When citizens met politicians – the process and outcomes of mixed deliberation according to participant status and gender

With deliberative democracy becoming increasingly incorporated into political processes, the instances of citizens and politicians meeting in deliberation, so-called mixed deliberations, is increasing too. While these are important steps towards more deliberative systems, the mixed deliberation setting nonetheless introduces certain risks regarding equality. Citizens and politicians have different deliberative capacities, which could affect the process of deliberation. Moreover, political inequalities such as gender gaps may be more evident in deliberations where politicians are present. This study focuses on a series of mixed deliberations in 2018 about a proposed municipal merger in Finland. Using content analysis of speech acts (N = 3404) as well as pre- and post-deliberation surveys (N = 225), we analyze patterns according to participant status (politician/citizen) and gender regarding dominance, deliberative discussion quality, and impact on internal and external efficacy. The findings show that politicians dominated the discussions and achieved a higher deliberative quality than citizens. The findings also revealed that women achieved higher deliberative quality in their speech acts.


Introduction
There has been a steady rise in the use of democratic innovations for improving the democratic quality within decision-making processes (see e.g. Bächtiger, Setälä, and Grönlund 2014;Farrell et al. 2020). One of the key normative requirements for a deliberation setting is equal participation (Karpowitz, Mendelberg, and Shaker 2012), and during the deliberations the expectation is that political inequality due to gender and social status should decline (Gutmann and Thompson 2004) as well as that no individual or group dominates (Gerber 2015, 112). However, within mixed settings (e.g. groups including participants from various backgrounds), there is a risk that certain individuals will dominate the discussion (e.g. Elster 2012, 6). These might be individuals of higher social standing or with higher knowledge about the specific topic being discussed (Bohman 2000;Flinders et al. 2016;Rosenberg 2007), or it might even be gender related (Karpowitz, Mendelberg, and Shaker 2012). In these circumstances, the quality of the deliberation might be negatively affected as the equal status of the participants during deliberation is challenged. Thus, instead of contributing to positive effects on the individuals participating, such as higher levels of political efficacy, the experiences derived from these types of events might even contribute to further solidifying negative sentiments towards politics. Deliberative events have occasionally received critique for failing to meet democratic principles such as inclusion and equality (e.g. Gerber, Schaub, and Mueller 2019). Additionally, the policy impact of democratic innovations has been criticized for being too weak (Fournier et al. 2011;Johnson 2015). To counter this criticism, deliberative democracy proponents have emphasized measures for safeguarding inclusion such as facilitation (Beauvais and Bächtiger 2016;Curato et al. 2020) and started inviting politicians to participate in deliberative events, often called mixed deliberation (Strandberg and Berg 2019) or mixedmembership deliberation, to close the gap between deliberative and representative decision-making processes (e.g. Farrell et al. 2020).
Previous research on the process and effects of deliberation has often concentrated on citizen deliberation (e.g. Bächtiger, Setälä, and Grönlund 2014;Elstub and McLaverty 2014) and elite deliberation (e.g. Bächtiger 2014), while mixed deliberation has received far less attention (cf Farrell et al. 2020;Strandberg and Berg 2019). The risk of including politicians in a deliberative event is primarily that they may dominate the discussions. Nevertheless, involving politicians in deliberation can increase the likelihood for events to be taken more seriously and reduce the distance between citizens and representatives (Setälä 2017;Farrell et al. 2020). Moreover, studies have shown indications of a gender gap in deliberation, adding to worries about deliberation not fulfilling ideals of political equality (Beauvais 2020;Mendelberg, Karpowitz, and Oliphant 2014). For example, women are less willing to participate in deliberation (Karjalainen and Rapeli 2015), and when they do, they tend to speak significantly less than men do (Karpowitz, Mendelberg, and Shaker 2012). Since inclusion and deliberative quality are central ideals in deliberative democracy theory (Young 2000), it is important to assess whether deliberative events live up to these, pay attention to inclusiveness beyond the recruitment procedure, and increase knowledge about the process and effects of mixed deliberation.
