Voting in referendums increases internal political efficacy of men but not women: evidence from Ireland's 2018 abortion referendum

ABSTRACT Will experience with direct democracy influence men's and women's political beliefs differently? Despite the closed gender gap in voter turnout, women remain less interested in politics and participate less frequently in non-voting activities than men. Scholars find women's lower sense of internal political efficacy as the origins of these gender gaps. In this paper, I examine whether the experience of direct participation in political decision-making alters women's feelings of internal political efficacy differently from it does men's. Building on the insights from the literature on the gendered psychological traits, I theorize that voting in referendums will promote men's internal political efficacy but not women's, because of women's greater susceptibility to the psychological costs of participation in referendums. Using an original panel survey conducted shortly before and after the 2018 abortion referendum in Ireland, I demonstrate the presence of the gendered effect in voting in referendums: While men reported increased internal political efficacy after voting in the referendum, women did not experience any meaningful change, even though the issue magnified women's psychological engagement with the vote. My findings suggest that differences in psychological dispositions between men and women create gendered reactions to citizen experience in the political arena.


Introduction
What would make women feel more politically efficacious such that their engagement in politics would equal men's? In most democracies, women tend to report significantly lower levels of confidence in their abilities to influence politics than similarly situated men (Gidengil, Giles, and Thomas 2008;Fraile and de Miguel 2019;Karp and Banducci 2008), suggesting the clear gender gap in internal political efficacy Wolak 2020). The gender gap in internal political efficacy has far-reaching implications for democratic representation, because a group's active political engagement at the mass level is a necessary condition for securing the group's representation in elected positions (Merolla, Sellers, and Fowler 2013). Fewer women than men run in elections, a product of their perception of themselves as not competent enough for political office, and this in turn, widens the existing gender gap in representation (Fox and Lawless 2011). In addition, women's lower sense of internal political efficacy has contributed to the persistence of the gender gap in political participation (Atkeson 2003;. Though the gender gap in electoral turnout has disappeared or even reversed in many democracies (Quaranta and Dotti Sani 2018), women participate less frequently than men in political activities other than voting, including contacting representatives, lobbying, and attending political rallies (Desposato and Norrander 2009;Ondercin and Jones-White 2011;Verba, Schlozman, and Brady 1995) that could be more effective in improving government responsiveness than electoral participation (Cleary 2007).
Existing work has focused on the role of women's representation in the political arena. While ample research demonstrates that the presence of other women in government boosts women' feelings of political efficacy (Atkeson and Carrillo 2007;Kittilson and Schwindt-Bayer 2012), few studies have paid attention to how women direct participation in politics alters their perceptions of political influence. However, with the increasing use of direct democracy across many countries, the question of how women respond to experience of direct participation in political decision-making has gained significance. In many political systems, ordinary citizens can directly participate in legislative processes through referendums or ballot initiatives. While the frequency and the scope of its usage vary considerably across these countries, direct democracy certainly plays an integral part in political decision-making in some countries (Altman 2010).
Will women's experience with direct democracy promote their sense of internal political efficacy? Previous studies theorized that direct democracy provides citizens with hands-on experience of political decision-making, and this experience could make them feel more efficacious in politics than before (Bowler and Donovan 2002;Smith and Tolbert 2004). However, some groups of citizens may find participation in direct democracy burdensome and challenging, and this in turn might offset the positive feelings they gain from the participatory experience. Building on the insights from the gendered political psychology literature, I expect that the psychological costs of making a complex policy decision will disproportionately affect women. Previous research demonstrates women's low self-confidence in their capabilities to participate in politics compared to men (Preece and Stoddard 2015;Wolak 2020) and their tendency to avoid conflictual environments (Atkeson and Rapoport 2003;Karpowitz and Mendelberg 2014). Drawing upon these findings, I hypothesize gendered effects of voting in direct democracy: while it can boost men's internal political efficacy after gaining experience in direct democracy, women will not experience the same boost. Specifically, I focus on the effects of voting in referendums, an important subtype of direct democratic institutions that are widely used across the globe.
