The end of the all-male party? Voter preferences for gender representation in political parties

ABSTRACT
 While parties have a large role in influencing the representation of women, much less is known of how voters perceive parties’ efforts to promote female candidates. Existing evidence from the literature suggests that, on an aggregate level, voters value female candidates at least to the same extent as male candidates. Meanwhile, evidence points both towards a general under-representation of women in politics as well as large differences between left-leaning and right-leaning parties when it comes to selecting female lead candidates and members of parliament (MPs). This study investigates voters’ preferences for gender representation inside political parties. Using a single vignette survey experiment in five European countries, I show that voters have strong preferences for equal descriptive representation of men and women in political parties and prefer women as lead candidates. Women have larger preferences for equal representation than men, while ideological differences are comparatively small.


Introduction
As one of the longest-serving female heads of government in the world, as well as the leader of the German Christian Democratic Party (CDU) between 2000 and 2018, Angela Merkel was the face of German conservative politics for almost two decades (Boussalis et al. 2021;Wiliarty 2010). However, the CDU has not experienced large increases in women's representation among members of parliament during that time, with the party never topping 25% women among their MPs. Meanwhile, parties across Europe have raised the share of women among their Members of Parliament considerably in recent decades. While a gender split between right-leaning and left-leaning parties exists in many countries such as between Labour and Conservatives in the United Kingdom, the SPD and CDU in Germany or the parties on the left and right in France in recent years, many right-leaning parties have recently also selected more and more female party leaders: The Front National in France, the CDU and AfD in Germany and the Conservatives in Britain have had women as party leaders in recent years. Overall, increases in women's representation occur unevenly across Europe, structured not only by political ideology and legislated quotas, but also by the kinds of positions women manage to reach.
This raises three fundamental questions: Does having women as members of parliament and in leadership positions affect the way voters perceive a party? In particular, is there a trade-off in the eyes of voters between a higher representation of women as parliamentarians and having a female lead candidate? Finally, do these effects hold for men and women or leftand right-leaning voters?
Parties have chosen different strategies to promote women inside the party: some make sure that female candidates are nominated in equal numbers or even in more electorally promising districts, like Labour in the UK (Nugent and Krook 2016;Wäckerle 2022). Others seem to have no intention of increasing the share of women among their candidates and MPs, but see female politicians reach powerful positions inside the party, like the AfD in Germany. France adopted a gender quota for the Assemblée nationale in 2000, which mandated parity among candidate by all parties, which ultimately significantly increased the share of women in parliament (Murray 2012). Finally, Spain introduced a legislative quota in 2007, although considerable increases in women's descriptive representation already took place before, due to high shares of female candidates among its left parties (Verge 2012).
Scholars have proposed different mechanisms that increase women's descriptive representation in political parties pointing to female leadership (Caul 2001;O'Brien et al. 2015), internal pressures (Valdini 2019) and challenger parties, which can force parties to adopt gender quotas (Weeks 2018). In this study, I propose an additional mechanism that could influence the nomination strategies of political parties: Voters' preferences for equal representation. While voters' gender preferences for individual politicians have been studied extensively (Schwarz and Coppock 2022), we lack evidence on how voters perceive political parties in terms of how they promote women as leaders and members of parliament.
I test the effect of female descriptive representation within a party on electoral support by running a vignette survey experiment in five European countries (France, Germany, Italy, Poland and Spain), covering a wide range of quota systems and historical evolution of women's representation. I show that voters indeed are more likely to express a likelihood to vote for parties with equal numbers of men and women among their MPs over imbalanced representation. Additionally, parties with a female lead candidate receive more support than those with a male lead candidate. These effects are larger for women. This study fills several research gaps: It extends findings on gender preferences of voters beyond the context of the US and UK, shows that gender preferences of voters are relevant for political parties and provides evidence that as parties nominate more women in recent years, party systems in Europe are becoming more congruent with their voters' preferences on women's representation. These findings provide an important missing piece for analysing parties' efforts to promote women and show that making sure parliamentary delegations are gender-balanced lies in the strategic interest of parties across the political spectrum.