The purpose of this study is to assess a series of mixed deliberative discussions (N = 225) about a proposed municipal merger in a Finnish municipality. We examine discussion dominance, the deliberative quality of the discussions, and what effects on efficacy taking part in the events has on the participants. More specifically, the study analyzes the influence of gender and status on discussion equality, quality of deliberation and effects on political efficacy in mixed deliberation. The deliberative discussions were held on three occasions in winter 2018 and consisted of small-n discussion groups all containing a mix of citizens and politicians.

Domination according to status and gender in mixed deliberations
The communicative process per sethat is, rational, fact-based argumentation between participantsarguably constitutes the core of the deliberative theory. Although well-organized deliberative events may counteract unwanted phenomena such as domination (e.g. Elstub and McLaverty 2014), the deliberative capacity of participants might nevertheless be unevenly distributed (Himmelroos 2011) and have a bearing on the discussions. Equality of participants means that all participants in the deliberation should have equal opportunity to make themselves heard in deliberation and that all arguments of deliberators are treated equally, regardless of who is making the argument (Fishkin 2009;Gastil 2008;Habermas 1998;Gerber 2015).
In the literature on deliberation, there are two main definitions of domination. Domination can relate to the distribution of the number of speech acts and speaking time (Fishkin 2009;Himmelroos 2011), but it can also be defined as the degree to which opinions change towards the views of the more socially advantaged participants in the groups (Fishkin 2009;Luskin et al. 2017). This article only relates to the former, which indicates that domination is present when speech acts are not proportionally distributed between persons or groups. The deliberative ideal of equal participation does not necessarily mean that one should expect absolute speaking equality in terms of number or length of speech acts (Mansbridge 1999, 232). More importantly, participants should have equal opportunities to speak (Gastil 1993). However, equality is not achieved if a person or a group is completely overrepresented in terms of the number of speech acts and length of talk (Gerber 2015, 112). Domination in the distribution of talk may have framing and agenda-setting effects that can give undue strength to some arguments over others (Gerber 2015, 112).
Moreover, some argue it might not be reasonable to expect deliberation to exhibit no patterns of domination because of differences in styles of expression (Gerber 2015, 112) or inequality carried by language and appearance (Lupia and Norton 2017). Likewise, in the context of mixed deliberation, the experience might influence the distribution of talk since politicians might be more accustomed to speaking in public and participating in political debates than the average citizen. Additionally, domination in discursive participation by politicians can be more pronounced in one-off deliberative events compared to reoccurring deliberation, where participants have time to grow more confidence (Flinders et al. 2016).
Sometimes, however, citizens of higher social status may dominate the discussions and thus negatively affect the equality in deliberations (Gastil 2009, 283;Polletta and Lee 2006, 701). Levine, Fung, and Gastil (2005) argue that such individuals might be even more inclined to leverage their status or skills when the issues of deliberation concern higher stakes to ensure that they get their way. Mixed deliberation can exacerbate these issues since politicians' higher status and familiarity with the deliberation issue as well as their debating skills can introduce political inequality in small group discussions with citizens (Moore 2012). Thus, some scholars have found indications of politicians dominating deliberative events (Flinders et al. 2016;Minozzi et al. 2015). Farrell et al. (2020) found that politicians, on average, spoke more frequently than citizens at the Irish Constitutional Convention. However, this assumption has also become challenged (see Farrell et al. 2020;Suiter, Farrell, and O'Malley 2016) as the dangers of domination are argued to be alleviated with good facilitation (Afsahi 2020;Elstub and McLaverty 2014) or that citizens do not perceive domination as a problem (Harris et al. 2020, 14). Nevertheless, we still expect those with higher social status, in this context politicians, to dominate the discussions. This leads us to the first hypothesis for our study: H1a: Politicians will dominate the discussions.