Using an original panel survey of Irish citizens before and after the abortion referendum in 2018, I test my theoretical expectation about the gendered effect of participation in referendums. In May 2018, Irish citizens voted in a national referendum on whether to legalize abortion. This was an opportunity for Irish women, who were historically excluded from the decision-making process, to have direct input on legislation that is so closely related to their rights. I conducted a two-wave panel survey of Irish adults shortly before and after the abortion referendum and analyze within-respondents changes in their sense of political efficacy following the referendum. A comparison of the two survey waves allows me to capture the overall effect of citizen participation in direct democracy as well as whether it appears differently between women and men. My findings demonstrate that the experience of voting on the abortion referendum increased men's internal political efficacy, but not women's. This finding implies that although direct democracy is designed to promote the inclusion of all social groups in political decision-making processes, its participatory effect might appear differently across groups, depending on a group's status in the political arena or psychological engagement with political processes. This finding offers important insights into understanding gendered effects of political institutions (Clayton 2015;Córdova and Rangel 2017;Kittilson and Schwindt-Bayer 2012).

Conditions to promote women's political efficacy
Political efficacy is the belief that one can influence the political processes. A long tradition in political science views it as an essential psychological underpinning for citizen engagement in politics (Abramson and Aldrich 1982;Finkel 1985). It influences people's decision to participate in politics by determining perception of how much their voice matters (Finkel 1985;Verba, Schlozman, and Brady 1995). The concept of political efficacy includes two dimensions: (1) external political efficacy, which refers to the belief that government is responsive to citizens' demands; (2) internal political efficacy, i.e. the belief that one is competent enough to understand and participate in politics (Niemi, Craig, and Mattei 1991).
Though gender inequalities in resources have largely disappeared, women continue to feel less efficacious in politics than men (Gidengil, Giles, and Thomas 2008;Karp and Banducci 2008). In particular, women report much lower levels of self-confidence in their abilities to understand and participate in politics compared to men, suggesting the clear gender gap in internal political efficacy Wolak 2020).
Scholars have sought to identify conditions that promote women's political efficacy. In particular, research has focused on whether the increased number of women in government induces attitudinal changes among women (Atkeson and Carrillo 2007;Barnes and Burchard 2013;Kittilson and Schwindt-Bayer 2012). The underlying theoretical argument in these studies is that large proportions of women in legislative bodies convey a sense that the political system is open to women's voices and will produce policy outcomes that are aligned with women's interests (Schwindt-Bayer and Mishler 2005). This process might enhance women's external political efficacy. Similarly, having the increased number of women in legislative bodies could weaken women's long-held belief that "politics is not for them," (Burns, Schlozman, and Verba 2001, 8), promoting their feelings of internal political efficacy.
These studies demonstrate how women's sense of political efficacy respond to the presence of other women in political decision-making processes. However, women's response to direct experience with these processes has been largely overlooked.
The gendered effect of participation in referendums on internal political efficacy A large body of work has investigated how citizens' exposure to direct democracy alters feelings of political efficacy (Bowler and Donovan 2002;Hero and Tolbert 2004;Smith 2002). For example, Bowler and Donovan (2002) claims that the availability of direct democracy offers avenues for influencing government policy-making, "ensuring that the public is consulted (or anticipated) in discussions about major policy issues." (376) and that this alone gives individuals a sense that the political system is responsive to their point of view. However, existing work on this topic has yielded mixed findings on this relationship; while some find positive associations between the exposure to direct democracy and political efficacy (Bowler and Donovan 2002;Hero and Tolbert 2004;Mendelsohn and Cutler 2000), other studies find that there is no evidence of such a relationship across different localities (Dyck and Lascher 2009;Schlozman and Yohai 2008).
The inconsistency in empirical findings in existing work might be in large part due to varying responses to direct democracy across social groups. Despite the considerable volume of research on the overall attitudinal effects of direct democracy, only a handful of studies consider how the attitudinal effects of direct democracy might vary across different social groups. 1 However, group heterogeneities are central to understanding the connection between experience with direct democracy and political efficacy, because group-based factors have crucial influences on how an individual responds to experience with political processes (Anderson and Guillory 1997). For instance, experience with direct legislation might generate more dramatic impacts on the ways individuals view their political role among some groups than others, depending on the group's political status or prior experience with the political arena as a group.
Building on these insights, I theorize that voting in referendums enhances men's internal political efficacy, but not women's. 2 I reason this theoretical expectation with gender differences in psychological traits that are firmly established in the literature: (1) gender differences in self-confidence in politics and (2) gender differences in the susceptibility to contentious environments. These differences could dampen women's psychological benefits they gain from experience in direct democracy in comparison to men. Below, I expand on these two logics.
