Introduction
Confronted by a widely perceived crisis of representative democracy, many countries around the world struggle with instilling trust into their citizens and evaluate how existing political institutions might be complemented by democratic innovations. These range from more direct forms of participation, such as referendums, to more deliberative ones, such as sortitioned citizen assemblies, also known as citizen panels (Smith, Reference Smith2009). The starting point for most of these attempts is the individual citizen and her/his personal disaffection. Yet, there is a second layer to the problem of democratic disaffection, or indeed an aspect that could become a new problem if purely individual-based solutions are adopted: the cultural group dimension. In this paper, we theorize and provide evidence that in certain contexts, this additional layer is also important for understanding citizen preferences for different forms of democracy.
To understand why, consider that in states with politically significant national or linguistic diversity, i.e., where sub-state communities nourish distinctive political identities along linguistic or national lines, we not only witness disagreement over how political decisions ought to be made to be perceived as legitimate, but also over who should make them – with the elites of national or linguistic communities competing for central and/or segmental decision-making authority (Mueller, Reference Mueller2024; Shair-Rosenfield et al., Reference Shair-Rosenfield, Schakel, Niedzwiecki, Marks, Hooghe and Chapman-Osterkatz2021). Decentralization and federalism have long been advocated as remedies for territorially divided societies (Erk, Reference Erk2008). While the nature and cohesion of their collective identity, their territorial concentration as well as their political power can vary, all such communities aspire to defend their interests politically. But to what extent this then translates into varying preferences for electoral, deliberative and direct democracy has so far not been investigated – neither within nor across culturally diverse countries.
Cultural majority/minority relations clearly matter for “classic” direct-democracy questions. In Belgium, for example, no referendum has been held since a Dutch-speaking majority outvoted the French-speaking minority in a consultation on the return of the King in 1950 (Deschouwer, Reference Deschouwer2012). Similarly, in Northern Ireland, Unionists who favour the region to remain part of the United Kingdom look with fear at the demographic change that will soon hand the numerical majority over to Republicans favouring unification with Ireland. This would then allow the latter to outvote the former in a referendum foreseen by the Good-Friday Agreement (Coakley, Reference Coakley2022). In Switzerland, 50.3% of voters famously refused to join the European Economic Area in 1992, with the German-speaking majority outvoting the French-speaking minority. Although in Switzerland, such sharp German-French divisions are rather the exception than the rule, they do occur from time to time – and when they do, become politically significant precisely for these reasons (Mueller and Heidelberger, Reference Mueller, Heidelberger, Schaub and Bühlmann2022; Stojanović, Reference Stojanović2021).
These examples all share a theoretically sound, common pattern: citizens judging the desirability of a specific form of decision-making also based on whether it (dis)advantages them as a national minority or majority. Despite its theoretical soundness and political significance, this issue has hitherto been mostly absent from both the literature on democratic preferences and innovation as well as from that on ethnic and territorial politics.
To be sure, a substantial body of empirical research has investigated whether and why citizens support different forms of democratic decision-making, in general, and how they would like important political decisions to be made, in particular (Bedock and Pilet, Reference Bedock and Pilet2023; Goldberg et al., Reference Goldberg, Lindell and Bächtiger2024). Such investigations build on broader work that analyses the preferences for different models – or at least conceptions – of democracy among citizens (Bengtsson and Christensen, Reference Bengtsson and Christensen2016; Coffé and Michels, Reference Coffé and Michels2014; Font et al., Reference Font, Wojcieszak and Navarro2015; Hibbing and Theiss-Morse, Reference Hibbing and Theiss-Morse2002; Landwehr and Harms, Reference Landwehr and Harms2020; Neblo et al., Reference Neblo, Estering, Kennedy, Lazer and Sokhey2010; Webb, Reference Webb2013; Yamaguchi et al., Reference Yamaguchi, Hata and Inoue2024). They pay particular attention to the winners and losers under these different forms of democracy (Bowler et al., Reference Bowler, Freebourn, Teten, Donovan and Vowles2023; Ferrín and Hernández, Reference Ferrín and Hernández2021; Vittori et al., Reference Vittori, Rojon and Pilet2024; Wu and Wu, Reference Wu and Wu2022; Werner, Reference Werner2020). But to what extent and how exactly such winner/looser expectations map onto cultural diversity and different forms of democracy has so far received only little attention. Even Hänni (Reference Hänni2017), who looks at the effect of perceived and real policy responsiveness among ethno-national minorities, remains at the general level of satisfaction with and support for democracy overall as well as the likelihood of ethnic protests.
Our aim here is to merge and take these debates forward by investigating whether, and if so why, citizens belonging to national or linguistic minorities have different preferences than majority members for how democratic decisions should be taken. This question is of both theoretical and normative relevance. Theoretically, if such group-level factors matter, the tendency in the literature to focus on individual-level attributes misses an important part of explaining democratic preferences – at least wherever there are in fact such groups, i.e., in states with politically significant national and linguistic diversity. Normatively, if substantively different democratic preferences do indeed exist across majority and minority groups, any reform of democratic institutions needs to take such divergences into account to deliver on greater inclusiveness (Gherghina et al., Reference Gherghina, Mokre and Miscoiu2021). Otherwise, increased legitimacy among majority members simply comes at the expense of lower levels among minority members, potentially even compounding existing inequalities (Hänni, Reference Hänni2017).
