that indicates whether each respondent mentioned a masculine or a feminine trait as a reason to like or to dislike each party—this yielded eight variables in all. Thus, for example, the first of these variables indicates whether a respondent mentioned a masculine trait as a reason to like the Democratic Party; the second indicates whether a respondent mentioned a masculine trait as a reason to dislike the Democratic Party, and so forth.

I estimate one probit models for each of these eight dependent variables. The independent variables are political knowledge, as assessed by the ANES interviewers (coded to run from zero for the least informed to one for the most informed),²¹ party identification (entered as a pair of dummy variables: one for Democratic identifiers and one for Republican identifiers, with independents as the reference category), and gender (entered as a dummy variable for women, with men as the reference category), as well as a dummy variable for each study year.²² I ran each model among all respondents who gave at least one mention, gendered or not, of the relevant type; thus, for example, the model for mentioning a feminine “like” about the Democratic Party included only respondents who mentioned some reason to like the Democrats.²³ This means that with this model I analyze the probability that a respondent mentions a masculine like, given that they had something positive to say about the Democratic Party.

Table 3 presents the results of these analyses for the Democratic Party. The cell entries represent the marginal effect of the independent variable on the probability of mentioning a gendered trait, with standard errors in parentheses.²⁴ The first row gives the impact of political knowledge on the probability of mentioning each sort of gendered trait for the Democratic Party. The effects here are clear. As I expect, political knowledge has a strong impact on viewing the Democratic Party in feminine terms (marginal effects of 0.182 and 0.104 for likes and dislikes, respectively, both p < 0.01), and no impact whatsoever on viewing the Democratic Party in masculine terms. This suggests that respondents who habitually pay more attention to politics are more likely to have absorbed the feminized discourse about the Democratic Party and more likely, therefore, to mention feminine traits as reasons both to like and to dislike the parties. Because the discourse around the Democratic Party does not emphasize masculine traits, highly-knowledgeable respondents are not systematically more likely than the less informed to receive masculine messages about the Democrats, and are therefore no more likely to mention masculine traits.

Table 3 Determinants of holding a gendered image of the Democratic Party, among respondents who mention any likes or dislikes, 1984–2004

The rest of Table 3 indicates that the likelihood of mentioning gendered traits about the parties is utterly unaffected by a respondent’s party affiliation and gender. Democrats and Republicans are equally prone to thinking about the Democratic Party in gendered terms, as are women and men.²⁵

Turning to images of the Republican Party, Table 4 presents results from the analogous probit models. Here the results for political knowledge are the mirror-image of those for the Democratic Party, as I expect. The most politically knowledgeable are much more likely than the least knowledgeable to mention masculine traits as something they like about the Republicans (marginal effect of 0.204, p < 0.01) and somewhat more likely to mention masculine traits as something to dislike about the Republicans (marginal effect of 0.071, p < 0.01). In contrast, and as expected, political knowledge has no substantive impact on the probability of mentioning feminine characteristics as reasons to like or dislike the Republicans.

Table 4 Determinants of holding a gendered image of the Republican Party, among respondents who mention any likes or dislikes, 1984–2004

Turning to respondent partisanship, here we see do see some small effects. Republican identifiers are somewhat more likely than others to mention masculine things they like about their own party, and somewhat less likely to mention masculine things they dislike about the party (marginal effects of 0.045 and −0.045 respectively, p < 0.01). Given the counterbalancing signs and relatively small sizes of the effects, however, these coefficients may simply be picking up the tendency of Republicans to mention more things they like (and fewer they dislike) about their party. In any case, this relatively limited variation does not suggest qualitative differences in gendered party images. Finally, and again as expected, there are only tiny differences between men and women, confirming that the gendering of the party images is something that both men and women pick up from the political discourse in similar ways.

Overall, then, these results are consistent with my expectations. Americans have absorbed the gendered discourses surrounding the parties, and associate stereotypically masculine and feminine traits with the Republicans and Democrats, respectively. We see a nuanced pattern of variation by political knowledge that suggest the gendered aspects of the public images the parties have been taken in by Americans insofar as they pay attention to politics. The lack of further variation among different types of respondents suggests that the gendered aspects of the party images are absorbed by everyone, and are not being rejected by some respondents based on their other political predispositions.

