Not Tough Enough? Gender Stereotypes and Vote Choice in Switzerland

ABSTRACT

While many studies examine the effect of gender stereotypes on electoral chances of women candidates in the U.S. little research investigates this relationship outside two-party systems and experimental designs. This article examines the relationship between gender stereotypes and vote choice for women candidates in the Swiss federal election of 2019. Our regression analyses show that voters holding male issue stereotypes are less likely to elect women candidates, while voters holding female issue stereotypes are more likely to elect women candidates. Interaction effects suggest differing effects depending on a respondent’s age, political knowledge, and the candidate’s party position.

KEYWORDS:

• Gender stereotypes
• vote choice
• Switzerland

Introduction

“Frankly, if Hillary Clinton were a man, I don’t think she would get five percent of the vote,” Donald Trump said in 2016 about his opponent Hillary Clinton (Gearan and Philipp 2016). With this statement, he implies that Clinton is an incompetent candidate and if it were a man-to-man contest, she would lose the election.¹ However, since she is a woman, she can use her gender to her political advantage. At the same time, he doubted whether she had enough “strength” and “stamina” - stereotypically male traits – to be president of the United States (Gearan and Philipp 2016), meaning that her female traits are a political disadvantage.² Examples such as these show how common stereotypical ideas about the characteristics, abilities, and attitudes of women and men in politics still are today. It is often assumed that these stereotypes about male and female politicians are not without consequences for their political success. In the U.S. this has led to extensive research about the effect of gender stereotypes on candidate evaluation, political decision making, and vote choice for a man or woman candidate (Bauer 2015, 2018; Hayes and Lawless 2015; Dolan 2010, 2014a, 2014b; Dolan and Lynch 2016; Fridkin and Kenney 2009; Holman, Merolla, and Zechmeister 2016; Huddy and Capelos 2002; Lawless 2004; McDermott 1998; Ono and Yamada 2020; Sanbonmatsu 2002, 2003; Sanbonmatsu and Dolan 2009).

But just as Donald Trump himself is not sure whether being a woman helps or hurts Clinton, the literature suggests differing results about the effect of gender stereotypes on vote choice as well.³ While some research finds a positive relationship between gender stereotypes and the propensity to vote for a woman candidate (Bos 2011; Brooks 2013; Hayes and Lawless 2015; Dolan 2010; Fridkin and Kenney 2009), other studies find a negative relationship between the two variables (Dolan 2010; Huddy and Capelos 2002; Knuckey 2018; Lawless 2004), contradicting results or even finding no substantial, consistent relation at all (Gordon, Shafie, and Crigler 2003; Bauer 2015; Hayes and Lawless 2015; Dolan 2014a, 2014b; Dolan and Lynch 2016; Johns and Shephard 2007; McDermott 1998; Sanbonmatsu 2002). Additionally, most studies investigate the U.S. and its two-party system, whereas little research has tested the effect of gender stereotypes on real-world vote choice in a multi-party system.⁴

Yet, research on more countries would be important, as some studies point to the fact that the effect between gender stereotypes and vote choice in the U.S. differs from the effect in other countries. Endo and Ono (2023) for example found that voters in Japan still hold more gender stereotypes than voters in the U.S. More importantly, apart from investigating countries other than the U.S. we believe that it is relevant to evaluate the effect of gender stereotypes outside of two-party systems. While in a two-party system, party affiliation is probably the main reason for voting for a certain candidate, in multi-party systems, there are several candidates of similar parties and with similar ideologies. This makes characteristics of these candidates more likely to affect vote choice – one of them being a candidate’s gender. Thus, multi-party systems can be viewed as a most likely case for gender stereotypes to affect vote choice. Lastly, to the best of our knowledge, most studies rely on experiments instead of real-world elections (see for example Schwarz and Coppock 2022 for a meta-analysis of 67 experiments). For multi-party systems in particular, mostly experimental research seems to have been conducted, and findings so far are inconclusive: While a Finnish study found varying effects of gender stereotypes on vote choice depending on whether the survey context was hypothetical or concerned real legislative elections (Lefkofridi, Giger, and Holli 2019),⁵ in Brazil, respondents preferred stereotypically female candidates (Lucciola 2022). Due to the potentially far-reaching implications and the lack of research focusing on countries outside the U.S. on multi-party systems, and on real-world elections, our research question is: To what extent did gender stereotypes affect vote choice regarding candidate gender in the 2019 federal election in Switzerland?

To answer our research question, we draw on individual-level data from the Swiss Election Study (Selects) from the Swiss federal election of 2019. Switzerland is in some respects a special case, as women in Switzerland only gained the right to vote in 1971 (Vatter 2020), around 50 years later than in most European countries (The Federal Assembly 2021). Despite above-average representation of women in parliament today,⁶ this late introduction of the right to vote for women could mean that gender stereotypes are still more predominant in Switzerland compared to other European countries. Therefore, Switzerland can be seen as a most likely case.

We measure gender stereotypes with male as well as female trait and issue stereotypes and link them with respondents’ self-declared vote choice in the 2019 Swiss State Council elections. The main advantage of this dataset is that it measures vote choices from real elections on the national level, while most of the research to date had to rely on experimental settings (see Dolan 2014a; Dolan and Lynch 2016). This is an advantage compared to such experimental studies in which respondents need to make choices in abstract settings without knowing additional information about the candidates. A rare bonus in our setting is that we can almost certainly exclude a social desirability bias since we manually code the candidates’ gender and do not explicitly ask for gender preferences. Therefore, our findings reveal information about the effect of gender stereotypes in the complex setting of a real election. Our logistic regression models indicate that male issue stereotypes lead to a lower probability of voting for a woman candidate, while voters holding female issue stereotypes have a higher probability of voting for a woman candidate. These effects do however differ depending on a respondent’s age, their level of political knowledge, and along the left-right placement of the candidate’s political party. Our results suggest that gender issue stereotypes exhibit a prevalent influence on vote choice for women candidates in multi-party systems. On the other hand, we find that trait stereotypes are not systematically related to women candidates’ election probability, implying that while issues still matter to voters, the candidates’ personal characteristics do not.

