Introduction
As instances of citizen participation beyond electoral democracy are spreading across the globe, assemblies play an increasingly important role in democratic practice today (Reuchamps et al. Reference Reuchamps, Vrydagh and Welp2023). Democratic assemblies can be defined as groups of citizens coming together to deliberate and sometimes also to decide public policy issues. As a tool of citizen participation, assemblies place relatively high costs on participants, as they require sustained physical presence and mental attention during a defined period. Turnout in democratic assemblies is therefore rather low. Studies of traditional assembly democracy, such as the New England town meetings (Bryan Reference Bryan2004), or the Swiss communal assemblies (Ladner Reference Ladner2002), report average participation rates between 5% and 10%. Turnout is also low in assemblies created in the wake of democratic innovations. In the participatory budgeting processes in Brazil, for instance, the highest participation in the yearly assemblies was reported at about 10% of a town’s total population in peak times (Baiocchi and Ganuza Reference Baiocchi and Ganuza2014: 35). These experiences fuel critiques of the core tenet of participatory democracy, namely the ‘romantic dogma’ (Warren Reference Warren1996: 243) that democratic participation is an attractive activity that people would naturally engage in if only they had the opportunity.
But aside from low turnout, assemblies also often suffer from participation inequality. This is a common problem in any form of political participation, but particularly so when turnout is low (Dacombe and Parvin Reference Dacombe and Parvin2021): citizens of lower socio-economic status, lower skills, less spare time and weaker political interest are less likely to attend assemblies. Proponents of deliberative democracy have therefore focused on opportunities for citizen involvement that are not fully participatory, but confined to complementing existing representative regimes (Bächtiger et al. Reference Bächtiger2018), such as deliberative mini-publics (Grönlund et al. Reference Grönlund, Bächtiger and Setälä2014). These often entail strategies of participant recruitment aiming to redress inequalities resulting from self-selection bias (Curato et al. Reference Curato2021). But in most instances, initial responses to invitations to participate in mini-publics were usually well below 10% (Fournier et al. Reference Fournier2011: 32; Caluwaerts and Reuchamps Reference Caluwaerts and Reuchamps2015; Jacquet Reference Jacquet2017). Recent evidence shows that, in spite of efforts to recruit inclusive groups of participants following this initial step, representation bias in assemblies persists: the better educated and the politically active are usually overrepresented (Karjalainen and Rapeli Reference Karjalainen and Rapeli2015; Boulianne Reference Boulianne and Hansen2018; Binnema and Michels Reference Binnema and Michels2022).
Thus, understanding the mechanisms of nonparticipation in democratic assemblies is crucial. Existing research on this topic (see the literature reviewed in the next sections) has mainly focused on individual-level predictors of willingness to participate or actual participation in such assemblies, isolating effects of sociodemographics, but also psychological, attitudinal and political variables. However, studies of the role of design features of assemblies in attracting (or deterring) potential participants are still scarce, particularly regarding the question of how such design features interact with individual-level predictors to determine citizens’ willingness to participate. Assemblies can be scheduled on different days, their duration can vary, matters put up for discussion can differ, as can deliberation style and atmosphere. All these aspects are likely to influence individual citizens’ decision to attend – and they are largely at the discretion of assembly organizers.
A better understanding of the impact assembly design has on participation is crucial for improving the inclusiveness of such assemblies, and that is the contribution this study seeks to make. The study was conducted in 2020 in three Swiss communities with institutionalized democratic assemblies, thereby ensuring a realistic environment for an investigation of individuals’ willingness to participate in such assemblies – a realism that is often limited in existing studies due to the use of fictitious examples (Abbott and Touchton Reference Abbott and Touchton2023: 4). In a first step, we use population survey data on a sample of 2,710 citizens to isolate the individual-level factors associated with the likelihood of citizens participating in these assemblies. Second, we use a conjoint experiment to understand the influence of assembly-related factors on participation. Third, we draw on insights from the two previous steps to identify those assembly-related features that have a particularly strong effect on the group of disengaged citizens who, otherwise, are less likely to participate.
Who participates (or not) in democratic assemblies?