Female representation within political parties
Political parties greatly influence the representation of women in politics. This process starts with recruiting potential political candidates. Already at that early stage, parties are more likely to approach and promote men than women Lawless 2004, 2010;Lawless and Fox 2010) and many parties have realized this step as an important barrier to the equal representation of women (Campbell, Childs, and Lovenduski 2006;Fortin-Rittberger and Rittberger 2015). Subsequently, parties are less likely to nominate women than men, a factor that is the key determinant of the share of women in parliaments . This nomination pattern is especially problematic because it gets reinforced as female candidates get nominated in less promising situations, such as constituencies the party is likely to lose (Ryan, Alexander Haslam, and Kulich 2010;Wäckerle 2022). Finally, even after reaching the parliament, women are less likely to be promoted to powerful positions in the party group (Barnes 2016;Barnes and O'Brien 2018) and less likely to get significant speaking time on the parliament floor (Bäck and Debus 2019;Fernandes, da Fonseca, and Won 2021).
Parties' efforts to focus on recruiting and nominating more women are often connected to external and internal pressure, such as changing electorates and the rise of new parties (McIlveen 2009;Weeks 2018). The promotion of women is often strategically tied to a calculation of costs (such as the replacement of and threat to male leadership) and benefits (such as reputation and international recognition) of including women in parties (Valdini 2019).
Additionally, researchers have pointed out a so-called "contagion" effect, which describes tactical adjustments made by parties to respond to competitors, for example through increasing the share of women among their candidates when other parties do so, too. These political processes can appear as macro-contagion effects on the party level or micro-contagion effects on the district level (Matland and Studlar 1996). Caul (2001) and Meier (2004) show how macro-contagion is an important driver of measures to increase women's representation. Wäckerle (2022) shows evidence of micro-contagion between constituencies for the Labour party and Borisyuk, Rallings, and Thrasher (2007) describe a cross-party micro-contagion effect on the local level in the UK. However, these effects are contingent on the salience of questions of equality among voters, as contagion occurs only when some parties act towards the representation of women and there is public pressure on other parties to follow suit (Cowell-Meyers 2017). In sum, parties' efforts to promote women as MPs are often strategic and connected to public pressures, rather than purely driven by the will to achieve parity between men and women.
Looking at party leadership, a substantial number of parties across the political spectrum have had female party leaders. O'Brien (2015) provides evidence that parties are most likely to first have female leaders in times of crisis, but are likely to stay in power if their party gains influence subsequently. While women might face tougher situations when they enter positions of power, they manage to have political careers similar to men (Escobar-Lemmon and Taylor-Robinson 2015; Jalalzai 2013). Female leadership at the top might also shatter the glass ceiling that has held women back from achieving offices (Jalalzai and Krook 2010) and can lead to more recruitment of female candidates (Cheng and Tavits 2011).

Voters' preferences for female representation
While the aforementioned biases in selecting and nominating women matter for those that aim to achieve political power in parties, it is unclear how the public judges the situation of women in political parties. When evaluating individual politicians, stereotypes are part of responses of many voters (Dolan 2010;Sanbonmatsu 2002). These stereotypes can take on different forms: voters may prefer men to women as politicians or vice versa. Alternatively, they might expect gendered behaviour from women and men and punish politicians for not conforming to these expectations (Bauer 2017;Bauer and Santia 2022). In list-PR, Devroe and Wauters (2018) find that women and men are judged similarly in terms of competence, but women are perceived to be more leftist.
Nonetheless, it is questionable how these preferences transfer to voting decisions: In a meta-analysis of 67 candidate choice experiments revealed that respondents in conjoint or vignette experiments prefer women to men overall (Schwarz and Coppock 2022). Women are preferred by an average of 2 percentage points, a result that is largely consistent across a multitude of studies. While many findings rest on data from the United States, Wüest and Pontusson (2017) present a similar result in Switzerland, as do Vivyan and Wagner (2015) in the UK.