Generally, our study contributes to research in at least two ways: First, our analysis further expands current literature on the effect of gender stereotypes on vote choice by investigating this relationship in another multi-party system, while most of today’s research still focuses on the two-party system of the U.S. Testing the effect in the democratically-unique Swiss multi-party system helps to further generalize the results on the effect of gender stereotypes, even apart from Finland. Second, our study evaluates real vote choices from actual elections instead of experimental settings, which remains an exception in today’s research on the topic at hand. Thus, we are able to investigate the effect of gender stereotypes in a real, complex setting, which could also help explain our findings.

The article is structured as follows: The next two sections review the existing literature on gender stereotypes and their effects on vote choice and develop our hypotheses. The following section describes the data, variables and statistical methods we employ. We then present and discuss our main empirical results. Finally, we summarize our findings, discuss their implications, and suggest areas where further research is needed.

Stereotypes

As humans, we all tend to view the world in a schematic, categorized way. It is people’s inferences and memories that guide their information-processing. These generic knowledge structures, in turn, drive people’s evaluations of and reactions toward others (Macrae and Bodenhausen 2001). One of these generic knowledge structures that guide people’s information-processing are stereotypes. They are commonly defined as automatically activated, general expectations about members of particular social groups, usually accompanied by a range of emotions, that can influence behavior (Amodio 2014; Bottom and Kong 2012; Fiske and Neuberg 1990; Macrae and Bodenhausen 2001). Or as Walter Lippmann ([1922] 1998) aptly stated: “For the most part we do not first see, and then define, we define first and then see.” Stereotypes thus influence knowledge processing in the context of social groups. Certain features are seen as typical for specific social groups, which means that the people holding these stereotypes exaggerate disparities between groups and underestimate variations within the group (Bordalo et al. 2016; Ellemers 2018; Kunda 1999). Unlike normal beliefs, stereotypes operate without conscious awareness (Amodio 2014; Kunda 1999; Macrae and Bodenhausen 2001). One contested presumption is that even low-prejudice persons experience this automatic activation of stereotypes.⁷ With a conscious effort, it is however still possible to suppress such stereotypical beliefs (Macrae and Bodenhausen 2001).

Generally, stereotypes are perceived as something negative, as they are deemed both personally and socially unacceptable (Amodio 2014). This does not necessarily have to be true since stereotypes can be helpful whenever quick estimates of someone’s probable behavior and abilities are needed, and they are also used as memory retrieval cues (Ellemers 2018; Fiske and Neuberg 1990). In this way, stemming from a mechanism of survival, these cognitive systems help structure the complex physical world (Amodio 2014; Fiske and Neuberg 1990; Lippmann [1922] 1998). Humans like to solve problems easily, which works best if they are put into a category and if this category is then used to prejudge a solution (Allport 1954). Whenever people are cognitively busy, they especially tend to fall back on stereotypes (Fiske and Neuberg 1990; Gilbert and Hixon 1991; Hayes 2011). Stereotypes therefore provide a maximum amount of information with minimal cognitive effort (Gilbert and Hixon 1991). While stereotypes can sometimes be accurate (Bordalo et al. 2016), most of the time, however, stereotypes oversimplify matters and lead to unjust assumptions and even inequalities (Ellemers 2018; Hastie 2016; Sanbonmatsu 2003).

This tendency to generalize perceived differences between groups and form stereotypical ideas about their characteristics and abilities is known for various social groups or categories such as race and gender (Bordalo et al. 2016; Ellemers 2018). Gender is a primary feature of perception. People immediately ascribe a set of characteristics to men and women, which contributes to the formation and persistence of gender stereotypes.⁸ In general, two types of political gender stereotypes can be distinguished: stereotypes regarding traits and stereotypes regarding issues. These are different for men and women (Huddy and Terkildsen 1993; Sanbonmatsu 2003; Turska-Kawa and Olszanecka-Marmola 2018). Trait stereotypes are expectations about characteristics that male or female politicians embody. While male politicians are stereotypically viewed as more rational, combative, assertive, and more ambitious than female politicians, women politicians are in turn evaluated as more sensitive, empathetic, warm, caring, compassionate, and consensus-building than their male counterparts (Dolan 2010, 2014b; Dolan and Lynch 2016; Haines, Deaux, and Lofaro 2016; Hastie 2016; Huddy and Terkildsen 1993; Johns and Shephard 2007; McDermott 1998; Schneider and Bos 2014; Turska-Kawa and Olszanecka-Marmola 2018). Issue or policy stereotypes, on the other hand, describe policy issues that are expected to be owned by either male or female politicians. While male politicians are deemed more competent in dealing with criminal matters as well as foreign and economic affairs, female politicians are deemed superior in the domains of education and health care (Dolan 2010, 2014b; Dolan and Lynch 2016; Huddy and Terkildsen 1993; Sanbonmatsu 2002, 2003). These expectations about the characteristics and issues of male and female politicians can in turn influence their chance of being elected.

Gender stereotypes and vote choice

Since women in politics stay alarmingly underrepresented in most countries, a thorough study into the causes seems pivotal. Scholars have long looked into the matter and found that aspects such as poor candidate quality or scarce campaign resources cannot explain female underrepresentation (Dolan 2010). Recent analyses of voter attitudes toward women candidates on the other hand present a rather different picture and suggest influences of gender stereotypes on vote choice (e.g., Dolan 2010, 2014a, 2014b; Dolan and Lynch 2016; Fox and Oxley 2003; Johns and Shephard 2007; Knuckey 2018; Lawless 2004; McDermott 1998; Sanbonmatsu 2002). Gender information can be conveyed through stereotypes, such as simply reading a candidate’s name on a ballot and immediately deriving world knowledge about social roles or personality traits of men and women (Canal, Garnham, and Oakhill 2015; Corbett 1991; Kunda 1999). As Samuel Popkin described back in 1991, “voters do reason about parties, candidates, and issues” (Kinsey and Popkin 1993, 569), with the reasoning being known as “low-information rationality.” This involves the use of information shortcuts such as gender or party stereotypes, which voters acquire knowledge about in everyday life and apply it to political decisions. Due to limited resources, people do not absorb all available information and, to make up for this incompleteness, they use previous experiences and stereotypes (Kinsey and Popkin 1993).