Existing work on educative effects of direct democracy suggests that institutions of direct democracy will make citizens feel more confident about their abilities to influence politics by increasing the supply of policy information and incentivizing citizens to seek out detailed information about the issue on the ballot (Bowler and Donovan 2002;Mendelsohn and Cutler 2000). Also, the availability of direct democracy might convey a message that citizens are trustworthy and competent to make a major policy decision (Frey 1997). Moreover, unlike legislative processes where groups that are resourceful and well represented have a greater voice than others (Gerber 2011), direct democracy offers all citizens opportunities to make direct input in policy-making. It implies that, under direct democracy, underrepresented groups can gain hands-on experience of political decision-making that otherwise would not be available to them. 3 This experience might make them feel more influential in politics than before.
However, women face several crucial psychological barriers that prevent them from gaining feelings of internal political efficacy after experiences of direct democracy. Participation in direct democracy is costly for voters in many ways and these costs might disproportionately affect women. Ample evidence indicates that voters find policy questions on ballot initiatives or referendums complicated and confusing to comprehend (Bowler and 2 For this paper, my theoretical expectation considers how voting in referendums will generate gendered effects on internal political efficacy but does not offer a clear prediction regarding external political efficacy. Separating these dimensions is particularly relevant when studying the gendered gap in political efficacy, because gendered dynamics might appear differently in each (Fraile and de Miguel 2019). 3 While direct democracy offers policy-making opportunities for underrepresented groups, this does not imply that all citizens gain equal influence. For example, members of interest groups will have a greater influence in the process of referendums or ballot initiatives than non-members (Boehmke 2002 Dyck and Lascher (2009) further argue that, because of the demanding nature of direct democracy, greater exposure to it may lead to a decline in political efficacy among low-information voters. Indeed, critics of direct democracy have cast doubts on voter competence to make informed policy decisions (Magleby 1984). Relatedly, existing work in psychology documents how the complexity in decisions overwhelms individuals and lowers confidence in their decision-making ability (Botti and Iyengar 2006). In addition, complexity in decisions is likely to reduce satisfaction and even generate anxiety, particularly in highstakes decision-making contexts, such as selecting romantic partners (D'Angelo and Toma 2017). A policy decision is clearly high-stakes: the decision has profound consequences, not just for the individual voter but for the society as a whole. These features associated with voting in direct democracy could make women particularly susceptible to its psychological pressures. Self-confidence plays a central role in determining whether an individual will take on a challenging task (Wolak 2020), such as voting on a policy matter. A vast literature in psychology has documented gender differences in self-confidence. Research finds that women show lower levels of confidence about their skills and abilities than men whereas men tend to exhibit over-confidence (Klayman et al. 1999;Kling et al. 1999;Soll and Klayman 2004). Studies also find women's tendencies to underestimate their intelligence, unlike men who often overestimate theirs (Beloff 1992;Furnham and Rawles 1995). Evidence from education research suggests that the gender gap in self-confidence also appears among school children when evaluating their mathematics abilities (Dahlbom et al. 2011;Jakobsson, Levin, and Kotsadam 2013).
The gender gap in self-confidence is particularly evident in politics, an area that is typically considered masculine. Research shows that women are significantly less confident in their capabilities to engage in politics than men, and this gender gap contributes to women's lower levels of political interest, attention, and internal efficacy than men (Pfanzelt and Spies 2019;Wolak 2020). Women also tend to exhibit less self-confidence in their qualifications for electoral politics, lowering their likelihood of running for office compared to equally qualified men (Fox and Lawless 2005;2011;Gaddie 2003). Studies find that the gender gap in self-confidence in politics inflates the gender gap in political knowledge, as women are more likely to say "don't know" at greater rates than men, who tend to guess the answers when they don't know them (Lizotte and Sidman 2009;Mondak and Anderson 2004).
Following this logic, women are less likely to feel confident about their capabilities make a decision in direct democracy settings than men. When faced with a complex policy decision in a referendum, women's relatively low self-confidence in politics (Pfanzelt and Spies 2019; Wolak 2020) leads them to hold self-doubts about their authority and capabilities to make a right decision, even when they possess the abilities to do so. This pattern is less likely to appear among men, who tend to exhibit higher self-confidence in politics.