While these studies predominantly focus on the evaluation of individual politicians, voters' judgment of parties is less well studied. Research has shown that voters often exhibit preferences for equal distributions and fairness (Sauermann and Kaiser 2010). In a similar vein, Miller and Vanberg (2015) provide experimental evidence that individuals tend to reject propositions that are unequal. These "social preferences" are important for understanding markets, collective action, cooperation and social norms (Fehr and Fischbacher 2002). Clayton, O'Brien, and Piscopo (2019) show that descriptive representation of women in a decision-making body increases the legitimacy that voters attribute to that decision.
Generally speaking, voters express preferences for a gender distribution in politics that is more balanced than the one in most countries today (Cowley 2013;Dolan and Sanbonmatsu 2009;Fernández and Valiente 2021). One tool that is specifically designed to raise the level of women in parliament are gender quotas and support for them is considerable among the public (Barnes and Córdova 2016; Batista Pereira and Porto 2020; Beauregard 2018). However, while voters might generally support measures constructed to overcome extreme gender inequality, measures that go beyond achieving this basic goal, such as 50% gender quotas might be considerably less popular (Espírito-Santo 2016).
In sum, the literature suggests that voters (a) moderately prefer female to male candidates in direct comparisons, (b) show general preferences for equal distributions in politics and (c) support measures that are aimed at raising descriptive representation, such as quotas. Given these findings, I suggest that voters hold similar preferences for equal representation when it comes to the party level: Hypothesis 1: Voters are more willing to vote for parties whose MPs are equally male and female over those with predominantly male or female MPs.
Even if voters prefer equal representation of men and women, it is unclear if voters only judge descriptive representation by the share of female MPs or also take female lead candidates into account. The literature has increasingly put more emphasis on "critical actors", powerful political actors that advance women's interests rather than the absolute number of female representatives Krook 2008, 2009;Schwindt-Bayer and Mishler 2005). In line with this theory, female lead candidates might have an out-sized influence on the representation of women and voters might recognize the importance of a woman reaching one of the most powerful positions in a party.
Additionally, voters might generally be unaware of the share of women in political parties. Findings by Sanbonmatsu (2003) and Stauffer (2019) suggest that individual voters are very unlikely to know the gender composition of parliaments. A female lead candidate might, therefore, be a useful information shortcut that voters use to overcome this lack of knowledge cancelling out the negative perception of a predominantly male group of MPs.
While much of the literature focuses on party leadership as the variable of interest, this study is set in an electoral context. Therefore, I use the term "lead candidate" instead of party leader below. In most cases, these two positions coexist. Exceptions are for example Prawo i Sprawiedliwość in Poland in 2015 and the Parti Socialiste in France in 2012.
Hypothesis 2: Voters are more likely to vote for parties with female than male lead candidates.
We have good reason to believe that women would have stronger electoral preferences for parties with higher representation of women. Women are more likely to support implementing measures to decrease the underrepresentation of women, such as quotas (Barnes and Córdova 2016;Cassese and Hannagan 2014). One key reason for this support is that women are personally more likely to benefit from electing parties with more women as more descriptively representative legislatures are thought to make more genderequal policies (Deckman and McTague 2015;Morgan and Buice 2013). Schwarz and Coppock (2022) show that, on average, support for women in politics is considerably larger for female than for male respondents.
Hypothesis 3: The gender of lead candidates and MPs has a larger effect on female than on male voters.
When it comes to ideology, evidence points towards more support for women in politics among left-leaning voters (Beauregard 2018;Erzeel and Caluwaerts 2015;Schwarz and Coppock 2022). Particularly in the US, there appears to be a large divide between Democratic voters, who are more likely to support female candidates and Republicans, who are more likely to support men (Saha and Weeks 2022;Teele, Kalla, and Rosenbluth 2018).
Hypothesis 4: The gender of lead candidates and MPs has a larger effect on leftleaning than on right-leaning and centrist voters.