As male trait stereotypes, such as assertiveness or ambition, are typically in line with agentic roles and qualities of a good leader, people holding male trait stereotypes are less likely to vote for a woman candidate, as they might see men as a better fit for the agentic role of a politician. The same holds true for female trait stereotypes such as warmth and compassion, which are in line with communal roles and contrast classical leadership qualities. People holding female trait stereotypes perceive women as unfit for roles requiring agentic traits and will not vote for them (Bauer 2015; Dolan 2014a; Huddy and Terkildsen 1993; Koenig et al. 2011; Lawless 2004; Richard and Smith 1998; Simon and Hoyt 2008). On the other hand and in accordance with the trustee model of representation, which states that elected political representatives act in the interest of a nation and according to civic nature (Rehfeld 2009), female stereotypes might prove beneficial for women candidates since women are typically seen as more consensus-oriented and receptive toward the will and needs of the electorate. Therefore, voters might also see women as better politicians when the voters are holding female trait stereotypes (Brooks 2013; Dolan 2014b; Dolan and Lynch 2016; Schwarz and Coppock 2022). To sum up, people holding male trait stereotypes are expected to have a lower probability of voting for a woman candidate. For people expressing female trait stereotypes, both negative and positive effects on women candidates’ election chances are cogitable. While some research found the proclaimed negative effects of male and female trait stereotypes (Huddy and Capelos 2002; Lawless 2004; Lefkofridi, Giger, and Holli 2019), other studies found positive or contradicting effects for trait stereotypes (Fridkin and Kenney 2009; McDermott 1998). From this, we derive our first set of hypotheses⁹:

H1:

Holding male trait stereotypes negatively correlates with voting for a woman candidate.

H2a:

Holding female trait stereotypes negatively correlates with voting for a woman candidate.

H2b:

Holding female trait stereotypes positively correlates with voting for a woman candidate.

While trait stereotypes can be claimed to operate homogeneously at all state levels, the effect of issue stereotypes heavily depends on whether the election is at the national or subnational level. As the stereotypically male policy areas are usually located at the national level and stereotypically female policy areas are usually located at state or subnational levels (Bauer 2015; Dolan and Lynch 2016), voters demand more issue competency in stereotypically male issue areas the higher a political office is (Huddy and Terkildsen 1993). While economic or defense policy are stereotypically viewed as male issue areas, they are also more often responsibilities of the national level, in Switzerland, but also in other countries (see Vatter 2020). Social policy and health policy, stereotypically female issue areas, are on the other hand usually located at the subnational level (Vatter 2020). Therefore, voters holding stronger gender issue stereotypes can be expected to vote for female candidates at the subnational level and male candidates at the national level. Hence, gender stereotypes might benefit or disadvantage women candidates, depending on the level of the election. This suggestion is confirmed by the finding that women either seek office more often on subnational levels or get elected into these subnational positions more often than into national level ones (Fox and Oxley 2003; Huddy and Terkildsen 1993). Some research is able to at least partially replicate the negative effect of male policy stereotypes on vote choice for women candidates at the national level (Dolan 2010, 2014b; Dolan and Lynch 2016; Lefkofridi, Giger, and Holli 2019) and even suggests that it is the most important gender stereotype for vote choice (Dolan 2010). Other research, however, finds positive or contradicting effects of male and female issue stereotypes (Dolan 2010, 2014b; Dolan and Lynch 2016; Fridkin and Kenney 2009; Sanbonmatsu 2002; Sanbonmatsu and Dolan 2009). Our last hypotheses can therefore be summarized as follows:

H3:

Holding male issue stereotypes negatively correlates with voting for a woman candidate on the national level.

H4:

Holding female issue stereotypes negatively correlates with voting for a woman candidate on the national level.

In addition to these linear effects, we investigate potential group effects by calculating interactions since it might be plausible that gender stereotypes do not influence vote choice for all population groups identically. We interact gender stereotypes with various characteristics of the voters: Earlier studies indicate that stereotypes only affect male voters (Johns and Shephard 2007; Ono and Burden 2019), while there is some more recent evidence that gender stereotypes mainly affect female voters (Schwarz and Coppock 2022). Therefore, we start by interacting gender stereotypes with voters’ gender. Regarding age, it could be expected that the effect of stereotypes is stronger for older generations and diminishes for younger generations (e.g., Devroe 2021). For education, it seems likely that the effect of stereotypes weakens with increasing education, as individuals reflect more on their own stereotypes (Devroe 2021). Existing literature also finds varying effects of gender stereotypes on vote choice for voters’ self-placement on the political left-right-scale (see Dolan 2014b; Dolan and Lynch 2016; McDermott 1998; Schwarz and Coppock 2022). Since stereotypes can be seen as a heuristic in candidate evaluation and their usage should therefore depend on other information (McDermott 1998), we also include an interaction of gender stereotypes with respondents’ political interest and political knowledge. Lastly, it could also be expected that the effect of gender stereotypes depends not only on characteristics of the voters but on characteristics of the candidates, too. Some studies suggest for example that gender stereotypes affect support for female Republican and Democratic candidates differently (Bauer 2018; Sanbonmatsu and Dolan 2009). In a last step, we therefore interact gender stereotypes with the left-right position of the candidate’s political party.