Several studies have also discussed a potential gender gap in deliberation (Caluwaerts 2013;Gerber 2015;Beauvais 2020;Karpowitz and Mendelberg 2014). For instance, studies indicate that women are less likely to turn up for deliberation events (see Dillard 2011, 58), that they are disadvantaged in speech participation and to participate proportionally less when they are in a minority position (Karpowitz, Mendelberg, and Shaker 2012). Analyzing data from a Finnish deliberation, a country that is generally renowned for its gender equality, Himmelroos (2017, 14-15) nevertheless found that women have less influence during deliberations than men. Moreover, research has shown that men's poor deliberative behavior (e.g. interrupting, dominating) undermines women's capacity to be effective deliberators (Afsahi 2020;Mendelberg, Karpowitz, and Oliphant 2014). Additionally, experimental research suggests that people are more willing to revise their opinions after hearing a man's counterargument than after hearing a woman's identical counterargument (Beauvais 2020). Despite these challenges, some studies suggest that the gender gap in talk distribution can decrease under the right conditions such as facilitated discussion (Karpowitz and Mendelberg 2014, 307;Showers, Tindall, and Davies 2015;Farrell et al. 2020), in deliberative polls (Siu 2009), or in evidence-driven deliberation (Levy and Sakaiya 2020). According to Jacquet (2017), participation in mini publics is rooted in individual perception, and as women have been shown to express lower levels of political self-confidence (Levy 2013), taking also onboard the main indication from the research, we expect men to be more dominating during the discussions. Hence, we present our second hypothesis: H1b: Men will dominate the discussions.

Deliberative quality in mixed deliberation according to status and gender
The definition and operationalization of deliberative quality are contested (Steenberger et al. 2003;Jaramillo and Steiner 2019;King 2009). However, when searching for high deliberative quality, scholars usually refer to discussions that stay on topic and feature rational argumentation, reciprocity, respect, sincerity and civility (Steenberger et al. 2003;Graham 2009;Kies 2010). Mixed deliberation is especially interesting since it induces differences in capacity. Siu (2009) demonstrates that the most informed participants produce higher quality argumentation in deliberation. Bächtiger and Beste (2017) argue that both politicians and citizens have the capacity to deliberate and score high on measures of deliberative quality when the conditions are right. However, deliberation among politicians reached higher levels of justification rationality and lower levels of respect compared to citizen deliberation (Pedrini 2014). Previous research has also found that politicians are better at deliberating than citizens in mixed deliberation (Gerber and Mueller 2018). An argument against this claim is that citizens can produce better deliberation compared to representatives since constituency and party ties do not bind them, nor do they have a need for strategic communication to get re-elected (Fishkin and Luskin 2005). Rosenberg (2014), on the other hand, questions citizens' ability to deliberate and argues that the average citizen has biased perceptions, relies on cognitive shortcuts and uses prejudicial thinking. Based on their higher levels of deliberative capacity and previous research findings, we expect politicians to express higher quality speech acts than other participants. A third hypothesis can thereby be formulated for our study: H2: Speech acts by politicians will have higher deliberative quality than those of citizens.
Turning to potential differences in deliberative discussion quality according to gender, previous findings are mixed. Several studies have found no gender gap in terms of deliberative quality (Klinger and Russmann 2015;Grunenfelder and Bächtiger 2007;Himmelroos 2017;). Nevertheless, a study by Afsahi (2020) indicates that women can display a higher capacity to engage in deliberative behavior compared to men. Likewise, women have been found to score higher on measures of deliberative quality compared to men (Jennstål 2019). On the other hand, some studies found that men had better deliberative skills compared to women (Caluwaerts 2013;Han, Schenck-Hamlin, and Schenck-Hamlin 2015;Ugarriza and Nussio 2016). Given the very inconclusive evidence, instead of another hypothesis, we have decided to formulate a research question regarding gender and deliberative quality: RQ1: What is the difference in deliberative quality in speech acts by men and women?
The effects of mixed deliberation on political efficacy according to status and gender The effects that deliberation is perceived to have on individuals taking part in deliberative discussions are usually related to enhancing citizens' competence for democratic participation (Fishkin 2009, 23-26). For instance, deliberating citizens have been found to increase their levels of social trust, trust for political institutions, political knowledge, as well as feelings of internal and external political efficacy. In this study, we focus on the two latter of these effects: changes in internal and external political efficacy. Internal efficacy is the belief that one is capable of effective political action and self-governance, while external efficacy is the belief that governing officials listen to the public and that there are legal ways to influence governing decisions (Niemi, Craig, and Mattei 1991).