Implicit Cognitive Connections Between Party and Gender

The gendered aspects of conscious party images may imply additional unconscious aspects as well. The interactions between conscious attitudes and unconscious—or implicit—attitudes and concepts is an active area of social psychological inquiry. While are obviously aware of our explicit thoughts, a surprisingly large amount of implicit cognitive activity occurs outside our awareness (Bargh and Morsella 2008; Dijksterhuis and Nordgren 2006; Greenwald and Banaji 1995; Wilson 2002). Implicit cognition affects our conscious thoughts—it would be of little interest otherwise—but we generally are not aware of those effects and have little control over them. In the racial realm, for example, people who have their race schemas primed, or cognitively activated, are more likely to judge ambiguous actions by an African American as aggressive, compared with people whose racial schemas are unprimed (e.g., Sagar and Schofield 1980). In the gender realm, there is a long line of research on the ways gender stereotypes influence attributions for success and failure. In male-dominated realms, for example, people tend to attribute men’s successes to ability and effort, and to attribute women’s successes to luck or the ease of the task (e.g. Deaux and Emswiller 1974).

The majority of political science research on implicit attitudes has focused on race, and has explored the ways white Americans associate policies like welfare and crime with race. This work has demonstrated that appeals to these implicitly racialized policies can mobilize racial predispositions to shape citizen’s attitudes and vote choices, and can do so without people being aware of the racial elements in their thinking. These effects can be quite powerful, and can be evoked by extremely subtle imagery and language—in one experiment, for example, a single reference to “inner city” criminals mobilized racial prejudice to shape crime opinions (Hurwitz and Peffley 2005); in another study a single picture of an African American candidate evoked racial considerations (Terkildsen and Schnell 1997).²⁶

So far I have shown that people ascribe traits to the Democratic Party that overlap with stereotypes about femininity, and to the Republican Party that overlap with stereotypes about masculinity. It is logically possible, of course, that people’s ideas about parties and about genders simply share some common traits, with no deeper connections. For example, though I might mention “red” as a feature of both apples and fire engines, this does not necessarily mean I have profound implicit associations between these two categories. In this case, my thinking about fire engines is probably not shaped much by what I know about apples, and vice versa. On the other hand, a wide range of shared features can be the explicit face of a more extensive set of implicit connections between the categories, as suggested, for example, by the research on racial stereotypes and crime and welfare policy. If the psychological connections between parties and gender stereotypes are like those between apples and fire engines, then the findings so far may imply relatively little for citizen’s political cognition about the parties and their candidates. On the other hand, if the shared traits are the conscious face of deeper, unconscious semantic connections between the two categories, then we might expect gender stereotypes to shape political cognition in ways more subtle and consequential than we have realized.

In this section of the paper I present experimental evidence about just these sorts of implicit links among a sample of college students. Specifically, I show that party images and gender stereotypes are not simply unrelated concepts with parallel content. Rather, implicit connections exist that connect the parties with ideas about gender.

To show this, I examine the effects of thinking about one or the other political party on the cognitive accessibility of ideas about masculinity and femininity. It is a well-established phenomenon in social psychology that accessing a concept, even implicitly, makes it more accessible in memory and therefore faster to access. Cognitive accessibility is the mechanism underlying many priming effects, in which exposure to a political issue makes that issue more accessible in memory, and therefore more likely to come automatically to mind subsequently in thinking about related issues.²⁷

Because cognitive accessibility is measured in milliseconds, we cannot assess it in the traditional survey context. However, we can measure it reliably with appropriate computer software. In this study, I used a lexical decision task (LDT) to measure cognitive accessibility; this is the standard approach developed by Fazio (1990) and employed by previous studies in political science (Valentino et al. 2002; Nelson et al. 1997). In the LDT, a series of letter strings are flashed on a computer screen one at a time in a random order. Some of these strings are words and some are nonsense letter combinations, and participants are asked to distinguish as quickly and accurately as possible between the two by pressing one computer key for words and a different key for non-words. Of the words in this study, five were stereotypically feminine or related to women (“feminine,” “housewife,” “librarian,” “nurse,” and “skirt”) and five were masculine or related to men (“doctor,” “janitor,” “masculine,” “razors,” “trousers”); these were mixed with twelve non-gendered filler words (e.g., “actual,” “tutorial,” “remorse”) and with 32 pronounceable nonsense strings (e.g., “catipal,” “igamine,” “raich”).²⁸ The computer recorded the length of time in milliseconds that participants took to classify each target string. The logic of this procedure is that respondents will be systematically faster to identify words that are relevant to concepts that have been recently activated; extensive research in social psychology has demonstrated the reliability and validity of this measure of cognitive accessibility (Fazio 1990; Wittenbrink 2007).