In this paper, we examine the importance of two types of group-level factors: citizens’ objective minority position, i.e., belonging to a national or linguistic community that constitutes a numerical minority in their country, and their subjective minority position, i.e., the perception that their cultural community constitutes a political minority. We assess and compare the relevance of these factors for citizens’ preferences towards three forms of democratic decision-making: (i) electoral, (ii) deliberative and (iii) direct. While democracy can be conceived in further varieties along different dimensions (Held, Reference Held2006), these three are ideal-typical forms of democracy regarding who holds decision-making power: elected representatives, sortitioned citizens, or the citizenry in its entirety, respectively.
Empirically, we draw on an original online survey conducted in late 2020 in three countries that comprise politically significant national or linguistic sub-state communities. In two of these, Belgium and Canada, the groups stand out numerically and regularly compete politically; while in Switzerland, language is occasionally politicized and, if so, becomes politically significant (Mueller and Heidelberger, Reference Mueller, Heidelberger, Schaub and Bühlmann2022). In each country, we surveyed the two largest language groups: Dutch- and French-speakers in Belgium, English- and French-speakers in Canada and German- and French-speakers in Switzerland.
The remainder of this article is structured as follows. Section 1 reviews the literature on citizens’ support for different forms of democratic decision-making. We then theorize possible causes for differences in support for electoral-representative, direct and deliberative democracy based on citizens’ objective and subjective minority positions. Section 2 describes our data, operationalization and method. Section 3 presents the results, which are discussed alongside the conclusions in section 4.
While we do find differences in democratic preferences based on respondents’ objective and subjective minority positions, they differ from our initial expectations. Instead of comparing the advantages of different forms of decision-making for one’s group, citizens rather seem to reason from the status quo, which they want to preserve if it advantages them and their group, and to alter if it does not. Furthermore, some group-effects are outweighed or at least equalled by individual-level factors – mainly satisfaction with the current functioning of democracy, economic well-being and political ideology. So, even in states with politically significant national or linguistic sub-state communities such as those analysed here, individual-level variables remain strong predictors of democratic preferences. Before being definitive, however, this insight needs to be verified in analysing whether and how citizens’ actual experience with alternative forms of democracy changes this picture and what will happen once such questions become politicized along cultural group-lines.
Theorising the determinants of democratic preferences: why minority position and perception could matter
In the wake of increasing criticism of the functioning of electoral representative democracy, political theorists (Mansbridge, Reference Mansbridge1983) and practitioners (Crosby et al., Reference Crosby, Kelly and Schaefer1986) have started to think of and implement new forms of democratic decision-making. Aimed at increasing citizen participation beyond elections, these are referred to as ‘democratic innovations’ (Smith, Reference Smith2009). While a broad variety of democratic innovations has developed, scholars often distinguish ‘direct’ from ‘deliberative’ innovations. By associating citizens more closely to decision-making, these go beyond the traditional distinction between majoritarian and consensus democracy (Lijphart, Reference Lijphart2012). Direct democracy aims at increasing the inclusiveness of the democratic process, often embodied by a referendum. Deliberative democracy, in turn, aims at improving reason-giving and debate quality, often embodied by a deliberative citizen assembly or panel.
Beyond their contemporary rediscovery as democratic innovations, deliberative and direct forms of democratic decision-making are in fact as old as democracy itself, with the first small city states in ancient Greece and India adopting direct and deliberative institutions (Hansen, Reference Hansen1991). It is only after the English, American and French revolutions that electoral representative institutions were adopted almost universally among modern democracies to accommodate both the large size of newly formed nation-states and modern political thought (Manin, Reference Manin1997). While further forms of democracy can be conceived along different dimensions (Held, Reference Held2006), representative, deliberative and direct democracy are the ones that most ideal-typically differentiate as to who holds decision-making power: representatives, sortitioned citizens, or the electorate in its entirety, respectively.
In the wake of the contemporary proliferation of deliberative and direct democratic innovations, an increasing body of empirical research has examined support for different forms of democratic decision-making (e.g. Pilet et al., Reference Pilet, Bol, Vittori and Paulis2023). We contribute to this literature by investigating in greater detail to what extent who should hold decision-making power is judged by citizens not only based on individual characteristics, but also based on their collective majority/minority positions.
Individual-level factors: satisfaction with electoral democracy, political ideology and economic background
The literature on citizens’ preferences vis-à-vis democratic innovations has hitherto focused on individual-level determinants. Three factors in particular were found to be relevant. First, personal satisfaction with the current functioning of electoral democracy matters. Beside very dissatisfied citizens, higher support for democratic innovations was sometimes also found among very satisfied ones (Bedock and Pilet, Reference Bedock and Pilet2023; Neblo et al., Reference Neblo, Estering, Kennedy, Lazer and Sokhey2010; Seyd et al., Reference Seyd, Curtice and Rose2018). Pilet et al. (Reference Pilet, Bedock, Talukder and Rangoni2024) showed that subjective dissatisfaction also extends to the objective ‘losers’ of representative institutions, i.e., voters of parties that are most of the time in opposition, as well as those who have a higher ideological incongruence with their parliament. Second, political ideology matters. While referendums were shown to attract higher levels of support among both leftist and radical right voters, citizen assemblies appeared to be above all endorsed by left-wing voters (Bengtsson and Mattila, Reference Bengtsson and Mattila2009; Jacquet et al., Reference Jacquet, Niessen and Reuchamps2022). Third, greater support for both referendums and citizen assemblies was found among lower-income citizens (Coffé and Michels, Reference Coffé and Michels2014; Neblo et al., Reference Neblo, Estering, Kennedy, Lazer and Sokhey2010).