Research and survey design
Observational studies of parties and voting decisions struggle with disentangling the specific effect of an attribute such as the gender of the lead candidate on vote choice. When observing real world parties, voters bring a set of preconceptions with them, based for example on partisanship, personal experiences, upbringing or socialization. Furthermore, respondents might adhere to social expectations, specifically when judging female representation and might be unwilling to state opinions that do not conform to equal representation. Some combinations of parties might also prevail in reality (e.g. parties with predominantly male MPs and a male lead candidate), while others are uncommon or not observed at all in a specific country (e.g. parties with predominantly female MPs and a male lead candidate). This makes it hard or impossible to estimate the effect of any of these attributes individually.
To tackle these issues, I use a single-profile vignette experiment that allows me to randomly vary the properties of a hypothetical party. Respondents are presented with a short paragraph describing a fictional party, which is defined by several attributes, each of which is randomly drawn from a set of levels. The design, therefore, guarantees complete randomization and all possible combinations of attribute levels could be shown to the respondent. Compared to a forced vote choice experiment (that instructs respondents to pick between two profiles of parties, resembling voting in a two-party system), a single vignette experiment allows respondents to judge parties individually, revealing nuanced preferences between multiple parties commonly found in a multiple party system. Additionally, I am interested in changes in degrees in support, arguing that a decrease in support for a party, or multiple parties, represents a meaningful quantity in a political system. Different forms of vignette and conjoint survey experiments produce revealed preferences that converge to behavioural benchmarks (Hainmueller, Hangartner, and Yamamoto 2015). In addition, vignette and conjoint experiments have high internal validity because they allow the researcher to fully control what respondents see and the randomization ensures that effects of single attributes can be isolated. Respondents also do not have to clearly state their preferences regarding the share of women in a party and can reveal them as part of a holistic assessment of the hypothetical party presented to them. This should alleviate concerns about social desirability biases. In recent years, conjoint and vignette experiments have been used extensively in political science (for an overview, see (Clayton and Anderson-Nilsson 2021)), for instance, for studying immigration (Hainmueller, Hangartner, and Yamamoto 2015), elections (Horiuchi, Smith, and Yamamoto 2018) and EU institutional reform (de Vries 2018).
The vignette design in this paper presents the respondent with a familiar task: Deciding how likely they are to vote for a party based on a set of limited information. While the kind of information commonly available to citizens is much broader than in this experiment (crucially involving information about policy preferences and legislative track record), voters should still be able to form an opinion on a hypothetical party based on limited, but highly relevant information about that party. Importantly, the survey is designed to limit associations to real parties, therefore, revealing preferences citizens have about political parties rather than simply reflecting partisan ratings.

Experimental setup
After making clear that all parties shown are purely hypothetical, I presented five vignettes that describe hypothetical political parties that run in an upcoming national election to respondents, each on a separate page. In a vignette design, certain attributes of the vignette that is presented to the respondent are randomly drawn from a set of levels. The levels and attributes for this survey are presented in Table 1. I not only vary the gender composition of the party, 1 but also their ideology and government participation to avoid any inferences made by the respondent about the ideological leaning of the party from the number of women among the MPs. Research has shown that voters estimate parties with female leaders to be more ideologically left than those with male leaders (O'Brien 2019) and any guesses about the ideology of a party based on its gender composition would be problematic for the results. This allows me to estimate the effect of an attribute holding every other attribute constant. One of the potential vignettes, translated to English, looked like this (the words in bold font were randomly drawn from the levels in Table 1): For an upcoming national election, party A nominated a woman as the lead candidate. Prior to the election, this right-leaning party was the prime minister party in government. The members of parliament for this party are predominantly female.