Research design, variables, and method

The case of Switzerland

With Switzerland being one of the last Western countries to implement women’s suffrage in 1971 and thus finally allowing all citizens aged 18 years and above to vote at the national level (see Vatter 2020), experience with women in political offices is still rather sparse. Following Jenning’s (2006) exposure theory, due to varying sex role socialization, it is suggested that perceptions of female politicians between societies with a long history of women in key political positions opposite to societies with a short history, such as Switzerland, differ. As the uneven gender distribution in political offices portrays (see Federal Statistical Office 2024; hereafter FSO), voters in this country are still getting used to having women in politics and thus, differences in sex role perceptions might persist. Despite the late start of women’s political representation and participation in 1971, over time, female representation in Swiss political offices has risen: While the first woman only became part of the executive, the Federal Council, in 1984, in 2019, three out of seven members of the Federal Council were female (in 2011 they were even briefly in the majority with four out of seven members; there were three women in 2023). The legislative, the Federal Assembly, consists of two chambers, the National Council and the Council of States. In the former, 42% of elected representatives in 2019 were women and in the latter, 26.1% women were featured (see FSO 2024).¹⁰ This is indicative of the fact that the Council of States remains a male-dominated domain and, as a result, gender stereotypes may still be more prevalent there as compared to the National Council where voters are more accustomed to female representation.

In our analyses, we target the Council of States. This is because the 200 seats of the National Council are distributed proportionally to cantonal size and in one canton alone up to 35 politicians can be elected. Party lists are commonly used for National Council elections, so gender might not be the primary electoral criterion. Moreover, with so many seats in some cantons, the elections are not very personalized and are more about the political party than about individual candidates. This is also shown by the fact that 43% of respondents in the dataset cast an unchanged party list. A possible gender effect here is therefore more due to the composition of the party lists than to the voters’ individual stereotypes. The Council of States, on the other hand, only has 46 seats with each canton having exactly two seats and each half-canton one seat. Elections for these two seats are held in a first-past-the-post election system,¹¹ which makes these elections much more personalized (Federal Chancellery 2024; Vatter 2020). The voters’ electoral decision therefore probably depends not only on the party – especially since voters in most cantons are eligible to elect two candidates but most parties only put up one candidate per canton – but also on characteristics of the candidates such as their gender. In combination with the aforementioned fact that female representation in the Council of States remains low, hence making gender stereotypes more probable, this makes the Council of States a most likely case and is therefore best suited for our research design.

As previously described, Switzerland has a multi-party system with many political parties which makes an influence of gender stereotypes on vote choice more likely. Different parties that are politically close to each other put up their own candidates and many candidates also run without party affiliation. It is easier to vote or not to vote for the candidate of Party A because of their gender if Party B makes similar political content and puts up a candidate of a different gender. With an effective party number of approximately 6.2 in the National Council after the 2019 parliamentary elections, Swiss voters had a large number of parties to choose from in international comparison.¹² Accordingly, it is plausible to assume that they would choose between different candidates from similar or no parties on the basis of the characteristics of the candidates themselves, such as their gender.

Data and variables

We analyze the Panel Survey of the Swiss Election Study (Selects) from 2019 (Selects 2022).¹³ Apart from gender stereotypes, the survey also asks respondents about their vote choice in the elections to the Federal Assembly in October 2019. With this data, our focus lies on national level election results. Due to the aforementioned reasons, we opt for vote choice in the first round of the election to the Council of States of 2019 as our dependent variable. However, we encountered the problem that in four out of 24 cantons with a majoritarian system, no relevant female candidates were nominated at all and that these cantons consequently have to be excluded from our analyses.¹⁴ It is a general observation in single member districts that parties tend to nominate more male than female candidates, as they are perceived to have better chances of getting elected (Matland and Studlar 1996). Figure A1 in the appendix displays significantly more male than female candidates for the Council of States, which once again underlines the relevance of our study. The Green Party and the Green Liberal Party are the only ones with more female than male candidates. This could mean that voters did not always have a female alternative that represented their political position, even though this is difficult to examine, as there were also a large number of candidates without party affiliation. Nevertheless, this needs to be kept in mind when interpreting our results.

Our dependent variable is the stated vote choice, for which we code the elected candidate either as male (0) or as female (1).¹⁵ Out of the 178 candidates in the survey, we coded 62 as female and 106 as male candidates. Out of the elected 42 candidates, 10 were coded as female and the remaining 32 as male. As most of the citizens were allowed to elect two candidates, we split them up into two separate observations and clustered the observations in the regression by respondent ID. Lastly, we keep the number of observations constant throughout our regression models. This left us with around 4,861 observations from about 2,783 different respondents.

Our independent variables consist of gender stereotypes. We hereby distinguish between trait and issue stereotypes as well as male and female stereotypes.¹⁶ In our data set, there are four items measuring trait stereotypes and five items measuring issue stereotypes. The items measuring trait stereotypes asked respondents whether they thought that women or men in politics were more likely to possess the following traits or whether there was no difference between men and women: ambition, ability to build consensus, decisiveness, and compassion. The items measuring issue stereotypes asked whether a man or a woman politician in the Federal Council was better suited to handle the following policy areas or whether it made no difference: defense and security, the economy, social policy, migration and asylum policy, as well as gender equality policy. Drawing on the stereotypical traits and issue areas identified by the literature (Dolan 2010, 2014a, 2014b; Fridkin, Kenney, and Serignese Woodall 2009; Haines, Deaux, and Lofaro 2016; Hastie 2016; Huddy and Terkildsen 1993; Johns and Shephard 2007; McDermott 1998; Sanbonmatsu and Dolan 2009; Schneider and Bos 2014; Turska-Kawa and Olszanecka-Marmola 2018) we code ambition and decisiveness as male trait stereotypes and compassion as well as the ability to build consensus as female trait stereotypes. Regarding issue stereotypes, we code defense and the economy as male and social and gender equality policy as female issue stereotypes. We do not use migration policy to code issue stereotypes.¹⁷ We coded each of the eight items as either 0 (respondent did not give a stereotypical answer, i.e. “no difference” or “man” in the case of a female stereotype, or “no difference” or “woman” in the case of a male stereotype) or 1 (respondent gave a stereotypical answer).¹⁸ Following the coding in previous studies (Dolan 2014b), we then build an additive index for each of the four stereotype-categories ranging from 0 to 2, indicating the number of stereotypical answers given.