Empirical accounts of low and declining turnout levels in participatory democracy are legion (Parvin Reference Parvin2018). In their book on stealth democracy, Hibbing and Theiss-Morse (Reference Hibbing and Theiss-Morse2002) argued that most people dislike politics, and therefore want participation limited to the rare circumstances when corrupt politicians need to be prevented from making decisions citizens dislike. In a similar vein, Diana Mutz (Reference Mutz2006) argued that people do not like political disagreement and seek to avoid engaging in conflictual debates if they can. More recent studies, however, are more optimistic about citizens’ participation in deliberative democracy. Neblo et al. (Reference Neblo2010) found that willingness to participate in deliberative events is more widespread than the stealth democracy thesis would lead us to expect, and that opportunities to participate in deliberative democracy are viewed, by many, as an attractive alternative to standard partisan and interest group politics. These USA findings were largely confirmed by studies in other national contexts, such as the UK (Webb Reference Webb2013), the Netherlands (Coffé and Michels Reference Coffé and Michels2014), Spain (Font et al. Reference Font, Wojcieszak and Navarro2015), and Finland (Christensen and von Schoultz Reference Christensen and von Schoultz2018).
But while it seems safe to assume that citizens are generally open toward participatory or deliberative democracy, it is clear that the willingness to actually participate in such events is not evenly distributed. Studies investigating this matter in a variety of national contexts (see for instance Neblo et al. Reference Neblo2010; Boulianne Reference Boulianne and Hansen2018; Pape and Lim Reference Pape and Lim2019; Gerber et al. Reference Gerber, Schaub and Mueller2019; Gherghina and Geissel Reference Gherghina and Geissel2020) have consistently established the importance of three sorts of individual-level predictors. First, these relate to aspects widely known to prompt civic volunteerism in general (Verba et al. Reference Verba, Schlozman and Brady1995: 16), namely resources (time, money and skills), engagement (interest in politics, concern with public issues, as well as a feeling of political efficacy), and recruitment (insertion in mobilization networks). Second, individuals have varying preferences with respect to the specifics of participating in deliberative events, which affect their likelihood to do so. Scholars (Karjalainen and Rapeli Reference Karjalainen and Rapeli2015; Jacquet Reference Jacquet2017; Jennstål Reference Jennstål2018) found that psychological factors such as willingness to expose oneself to conflicting views, public meeting avoidance, as well as personality traits are important predictors of an individual’s willingness to participate in deliberative events. Third, Neblo et al. (Reference Neblo2010: 572) emphasized the influence of perceptions of democracy: depending on people’s preferences over good democratic practice and processes, they will be more or less eager to participate in a particular deliberative event.
Based on existing scholarship, we can therefore formulate three hypotheses on relevant individual-level determinants of participation in democratic assemblies:
H1: Participation in assemblies depends on citizens’ civic resources, engagement and recruitment. Individuals with more resources, higher levels of political engagement, as well as stronger insertion in recruitment networks will participate more frequently.
H2: Psychological factors play a role for citizens’ participation in assemblies. Individuals with high aversion to conflicts and public meetings are less likely to participate.
H3: Perceptions of democracy are important. Citizens who harbor positive views about democracy in general and of assemblies in particular, participate more often.
The varying appeals of democratic assemblies
Uneven participation in deliberative democracy leads to unintended effects: if those who participate tend to be politically privileged, deliberation might amplify political inequalities (Sanders Reference Sanders1997). Given these concerns, there is a vivid scholarly interest in the design features of deliberative events that could make them more attractive to citizens in general, and to otherwise underrepresented groups in particular (Gherghina et al. Reference Gherghina, Mokre and Mişcoiu2021). Empirical study of the effects of deliberative design features on participant recruitment is cumbersome, however, which is why such studies are still scarce. Indeed, causal analysis of the impact of single design features would require examining a large variety of different deliberative events in comparable contexts. The handful of existing studies on the topic have therefore turned to experimental designs.
Some studies have analyzed citizens’ attitudes and preferences towards instances of deliberative democracy in general. Using a conjoint experiment built into an online panel population survey in Finland, Christensen (Reference Christensen2020) finds that, with respect to design features of deliberative processes, citizens care about transparency, face-to-face interaction (rather than online), the availability of expert advice on complicated issues, and that they prefer processes that do not require too much temporal investment from participants. A similar conjoint experiment in Germany (Goldberg and Bächtiger, Reference Goldberg and Bächtiger2023) confirmed citizens’ preferences for face-to-face interaction (rather than online formats) and found that citizens value consensus over narrow majority decision-making in deliberative venues. Moreover, the two studies both find that citizens prefer deliberative processes that have an advisory rather than a decision function in the policy-making process. However, findings differ regarding the deliberative body itself: while Christensen’s respondents did not significantly prefer inclusive over non-inclusive groups, Goldberg and Bächtiger found large groups to be more attractive.