After showing the vignette, I asked respondents to indicate whether they would vote for the hypothetical party with a voting propensity scale ranging from 0 (very unlikely) to 10 (very likely). This scale is used to estimate electoral support for these hypothetical parties and changes in voting propensity reveal preferences for attributes of parties in the electoral context. The mean of this variable is 4.95 and the standard deviation is 2.65.  (2022) cite 67 studies on gender preferences by voters of which 28 are done in the US and 10 in the UK. This research project extends these findings to new countries and partycentred electoral systems. While much of the literature has been developed in candidate-centred systems, parties are the central political actors especially in Germany, Poland, Italy and Spain.
The survey was conducted with Dalia Research and field time was 30 November 2018-13 December 2018. 2 In each country, about 800 votingage citizens rated five vignettes each, adding up to a total of 20,880 decisions. I had the English survey translated into German, French, Spanish, Italian and Polish and checked the translations with the help of political scientists that are native speakers of these languages. The exact wording of all questions that were presented can be seen in Appendix J. Dalia Research contacts respondents on a network of apps and websites and invites them to participate to their survey. Respondents are recruited for the survey after opting into taking part in research in exchange for non-monetary incentives. Zhang, Antoun, and Yanna Yan (2020) find that respondents recruited in such opt-in panels are providing high-quality responses. Screening questions ensure that the pool of respondents match the voting-age population in regard to age, education and gender. The resulting sample is consistent with the broader public both in terms of demographics as well as political preferences. More details on the sample can be found in Appendices A and B. The research project has received ethics approval from the Ethics Committee of the Faculty of Management, Economics and Social Sciences at the University of Cologne.

The representation of women and electoral support for political parties
Using respondent-level cluster-robust standard errors and guaranteeing full randomization, I can obtain the average marginal component effect (AMCE), which represents the effect of one attribute level (compared to a baseline), keeping all other attributes constant (Hainmueller, Hopkins, and Yamamoto 2014). However, I present the plots in the main section of this study using marginal means (MMs) rather than AMCEs, as MMs can be interpreted as predicted values of the dependent variable and have the benefit of requiring no arbitrary choice of reference category (Leeper, Hobolt, and Tilley 2018). The AMCE results can be found in Appendix C. Also, marginal means facilitate the analysis of subgroup preferences. In subsequent plots, I present the marginal mean estimate as well as 95% confidence intervals. Vignette analysis is done with the cregg package (Leeper, Hobolt, and Tilley 2018). Figure 1 shows the main result of the marginal mean analysis. I show pooled results of all 20,880 vignettes from all five countries with respondent-level clustered standard errors. While there are no statistically significant Figure 1. Marginal means of electoral support for a party. Notes: Marginal means for all 20,880 vignette ratings with 95% confidence intervals; outcome is on 0 ("very unlikely") to 10 ("very likely") voting propensity scale. effects of government status, centrist parties are preferred to both leftleaning and right-leaning parties in that order.
The strongest effect in the survey exists for parties whose members of parliament are predominantly male. Those parties lost 0.74 points on the approval scale compared to parties with equal numbers of male and female MPs, which corresponds to 0.28 standard deviations on the dependent variable. Parties with predominantly female MPs are also less popular than parties with equal numbers of men and women. These findings support Hypothesis 1 and underline preferences for descriptive equality between men and women at the party level (Schwarz and Coppock 2022).
Respondents prefer parties with female over those with male lead candidates by 0.26 points on a scale from 0 to 10 (this corresponds to 0.10 standard deviations on the dependent variable). While this effect substantively smaller than the effect of MP gender, it is statistically significant. I, therefore, find support for Hypothesis 2.