Finally, we employ some control variables, namely the respondents’ gender, age, education, income, their position on the left-right-scale, their political interest, whether the person voted for was an incumbent, the political left-right placement of the candidate’s party, as well as the canton the respondent voted in.¹⁹ For a more detailed overview of our measurement of the control variables, see Table A4 in the appendix. Methodologically, we employ logistic regression with the elected candidates’ gender as the dependent binary variable. We cluster the observations by respondent ID, as most respondents were allowed to elect two candidates. In our empirical analysis, we first calculate a separate logistic model for each of the four stereotypes, before calculating a full model with all four stereotypes simultaneously.

Results

Before turning to our regression models, we first briefly discuss the distribution of gender stereotypes in Switzerland. Figure 1 depicts the distribution of gender stereotypes for all four categories: male trait, female trait, male issue, and female issue stereotypes. First, it is noteworthy that in two of the four categories, namely male stereotypes, a majority of respondents do not exhibit any gender stereotypes. Nevertheless, depending on the type of gender stereotype, between 45 and 65% of respondents exhibit at least one stereotypical idea about the characteristics and abilities of men and women. Female trait stereotypes are particularly prevalent. Approximately one-third of the respondents exhibited the maximum number of stereotypes in this category. Regarding differences by the gender of the voter (see Figure A5 in the appendix), we find that women show a much smaller number of male issue stereotypes and somewhat less male trait stereotypes. Their female trait and issue stereotypes, however, are just as pronounced as they are for male voters. So, while not all Swiss people hold stereotypical ideas about men and women in politics anymore, there is still a considerable amount of gender stereotyping embedded in people’s minds – in men’s as well as in women’s. Additionally, these descriptive findings are a good indication that many respondents did not answer in a socially desirable way, as a significant amount of gender stereotypes were still expressed.

Figure 1. Distribution of Gender Stereotypes in Switzerland.

Source: Selects (2022), own calculations and depiction.

Next, we begin with empty regression models that encompass only the four stereotypes and cantonal fixed effects, but no control variables. The results of these regression models can be found in Table A6 in the appendix. They show that only male issue stereotypes affect the probability of voting for a woman candidate. People who hold male issue stereotypes and thus believe that male politicians are better suited to handle policy areas such as defense and security or the economy are less likely to vote for a woman candidate in the Swiss elections to the Council of States (p < 0.001).

In a second step, we add control variables. The results of our logistic regression analyses in Table 1 show how these gender stereotypes relate to vote choice when controlling for gender, age, education, income, the political position and political interest of the respondents, as well as for the incumbency status and party position of the candidates. Columns 1 to 4 show that none of the four types of stereotypes exhibit any effect on the probability of voting for a female candidate without controlling for the other kinds of stereotypes. In the full model in column 5 that includes all four stereotypes simultaneously, we find no effect of neither male nor female trait stereotypes. Stereotypes about typically male or female traits hence do not seem to impact the likelihood of electing a woman candidate, which means that we have to reject hypothesis H1, H2a and H2b.

Table 1. Logistic Regression Models.

	(1)	(2)	(3)	(4)	(5)
Male trait stereotype	-0.036 (0.053)				-0.026 (0.063)
Female trait stereotype		-0.017 (0.044)			-0.030 (0.055)
Male issue stereotype			-0.070 (0.051)		-0.129* (0.058)
Female issue stereotype				0.051 (0.045)	0.126* (0.055)
Gender (female)	0.154* (0.075)	0.156* (0.075)	0.142 (0.076)	0.157* (0.075)	0.132 (0.076)
Age	-0.000 (0.002)	-0.000 (0.002)	-0.001 (0.002)	-0.001 (0.002)	-0.001 (0.002)
Secondary education	0.005 (0.094)	0.006 (0.095)	0.003 (0.095)	0.003 (0.095)	0.001 (0.095)
Tertiary education	0.173 (0.101)	0.172 (0.101)	0.166 (0.101)	0.175 (0.101)	0.171 (0.101)
Income	0.021* (0.010)	0.021* (0.010)	0.020* (0.010)	0.021* (0.010)	0.020* (0.010)
Left-right self-placement	-0.002 (0.019)	-0.002 (0.019)	0.002 (0.020)	-0.001 (0.019)	0.008 (0.020)
Political interest	-0.058 (0.060)	-0.058 (0.060)	-0.059 (0.060)	-0.059 (0.060)	-0.059 (0.060)
Incumbency	-5.776*** (0.228)	-5.777*** (0.229)	-5.775*** (0.228)	-5.777*** (0.229)	-5.787*** (0.229)
Position of candidate’s party	-0.044*** (0.002)	-0.044*** (0.002)	-0.044*** (0.002)	-0.044*** (0.002)	-0.044*** (0.002)
Constant	0.836** (0.256)	0.833** (0.255)	0.876*** (0.258)	0.784** (0.256)	0.854*** (0.259)
N	4861	4861	4861	4861	4861

Note: For education, the baseline category is “primary education.” Regarding the position of the candidate’s party, a lower value means a position further to the left and a higher value means a position further to the right. Models are estimated with cantonal fixed effects. Standard errors in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001. Source: Selects (2022), Lehmann et al. (2022), own calculations.