A second set of studies focuses more particularly on citizens’ willingness to participate in deliberative events with variable features. Neblo et al. (Reference Neblo2010) included a vignette experiment in which they manipulated a number of design features of hypothetical deliberative events potentially relevant to citizens’ decision to participate or not. They found significant effects only for the type of participants and financial compensation: respondents’ willingness to participate is significantly higher when the session includes their member of Congress (vs. only citizens vs. local officials), as well as when financial compensation is provided (25 USD vs. none). Willingness to participate turned out to be insensitive, however, to length (1 hour vs. 1 day), the mode of the deliberative session (face-to-face vs. online), or to the topic (unspecific vs. immigration policy). Collins’ (Reference Collins2021) recent experimental study on the willingness to attend school board meetings focuses on the effects of different types of deliberation. The finding was that exposure to participatory and deliberative meetings (as compared to closed and non-deliberative meetings) increases willingness to attend school board meetings and trust in local officials. Finally, the conjoint experiment conducted by Abbott and Touchton (Reference Abbott and Touchton2023) examined the role of institutional design features of participatory budgeting on people’s interest to actually participate. They found political importance to play a role: the larger the budget, and the more binding the decision to be made, the higher citizens’ interest in participating. Initiation by non-partisan and non-governmental entities (rather than the mayor’s office or the city government) was also preferred, as were topics with potential benefits to the whole community (vs. individual benefits). Online meetings (rather than face-to-face) were also beneficial, while the length of a meeting decreased respondents’ expressed likelihood to participate. However, respondents’ interest in participating was found to be indifferent to the type of deliberation (with or without vote, big or small groups), or to the topic (variety of policy fields).
In sum, these studies point to five assembly design features that can be assumed to affect individual decisions about participating. These are (1) political importance (e.g., in terms of bindingness, budgetary volume, or thematic areas); (2) ease of participation (e.g., length and/or accessibility of meetings), (3) individual incentives for participating (e.g., financial compensation or other benefits), (4) composition of the group that is participating (e.g., inclusiveness), (5) type of deliberation (e.g., meeting style). Based on these insights, we formulate the following hypotheses on how features of design influence individual participation in democratic assemblies:
H4: The greater the political importance of an assembly, the more likely individuals are to participate.
H5: The lower practical hurdles to participation in an assembly, the higher the willingness of individuals to participate.
H6: Individual incentives increase the willingness to participate.
H7: Individual decisions to participate in assemblies are more likely when the deliberating group is more diverse.
H8: Compared to adversary meeting styles, consensual meeting styles increase the probability of an individual participating in assemblies.
The effects of deliberative design features on an individual’s willingness to participate are, arguably, unlikely to be uniform across individuals. In his study on the impact of meeting styles on participation in schoolboards, Collins finds that racial and ethnic groups react differently to variations of meeting styles (2021: 801). More particularly, the boosting effect of open and deliberative meeting styles (as compared to closed and non-deliberative meetings) was stronger among people of color (i.e., Asian, Latinx and Black respondents) than among white respondents. While the explanations for these racial and ethnic group effects obviously lie in the specifics of the USA context, the general conclusion here is that features of deliberative or participatory democracy which enhance participation generally can be hypothesized to have a stronger effect on marginalized groups:
H9: Design features of democratic assemblies which enhance participation have stronger effects on marginalized groups.