We can learn more about the interdependence between the gender of lead candidates and MPs by looking at the interaction of the two variables. Figure 2 shows the interaction between lead candidates and the share of women among the MPs (this graph only shows the relevant portion of the analysis, for a full version see Appendix F). Importantly, among parties with predominantly female MPs, the gender of the lead candidate does not have an effect on the electoral support for the party. Meanwhile, when the MPs of the party are predominantly male, parties with female lead candidates are preferred to those with male lead candidates. A similar effect exists for parties with equal gender representation among MPs. Marginal means for all 20,880 vignette ratings with 95% confidence intervals; outcome is on 0 ("very unlikely") to 10 ("very likely") voting propensity scale. This effect is substantially large: While parties with predominantly male MPs and female lead candidates are regarded almost as positively as those with predominantly female MPs, those with male lead candidates are regarded almost half a point more negatively on the voting propensity scale (0.15 standard deviations on the dependent variable). The contrast between the configuration that is rated the worst (male MPs and male lead candidate) and the one rated the most positively (equal male and female MPs and female lead candidate) is 1.09 on the voting propensity scale or 0.41 standard deviations on the dependent variable. For the overall electorate, running a female lead candidate has the strongest positive effect precisely in those situations in which voters would punish low female descriptive representation in the party.
Appendix D outlines tests for attention paid to the survey and potential priming of the importance of candidate gender, especially since every respondent was shown multiple vignettes. There are no vignette ordering effects, respondents paid sufficient attention to all vignettes and the effects shown above would even be larger if only the most attentive respondents were included. This also alleviates concerns that respondents might be primed by answering the first vignette to pay more attention to the descriptive representation in the following ones.
Appendix E shows this main model for the different countries in the survey. Germans are least willing to vote for right-leaning parties, while respondents in Italy and Poland prefer right-leaning to left-leaning parties, in line with recent election results. Running a female lead candidate results in a more positive evaluation in France, Italy and Spain, whereas there are no differences in Germany and Poland. Parties with predominantly male MPs are rated the most negatively in all countries except for Poland, where respondents prefer equal representation to both the overrepresentation of women and men.
Finally, Appendix I shows that it is unlikely that the rating of some randomly generated party profiles that resembled real parties influenced the results: Effects for those party profiles are virtually the same as for others. This addresses a key warning raised by (Clayton and Anderson-Nilsson 2021): Studying political gender preferences of voters might be biased because it takes candidate (or party) profiles into account that are extremely unlikely to exist in real life.

Unlikely allies for women in politics? Men and right-leaning voters
Hypotheses 3 and 4 are dealing with the analysis of two key variables that might influence voters' attitudes towards representation: gender and ideology. The left panel in Figure 3 shows the marginal mean of voting for a party for men and women, zoomed in on the effect of lead candidate and MP gender (a full version of this plot can be found in Appendix G). 3 While both men and women prefer female to male lead candidates, the positive effect of running a female lead candidate is stronger for women. This is confirmed looking at a regression model that interacts the attributes from the design with respondent gender and shows a significant interaction effect (see Table G1 in the Appendix). Tables G2, G3 and G4 in the Appendix show that these results are robust to controlling for vignette number, duration of expression and country.
Similarly, both men and women prefer equal over unequal gender representation among MPs. However, women rate parties with predominantly female MPs almost as positively as those with equal representation while men dislike predominantly male and female parties to a similar extent. As Table G1 in the Appendix shows, the negative effect of predominantly male MPs for female respondents (−0.937) is almost cut in half for male respondents (−0.532). Meanwhile, the moderate negative effect of predominantly female MPs for female respondents (−0.196) is more than doubled for male respondents (−0.458). While women have more clearly defined preferences in terms of descriptive gender representation, men also strongly prefer equal to unequal representation. These findings support Hypothesis 3 and underline that, while there is support for gender equality among men, this support is much stronger for women (Barnes and Córdova 2016;Deckman and McTague 2015;Morgan and Buice 2013). The effects for left-and right-leaning as well as centrist voters are shown in the right panel of Figure 3, again zoomed in on the gender composition. Compared to the results for respondent gender, I find considerably smaller differences between ideological groups in party support based on gender composition (see Model 2 in Table G1 in the Appendix). All voters prefer equal over unequal representation, and the interaction effect between lead candidates and respondent ideology is not statistically significant. The only statistically significant effect for the ideology interaction is that left-leaning voters are considerably more negative towards parties with predominantly male MPs than other voters, which matches the literature that find stronger support for gender equality among left-leaning voters (Beauregard 2018;Schwarz and Coppock 2022). In this limited scope, I find support for an ideologically driven effect, but overall there are comparatively small ideological effects which do not lend broad support for Hypothesis 4.