The feminine and masculine words were chosen to be gender-related, but not possibly associated directly with either of the parties. This rules out most gender-related traits, and it ensures that insofar as thinking about the parties makes these gender-relevant words more accessible, that this must be due to cognitive links between party and gender concepts, and not simply because the target words are themselves linked directly with the party. That is, if response times to a feminine trait word like “compassionate” are reduced by thinking about the Democratic Party, this could simply be due to the direct association of compassion with the Democrats. On the other hand, words like skirt—which have a clear link with gender but no plausible direct connection with politics—should be made accessible by thinking about the Democrats only insofar as ideas about the Democrats and about gender are linked implicitly. I did include two gendered traits—“masculine” and “feminine”—given their obvious face validity as measures of gender associations for the parties. In any case, the results presented here are substantively unaffected by the exclusion of these two items.

Therefore, insofar as implicit cognitive connections exist between the Democratic Party and femininity, I expect that thinking about the party should facilitate recognition of feminine words. Similarly, an implicit connection between the Republican Party and masculinity would lead thinking about the Republican Party to facilitate the recognition of masculine words. The lexical decision task was embedded in a web-based survey on political attitudes and political advertising.²⁹ The study was completed between December 2008 and February 2009 by 195 undergraduate students at a large and diverse state university, who were recruited from three lower- and mid-level political science courses in return for extra course credit.³⁰ At the beginning of the survey, all participants were shown an identical pair of nonpolitical television advertisements.³¹ Next, participants were randomly assigned to one of three conditions: control, Democratic or Republican. Those in the control condition immediately completed the lexical decision task. Those in the Democratic or Republican conditions were induced to think about the Democratic or Republican Party, respectively, by answering the standard ANES likes and dislikes questions about the party, and then completed the lexical decision task. After the LDT, respondents completed whichever likes and dislikes they had not already answered, followed by additional political and demographic questions. (The structure of the experiment is depicted in Table 5.) Importantly, by using the likes and dislikes battery I was able to induce respondents to think about one of the parties without introducing anything specific about the party. This ensures that any implicit connections between party and gender that I find exist already for participants and are not simply the product of the experimental stimulus.

Table 5 Summary of experimental conditions

To measure the implicit associations between femininity and the Democratic Party, I compare the average reaction time for feminine words between respondents in the control and Democratic conditions.³² There is enormous individual variation in reaction times to all words, so to maximize statistical power I estimate this difference with a regression model that includes each individual’s average reaction time for the neutral words as a covariate, plus a dummy variable for the experimental condition.³³ The coefficient on the condition dummy is the direct estimate of the effect on reaction time to feminine words of thinking about the Democratic Party, and is therefore my indicator of an implicit cognitive connection between the party and femininity. To estimate the implicit association between masculinity and the Republican Party I conduct an analogous analysis of masculine-word reaction times, comparing the control and Republican Party conditions. Finally, in order to put the estimated effects in context, I scale the results to reflect the neutral-word reaction times of an average respondent. Figure 3 presents the results of this procedure; the underlying regression models are presented in the Appendix (Tables 10 and 11).