Empirical context: communal assemblies in Switzerland
Our study examines citizens’ willingness to participate in democratic assemblies by focusing on the case of communal assemblies in Switzerland. Switzerland is widely known for its vibrant referendum democracy at the national level (Qvortrup Reference Qvortrup2024), and its local government system is also characterized by wide-ranging and meaningful instruments of direct democracy (Ladner Reference Ladner2002). The autonomy of Swiss communities (the so-called Gemeinden/communes/comuni) – both politically and economically – is high in international comparison, and encompasses statutory powers, competencies in a wide range of policy fields, as well as far-reaching fiscal autonomy. Swiss local democracy comprises both representative institutions (i.e., a directly elected government), as well as institutions of direct democracy such as referendums (government decisions subjected to a popular vote) and initiatives (through which citizens can propose new legislation). In addition, around 80% of the roughly 2,200 Swiss communities have organized local government so that the legislature is not a representative institution (i.e., a council with directly elected members), but the supreme political authority is an assembly of all citizens which meets twice (or more) per year.Footnote 1 This communal assembly decides local laws and regulations, approves the community’s budget, and acknowledges accounts given by the communal government. Very similar to the New England town meetings in their functions and organization, Swiss communal assemblies qualify as what Lafont (Reference Lafont2020) calls ‘participatory deliberative democracy:’ citizens have the right to attend, as well as to discuss policy proposals and make binding decisions. With roots in the pre-modern era, Swiss communal assemblies are a legacy institution that is well-known and highly valued by citizens – even by those who do not regularly attend (Rochat Reference Rochat2020).
More specifically, this study focuses on the assemblies of three communities in central Switzerland: Glarus (12,500 inhabitants), Glarus South [Glarus Süd] (9,600 inhabitants) and Glarus North [Glarus Nord] (18,800 inhabitants). Assemblies in these three communities bear the same characteristics as those in the rest of the country, and can therefore be seen as representative of the nearly 1,800 of the overall 2,200 Swiss communities that use the assembly system (Rochat Reference Rochat2020). The assemblies are usually held twice a year, but the communal government may convene additional assemblies in case of urgent matters that it cannot decide in its own competence. Citizens aged 16 or older and residing in the community have the right to attend the assembly without prior registration or notification and can participate in decision-making on the matters that are on the agenda. The agenda is defined by the communal government, but citizens also have the right to suggest agenda items. Decisions in the assembly are made by simple majority and through open voting (e.g., raised hands). Participants also can ask questions of the government, debate matters on the agenda, and propose amendments. Turnout in the assemblies of the three communities averages around 4 to 5%, and corresponds to an attendance of approximately 380 to 500 citizens per assembly (Rochat and Kübler Reference Rochat and Kübler2021: 27). This is comparable to the nationwide mean of assembly turnout in communities of similar size (Ladner Reference Ladner2002: 823). Hence, the overwhelming majority of the citizens of the three communities under scrutiny usually choose not to participate. Why is that? This is what we seek to clarify in this study.
Research design
Our examination of individual citizens’ likelihood of participating in the communal assemblies proceeds in three steps. First, we use data from a population survey to identify the determinants of assembly participation at an individual level. Second, we focus on the impact that different design features of assemblies have on citizens’ intent to participate, drawing on the results of a conjoint experiment built into the survey. Third, we focus on the combined effect of assembly design features and individual marginalization by examining whether and how the intent of underrepresented groups to participate in an assembly is influenced by assembly design features.
All analyses were performed in R, using the packages cregg and survey.
Data collection
In fall 2020,Footnote 2 the statistics office of the canton of Glarus drew a random sample of 9,000 eligible voters from the official residents’ registers in the three communities under scrutiny – the sample corresponded to roughly one-third of all eligible voters. The state chancellery of the canton of Glarus then sent these individuals a postal invitation to participate in a self-administered survey (online or by filling in a paper questionnaire). To comply with legal requirements on data protection and research ethics, the incoming questionnaires were handled by an independent research team (see also section A5 in the online supplemental material). Two-thousand seven hundred ten valid questionnaires were returned (response rate of 30.1% – about one-tenth of all eligible voters). The sample includes voters from all three communities, aged from 16 to 97 years. Compared to the census data, male respondents, people in the oldest age cohort and residents of Glarus South are slightly overrepresented. The data were therefore weighted before the analysis (see section A1 in the online supplemental material for details).
Individual-level survey data: variables and operationalizationFootnote 3
The main dependent variable of interest is an individual’s self-reported participation in citizen assemblies in the three communities under scrutiny. Previous studies on participation in regular political decision-making processes in the Swiss context have identified three main patterns of participation: some citizens (almost) always participate, others never participate at all, and in between are the so-called selective participants, that is, citizens who sometimes participate and sometimes do not (Tawfik et al. Reference Tawfik, Sciarini and Horber2012). We, therefore, asked the respondents how often they had attended communal assemblies in the five years prior to the interview, with three answer categories to choose from: ‘(almost) always,’ ‘sometimes,’ and ‘never.’ Descriptives show that participation in assemblies correlates strongly with political participation at other territorial levels (Table A4 in the online supplemental material). Those who reported that they ‘(almost) always’ attended the communal assemblies in the five years prior to the survey also reported participating more frequently in elections as well as in referendums at higher state levels, compared to those who reported attending ‘sometimes’ or ‘never’.