While the main hypotheses of this paper are geared towards group-level factors, we control for these individual-level factors in our analyses and expect similar results. Doing so allows us to both re-examine the findings of existing scholarship with new data at the sub-national level and, more importantly, to assess the relative importance of group-level factors.
More recently, scholarship has moved beyond citizens’ democratic preferences in general by also scrutinising how the outcome of democratic innovations (or its anticipation) influences their support. It was shown, for example, that such support is higher among citizens who think that the referendum (Brummel, Reference Brummel2020; Werner, Reference Werner2020) or citizen assembly (Pilet et al., Reference Pilet, Bol, Vittori and Paulis2023) results in a decision in line with their own policy preferences. Or, as Werner (Reference Werner2020, p. 312) puts it: “If I’ll win it, I want it”. In fact, comparing reasons of support, Landwehr and Harms (Reference Landwehr and Harms2020) show that intrinsic motivations (i.e., referendums valued as democratic procedure) are often beaten by instrumental motivations (i.e., referendums valued because they produce favourable results). This echoes a larger body of literature on winner-loser effects on democratic preferences (Anderson et al., Reference Anderson, Blais, Bowler, Donovan and Listhaug2005), which finds government voters to be foremost driven by government responsiveness (Bowler, Reference Bowler2017) and winning voters to favour majoritarian electoral systems (Ferrín and Hernández, Reference Ferrín and Hernández2021) or, in autocratic regimes, the continuity of the regime (Wu and Wu, Reference Wu and Wu2022).
Since we are interested in citizens’ democratic preferences in general, we do not take into account outcome effects in this study. However, if one considers that the minority position of a national or linguistic community might influence the democratic preferences of its members by anticipation, precisely because they expect to be outnumbered through some innovations but not others (see also Hänni, Reference Hänni2017), the question we examine also speaks to the relevance of rational outcome anticipation for democratic preferences. Or, to paraphrase Werner (Reference Werner2020): if we (as a group) may lose it, then I (as a person) do not want it.
Group-level factors: minority positions and perceptions
On the group-level, most research on democratic preferences has so far been directed towards the political culture or experience of sub-state communities. Bühlmann et al. (Reference Bühlmann, Vatter, Dlabac and Schaub2013), for instance, have shown French-speaking Swiss cantons to lean towards the liberal-representative model, whereas German-speaking cantons lie closer to the direct-democratic ideal – up to and including the citizen assemblies still practised widely at local or even cantonal levels (Vatter, Reference Vatter2024). Although it is not per se the objective of this study to unravel the effects of personal or collective experience on democratic preferences, our case selection partly accounts for this by comparing across countries with different levels of experience with different forms of democracy.
As for the minority-majority position of communities or entities, scholarship has studied the effects of referendums on minority rights (Bochsler and Hug, Reference Bochsler and Hug2015; Bolliger, Reference Bolliger, Freitag and Wagschal2007; Vatter, Reference Vatter2011) and looked at the attitudes of disadvantaged groups (without group identity) vis-à-vis deliberative-democratic innovations (Talukder and Pilet, Reference Talukder and Pilet2021). Several articles also focus on attitudes among minority members resulting from policy responsiveness and descriptive representation (Hänni, Reference Hänni2017) or the effect of autonomy and power-sharing institutions on national pride and conflict (Bühlmann and Hänni, Reference Bühlmann and Hänni2012; Juon, Reference Juon2025). However, less attention has been paid to the opinion of national or linguistic sub-state communities vis-à-vis different forms of decision-making and its determinants.
Yet, such communities can be politically highly significant and challenge national decision-making (Keating, Reference Keating1996) – not only if they disagree with the content of a decision, but also with the way it has been taken. Beyond the political importance of such groups, the question also deserves attention for theoretical and normative reasons. Theoretically, if national and linguistic communities do indeed have different democratic preferences based on their majority-minority position, then not accounting for them leaves out a significant part in the story of democratic preferences and political attitudes. Normatively, in view of the rising number of democratic innovations across countries worldwide (OECD, 2021), potentially diverging opinions of national and linguistic communities need to be taken into consideration if one wants to fulfil democratic inclusiveness (Gherghina et al., Reference Gherghina, Mokre and Miscoiu2021). Taken together, our study contributes to the literatures on democratic preferences, innovation and reform by testing for the importance of national and linguistic diversity for democratic preferences. Doing so, it also speaks more broadly to the literatures on ethnic and territorial politics in that existing societal divisions could in fact be further aggravated through democratic reforms that operate along the simple majoritarian logic of “one person, one vote” (Juon, Reference Juon2025).
Hypotheses
Departing from the two trends in the literature to investigate outcome-oriented determinants at the individual-level (Werner, Reference Werner2020) and cultural considerations at the group level (Bühlmann et al., Reference Bühlmann, Vatter, Dlabac and Schaub2013), our main objective is to combine them by examining how outcome-anticipations at the group-level shape citizens’ individual preferences regarding democratic innovations. Following sociological theories of multinational and multilingual states (Keating, Reference Keating2001; Lipset and Rokkan, Reference Lipset, Rokkan, Seymour and Rokkan1967), we expect national and linguistic sub-state communities, despite their constructed nature (Anderson, Reference Anderson1983), to be a relevant frame of reference for many citizens (Zuber, Reference Zuber2013). Accordingly, citizens should not only reason about the outcomes of political processes in individual terms (“me”), but also in the collective terms of their community (“us”). Following rational-choice theories of institutional and democratic preferences (Hall and Taylor, Reference Hall and Taylor1996; Junius et al., Reference Junius, Matthieu, Caluwaerts and Erzeel2020), we thus expect citizens to anticipate the consequences of the type of decision-making for their specific group.