Fig. 3

Implicit party-gender connections. Figures show effect of thinking about a party on reaction times to gendered words. For example, the left panel shows that respondents who thought about the Democratic Party first, in the Democratic Party condition, identified feminine words an average of 51.1 ms faster than those in the control condition, who did not first think about a party. Based on regression models described in the text and presented in the Appendix

The results indicate that there is, in fact, an implicit cognitive connection between the Democratic Party and femininity among the study participants. As depicted in the left panel of Fig. 3, average response time to the feminine words was reduced by 51 ms (t = 2.27, one-sided p = 0.012). For the Republican Party and masculinity, the results are also consistent with expectations, although the effect is smaller. Thinking about the Republican Party reduces average reaction times for masculine words by about 20 ms (t = 1.69, one-sided p = 0.046). These findings suggest that bringing the Democratic Party to mind—by asking participants what they like and dislike about the party—makes more accessible their ideas about gender and femininity. Conversely, thinking about the Republican Party appears to activate ideas about masculinity.

The estimate of the strength of the Democratic-feminine association is quite a bit larger than the estimated Republican-masculine connection; however, the difference between the two estimates does not itself achieve statistical significance (two-sided p ≈ 0.20). I hesitate, therefore, to go too far in interpreting this difference. Nevertheless, if these estimates do reflect a real difference in the underlying cognitive connections, we might understand this difference in two related ways. First, the political discourse over the past several decades may simply have been more effective in linking the Democratic Party with femininity than in connecting the Republicans with masculinity. Second, messages in the political environment that evoke femininity may be more psychologically salient than masculine messages, because they stand out more in a political realm that as a whole is symbolically—and often literally—male. This masculine baseline could make the gendered aspect of references to femininity implicitly stand out, while many references to masculinity might feel simply “political” and not so much about gender. Such a pattern would be consistent with work in social psychology showing that non-prototypical members of a category are generally more salient than prototypical members.³⁴ Despite these possible partisan differences, the broader finding is that when participants thought about the political parties their ideas about gender were activated unconsciously. This activation encompassed not just obviously politically-relevant gender attributes such as compassion or toughness, but also gender associations that have little or no explicit political relevance such as “skirt” and “nurse,” “razors” and “janitor.”

Additional research is clearly necessary to explore the nuances of the implicit associations uncovered in this study, and to assess how they generalize to the American public as a whole. This study suggests that citizens’ gender schemas may be activated cognitively when they think about the political parties; if they are, those schemas would then be accessible and therefore likely to play a hidden role in making sense of ambiguous political phenomena. Since the meaning and interpretation of most political figures, policies, and actions are to some extent ambiguous, there is much potential for implicit party-gender links to affect political cognition. For example, is a new economic proposal prudently cautious, hopelessly timid, or recklessly aggressive? Or does the use of cruise missiles to strike at distant, hidden enemies represent a cowardly unwillingness to confront our foes, or a sensible decision not to risk needless danger? If thinking about the political parties activates citizens’ gender stereotypes, then those stereotypes could influence the interpretation of those sorts of policies, and could do so differently depending on the party affiliation of the leader who proposes them. Similarly, people might bring somewhat different standards to bear when evaluating leaders from different parties, and might make different baseline assumptions about the sorts of traits they possess. In short, these initial experimental results suggest the importance of additional research on subtle ways that gender stereotypes can shape political judgments that extend beyond explicit questions of gender.

Conclusion

Taken together, my findings suggest that ideas about the two political parties are mapped onto ideas about the two genders, both in the images citizens consciously hold of the parties and in the implicit connections between these images and their gender concepts.³⁵ This gives us a richer understanding, of course, of the cognitive underpinnings of partisan attitudes. In addition, my findings suggest several avenues for additional research.

The connections I find between party and gender stereotypes have implications for how we conceptualize and study the intersections between the two. There is a growing literature on the respective roles played by gender and party stereotypes in shaping citizens’ impressions of candidates, which builds on a long line of experimental work demonstrating that gender stereotypes shape citizens’ perceptions of candidates’ traits and issues positions, especially for female candidates and especially in the absence of partisan cues. However, some recent work suggests that partisan stereotypes can trump gender stereotypes when both cues are available; others find that partisan and gender cues can interact in more complex ways (Hayes 2009a; Dolan 2004; Huddy and Capelos 2002; Koch 2002). This seems especially true for female Republican candidates, whose party and gender cues in some sense conflict (McDermott 1997).