However, in the full model, when including all four stereotypes simultaneously, we find a negative effect of male issue stereotypes and a positive effect of female issue stereotypes on the probability of electing a woman candidate to the Council of States (both p < 0.05). The fact that these two types of stereotypes are positively correlated could explain why their opposing effects only become visible once both are included in a model, while they remain masked without controlling for all stereotypes. Regarding the effect size, holding 2 instead of 0 male issue stereotypes decreases the probability of electing a woman candidate by about 2.9 percentage points, while holding 2 instead of 0 female issue stereotypes increases the likelihood of electing a woman candidate by about 2.8 percentage points. The finding on male issue stereotypes is in accordance with our hypothesis H3 and means that people with more stereotypical ideas about male political issue areas are less likely to vote for a woman candidate in the election to the Council of States. This is also in line with current literature (Bauer 2015; Dolan and Lynch 2016), which proclaims that stereotypically male issue areas are more often located at the national level, while stereotypically female issue areas are usually located at the subnational level. Thus, holding issue stereotypes increases the likelihood of voting for a woman candidate at the subnational level and decreases it at the national level. In Switzerland, defense policy and economic matters are responsibilities of the state, while social policy (including health policy) tends to be located at the cantonal level (see Vatter 2020). People holding male issue stereotypes thus consider men more capable of dealing with issues at the national level in Switzerland. As the Council of States is also located at the national level, these people are more likely to vote for male candidates. However, these considerations are not in line with our finding on female issue stereotypes, for which we found a positive effect of electing a woman candidate to the Council of States. Hence, we find the opposite of what we expected in hypothesis H4. In this respect, it may be that respondents also view gender equality and social policy as a national issue and are more likely to vote for women if they believe that they are particularly good at dealing with these policy areas. Or that they would like to see women who are more fit for dealing with regional issues also in national political positions. In conclusion, we can confirm our hypothesis H3 on male issue stereotypes but have to reject our hypothesis H4 on female issue stereotypes as well as the two hypotheses H1 and H2 on trait stereotypes.

Lastly, it might also seem plausible to expect male and female stereotypes to reinforce one another and the effects to be stronger when someone holds both, female and male, instead of only one stereotype. To test this, in appendix A7, we interact female trait stereotypes with male trait stereotypes and female issue stereotypes with male issue stereotypes. In both cases, the interaction coefficient is not statistically significant. Hence, male and female stereotypes seem to operate independently of one another and neither reinforce nor diminish each other’s effect.

Interactions

In a third step, we interact the four stereotypes with different characteristics of the voters to investigate whether the effect of gender stereotypes on vote choice differs among social groups. Table A8 in the appendix lists the results of all interaction models. For the respondent’s gender, their education, their left-right self-placement, and their level of political interest we do not find any statistically significant interaction terms. However, we find statistically significant interactions with the respondent’s age, their level of political knowledge, and the left-right placement of the candidate’s political party. We will now briefly discuss the five statistically significant interaction coefficients we find. The graphs displaying the marginal effects of the stereotypes along their moderators can be found in Figures A9 to A13 in the appendix.

First, we find three significant interactions between the respondents’ age and their stereotypes. The first interaction term concerns male trait stereotypes (p < 0.01). In the previous regression models, we did not find any effect of male trait stereotypes on the probability of electing a woman candidate. However, the interaction model shows that there are opposing effects of male trait stereotypes for young and for old voters. While for young voters we find a statistically significant negative effect of male trait stereotypes on electing a woman candidate, for old voters the effect even becomes significantly positive. Regarding substantiality, moving from 0 to 2 stereotypes decreases the probability of electing a woman candidate by about 10.2 percentage points for an 18-year-old voter, while it increases the likelihood by around 7.3 percentage points for an 88-year-old voter. Second, we also find a significant interaction term between age and female trait stereotypes (p < 0.05) for which we did not find any effect in previous models either. Here, the effect is reversed to male trait stereotypes. Young voters thus exhibit a significant positive effect of female trait stereotypes on the likelihood of electing a woman candidate (the likelihood increases by about 6.8 percentage points when moving from 0 to 2 stereotypes for an 18-year-old), while older voters experience a significant negative effect (the probability of electing a woman candidate decreases by about 6.3 percentage points when moving from 0 to 2 stereotypes for an 88-year-old voter). Perhaps older people do not deem assertiveness or ambition traits that are well suited for politicians and therefore turn to women candidates when they believe that male politicians are more prone to having these characteristics. On the other hand, older people also refrain from electing women candidates when they deem women consensus-oriented and compassionate. Lastly, we also interact female issue stereotypes with the age of the voter (p < 0.05). Our interaction model shows that the previously found positive effect only holds true for voters that are around 55 years or older, while there is no relationship between female issue stereotypes and voting for a woman candidate for younger voters.

Second, we also find a significant interaction term between a voter’s level of political knowledge and male trait stereotypes (p < 0.05). We find that male trait stereotypes exert a negative effect on electing a woman candidate, but only for respondents low in political knowledge. For people who answered no question on political knowledge correctly, the probability of electing a woman candidate drops by about 9.2 percentage points when they hold 2 instead of 0 male trait stereotypes. This finding is consistent with the notion that gender stereotypes generally function as heuristics. Those with more political knowledge rely less on such heuristics, so the effect of gender stereotypes disappears with increasing political knowledge.

Lastly, we find a statistically significant moderation by the left-right placement of the candidate’s political party on the effect of male issue stereotypes on the likelihood of electing a woman candidate (p < 0.001). For candidates that run for a party on the political left, male issue stereotypes of the voters have a positive effect on electing a woman candidate. For candidates from the party furthest to the left in our dataset (Ensemble à Gauche), having 2 instead of 0 male issue stereotypes increases the voters’ likelihood to elect a woman candidate by about 7.9 percentage points. Regarding candidates on the very right of the spectrum (Swiss People’s Party), on the other hand, holding 2 instead of 0 male issue stereotypes decreases the probability of electing a woman candidate by 13.8 percentage points. Thus, while we find a negative effect of male issue stereotypes on the probability of electing a woman candidate in our full sample, we can now conclude that this effect only holds when it is about candidates on the right of the political spectrum, while women candidates on the political left even profit from male issue stereotypes.