To begin with, referendums are a priori an instrument of the majority, since they are essentially based on a ‘one person, one vote rule’ (Altman, Reference Altman2010). This does not per se rule out centripetal effects, especially if held regularly and open to minority groups themselves (Stojanović, Reference Stojanović2021, pp. 66–72), nor the possibility of additional thresholds in view of minority protection (e.g. Freiburghaus and Vatter, Reference Freiburghaus and Vatter2024). But it does open the door to a tyranny of the (simple) majority, and even a highly legitimate one at that since “the people” will have spoken.
Deliberative citizen panels, in turn, aim to give equal weight to all salient subdivisions of society by composing them through socio-demographically stratified sortition (Goldberg et al., Reference Goldberg, Lindell and Bächtiger2024). To be sure, even in such assemblies minorities can be outnumbered, ignored and outvoted. However, the emphasis on consensus-oriented deliberation rather than head-counting and the priority of arguments and reason over status and affiliation operate against a purely majoritarian logic.
Finally, electoral democracy in both its majority/plurality and proportional variant relies on competitive elections with decisions taken by simple or compound majorities after adversarial campaigns and debates. Although this accords a larger weight to the majority in the final decision-making even in proportional systems (Hänni, Reference Hänni2017), minorities still have the possibility to voice their concerns throughout the decision-making process and pre- or post-electoral alliances can be more or less inclusive. Depending on the precise contours of political systems, groups may also be given additional power-sharing protection (e.g., through executive parity or alarm bell procedures as in Belgium, see Deschouwer, Reference Deschouwer2012). Finally, even majoritarian features such as simple plurality elections may work in favour of cultural minorities if these are territorially concentrated, e.g. in Quebec or Scotland (Mueller, Reference Mueller and Griffiths2023).
Nevertheless, in what follows, we believe it reasonable to assume that most citizens will be either unaware of the additional intricacies of different democracy models, or base themselves on the type and experience of their country and region. In either case, the difference regarding electoral democracy in Canada (with plurality voting and single-party governments) versus Switzerland and Belgium (list-proportionality and multi-party governments) disappears as minorities are adequately represented in all three. Assuming further that in states with politically significant national and linguistic diversity, group-identities matter and that citizens reason strategically about established or new forms of decision-making, we firstly hypothesize that:
H1: Citizens belonging to an objective national or linguistic minority in a state have, compared to members of the objective majority, (a) lower support for direct, (b) higher support for deliberative and (c) no difference in support for electoral forms of democratic decision-making.
When thinking about the mechanism through which majority/minority positions influence citizens’ preferences, it also seems reasonable that what matters is not so much citizens’ objective but rather their subjective minority positions, i.e., their perception that their linguistic group does indeed constitute a political minority. Accordingly, we hypothesize that:
H2: Citizens belonging to an objective national or linguistic minority, which they also perceive to form a political minority, have, compared to members of the objective majority, (a) lower support for direct, (b) higher support for deliberative and (c) no difference in support for electoral forms of democratic decision-making.
When objective and subjective minority-majority positions are diametrically opposed, it seems furthermore plausible that yet again different preferences ensue. In this case, numbers do not match power, either because the majority has too little or the minority too much influence. Notably citizens belonging to an objective minority but which they do not subjectively perceive to be minoritized (satisfaction despite numerical inferiority) should exhibit greater support for existing electoral institutions compared to other forms that might alter this situation. Among the two innovations on offer, they should however prefer deliberative over direct forms of decision-making because their numerical minority persists. We thus hypothesize that:
H3: Citizens belonging to an objective national or linguistic minority but which they do not perceive as a political minority have, compared to those who perceive their minority community also as a political minority, (a) higher support for electoral, (b) lower support for deliberative and (c) no difference in support for direct forms of democratic decision-making.
Conversely, citizens belonging to an objective majority but who do not perceive their community as such (dissatisfaction despite numerical superiority) might be at odds with the electoral institutions currently in place. Direct or deliberative forms of decision-making offer to alter this situation. However, they should prefer direct over deliberative forms of decision-making because of their numerical majority status. We hypothesize thus that:
H4: Citizens belonging to an objective national or linguistic majority but which they do not perceive as a political majority have, compared to those who do perceive their majority community also as a political majority, (a) higher support for direct, (b) lower support for electoral and (c) no difference in support for deliberative forms of democratic decision-making.
Finally, if the results of previous group-level studies on the importance of political culture, tradition and experience hold, the hypothesized effects should be moderated by the degree of experience that countries and minorities have with representative, direct and deliberative forms of decision-making – in two respects. On the country-level, preferences for direct forms of decision-making should on average be higher in states that regularly hold referendums (i.e., Switzerland), those for deliberative forms of decision-making should be higher in countries that regularly organize citizen assemblies (i.e., Belgium) and those for electoral-representative forms of decision-making should be higher in countries which organize no or far fewer referendums and citizen assemblies (i.e., Canada).