In addition, the process of referendums creates a highly conflictual political environment that serves as an additional psychological barrier to women. Before a referendum, citizens often witness intense public debates and campaigns over the issue on the ballot, debates that tend to magnify pre-existing social divisions and conflicts (Donovan and Tolbert 2013;Pantoja and Segura 2003). These public discourses are likely to generate a political environment marked by disagreements and hostility.
Existing research offers reasons to believe that these conflictual social contexts will have disproportionate impacts on women. A large body of research in political science attributes women's low levels of political ambition relative to men (Fox and Lawless 2004;Preece, Stoddard, and Fisher 2016) to gender differences in psychological traits (Kanthak and Woon 2015;Oliver and Conroy 2018;Schneider et al. 2016). In particular, studies show that women tend to avoid competitive and conflictual situations in politics, such as running a campaign (Kanthak and Woon 2015) or engaging in an interpersonal political disagreement (Coffé and Bolzendahl 2017;Ulbig and Funk 1999;Wolak and McDevitt 2011).
There is ample research showing that women tend to feel more insecure and more likely to hesitate to express their opinions in a contentious environment than men. For example, a study of Irish citizens (Miller, Wilford, and Donoghue 1999) shows that women tend to stay away from a political argument because they do not believe they can persuade others in a political discussion. Another study finds that women are less likely to engage in political learning in a conflictual environment, such as partisan politics, unlike men who display a greater willingness to seek out more information in the same environment (Wolak and McDevitt 2011). Women also tend to avoid disagreeing with others during political discussions, doubting their ability to successfully challenge others' opinions (Atkeson and Rapoport 2003;Karpowitz and Mendelberg 2014;Mansbridge 1983). Psychological research explains these gendered patterns with differences in social roles for men and women. As women are accustomed to social roles that value cooperation and consensus, while men are socialized to differentiate themselves (Eagly, Wood, and Diekman 2000;Eagly and Karau 2002), women tend to feel more uncomfortable with conflicts and disagreements than men. Karpowitz and Mendelberg (2014) note that women are less motivated to express controversial views, because "it is deemed masculine territory (34)." The insights from the literature imply that psychological pressures to cast a vote on a policy matter in a contentious setting will fall more heavily on women than men.
Based on these insights from the literature, I expect that women are less likely than men to enhance internal political efficacy after voting in referendums. Women have lower self-confidence and greater avoidance of conflictual environments than men, as demonstrated in the literature, making them more susceptible to the psychological costs of voting in a referendum. These costs in turn could make women doubt their political authority and abilities to influence politics, whereas men are less influenced by these costs. These considerations lead to the following hypothesis about the gendered effect of voting in referendums: Gendered Effect Hypothesis: Voting in referendums will increase men's feelings of internal political efficacy but not women's.

Case and research design
To test the gendered effect of voting in referendums on political efficacy, I conducted an original panel survey in Ireland before and after a national referendum on 25 May 2018, when citizens voted on whether to repeal the constitutional ban on abortion.
The Irish case offers a conservative test for the hypothesized gendered effect of voting in referendums. The 2018 referendum was unusual in its bottom-up nature, in that civil society largely drive the campaigns on both sides (Farrell, Suiter, and Harris 2019). In particular, women drove the discussions and actions in the campaign process, while many men considered themselves outsiders from this issue and stepped back. The press reported that many women felt they had to "do something" because of their strong feelings about this issue. 4 Both the relevance of the issue to women's rights and women's active involvement in the process give reasons to expect a positive effect of voting in the referendum on women's political efficacy. Thus, if findings based on the abortion referendum confirm my hypothesis that the effect will not appear among women, that pattern is likely to be more pronounced in other cases of referendums. 5

The abortion referendum of 2018
Ireland banned abortion in 1983 with the Eighth Amendment to its Constitution. Debates about the ban have been frequent ever since, prompted by several controversial cases, such as "Case X" in 1992 and the 2012 death of Savita Halappanavar. However, abortion rights remained extremely limited in line with Ireland's long-held conservative norms under the influence of the Catholic Church.