We conclude that the effect of gender stereotypes, even though fairly stable among different genders, educational levels, left-right self-placement and levels of political interest, varies along a respondent’s age, their level of political knowledge as well as the candidate’s political position. With regard to our main effects of male and female issue stereotypes, we were able to refine these findings and show that the effect of female issue stereotypes only holds true for older voters, while it is not statistically significant for young voters. Regarding our main effect of male issue stereotypes, we find that it only holds for candidates on the political right, while it is even reversed for candidates on the political left. Lastly, we found some effects of male trait stereotypes and female trait stereotypes for certain age groups and levels of political knowledge. The age of the voters seems to be especially decisive for the effect gender stereotypes have on the vote choice for women candidates, as some stereotypes even operate in reverse order depending on the voter’s age. However, our cross-sectional design cannot clarify whether this is a generational effect or a life-cycle effect. Further analyses with panel data are necessary for this.

Conclusion

Drawing on individual-level data from the Swiss Election Study (Selects) from 2019, this article explores the linkage between gender stereotypes and vote choice for women candidates in a multi-party system. While the effect of gender stereotypes on vote choice for women candidates has often been studied in two-party systems, mostly in the context of the U.S. how this relationship behaves in multi-party systems has, apart from Finland (see Lefkofridi, Giger, and Holli 2019), not been analyzed comprehensively before. However, a stronger relationship could be assumed here, since the party is less deterministic in influencing vote choice due to the larger selection of politically similar parties in a multi-party system.

We test four types of gender stereotypes and find that there is a significant negative effect of male issue stereotypes on vote choice for women candidates, while there is a significant positive effect of female issue stereotypes on electing women candidates. However, both these effects only hold true when controlling for the other three stereotypes. Thus, gender stereotypes are not per se bad for women candidates. Second, we tested whether the effects of these gender stereotypes vary along a number of characteristics of voters as well as candidates. We find that the negative effect of male issue stereotypes only holds for candidates with a political position on the right of the political spectrum, while the effect even becomes positive for candidates on the left. We also find that the positive effect of female issue stereotypes only holds for older voters. Additionally, while we did not identify a direct effect of trait stereotypes on vote choice, we did however find some effects of trait stereotypes for certain subgroups. We find that young voters holding male trait stereotypes are less likely to elect a woman candidate while for older voters, male trait stereotypes have a positive impact on electing women candidates. The reverse is true for female trait stereotypes. Lastly, we find that the negative effect of male trait stereotypes on electing women candidates only holds for voters with a lower level of political knowledge. These analyses lead us to three broader conclusions. First, our results show that political gender stereotypes are still widespread even more than 50 years after political gender equality was achieved in Switzerland. Second, our study shows that not all gender stereotypes necessarily reduce electoral chances of women candidates, but that they have a relevant influence on them. Third, we show that the effects of gender stereotypes vary across the socio-demographic characteristics of voters, especially age, as well as across the political position of the candidates. Therefore, no generalized conclusions should be drawn about the effects of gender stereotypes.

This article does not imply a causal interpretation of the relationship between gender stereotypes and vote choice, as we only use cross-sectional data. Employing time series data would have the further advantage that with the many interactions between gender stereotypes and age, it could be investigated whether these effects are life-cycle effects or whether some generational shift is currently happening. Additionally, we draw on data from a panel study and results should be interpreted carefully due to panel attrition. Apart from Lefkofridi, Giger, and Holli (2019), this is one of the first studies we know of to investigate the relationship between gender stereotypes and vote choice in a multi-party system; hence, further research is needed to generalize our findings. Switzerland was one of the last countries in Europe to adopt the right to vote for women and it is therefore plausible to assume that gender stereotypes are still more prevalent in Switzerland compared to other European countries. Compared to the National Council, the Council of States also posits a most likely case due to their low female representation and the concomitant higher likelihood of gender stereotypes there. Accordingly, as we only investigated the Council of States, our findings are limited to this part of the Swiss legislative. As both, Switzerland and the Council of States are a most likely case, our coefficients are probably a more liberal estimate of the effect of stereotypes on vote choice. Furthermore, the national election of Switzerland in 2019 was special in the sense that more female candidates stood for election and voters also elected more women than previously (see Giger et al. 2022). Hence, it remains open to what extent our results are generalizable both time- and context-wise. We thus suggest that further research investigates the relationship between gender stereotypes and vote choice in other European multi-party systems and on the subnational level, too.

Despite these shortcomings, our paper is relevant from both a societal as well as from a scientific point of view. From a scientific perspective, this is one of the first studies to look at the effect of gender stereotypes outside of a two-party system. Additionally, we use data from real-world elections instead of experimental data and show that even in the complex, high-information setting of a real election, gender stereotypes still matter. Socially, this research is relevant because gender stereotypes are still prevalent – our analyses show that 45 to 65% of respondents hold at least one stereotypical idea about the abilities and characteristics of men and women – and it is therefore important to know whether and how they influence the electoral chances of women candidates. Although gender stereotypes do not necessarily diminish women’s electoral chances, a societal discourse on these issues seems necessary. Our study is intended to provide the impetus to revisit the issue of gender stereotypes in politics and to give more thought to them and their effects.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental data

Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/1554477X.2024.2338587

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Notes on contributors

Sophie Ruprecht

Sophie Ruprecht is a PhD candidate and research assistant at the Institute of Political Science, University of Bern, Switzerland. Before, she studied political science at the University of Bern, which earned her a master’s degree in Swiss and comparative political science from the University of Bern. Her main research interests concern gender stereotypes, voting behavior, social acceptance of renewable energy, energy politics, communication and information effects.

Alina Zumbrunn

Alina Zumbrunn is a PhD candidate and research assistant at the Institute of Political Science, University of Bern, Switzerland. Before, she studied political science at the University of Bern and the Central European University in Budapest, which earned her a master’s degree in political science from the University of Bern. Her main research interests concern gender stereotypes, voting behavior, the rural-urban divide, political sociology and political psychology.