On the group-level, the hypothesized effects should be stronger for forms of decision-making that minorities have experience with, i.e., the negative effects of referendums should be stronger for French-speakers in Switzerland because they experience more regularly what it means to be outvoted, and the positive effects for citizen assemblies should be stronger for French-speakers in Belgium because they have had more exposure to their consensual nature. To account for this, beyond analyses on the aggregate level to test our main hypotheses, we will also run models that control for cross- and within-country differences to further look into such potential variation. The nature of this cross- and within-country comparison is purely explorative, however, since an actual analysis of the extent to which experience with different forms of democratic decision-making influences citizens’ attitudes towards them would not only require more countries but also a direct measurement of an individual’s actual degree and type of experience with the different forms.
Case selection, data and method
Case selection
We expect the hypothesized effects to be most relevant in states where national and linguistic sub-state communities are institutionally or socially structured on the sub-state level and regularly compete politically in the overall state. Institutional and social structures below the state level provide communities with their own public spaces and tools to develop and maintain a distinct political consciousness. At the same time, the more there is political competition at the national level, the more overall decision-making is thought of in terms of zero-sum games between sub-state communities and the more salient one’s group identity. If the hypothesized effects hold, it is in such ‘typical cases’ that we would expect to see them occurring most clearly. With our study being the first to test the effect of national-linguistic diversity on democratic preferences, it is in such typical cases that one would like the expectations to be tested first (Gerring, Reference Gerring2006).
Belgium, Canada and Switzerland all fit this profile very well. First, because their federal structure guarantees an awareness of sub-state spaces, identities and ensuing dynamics. Second, because their politics is more (Belgium and Canada) or less (Switzerland) dominated by cultural considerations structured along linguistic lines (see also Mueller et al., Reference Mueller, Dardanelli, Reuchamps, Bundi, Lecours and Niessen2024). Third, in all three cases the territorial delineation of language groups, coupled with federal structures and historical legacies, provides them with additional meaning and to some extent also reifies them (e.g., the Assemblée nationale of Quebec, or the Loterie Romande in Switzerland). All this makes it more likely for democratic preferences to be thought of in group-level terms, which in turn increases the chances that relative acceptance levels co-depend on majority/minority status and/or perceptions.
More specifically still, we focus on the two linguistic groups per country that stand out numerically and politically. We chose to study these in particular because they have the highest potential for national political competition between language groups to be perceived in terms of zero-sum games, rather than in terms of coalitions that could be formed across multiple communities (Duchacek, Reference Duchacek1988).
While the federal arrangements in all three countries comprise more than two sub-state entities and communities, all three regularly exhibit at least some form of political competition between the two largest language groups. In Belgium, Dutch- and French-speakers have been politically competing since almost the birth of the country in 1830. To pacify the struggle, the state was progressively decentralized and transformed from a central into a federal state along linguistic (and partly territorial) lines from 1970 until today (Deschouwer, Reference Deschouwer2012). Canada has witnessed similar developments with French-speakers in Québec seeking increased political autonomy for their province and, at different moments in time, even independence (Gagnon and Lachapelle, Reference Gagnon and Lachapelle1996). The competition between French- and English-speakers has furthermore twice resulted in very opposing stances in national referendums. In Switzerland, finally, although the federal arrangement is not set-up on a linguistic basis, and although national identification does not necessarily run along linguistic lines (Dardanelli, Reference Dardanelli, Requejo and Badia2012), language still matters politically – with specific rights being entrenched for each linguistic group in terms of public broadcasting, representation and regional cooperation, and with frequent political disagreements at both the federal level and in bilingual cantons (Stojanović, Reference Stojanović2021).
All three countries are thus typical cases for our research question by virtue of their politically significant national or linguistic sub-state communities. However, they have diverging experiences with different forms of democratic decision-making. This allows us to integrate the importance of experience with different forms of democracy in further analyses and to exploratively test if the effect of national-linguistic diversity on democratic preferences is indeed altered by different democratic experiences.
In Belgium, although it could be seen as an unlikely place for democratic innovations given its strong partitocratic grip and long history of linguistic tensions that have scared off most referendum proponents, numerous deliberative mini-publics have been organized at all political levels since the early 2000s. By now, half of Belgium’s constituent units have institutionalized deliberative democracy: the German-speaking Community with its Permanent Citizen Dialogue, Brussels and Wallonia with mixed deliberative committees bringing together regional MPs and randomly selected citizens (Vrydagh et al., Reference Vrydagh, Devillers, Talukder, Jacquet and Bottin2020).
Switzerland, in turn, has a long democratic tradition that combines representative with direct forms of democracy. Binding referendums, initiated either by the people or in reaction to government-proposed legislation, regularly occur at federal, cantonal and local levels – albeit with varying frequency sub-nationally (Vatter, Reference Vatter2024). Small German-speaking municipalities as well as two German-speaking cantons also have popular assemblies next to or instead of parliaments to legislate, where a mix of direct and deliberative democracy takes place publicly, whereas local parliaments are the norm in French-speaking Switzerland (Linder and Mueller, Reference Linder and Mueller2021). More recently, a series of deliberative mini-publics has been organized at local level (Kübler et al., Reference Kübler, Rochat, Yun Woo and Van der Heiden2020).
In Canada, finally, several provinces such as British Columbia and Ontario have pioneered the use of deliberative citizen assemblies – in particular for the reform of electoral systems (Warren and Pearse, Reference Warren and Pearse2008). After public significance in the early 2000s, the country has however not moved towards institutionalising such devices. As for referendums, they are foreseen in the Canadian constitution but are in practice only seldomly used.