As specified in hypotheses H1–H3, independent variables at the individual level relate to civic resources, political engagement and recruitment (H1), relevant psychological factors (H2), as well as perceptions of democracy (H3). With respect to H1, our measurement of resources and skills taps into civic resources more generally (education, as well as internal political efficacy), but also into specific constraints regarding assembly participation (household with small children). Assemblies in the three communities under scrutiny are usually held in the evenings, which makes it difficult for parents of small children to attend. Political engagement is measured by motivational factors such as interest in local politics, emotional attachment to the community, as well as homeownership which ties an individual’s assets to the community and therefore raises his or her stake in local policies. Insertion in recruitment networks is measured by memberships in local associations, as well as by the age-weighted duration of residence in the community. As regards H2, psychological factors relevant to an individual’s willingness to participate in an assembly were operationalized as a score on their perceived uneasiness to speak up in front of other people (public meeting avoidance), as well as a score on their perceived uneasiness about conflictual assemblies (conflict avoidance). Operationalizing H3, our gauge of respondents’ perceptions of democracy included their view on the overall political system’s responsiveness to citizens’ demands (external political efficacy), as well as their level of trust in the local government. In addition, we were also interested in their views on communal assemblies as a democratic institution. More precisely, we asked respondents to what extent they agreed or disagreed with a statement referring to a common and frequently voiced criticisms of such assemblies, namely that of a very low turnout and biased representation. Finally, the sociodemographic variables age and gender are used as controls.
Measuring the impact of assembly design features: conjoint experiment
To identify the effects of assembly design features on respondents’ willingness to participate, we carried out a conjoint experiment (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2014). We presented respondents with two descriptions of fictitious assemblies in their community and asked them to choose which of the two assemblies they felt more inclined to attend (choice-based conjoint analysis). This was repeated twice for each respondent, so that each respondent was presented with two pairs of two fictitious assemblies, with elements of description that were randomly varied (see Table A7 in the supplemental material).
These elements of description – called attributes – and their substance – called attribute levels – were defined to operationalize hypotheses H4–H8 on how assembly design features influence individual likelihood to participate (see Table 1). Of course, attributes and attribute levels needed to be chosen so as to be plausible in the empirical context under scrutiny. This means that assembly features that are legally regulated cannot be varied. For instance, competencies of the assembly, but also meeting modalities (e.g., online vs. face-to-face) were not part of the choice universe, as these are defined by law. Similarly, as the right to participate in a communal assembly is legally restricted to residents who hold Swiss citizenship, the range of plausible criteria to vary the degree of assembly diversity is also limited.
Choice experiment: attributes and levels
The political importance of a given assembly (H4) was operationalized by varying the most important item on the assembly’s agenda. The ‘annual financial statement’ is arguably an agenda item of very little importance: the assembly can simply approve or disapprove the government’s accounts of the past year. Consequences of either decision are nil: the money has already been spent and should the accounts be disapproved by the assembly, consequences for the government are largely symbolic as it is then simply obliged to present an improved account at the following assembly. Four other attribute levels describe realistic matters of communal decision-making that have direct and tangible consequences for most citizens. More precisely, these were ‘a tax rate increase by 5%’, ‘overall revision of land use planning’, ‘revision of the communal constitution,’ as well as ‘50,000 CHF credit for road maintenance’.
Practical hurdles to assembly participation (H5) were specified using three attributes. First, the expected duration of the assembly is obviously crucial: the longer an assembly, the more temporal resources it requires from participants.Footnote 4 Second, the distance to the assembly location is also relevant: depending on how far participants need to travel, they will find it more or less cumbersome to actually attend an assembly.Footnote 5 Third, the scheduling of the assembly was varied between weekday evenings and weekends, assuming that weekdays would be more practical for most citizens and hence preferable.
Individual incentives (H6) were operationalized in referring to actual practice by Swiss communities regarding amenities of assembly participation. While ‘no amenities’ is the obvious baseline which is also current practice in the three communities under scrutiny, offering drinks and nibbles after the assembly (a so-called apéritif) is a popular tradition elsewhere in Switzerland, and some rare communities even offer gifts (most frequently in the form of vouchers) to assembly participants with the explicit aim of increasing turnout.