In 2016, the Parliament established the Irish citizens' assembly to enable ordinary citizens to deliberate on several major issues in Ireland, including abortion. The Assembly consisted of a government-appointed chairperson and 99 citizens who were randomly chosen to broadly represent Irish society. From November 2016 to April 2017, the assembly gathered to hear information on both sides of the abortion debate and to deliberate on its solution. In June 2017, the Citizens' Assembly submitted a report that included a call for the referendum on the abortion ban. This call has led to the government's announcement in January 2018 to hold a referendum on the issue in May 2018 (Suiter 2018).
The abortion referendum was an opportunity for many women, especially among younger generations, to be politically active for the first time, and many were. Turnout overall was very high: 64.1% of eligible voters voted in the referendum, and 66.4% of them voted to repeal the abortion ban. In previous years, male politicians dominated the legislative processes for the abortion issue in this country, joining many other advanced democracies. 6 This referendum, however, gave Irish women an opportunity not just to vote for the issue but also to lead its legislative process.
For several months before the vote, heated public debates and a largescale campaign from both pro-choice and pro-life groups prevailed in Irish society. These campaigns were highly negative, magnifying the divisions and conflicts in Irish society. 7

Panel survey
In order to test the suggested hypothesis, I focus on measuring the changes in internal political efficacy before and after voting in the abortion referendum. 8 A two-wave panel survey of Irish adults before and after the abortion referendum allows me to estimate within-respondent changes in the sense of political efficacy following the referendum. Panel data enables comparisons of the same individual at different times that provide direct evidence of changes, while holding other individual-level characteristics that also affect political efficacy constant. For this survey, the online panel provider 6 https://www.nytimes.com/2018/05/24/world/europe/ireland-abortion-referendum.html. 7 https://foreignpolicy.com/2018/05/24/irelands-nasty-no-campaign/. 8 This paper focuses on testing the hypothesized gendered effect on political efficacy, and it does not directly test the assumption that gender differences in psychological traits will lead to gendered responses to voting in referendums. Testing this assumption will be a promising area for future research.
Qualtrics Inc. recruited 534 voting-age Irish citizens. Qualtrics offers better demographic and political representation than other commonly used online samples, such as Amazon Mechanical Turk and Facebook (Boas, Christenson, and Glick 2020). Due to the difficulty in recruiting respondents aged 65+ and those with low levels of education, my sample includes fewer old respondents and more educated respondents than a nationally representative sample. To address this issue, I conducted an additional analysis with post-stratified weights on age and education. The results appear in Appendix D.
Both surveys asked the respondents the standard political efficacy questions from the American National Election Studies. I estimate the effects on internal and external political efficacy separately, because my theoretical expectation focuses on gendered effects on internal political efficacy, not the effects on overall political efficacy. In measuring levels of political efficacy, I used three different questions to measure the respondents' internal and external political efficacy, respectively. 9 Each item is measured on a 9point scale, and I rescaled the measures so that higher values indicate a greater sense of political efficacy. Next, following previous research on political efficacy (Chamberlain 2012;Wolak 2018), I created an additive measure of external and internal political efficacy by summing the responses to individual items and then rescaling it from 0 to 1 to make interpretation of the results more intuitive. The pre-referendum survey was fielded from May 15 to May 24. The recontact survey that asked the respondents the same set of questions was fielded from May 29 to June 8. The first survey had 789 complete responses. The recontact rate was 67.7%, reducing the sample size to 534. A short time gap between the waves of an average of 11 days minimizes the possibility that other time-specific factors drive changes in respondents' political efficacy levels. During the gap between the two surveys, the abortion referendum was the single most salient issue during the time of two surveys and there was no unexpected political scandal between the two surveys. 10 Also, while there is a host of individual characteristics that could also affect political efficacy, such as political interest, knowledge in politics, and partisanship, these characteristics do not vary substantially within a single person over a short period of time. Thus, observed changes in responses will suggest that the referendumeither a simple exposure to or experience in the referendumshifted feelings of political efficacy, separate from other events.