In sum, all three countries have some experience with the three different forms of democracy studied here, but to varying degrees. Switzerland is by far the country with the strongest direct-democratic pedigree, Belgium stands out for its experience with deliberative democracy. Canada is located somewhere in the middle, with neither a strong direct nor deliberative tradition, despite at least some practice of both.
Data
We studied citizens’ opinions in the three countries with an online survey conducted in October/November 2020 (first and only wave), whose broader objective was to study democratic preferences and political attitudes in multi-lingual countries.Footnote 1 The survey was identical in each country and we made sure to reach a representative sample among the two largest linguistic communities. Table 1 provides a breakdown of key socio-demographic characteristics – age, gender, education – for each language group. Despite minor variation, all six groups are about identically composed along those dimensions. What is more, we are less interested in understanding the views of, for instance, ‘the Walloons’, as we are in understanding the determinants of such views within and across groups and countries.
Table 1. Sample properties by group and key socio-demographics
Variables and method
To operationalize our dependent variables, respondents were invited to position themselves on a scale from 0 (fully disagree) to 10 (fully agree) on whether important political decisions should be taken by (i) elected politicians, (ii) randomly selected citizen panels, or (iii) referendum. We framed the question around ‘important political decisions’ to capture respondents’ democratic preferences for major decisions. In the absence of strong correlations between the three variables,Footnote 2 we concluded that the three forms of democracy were captured distinctively and that they could be analysed as such, without restrictions on combinations of high/low values.
To operationalize our two independent variables, we proceeded as follows. We captured respondents’ objective minority positions by asking for their mother tongue. We only retained respondents whose mother tongue was one of the two main languages of the country (i.e., French vs. English, Dutch or Swiss-German). To assess respondents’ subjective minority position, we asked to what extent they thought French-speakers were influential in national politics on a four-point scale. For French-speakers – who constitute an objective minority in all three countries – we used this question as direct measurement of subjective influence. For non-French-speakers, we used the question as indirect measurement by reversing the scale.
Online Appendix 1, which provides an overview on the distribution of objective and subjective majority/minority positions among our respondents, shows that 69% of French-speaking Canadians, 69% of French-speaking Swiss and 64% of French-speaking Belgians somewhat or strongly agree that in their country, French-speakers have little influence at federal level. Conversely, 70% of English-speaking Canadians, 70% of German-speaking Swiss and 77% of Dutch-speaking Belgians disagree. To us, these relatively uniform perceptions among both French- and non-French-speakers indicate that our question was understood similarly across both groups and countries. Furthermore, for about a third of all respondents, the subjective majority/minority perception differs from their objective majority/minority position, so there appears to be meaningful variation to investigate the hypothesized interaction effects (H3 and H4) as well.
To re-assess the validity of the individual-level factors found relevant in previous research and to understand the relative importance of our group-level variables of interest, we also captured respondents’ satisfaction with the current functioning of democracy, their political ideology and their personal economic situation. Thus, we asked respondents to indicate to what extent they were satisfied with the current functioning of democracy at the federal level in their country (on a scale from 0 to 10). To capture their political ideology, rather than asking about their political self-positioning, which is subject to evaluation biases, we asked respondents “which party do you usually vote for at federal elections, or which party is generally closest to you?”, and then classified parties according to their socio-economic left-right placement (see Appendix 3). Finally, to account for their economic situation, we asked respondents “how do you feel about your household’s income nowadays?”, with responses ranging from 1 (finding it very difficult on present income) to 4 (living comfortably on present income).
To account for the potentially heterogenous effects of respondents’ political efficacy and socio-demographics, we also asked for their political interest, external political efficacy, education, age and gender. All variables and their operationalization are listed in Table 2. Descriptive statistics can be found in Appendix 2.
Table 2. Variables included in the analyses and their operationalisation
Note: ‘Do not know’-answers were treated as missing values for all variables.
To test our hypotheses, we used Ordinary Least Square (OLS) regression analyses. To disentangle the effect of our main independent variables, the effect of the other independent and controls variables and to account for cross- and within-country variation, we built models progressively in three steps. In each step, we ran the regressions for each of the three dependent variables capturing citizens’ democratic preferences (DPR), i.e., their support for decisions taken by (1) elected politicians, (2) sortitioned citizen panels and (3) referendums.