The nature of the deliberating group (H7) was operationalized by varying the kind of company a participant could expect at the assembly. The baseline was defined as ‘nobody’ – meaning that respondents expected to have no personal relations with other assembly participants – and being accompanied by ‘family members’ or ‘friends’ was assumed to be more attractive, with ‘neighbors’ or ‘work colleagues’ in between.
Finally, the meeting style (H8) was operationalized as a variation of the atmosphere participants can expect at the assembly. ‘No discussion’ describes a situation where participants had not felt a need for debate. ‘Minor disagreements’ expresses initial consensus, whereas ‘agreement after lively debate’ denotes an atmosphere in which major differences of opinion existed but were resolved after debate – a deliberative consensus so to speak. ‘Open dispute’ describes a very adversarial meeting style.
The purpose of conjoint analysis is to single out the (positive or negative) causal effects of individual attribute levels on the outcome. To measure these effects, Hainmueller et al. (Reference Hainmueller, Hopkins and Yamamoto2014) propose the calculation of average marginal component effects (AMCEs). These correspond to the difference, averaged over all respondents, of the outcomes with a given attribute level compared to the results with a reference level. As a relative measure, however, AMCEs say nothing about the underlying absolute preferences for or against a particular conjoint profile. This can be a problem when comparing the effects of different subgroups (Leeper et al. Reference Leeper, Hobolt and Tilley2020). Differences between AMCEs of different subgroups should not be confused with differences in the underlying preferences of those subgroups. The average preferences of respondents for or against conjoint profiles with specific attribute levels can instead be adequately represented with marginal means and used for descriptive purposes. Furthermore, using marginal means instead of AMCEs eliminates the need for reference categories. We therefore follow Leeper et al. (Reference Leeper, Hobolt and Tilley2020) and base our conjoint analysis on marginal means.
Like the correlational analyses, the conjoint analyses were carried out with weighted data. However, there is a debate in the literature as to whether weights should be used in survey experiments (Miratrix et al. Reference Miratrix2018). As an additional robustness check, the conjoint analyses were therefore conducted with unweighted data (see Figures A4 and A5 in the supplemental material). They show that the calculated effects are robust and are not due to bias caused by survey weights.
Assembly design features and individual predictors: combined effects
In a third step, we combine the insights from the two preceding analyses. The goal is to clarify whether and how specific elements of assembly design are valued differently by under- and overrepresented segments of the citizenry. For this purpose, we divide the sample into two groups. The first group contains respondents who tend to belong to the underrepresented strata. The second group consists of respondents who tend to belong to the overrepresented strata. This classification is based on the regression model of participation predisposition previously estimated. Respondents with a low expected probability of assembly participation are categorized in the first group, persons with a high probability of participation in the second group. We will thus refer to these two groups as the ‘engaged’ and the ‘disengaged’ citizens – echoing the terminology used by Parvin (Reference Parvin2018). This categorization shifts the focus away from (reported) participation frequency to expected participation probability. Finally, we use conjoint analysis to compare the effects of different assembly design features on participation intention across the two subgroups.
Individual predictors of assembly participation
We first examine the effect of individual-level variables on self-reported participation in communal assemblies. Distinguishing between respondents who reported having participated (almost) always, sometimes, or never in a communal assembly in the five years prior to being surveyed, Table 2 shows the distinctive features between these three groups according to the independent variables previously identified.
Self-reported participation in communal assemblies five years prior to the survey: descriptives of individual features
Note: Differences between groups significant at
*P < 0.05,
**P < 0.01,
***P < 0.001.
Chi-square tests were performed for categorical, Kruskal-Wallis tests for ordinal and analysis of variance for continuous variables. The data are weighted.
As can easily be seen in Table 2, the three groups are distinct in all (but one) features that were examined, with differences playing out as expected. Participation in communal assemblies is positively associated with factors known to affect civic engagement more generally, that is: civic resources and skills, motivation, as well as membership in recruitment networks. Psychological factors also play a role: public meeting avoidance and conflict aversion are associated with lower participation in communal assemblies. Perceptions of democracy are also distinct in the three groups: respondents who participate more often are also those with more positive assessments of the political system or of communal assemblies. There is one exception to this observation, however: levels of trust in local government are not significantly different between the three groups. Those who never participate in communal assemblies do not trust the local government less than those who report to have participated in communal assemblies sometimes or (almost) always.