Findings
I begin with estimating mean differences in levels of political efficacy before and after the referendum. My hypothesis suggests the presence of gendered 9 See Appendix B for the exact wording of the questions. 10 https://www.nytimes.com/2018/05/25/world/europe/ireland-abortion-referendum-explainer.html. responses in the experience of voting in direct democracy. In order to directly test this hypothesis, the data need to only include the respondents who voted in the referendum. Thus, I first compare means of pre-and post-referendum levels of political efficacy of the respondents who answered that they voted in the abortion referendum. Among 534 respondents who completed both pre-and post-referendum surveys, 437 answered that they voted in the referendum and the remaining 97 answered they did not. To ensure that gender proportions are balanced across voters and non-voters, I estimated the likelihood of voting with respondent demographics. The results demonstrate that the respondent's gender did not predict the likelihood of voting in the referendum. 11 Figure 1 plots mean differences in internal political efficacy before and after the referendum, disaggregated by respondent gender. The dots in the plot indicate post-referendum−pre-referendum efficacy levels, estimated using one-sided paired t-tests. Among the 437 respondents who answered they voted in the referendum, men's internal political efficacy increased by 2.2 percentage points and the effect was statistically significant at the p < 0.05 level. However, there was no difference in internal political efficacy of women who voted. This result offers some suggestive evidence that the experience of voting in direct democracy increased men's sense of internal political efficacy, but not women's, confirming the presence of gendered effect. As a comparison, I report the t-test results using all respondents (N = 534), including both voters and non-voters. When non-voters are included, there was no change in levels of internal political efficacy post-referendum for men or women. Together, the t-tests suggest that voting in the referendum brought a moderate increase in internal political efficacy only among male respondents, but not among female respondents. Also, when non-voters are included, the positive effect for men disappears.
An alternative way to evaluate (1) how voting in the referendum altered internal political efficacy levels and (2) how the effects of voting in the referendum vary by respondent gender is to model change in internal political efficacy levels after the referendum with these predictors (i.e. voting in the referendum and gender) using regression models. Table 1 presents the results of regressions models. Columns (1) through (3) report the results of the models using Change in Internal Political Efficacy as the outcome variable. The variable was constructed by subtracting the pre-referendum internal political efficacy values from the post-referendum internal political efficacy values. The variable takes the value from −0.625 to 0.792. As a comparison, models (4) through (6) report the results using the Change in External Political Efficacy (i.e. External Political Efficacy post -External Political Efficacy pre ) as the outcome variable.
In addition to an indicator of whether a respondent voted in the referendum (Voted), I examine an interaction effect between a respondent gender and experience of voting. This interaction effect allows me to directly test my hypothesis about gendered effect of voting in direct democracy. I control for the effects of other demographic variables that might be associated with change in efficacy levels, including marital status, household income, religiosity, age, education levels. The models also control for political party one supports and support for abortion, all measured in the pre-referendum survey. The models also include an indicator of whether a respondent voted to repeal the ban (Voted to Repeal) as a regressor, because voting for the winning outcome might have a separate effect on Change in Political Efficacy (Anderson and Guillory 1997;Blais and Gélineau 2007). The values of Voted to Repeal for the respondents who did not vote are coded as 0. Finally, the outcome of these models, Change in [Internal/ External] Political Efficacy, might be subject to a ceiling effect driven by the respondents who scored high on efficacy questions in the pre-referendum survey. This ceiling effect might lead to an underestimation bias. Addressing this concern, columns (2), (3), (5) and (6) include initial levels of internal political efficacy measured in the pre-referendum survey as a regressor.
The results confirm the gendered effects of voting in the referendum. Among women, voting in the referendum did not lead to a relative gain in efficacy levels, as indicated by null effects of Voted across all models. By contrast, voting in the referendum did make a difference among men: Men who voted reported a greater increase in internal political efficacy levels than others. The positive and statistically significant coefficient estimate of the interaction term, Male and Voted in column (2) of Table 1 demonstrates this gendered effect. 12 Using the results in column (3), I calculated marginal effects of Voted for men and women and display them in Figure 2, along with 95% confidence intervals. This plot shows that after the referendum, the level of internal political efficacy for men who voted increased 6 percentage point more than others, while the same effect did not appear among women who voted. This effect size represents 42% of one standard deviation of the outcome variable, Change in Internal Political Efficacy. This effect size is substantively meaningful given the short time gap between the two waves and a stable nature of political efficacy, as suggested by previous research (Chamberlain 2012;Iyengar 1980). My finding suggests that even over a short period of time, voting in a referendum generated a meaningful change in men's internal political efficacy, a feature that is not easily movable.