First, as baseline models, we regressed citizens’ democratic preferences on their linguistic majority-minority position (LMM), interacted with their perceived group efficacy (PGE). Second, to take into consideration the other individual-level independent and control variables found relevant in the literature and to account for unobserved cross-country differences (CTR), we added these to obtain the final models based on which we test our hypotheses. Finally, to further explore within-country differences, we interacted the country variable (CTR) with respondents’ linguistic majority-minority positions (LMM). The equations for these three steps are as follows:
Step 1 – Baseline models:
$$DP{R_{1 - 3}} = \;{b_0} + \;{b_1}LMM + {b_2}PGE + \;{b_3}LMM:PGE + e \quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad$$
Step 2 – Full models:
$$\eqalign{DP{R_{1 - 3}} = \;{b_0} &+ \;{b_1}LMM + {b_2}PGE + \;{b_3}LMM:PGE + e \cr &+ \;{b_{4 - 6}}Further\;IVs + {b_{7 - 11}}Controls + {b_{12}}CTR + \;e}\quad\quad\quad\quad\quad\quad\quad\quad\quad$$
Step 3 – Full models with country and group interactions:
$$\quad\eqalign{DP{R_{1 - 3}} = \;{b_0} &+ \;{b_1}LMM + {b_2}PGE + {b_3}CTR \cr &+ \;{b_4}LMM:PGE + {b_5}LMM:CTR + {b_6}PGE:CTR + {b_7}LMM:PGE:CTR \cr &+ \;{b_{8 - 10}}Further\;IVs + {b_{11 - 15}}Controls + e}$$
Since the number of observations varied between baseline and full models and to ensure that potential effect changes are not due to lower sample sizes, we ran the baseline models by restricting their sample to that of the full models. Regression diagnostics were conducted and only heteroscedasticity was detected in some models, which is why we report robust standard errors for the respective models.Footnote 3
Findings
Before turning to the regression analyses, a glance at the descriptive distribution of respondents’ preferences for the three forms of democratic decision-making, summarized in Figure 1, appears useful. Across all respondents taken together, referendums are the most popular form of decision-making with an average support of 6.1 on the 0–10 scale, followed by elected politicians with an average support of 5.4. Sortitioned citizen panels only score 4.5, which is even below the scale’s mid-point of 5. This tendency is more or less robust across countries, with the exception of Canada where elected politicians remain preferred over referendums. Noteworthy is also the substantively higher support for referendums in Switzerland and for citizen panels in Belgium. This might be explained by the greater experience with referendums in Switzerland and with citizen panels in Belgium, as discussed above.
Figure 1. Mean and std.-dev. for the dependent variables by country and objective minority position.
Note: Country and overall averages are not weighted here and should be interpreted accordingly.
If we then look at these preferences grouped by respondents’ objective minority position, further tendencies appear. While the linguistic majorities in Canada and Switzerland tend to favour referendums more than their respective linguistic minorities, the opposite can be observed in Belgium. In turn, minorities are more in favour of electoral democracy than majorities in Canada and Switzerland – and again the opposite is the case for Belgium. Citizen panels, finally, gain more support by the linguistic majority in Switzerland than by the linguistic minority. Here, the opposite is true in both Belgium and Canada. While these are only descriptive observations whose statistical significance is scrutinized in the regression analyses below, they already suggest that respondents’ objective minority position is, on its own, an insufficient predictor of citizens’ democratic preferences.
When turning to the regression analyses, whose results are summarized in Table 3 and whose predicted values are plotted in Figure 2, we can already see in the baseline models (1, 2 and 3 – comprising only the main independent group-level variables) that citizens’ objective minority positions do not always run in parallel with their subjective minority perceptions – which explains the diverging democratic preferences. Elected politicians are above all supported by citizens with high perceived group efficacy regardless of their objective linguistic majority/minority status. Among numerical minorities, a high level of perceived group influence allows them to catch up, so to speak. Citizen panels, in turn, are above all supported by the members of the majority who also perceive their group as influential as well as by minority members who do not. The first have nothing to fear, the latter everything to gain from this particular democratic innovation. A similar interpretation can be made for referendums.
Table 3. OLS regression results for respondents’ democratic preferences (support for)
Note: * p < 0.05, ** p < 0.01, *** p < 0.001.
Figure 2. Predicted values for the OLS regression models.
Note: Confidence intervals = 95%. Y-scale restricted from 2 to 8 to optimize visualization.
When looking at the full models (4, 5 and 6 – comprising the main independent group-level variables as well as the individual-level and country controls), we see these tendencies validated. It appears, furthermore, that citizens’ democratic preferences are highly correlated with several individual-level variables. Their satisfaction with the current functioning of democracy positively influences their support for elected politicians, but negatively impacts their support for both citizen panels and referendums. The same goes for citizens’ economic situation, which – if difficult – decreases support for citizen panels and referendums alike. Concerning political ideology, compared to the voters of centre-parties, higher support for citizen panels is found among voters of far-left parties, while higher support for referendums is found among far-left, far-right, green and non-aligned voters. A glance at the other control variables shows higher support for elected politicians among citizens with a high political interest, which also correlates with lower support for citizen panels. Educated and older citizens exhibit lower support for citizen panels. Age is furthermore negatively correlated with support for elected politicians. These findings are in line with the existing literature and thereby increase our confidence in the external validity of our data, models and other results.
Taken together, these findings only partially corroborate our hypotheses. Regarding H1 and H2, where we expected objective and subjective minority members to be more supportive of citizen panels but not of referendums, we see that both democratic innovations gather stronger support among minority members if they have a low perceived group efficacy. Similarly, while we expected in H3 that the members of linguistic minorities who do not perceive their group as a minority are supportive of elected politicians and, to a lesser extent, of citizen panels, we see that they like referendums still better than citizen panels. As for H4, while we expected the members of linguistic majorities who do not perceive their group as a majority to be above all supportive of referendums and much less of elected politicians, we see that they are quite equally supportive of both. In sum, we do not find consistent evidence for our hypotheses.
One should furthermore note that the full models return a substantively higher R 2 than all the baselines, meaning that they explain much more of the variation in the dependent variables. Since the models rely on different variables and degrees of freedom, we cannot immediately conclude that the further independent variables (capturing individual-level variation) have a higher explanatory power than our main independent variables (capturing group-level variation). Instead, as a further test, we regressed respondents’ preferences for elected politicians, citizen panels and referendums in a multi-level null-model on their linguistic group and country (see Appendix 4). For elected politicians, we found 4.1% of the variance on the group-level (language group and country); for citizen panels, 1.9%; and for referendums, 5.4%. Taken together, this confirms that the amount of variation explained at group-level is substantively lower than at individual-level.