My findings suggest that we should not think of party and gender stereotypes as independent alternatives, but rather as two sets of stereotypes with important links. This is true in the narrow sense that party stereotypes may contain implicit gendered content. More broadly, my results speak to the intersectionality of gender and party stereotypes: the ways in which gender and party categories may each derive their meanings in part from their relationship with the other (on intersectionality in political science research, see Hancock 2007). Thus, for example, a candidate’s party affiliation might influence voters’ perceptions of his or her enactment of masculinity and femininity. Voters may make inferences about a candidate’s compassion or strength not in reference to some generic ideal, but rather against expectations created by candidate’s gender and party. In short, Republican candidates—male and female alike—might be judged against a baseline expectation that they are relatively masculine, and Democrats against a more feminine baseline. This gives us added perspective on coded—and not-so-coded—appeals that assert or question a candidate’s masculine credentials. For example, Arnold Schwarzenegger’s repeated references to Democrats as “girlie-men”³⁶—along with more subtle evocations of that basic point by others—might carry a strong political punch insofar as they evoke not simply the feminine traits consciously associated with the Democrats, but also much wider range of negative characteristics our culture associates with effeminate men.

A second question for further research concerns the net electoral effects of the gendering of the political parties. Studies of the effects of candidate gender are suggestive here. Gender stereotypes have mixed effects on the evaluation of female candidates. On the one hand, they are judged to be worse decision makers and weaker leaders, as well as less competent on and less interested in issues of foreign policy and the economy. On the other hand, female candidates are also viewed as more honest and more compassionate, and are believed to be more interested in and trustworthy on “compassion” issues such as health care, education, and those that affect women and children. This means that the public issue agenda matters; in 1992, for example, the Hill-Thomas hearings and other factors led voters to favor outsiders, and women in particular (Delli Carpini and Fuchs 1993; Duerst-Lahti and Verstegen 1995; Sapiro and Conover 1997; Dolan 1998; Kim 1998). More broadly, candidates may make strategic choices about their self-presentation, choosing to highlight or downplay gender characteristics depending on the electoral environment (Kahn 1993; Iyengar et al. 1997).

Different issue agendas and different constructions of the problems we face should affect the degree to which citizens feel a need—conscious or subconscious—for symbolically masculine leaders. The masculine image of fatherly protection may be more appealing in times of external threat and in times when people feel insecure about changing gender relations within society. Kristin Hoganson argues, for example, that shifting gender relations in the family, the workplace, and in politics conspired to make a form of potent, aggressive masculinity particularly politically salient at the turn of the twentieth century and contributed to American involvement in the Spanish-American War. After the subsequent Philippines war turned into a bloody, cruel quagmire, however, this aggressive masculinity came to seem reckless and dangerous, increasing the appeal of those who promised an end to the war and to take a more peaceful approach (1998). The obvious parallels with modern developments in Afghanistan and Iraq present a fruitful area for additional research on the ways that partisanship and masculinity and femininity play out politically against different policy backgrounds.

Given the masculine associations of leadership, we might expect that on balance the masculinization of the Republican Party and feminization of the Democratic Party may have conferred advantages to the Republicans, at least at the presidential level. However, cultural ideas about masculinity and femininity and about their connections with politics are complex enough that Democratic candidates may have more latitude than simply to try to out-man the Republican Party. For example, while observers have commented on Barack Obama’s relatively feminine appearance and approach, he does not seem to have suffered from this image as much as his recent Democratic predecessors. Of course many factors shaped Obama’s image and his ultimate success, including his unique status as the first African American major-party nominee and the deep public anger over the Iraq war and other failings of the Bush presidency. This broader context, however, may have helped Obama to project an image not of effeminacy, nor of aggressive masculinity, but rather of moral and controlled manliness; an image that may also have helped him counter stereotypes of black violence (Cooper 2008). This sort of reshaping of the terms of the connection of masculinity and politics may have helped the Democrats win the White House in 2008; it suggests that we explore further the ways that that sort of implicit gender context shapes broader political cognition and behavior.

Appendix

Table 6 Masculine (+) traits

Table 7 Masculine (−) traits

Table 8 Feminine (+) traits

Table 9 Feminine (−) traits

Table 10 Impact of thinking about Democrats on feminine-word reaction times

Table 11 Impact of thinking about Republicans on masculine-word reaction times