Previous findings about the effects of deliberation on participants' internal and external efficacy are rather inconclusive. Several studies have found deliberation to have positive effects manifested as increased levels of efficacy (Grönlund, Setälä,  Grönlund , Setälä, and Herne (2010, 99) note that external efficacy might increase since participants become aware of the complexity of political issues and gain a greater understanding of the actors and procedures of representative democracy. Setälä (2017) also argues that mixed deliberation gives citizens a chance to interact with politicians, which adds weight to the process and can help to increase their political efficacy (Setälä 2017 see also, e.g. Ryan 2014; c.f., Geissel and Hess 2017). Tentatively, this might be due to the simple fact that the politicians can explain things hands on in the deliberations. On the other hand, it is entirely possible that mixed deliberations reveal that politicians do not live up to expectations that could result in lower external efficacy. Flinders et al. (2016), for instance, found no effects on efficacy in their study of mixed deliberation. Moreover, internal political efficacy might increase among participants in deliberation as they 'practice' taking part in a political process themselves (Gastil 2000, 358;Grönlund, Setälä, and Herne 2010, 98). This mechanism might also be especially relevant in mixed deliberation since the presence of politicians might make this 'practice' even more real, and citizens may directly see that they are equal to politicians in the process. However, Mutz (2006, 358), argues that being challenged and confronted with opposing views may cause doubts in one's position and beliefs. Since politicians in mixed deliberation often have a higher deliberative capacity than citizens (i.e. prior knowledge and experience in argumentation), it is plausible that opposing, well argued, views from politicians cause citizens to doubt their own capacity even more.
Considering, finally, effects of mixed deliberation on efficacy according to status and gender, previous research is rare. While several studies consider mixed deliberation and efficacy, they seldom specifically compare how efficacy for citizens and politicians respectively are affected by deliberation, although an underlying citizens' perspective is usually discernible. There is also a research gap about differences in deliberation effects due to gender. Karpowitz and Mendelberg (2014, 193) suggest that women's sense of efficacy is more sensitive compared to that of men. However, Himmelroos, Rapeli, and Grönlund (2017) found no gender differences in changes in levels of internal efficacy due to deliberation. The rather scarce empirical research and inconclusive evidence give us reasons to formulate a second research question instead of hypotheses regarding the effects according to status and gender of mixed deliberation on efficacy: RQ 2: How do the effects of deliberation on internal and external efficacy differ according to gender and status?
Context, design, methods and data The context In the fall of 2017, the municipality of Korsholm in Western Finland started negotiating about a possible municipal merger with the neighboring city of Vasa. The purpose of the negotiations was to present an agreement for the municipal council to decide on. Korsholm has about 19,000 inhabitants, of which a majority have Swedish as their native language (69%) and a minority of native Finnish-speaking citizens (29%). In Vasa, with 68,000 inhabitants, these language relationships are the opposite, as 69% are Finnish-speakers and 23% Swedish-speakers.
In the fall of 2017, after a heated and polarized debate, the council in Korsholm voted in favor of initiating negotiations with Vasa (26 votes in favor, 17 opposed). The merger issue was especially contested for several reasons. The most crucial issue was that a merger would turn the Swedish-speaking majority in Korsholm into a minority in a new, merged municipality. The merger was thus at the core of the sensitive language debate in bilingual Finland. It was also an issue that directly affected all citizens in the municipality, everyone had a personal connection to it. The issue strongly divided both citizens and politicians, which spurred impassioned debates in traditional and social media. In fact, in a separate survey distributed to all citizens in Korsholm (N = 6686, response rate 43.4%), opinions about the merger, on a scale between zero (totally against) and ten (totally in favor), was bimodal (skewness 0.06 and kurtosis -1.60). About 48% were against the merger and 44% in favor, while the remaining 8% did not express any preferences in neither direction. For many citizens, the mere decision to start negotiating stirred up emotions. The issue received a lot of media coverage citing the conflict between proponents and opponents of a municipal merger.

Design
For the purposes of this study, we conducted a series of three deliberative discussions, with mixed groups containing both citizens and politicians (N = 225), together with the municipality. The deliberative discussions were part of the official municipal hearings of citizens prior to negotiations about the merger with Vasa. The details of the deliberative discussions are described next.