Appendix H shows that only supporters of very few parties actually prefer predominantly male MPs to equal representation (voters of UDI in France, PSL in Poland and FdI in Italy, although not to a statistically significant degree). Conversely, even voters of right-leaning parties such as PiS in Poland or CDU/CSU in Germany prefer equal representation, as do voters of centrist and left-leaning parties.

Conclusion
Voters strongly prefer equal representation of men and women when making their vote choice. In all subgroups (men, women, left-and right leaning), as well as all countries in this study, voters prefer an equal share of men and women among MPs of a hypothetical party to parties that have predominantly male or female MPs. Additionally, female lead candidates increase the likelihood to vote for a party and can even out the negative effect of predominantly male MPs. These effects are generally stronger for women than for men.
These findings support the importance of gender balance within political parties for voters: It is very likely that left-leaning parties benefit from being perceived more inclusive towards women as voters across the political spectrum have preferences for equal representation. Additionally, the situation that is judged the most negatively by voters is an all-male party with both male lead candidates and predominantly male MPs. Finally, the survey found that female lead candidates are not only important for promoting women within the party but have a substantial impact on how voters judge political parties.
From a party politics perspective, there are good reasons to expect parties to not simply reflect public preferences on women's representation: Parties are gendered institutions and power within these institutions has traditionally been in the hands of men O'Brien 2015). Parties might also be vary of nominating more women as they might think that in the gendered workplace of a parliament, they are less likely to be able to maintain the necessary continuous time investment (Iversen and Rosenbluth 2010). Even besides incumbency, decision-making structures such as party conventions are far from efficient markets that make sure voter preferences are reflected in candidate nomination (Rahat, Hazan, and Katz 2008).
As this survey design relied heavily on randomization of attributes and hypothetical parties, these findings cannot be extrapolated easily to existing parties. Clayton and Anderson-Nilsson (2021) warn against using experiments with hypothetical scenarios to study biases that might only manifest in real-world behaviour. This study design aimed to uncover underlying preferences of voters, which are independent of real political parties. In the trade-off between preferences getting drowned out by partisanship versus revealed preferences not being perfectly applicable to real-world settings, this study chooses the latter. Clearly, combining these findings with a similar study that uses a setup that is closer to a real-life electoral setting would be very important in understanding this subject even better.
Meanwhile, there are good reasons to take these results seriously on their own. In campaigning, parties can activate these general preferences and groups within parties that promote women's representation should be more vocal about the importance of women's representation for electoral success. Additionally, lower excitement about the pool of candidates may lead to less willingness to campaign for a party, motivate friends and family to vote or splitting a ticket in races with multiple candidates or tiers.
Social desirability might also play a role in the findings presented above: Some respondents might not want to state that they prefer male to female politicians (or female to male politicians) when asked directly. However, online vignette experiments are designed to minimize such an effect, as respondents anonymously rate complete profiles rather than explicitly state their preferences to an interviewer. In a similar vein, there might be concerns about a priming effect, as the vignette used in this experiment provides more information about the gender composition of parties than political positions. Here, future research could test whether additional information, for example of policy positions, attitudes on women's issues or other factors potentially interesting to respondents might increase or decrease the importance of descriptive representation in the eyes of voters.
Future research could also focus on parties' strategies of promoting women: Are parties aware of the dynamics described in this study and do they nominate female MPs and lead candidates at least partly with voters' preferences in mind? How do these findings translate to actual parties? Additionally, the underlying mechanism that drives the preferences described in this paper remains open for research: What are the reasons for voters to express these preferences? Is it a general feeling of fairness or are voters thinking their preferences will be more represented in a genderequal party? And finally, how are these processes different for right-leaning versus left-leaning parties? The findings presented in this paper lay the groundwork for answering these research question by establishing voters' preferences for equal representation of men and women in political parties.