The results of two separate analyses offer additional support for my theoretical argument. First, the null effect of the interaction between Voted and Male on Change in External Political Efficacy. is consistent with my claim that psychological differences between men and women drive gendered effect of voting in a referendum. While internal political efficacy considers subjective political competence, external political efficacy is related to perceptions of political system (Lassen and Serritzlew 2011;Niemi, Craig, and Mattei 1991). I suggest that women's lower self-confidence relative to men drives the gendered reactions, and thus, this mechanism only applies to internal political efficacy, but not to external political efficacy. In line with this expectation, columns (5) and (6) of Table 1 show that there is no gendered effect of voting in the referendum for external political efficacy. Second, models including the variable, Voted to Repeal suggests that voting for the winning outcome improves men's internal political efficacy, but not women's. In columns (3) and (6) of Table 1, I report the results of the models including the interaction term of Voted to Repeal and Male. The results show that Voted to Repeal itself does not have an effect on Change in Internal Political Efficacy, but the interaction of Voted to Repeal and Male has a statistically significant and positive (5.5 percentage point increase) effect. In addition, while Voted to Repeal has an overall positive effect for  Table 1. Solid lines represent 95% confidence intervals.
Change in External Political Efficacy, there was no gendered effect, as shown by the null interaction effect between Voted to Repeal and Male (See columns (5) and (6) of Table 1). These findings suggest that even when their vote choice aligns with the winning outcome, women's experience of voting in a referendum only improves their perception of government responsive, but not their internal self-efficacy in politics. Among men, the same experience boosts both the perception of government responsiveness and self-efficacy.
Together, my results suggest that voting in direct democracy influences feelings of internal political efficacy differently by gender. As hypothesized, while men reported an increased sense of internal political efficacy after voting in direct democracy, that effect was absent among women.

Discussion and conclusion
The continued expansion of direct democracy in legislative processes across many governments renders it imperative for researchers to study citizens' responses to experience of direct participation in policy-making processes. This paper investigates a previously overlooked question in this research: how members of marginalized groups respond to experience of voting in referendums, and how that experience alters their existing political beliefs. I examine how voting in referendums generates differential impacts on men's and women's feelings of internal political efficacy. Building on the previous findings of gender differences in psychological orientations with politics, I expect men's and women's responses to voting experience in referendums to vary considerably. In particular, I predict that the increased feelings of internal political efficacy after voting in referendums would apply to men but not to women.
Though research suggests experience with direct democracy makes citizens feel more trusted and politically efficacious (Bowler and Donovan 2002), my findings demonstrate that such experiences do not benefit women to the same extent as men. My analysis of a panel survey of Irish citizens before and after the historic abortion referendum confirms the presence of gendered association in citizen participation in referendums: while men reported increased internal political efficacy after voting in the abortion referendum, women did not experience any change in their sense of internal political efficacy, even in a case where women played an exceptionally active role during the campaign. Thus, this gendered effect is likely to be more evident in other instances of direct democracy where women's psychological engagement with the issue on the ballot will be lower than when the vote concerns abortion.
My findings carry important implications for political institutions and women's political engagement. Despite their increasing access to resources, women still report lower levels of political efficacy than men, leading to a persistent gender gap in political engagement and representation. Previous research has demonstrated that inclusive political institutions, such as proportional representation electoral systems or legislative gender quotas (Barnes and Burchard 2013;Kittilson and Schwindt-Bayer 2010), might enhance women's feelings of political efficacy by conveying the system's openness. Direct democratic institutions that are designed to allow all citizens to influence policy-making equally could send a similar signal of openness to underrepresented groups like women. However, my findings suggest that the psychological costs of participation that disproportionately affect women could counteract the empowering effects of an inclusive political institution. These findings add to the recent work highlighting the importance of understanding gender differences in psychological tendencies in closing the gender gap in political engagement Wolak 2020).
This paper opens up new avenues for future research. While I expect psychological differences between men and women that previous research demonstrates to generate the gendered effect of voting in a referendum, this study does not directly test this mechanism. Future research should directly test whether women do feel more psychologically burdened than men when voting in a referendum. In addition, although this paper offers evidence for the gendered effect of voting experience in a referendum, it does not examine how long this effect lasts. I note that the boost in men's internal political efficacy might disappear in the long run. This possibility introduces a promising avenue for future research. Future research should also explore the presence and the direction of cumulative effect of experience with direct democracy.

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