Procedure and participants of the deliberative discussions
Since the deliberative discussions had an official status, we did not enforce any experimental treatments in the deliberative discussions, and all of them were thus conducted in small-n groups with facilitation and discussion rules. However, we used a pre-test posttest experimental design in the small-n deliberation groups (see Gribbons and Herman 1997). Participants thus answered a pre-test questionnaire prior to deliberation (abbreviated to T1 henceforth) and immediately after deliberating (T2 henceforth). Besides basic demographics, these surveys contained items on internal and external efficacy (see Appendix for details). By studying pre-test to post-test changes regarding efficacy, we can answer RQ2 regarding how mixed deliberation affected efficacy according to status and gender. The T2 survey also contained items on self-reported discussion quality which will be one part, objective measures of discussion quality being the other, of our testing of hypothesis twothat is politicians' speech acts will be of higher deliberative qualityand the second research question on whether there are differences in discussion quality according to gender.
The deliberative discussions were held on three occasions, on 31 January (N = 92, divided into 8 groups), 13 February (N = 94, 10 groups) and 20 February (N = 39, 4 groups) in 2018. Since the municipality is geographically vast, these events were held in different parts of the municipality to give all of the municipality's citizens a chance of attending an event. Each event was open to all interested; there were no random samples of invited citizens. This resulted in some overlap between the events so that some citizens attended several events and likewise for the politicians. As to be expected, when no random sample was used as a basis for inviting participants to deliberate, the sample of citizens deliberating was skewed compared to the population in general (see Appendix B). This skewness is important to keep in mind when analyzing the actual deliberations since it sets the premises unequally. Women, for instance, were always in minority within the deliberation groups as were politicians.
The template was the same for each deliberation event: after a brief welcoming speech and information about the process of municipal negotiations of the merger held by the municipality's chief executive officer, participants at each of these three events were randomly allocated to the small-n groups (event one: 8 groups, event two: 10 groups, and event three: 4 groups). We used stratified randomization to ensure that each small-n group contained a mix of politicians and citizens (Suiter, Farrell, and O'Malley 2016;Setälä 2017;Flinders et al. 2016). The typical ratio between citizens and politicians was 1:3 in each discussion group, as was the ratio of women to men. Each small-n group had an assigned facilitator and participants were initially given time to answer the T1 survey, to read a brief information package about the merger, and to get acquainted with the rules of discussions. Discussions lasted for about two hours after which the groups summarized the most important aspects brought up in a discussion regarding the proposed municipal merger and each participant then answered the T2 survey. The rules and tasks for the facilitators were designed to steer the deliberation processes as close to the normative ideals of deliberation discussion as possible (e.g. Landwehr 2014). The rules essentially supported the ideals of reasoned justifications, reflection, sincerity and respect, whereas the facilitator made efforts to ensure reciprocity, inclusion and equality of discussion. The facilitators were graduate political science students who were trained for the task in two sessions. Some of them also had previous experience in facilitating deliberative discussions.

Measures
Since there are several different ways through which the study will test the hypotheses and answer its research questions, we provide a summary of the central elements of the design of the study discussed thus far and demonstrate how these elements will be measured (see Appendix A and C for details). Most of the measures were straightforward, such as counting the number of speech acts, measuring their length in the number of words, and noting whether the speaker was a citizen, politician, man or woman. Discussion quality, on the other hand, was a more challenging aspect to measure. To measure deliberative quality (coding scheme in Appendix A), we relied on an adaptation of well-established measures of discussion quality (see, e.g. Jensen 2003;Stromer-Galley 2007;Graham 2009;Steenberger et al. 2003). In earlier work on discussion quality (e.g. Strandberg and Berg 2013, 136;Strandberg and Berg 2015, 172), this scheme has achieved high inter-coder reliability with .93 in percentage-agreement and Krippendorff's alpha for sub-measures ranging from 0.7 to 1.0. Internal and external efficacies were measured using standard survey items as both summarized efficacy scales and as single survey items. We opted for also studying single efficacy items in order to explore in detail which aspects of internaland external efficacy that taking part in mixed deliberation potentially affected (see Table 4 for items).