When exploring the variation at group-level further by looking at the full models with country and group interactions (nos. 7, 8 and 9 – interacting respondents’ linguistic majority/minority positions by country), we can see that the aforementioned tendencies are robust regarding citizen panels. Regarding support for elected politicians, however, we see less difference in support among minority members in Switzerland as well as among majority members in Belgium – regardless of their perceived group efficacy. When it comes to referendums, we see perceived group efficacy moderating the support among majority members to a lesser extent in Switzerland.
Discussion and conclusion
This article has delved into the determinants of citizens’ preferences for different forms of democratic decision-making by theorising and testing group-level attributes in three countries with politically significant linguistic minorities. Our goal was to enrich the literatures on democratic preferences and innovation, which have so far focused mostly on individual-level characteristics, as well as those on cultural diversity. We did so with original survey data collected in Belgium, Canada and Switzerland – three federal countries with politically significant national or linguistic sub-state communities, two of which stand out numerically and occasionally, if not regularly, witness political competition. Although these cases can be considered typical, our results provided only partial evidence for our hypotheses.
We found that citizens who subjectively perceive their political community as influential support any form of democratic decision-making – electoral, deliberative, or direct – if they also belong to an objective majority. Members of objective minorities, in turn, more strongly oppose deliberative or direct forms of democracy, but exhibit greater support for existing electoral institutions when they do not perceive themselves as political minorities, compared to when they do. Put differently, for citizens belonging to an objective majority, how political decisions are taken does not seem to matter much, provided they feel their community to be politically empowered. Members of an objective minority, however, only embrace newer forms of democracy if they perceive their group to be non-influential.
This can also be read as a sign that, overall, the democratic integration of the francophone minority in all three countries studied here is a success, no matter if frequent referendums are held (Switzerland) or elections dominate (Canada and Belgium) – possibly also because all three are multilingual federations. At the same time, these findings not only reveal the astonishing importance of country-level institutions, practices and traditions even in deeply divided societies, but also how democratic specificities such as FPTP in Canada or referendums in Switzerland can themselves become part and parcel of a shared, cross-linguistic and thus civic political culture.
Albeit different from our initial expectations, these tendencies are still in line with the sociological and rational-choice theories that guided our hypotheses: group-level factors appear to matter in that there seems to be strategic thinking along community lines. This thinking, however, is more general than we expected. Instead of comparing the advantages of different forms of decision-making for one’s group, citizens rather seem to reason from the status quo, which they want to preserve if it advantages them and their group. Only if they expect to be favoured by newer forms of decision-making do they support them. Once they think the status quo disadvantages them, they want it to be altered by any other form of decision-making – regardless of how advantageous it might be for them as a group.
Although the detected group-level effects are noticeable, they become outweighed or at least equalled by the individual-level factors found relevant in the literature: mostly satisfaction with the current functioning of democracy as well as economic well-being, and, in part, political ideology. Thus, even in states with important sub-state politics and politically significant national or linguistic sub-state communities, individual-level variables remain strong predictors of democratic preferences, while the explanatory power of group-level factors – although relevant – remains more limited. Interpreted positively, people in these countries are anything but caught in a communitarian mindset.
That being said, one should consider that our respondents were asked about their attitudes towards electoral, deliberative and direct-democratic institutions in general and regardless of their personal experience with either democratic innovations or the status quo. Nevertheless, in the only country with substantial direct-democratic experience among the three, Switzerland, average levels of support were observed as hypothesized: larger for referendums among objective majority members and without subjective interaction effects. Hence, it may well be that, for the expected effects to manifest, non-representative forms of democratic decision-making have to be actually used and, perhaps, even politicized over a longer time along group lines. Otherwise, the risk of being outvoted remains too hypothetical to have traction.
Taken together, these results add an interesting piece of evidence to the existing literatures, namely that even in countries with politically significant national and linguistic sub-state communities, and even if they find themselves in a bipolar position which is, per se, prone to competition, citizens’ democratic preferences seem as much driven by individual-level considerations as by group-level factors. This has two concrete policy implications. First, even in such contexts, democratic preferences seem to substantially transcend political communities and so would attitudes towards potential democratic reforms. There is thus potential to build cross-group consensus. Second, since group-level factors were still found to be relevant to some extent for some democratic forms and contexts, attention to them is warranted when deliberative or direct-democratic reforms are implemented in such countries – making sure that all segments of society feel respected by the decision-making rules that are adopted. Looking ahead scientifically, our findings call for a deeper look into whether and how citizens’ actual experience with existing and more innovative forms of democracy changes this picture over time – especially if and when politicized along group-lines.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S175577392510009X.
Acknowledgements
We would like to thank Coree Brown Swan, Rahel Freiburghaus, Sergiu Gherghina, Vincent, Jacquet, Michael Meffert, Jonathan Philips, Sebastien Rojon, Jonah Schulhofer and Wouter Veenendaal for their helpful comments on earlier drafts of this article. We are furthermore grateful for the constructive remarks of three anonymous reviewers. Any remaining imperfections are of course our own responsibility.
Funding statement
Research funding was provided by a University of Lausanne ‘Crédit de démarrage’-grant. Christoph Niessen was furthermore funded by a postdoctoral research grant from the Research Foundation – Flanders (FWO, n°1274924N).
Competing interests
We have no conflicting interests to report.
Open access