Notes

1. Authors are listed alphabetically, as they contributed equally.

2. For a more thorough analysis of Clinton’s candidacy and gender effects in politics, we recommend Manne (2018) and Sorrentino, Augoustinos, and LeCouteur (2022).

3. While we are fully aware that there are more than two genders, we hereafter stick with the binary classification of gender, “male” and “female”. This simplification stems from the fact that voters generally hold stereotypes about female and male politicians, and that they classify the politicians’ names and images on the electoral lists accordingly into these two groups.

4. The Finnish study of Lefkofridi, Giger, and Holli (2019) poses a notable exception, as the authors evaluated both experimental and real-world data in another European multi-party system.

5. Despite the lack of extensive research on the influence of gender stereotypes on vote choice outside of two-party systems and experimental settings, our study leans on Lefkofridi, Giger, and Holli (2019) who first ventured into this new setting by using real-world data from 2011 in Finland, another multi-party system. We however depart from their study in three key aspects: By looking at Switzerland, we investigate a most likely case, whereas Finland can be seen as a “least likely” case due to its higher degree of political equality. This enables the coverage of a wider range on the spectrum of multi-party systems, allowing for more generalizable statements. Second, our data is from the 2019 national elections, hence being able to provide more current knowledge. And third, apart from including only issue stereotypes, we also examine trait stereotypes.

6. In total, 39% of Swiss federal parliamentarians are women, compared to a European average of 30% (UN Women 2020).

7. Bauer (2015) for example concluded that especially for female stereotypes, automatic stereotype activation is not the case.

8. Female politicians are judged differently than women, as they are usually seen as a subtype of women in general. This leads them to suffer from the so-called “double bind” (see Costa 2021): the need to fulfill two opposing stereotypes separately. On the one hand, they should fulfill feminine stereotypes, on the other hand, they should be competent – a trait that is stereotypically associated with masculinity instead of femininity. The difficulty of this balancing act is also reflected in empirical evidence which shows that female politicians are neither ascribed stereotypically male qualities nor that they can benefit from stereotypically female qualities (Schneider and Bos 2014). However, this picture does not emerge in our data: Figure 1 later in the text shows that female politicians are attributed both, more stereotypical qualities and more stereotypes about policy fields than male politicians.

9. While H3 and H4 theoretically stress the fact to be different depending on the level of the political office, H1 and H2 should hold irrespective of the level of political office. In the course of this study, we are however only capable of testing effects on the national level, therefore, all hypotheses aim at national-level effects.

10. In international comparison, female representation in the Council of States is thus still below the European average of 29.1% in upper chambers (UN Women 2020).

11. As the Cantons of Jura and Neuchâtel elect their representatives for the Council of States in a proportional representation system, their elections focus more on the parties than on the candidates. Therefore, we decided to exclude observations from these two cantons from our analyses.

12. Own calculations based on data from the Federal Statistical Office (FSO) and the calculation method according to Ladner (2004, 67–96).

13. The panel consists of three waves, which were conducted from May to December 2019, starting off with about 7,900 respondents in the first wave and ending with around 4,700 respondents in the last wave. We employ variables from all three waves.

14. These are the cantons Appenzell Ausserrhoden, Glarus, the Canton of Solothurn and Schaffhausen.

15. We do this manually and based on our own evaluation of the candidates’ name, as this is the way the voters most likely assessed the candidates’ gender, too. We acknowledge that our assessment of a candidate’s gender might not be in line with their self-identification. However, stereotypes are activated based on a voter’s perception of the candidates and not based on the candidates’ self-identification. This is why we opt for this operationalization.

16. These stereotypes correlate to some extent with correlations between 0.246 (male issue stereotypes and female trait stereotypes) and 0.5355 (male trait stereotypes and female trait stereotypes). Nevertheless, stereotypes can be considered as separate variables with these levels of correlation.

17. In the literature, migration is typically coded as a male policy stereotype (Dolan 2014a, 2014b; Dolan and Lynch 2016; Fridkin, Kenney, and Serignese Woodall 2009; Sanbonmatsu and Dolan 2009). Our empirical analysis however showed that 15.8% of respondents thought of migration as a female policy issue, while only 13.2% of respondents thought of it as a male policy issue. Therefore, we decided not to use migration for either of the two policy stereotypes. A separate analysis of migration stereotypes on vote choice for a woman candidate can be found in Table A2 in the appendix. We find that people who believe migration to be a male policy issue are less likely to elect a woman candidate, even though this effect diminishes significantly when adding the control variables. People who believe migration to be a female policy issue, on the other hand, are neither less nor more likely to elect a woman candidate.

18. We repeat our analysis with a stricter coding of gender stereotypes, in which only“no difference” counts as no gender stereotypes while respondents with inverse stereotypes (i.e. answering“man” for female stereotypes and“woman” for male stereotypes) were excluded from the analysis. Results (available upon request) remain substantially the same, however, in some cases the level of statistical significance diminishes slightly.

19. We code the political left-right placement of the party of the candidate using the Manifesto Project Database (Lehmann et al. 2022). However, this dataset only contains a coding of the 11 largest of a total of 31 parties to which the candidates in the survey belong. Due to the resulting missingness, the cantons of Basel-Stadt and the Canton of Schwyz are excluded from our data set. Together with the exclusion of Jura and Neuchâtel due to their electoral system as well as Appenzell Ausserrhoden, Glarus, the Canton of Solothurn and Schaffhausen due to their lack of female candidates, this leaves us with 18 cantons to analyze. To compare whether the excluded cantons are fundamentally different from the cantons that remain in our sample, Figure A3 in the appendix depicts the distribution of gender stereotypes in the excluded cantons. Comparing this graph with Figure 1 shows that gender stereotypes are distributed very similarly in the two groups, which means that the exclusion of these eight cantons is unlikely to systematically bias our results.