Findings
We present our findings in the order of the hypotheses and research questions. We thus start with analyzing dominance in the discussions (Hypothesis H1a and H1b). Table 1 presents the share of all speech acts and the share of all words spoken in the deliberations according to gender and status.
The findings show that politicians are overrepresented, compared to their share of participants in the deliberation, in speech acts and in the share of the amount of words of all speech acts. Conversely, citizens are underrepresented regarding these two aspects. Hypothesis H1a is thus confirmed. It is, however, important to note that one explanation for why politicians dominated was that they were often on the receiving end of questions from the citizens. Turning to gender, there were no major differences regarding dominance, which means hypothesis H1b can be rejected. Thus, our findings were in contrast to Himmelroos' findings from a deliberation in Finland (2017,(14)(15). A separate analysis of the self-reported assessments (measured in the post-discussion survey) of equality in the discussions shows no significant differences according to gender nor according to status. Hypotheses 1a and 1b are thus not confirmed in light of how participants experienced dominance in the discussions.
We now turn to analyzing the discussion quality of the deliberative discussions with a focus on whether politicians achieved a better deliberative quality in their speech acts (Hypothesis 2). Additionally, we explore whether there were any differences in deliberative quality according to gender (RQ1). Deliberative quality is operationalized using both objective measures of discussion quality (Graham 2009;Jensen 2003;Kies 2010;Papacharissi 2004) from content analysis of the transcriptions of the deliberations (Table 2) and self-reported assessments by the participants (reported in text). The findings largely confirm Hypothesis 2 since there were several statistically significant differences in deliberative discussion quality whereby politicians achieved a higher quality. The largest differences are in external validations and justifications. Turning to explore gender differences, we see an overarching pattern that women tend to achieve a higher deliberative quality in their speech acts. The only exception is for the use of justifications where a higher share of men used qualified or sophisticated justifications in their speech acts. Men also displayed a significantly higher share of negative reactions, negative tone, incivility and impoliteness than women did. We continue by looking at how the participants themselves experienced the discussion quality, which yielded only one significant difference according to statuscitizens felt that new facts had been brought up during the deliberations to a higher extent than politicians didand no significant difference at all according to gender. Therefore, hypothesis 2 is only supported in light of the objective measures of discussion quality but not in light of the self-reported measures. The same assertion can be made regarding differences according to gender, which only manifested themselves in the objective measures of deliberative quality.
We end our presentations of the findings by analyzing the effects of deliberation on the participants' feelings of internal and external efficacy according to status and gender. Although deliberation has generally been found to increase especially internal efficacy (see overview in Boulianne 2019), we did not formulate any hypotheses regarding whether such effects differ according to either status or gender. Rather, RQ2 is concerned with exploring whether such differences exist or not. We start, then, by looking at the differences in pre-to post-deliberation feelings of internal-and external efficacy according to status in Table 3. As shown in the efficacy scales in Table 3, neither the citizens nor the politicians experienced any significant changes in efficacy. Looking at the single items measures, the politicians' feeling of internal and external efficacy were not significantly affected by deliberating. For citizens, however, two significant changes were observed: firstly, there was a slight decrease in the feeling that oneself knows more about politics than most people. Secondly, there was an increase in feeling that people can exert influence through voting. We now focus, finally, on the same changes according to gender (Table 4).
The findings according to gender show that men agreed significantly less with the statement that politics sometimes seems so complicated. There were no other changes for men. Two significant changes were observed for women: they agreed less with the statement that they have no say on what the municipal council or executive board decide, and they agreed less with the statement that the people's opinions are taken into account through the parties' decision-making. The efficacy scales did not change significantly for men or women.

Discussion and conclusions
In this study, the purpose was to study equality in deliberation according to status (i.e. citizen or politician) and gender. Dominance, deliberative discussion quality, internal, and external efficacy were in focus. Our context was a series of mixed deliberations about a proposed municipal merger in the Finnish municipality of Korsholm, which arguably was a very a contested issue at the time of data collection in 2018. As discussed in the Note. The efficacy scales have been recoded so that a higher value corresponds to higher level of efficacy. The single items range between 0 and 3. *p < .05 (Paired-samples t-test).
design section, the sample of deliberators was skewed compared to the general population in the municipality in question. Interestingly, the skewed sample provides an interesting setting regarding status and gender, our two focus variables, whereby both politicians and women were in a clear minority in the deliberation groups. Sunstein (2002; see also Farrar et al. 2009;Gastil 2000), for instance, has shown how people who are in the minority tend to alter their behavior according to the group majority. Hence, we will consider this when discussing our main findings next. When examining dominance in the deliberations, we found clear differences according to status but not according to gender when looking at the objective measures of dominance. We thereby confirmed hypothesis 1a that politicians would dominate but not hypothesis H1b that men would dominate. We already pointed out that politicians often found themselves answering questions from citizens, which partly explains why they dominated. We also showed that although politicians dominated discussions, the deliberative quality of their speech acts was higher than those of the citizens, hence raising the discursive quality of the deliberations. Thus, our findings partly support earlier findings that politicians dominate in mixed deliberations (e.g. Flinders et al. 2016;Minozzi et al. 2015) while nevertheless indicating that good facilitation can partially overcome negative aspects of domination (c.f. Afsahi 2020;Farrell et al. 2020;Suiter, Farrell, and O'Malley 2016).
The fact that men did not dominate is interesting given that they constituted over 70% of the deliberators, and earlier research, including research on deliberations in Finland (e.g. Himmelroos 2017), would suggest that men often dominate. Since all discussions were set up with moderation and deliberative rules, maybe these procedural safeguards helped curtail dominance to some extent (e.g. Karpowitz and Mendelberg 2014;Showers, Tindall, and Davies 2015). Nevertheless, this is not the whole explanation since the Finnish deliberation analyzed by Himmelroos (2017) also used these procedural safeguards. A plausible explanation is one provided by Karpowitz, Mendelberg, and Shaker (2012) whereby gender equality in deliberation, even when women are in a minority, works best in homogenous and unanimous discussion groups. When we apply this to our deliberations, it seems a good fit as our discussion groups were unanimous in terms of an overwhelming share of participants being against the proposed municipal merger.
Our findings regarding objective measures of deliberative discussion quality confirmed hypothesis 2 whereby, as we discussed earlier, politicians achieved higher quality than citizens did. Since politicians have more prior knowledge on the issue and are used to deliberating and justifying their positions in their work in the municipal council (c.f Gerber and Mueller 2018;Siu 2009), this was hardly surprising. Regarding differences in discussion quality according to gender, we discussed earlier that existing research findings were inconclusive (e.g. Caluwaerts 2012;Himmelroos 2017;Jennstål 2019;. In our analyses, the main pattern was that speech acts by women deliberators were of higher quality than those of men. Men, on the other hand, tended to score higher in measures that signal opposition and negativity than women did (c.f Afsahi 2020). Nevertheless, it should be noted that the overall level of negative elements in the deliberation quality was low, possibly due to facilitation (Beauvais and Bächtiger 2016;Curato et al. 2020). When judging discussion quality from self-reported measures, citizens reported having heard about new facts during the discussion to a significantly higher extent than politicians did, but no other differences were found. Thus, even though it has been argued that high-quality deliberation is only possible about issues that participants do not care strongly about (e.g. Parkinson 2006), the findings of this study show that such issues do not inevitably mean poor deliberative quality either.
Our final empirical focus, regarding internal and external efficacy, generally did not reveal large changes due to deliberation, which is perhaps to be expected given that the deliberations were one-off events lasting two hours. While the differences according to status did show two significant changes for citizens and not politicians, the direction of the changes was somewhat surprising. Firstly, after deliberation, citizens felt even less that they knew more about politics than most people do. Potentially, to explain this, a municipal merger issue is so complex that the deliberation on it made citizens realize that their own knowledge of politics is limited. Secondly, in the post-deliberation measure, citizens increasingly agreed that people could exert influence through voting. When analyzing deliberation's impact on efficacy according to gender, we found that both men and women increased their internal efficacy. Men agreed less with the statement that politics is sometimes too complicated, and women agreed less with the statement that they have no say on what the municipal council decides. Women also changed their opinion on one item concerning external efficacy whereby they agreed less with the statement that people's opinions are taking into account in the parties' decision-making. On balance, thus, our findings hardly serve to provide any clear evidence to the already inconclusive research regarding effects on efficacy in mixed deliberations (c.f. Flinders et al. 2016;Geissel and Hess 2017). Moreover, as the deliberations were held over a highly contentious issue (Setälä et al. 2020;Strandberg and Lindell 2020), the findings may not be generalizable beyond contentious issues.
As deliberative discussions are increasingly incorporated into political processes and institutions, the frequency of mixed deliberations is bound to increase. This development means that it also becomes increasingly important to gain more knowledge about mixed deliberation in order to find optimal ways for conducting mixed deliberation and achieve equality in these deliberations. The present study focused on two aspects of equality in mixed deliberation, status as a politician or citizen as well as gender. Our findings regarding dominance, discussion quality and effects on efficacy mainly suggest that mixed deliberation between citizens and politicians is bound to have some differences between citizens and politicians, although organizers employ facilitation and discussion rules. Deliberative mini-publics, where a representative random sample deliberates, are even likelier to display demographic equalitynot only in terms of who takes part but also in terms of what happens in the deliberations. Nevertheless, systematic comparisons of mixed deliberations under various circumstances are needed to establish this.
Some remarks on limitations with our study are due as well. First, the deliberation events studied here were not pure academic experiments as they featured a connection to actual decision-making by being part of the official merger enquiry of Korsholm. While this strengthens the external impact of the events and perhaps the external validity of our findings, it limits our possibility to employ an experimental design to causally test the effects of mixed versus non-mixed deliberation. Instead, our purpose here was to study equality within mixed deliberation according to status and gender. We had to take into consideration that each citizen had to be given the same opportunity to meet politicians and expected to discuss with politicians and citizens rather than only with other citizens. Participants in the study were self-selected, which resulted in a disproportional representation of the citizens in the municipality. However, we tried to minimize systematic differences between the small-n groups by using stratified randomization. Moreover, as mentioned, the controversial nature of the issue discussed can mean that our findings may not generalize beyond contentious issues.
We end with a final reflection on the central premise of mixed deliberation, that citizens and politicians mix in deliberative discussion and how that relates to equality in deliberation. We would argue that it is unrealistic to expect perfect equality between citizens and politicians in mixed deliberations. Differences between citizens and politicians are bound to exist due to the premises from which these two groups enter a deliberation and these need to be taken into account when designing such deliberations. Citizens seldom have previous experience with debates and argumentation, have less knowledge about the issue, and feel less confident in themselves, whereas politicians are already accustomed to debating and arguing their point, have in-depth knowledge, and are more confident in their own political competence. This, however, is not necessarily a negative thing. One should keep in mind that the mere opportunity for citizens to meet politicians and discuss an issue of relevance has democratic value in itself (Button and Mattson 1999;Strandberg and Berg 2019). Mixed deliberations and equality therein are nevertheless clear avenues for future research to explore systematically in order to learn more about key factors for functional deliberation between citizens and politicians. Such research would be important steps towards a functional coupling of deliberation and decision maker and, thus, arguably, key components in a functional deliberative system.

Notes on contributors
Kim Strandberg is a professor of political science and political communication at Åbo Akademi University. His research concerns democratic innovations and especially deliberative innovations. His and his co-authors' article about deliberation and polarization won the International Political Science Association's 2020 Meisel-Laponce Award for the best article in the last four years.
Janne Berg (Ph.D.) is a university lecturer (acting) at Åbo Akademi University. He has published several articles about participatory and deliberative democratic innovations. Methodologically, his expertise is in content analysis and experimental studies.
Thomas Karv (Ph.D.) is a junior post-doc researcher at Åbo Akademi University. His thesis concerned public trust for supra-national entities. He has also taken part in conducting experiments in deliberative democracy at the local level.
Kim Backström is a doctoral student in political science at Åbo Akademi University who focuses on legitimacy and how it can be generated in various unfavorable contexts. He is an expert in content analysis using the discourse quality index and also familiar with survey data.

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

Funding
This work was supported by Academy of Finland [294411] and Högskolestiftelsen i Österbotten [OstroInvolve 2.0].