Further differences between the three groups exist at the level of the sociodemographic control variables. Respondents who declared that they had participated (almost) always in communal assemblies are older than those who participated selectively, who in turn are older than those who never participated. Furthermore, the share of men is higher among those who participated (almost) always than in the other two groups. The group of those who declared they had never participated contains the highest share of women.
In the next step, the effect of the various independent variables is tested in a multivariate setting. Following our hypotheses, we estimated five proportional odds models (Table 3). The dependent variable is self-declared participation in communal assemblies in the five years prior to the survey, where 1 stands for ‘never’, 2 for ‘sometimes’, and 3 for ‘(almost) always’. The first model included the control variables only, the other models estimated the influence of predictors relating to civic engagement (H1), psychological factors (H2), as well as perceptions of democracy (H3). Finally, a full model was estimated.
Predictors for self-reported participation in communal assemblies in the 5 years prior to interview (ordinal logistic regression)
Note: Table entries are unstandardized proportional odds model coefficients, with standard errors in parentheses. Data are weighted.
The controls-only model confirms that age and gender play a role: older and male respondents are more likely to participate in communal assemblies. Model 1 confirms that factors known to spur civic engagement more generally also play out in favor of participation in communal assemblies. Civic resources and skills (internal political efficacy), as well as motivation (interest in local politics, attachment to the community, homeownership) are all positively associated with participation in assemblies, as is insertion in recruitment networks (years of residence, membership in organizations). Interestingly, however, tertiary education does not show a significant effect in the multivariate model, and neither does living in a household with small children. While the influence of tertiary education is probably explained away by internal political efficacy (citizens with better education are also those with higher political sophistication), living in a household with small children most likely interacts with age.
Model 2 emphasizes the importance of psychological factors in participation in communal assemblies. Respondents who do not like public meetings, or who feel uneasy about potential conflicts that emerge in communal assemblies are less likely to participate. Model 3 corroborates the idea that respondents’ perceptions of democracy in general, and of deliberative democracy in particular, also have an effect on assembly participation. A favorable view of the democratic responsiveness of the wider political system (external political efficacy) is positively associated with assembly participation, while a critical attitude towards the communal assembly has a negative effect. And, as in the bivariate analysis, assembly participation is not significantly associated with levels of trust in local government.
The full model, finally, shows that these effects are fairly robust. Only two variables (emotional attachment to the community as well as external political efficacy) are no longer significant in the full model. In general, therefore, there is sufficient evidence to corroborate hypotheses H1–H3 about individual-level predictors of participation in communal assemblies. These results confirm the general assumptions, made at the outset, that people who regularly participate in communal assemblies significantly differ from those who participate less frequently, or who prefer to stay away altogether. This means that the lower the turnout, the more an assembly is likely to suffer biased representation. In particular, there will be an overrepresentation of older, male citizens with higher civic resources, who are more interested in local politics, have lived in the community for a long time and are well integrated in associational networks, who have psychological predispositions that enable them to better cope with public meeting exposure and potential conflicts, and whose preferred models of democracy are congruent with the specificities of assembly democracy. And in corollary, there will be an underrepresentation of younger citizens, women, those with fewer civic resources, who are less interested in local politics, have recently moved there and are distant from associational networks, who feel uneasy in public meetings and are wary of conflictual situations, and who are doubtful of the communal assembly system as a democratic model.
The effect of assembly design features
The preceding analyses have given us an idea of individual predispositions on citizens’ likelihood to participate. But what about the assembly-related design features that increase (or decrease) participation? The conjoint experiment provides an answer to this question (Fig. 1).
Effects of assembly design features on likelihood to participate: conjoint analysis on full sample.
Notes: The squares represent the marginal means. The horizontal lines drawn through the squares are the 95% confidence intervals. The calculation is based on weighted data. NObservations = 9928, NRespondents = 2561. The data is clustered. See Table A8 for full results.
First, citizen’s intent to participate in an assembly clearly increases with the political importance of the main agenda item (H4). As expected, an assembly where the (politically unimportant) annual financial statement is the main agenda item depresses intent to participate most strongly. All other items generate higher willingness to participate, especially tax rate increases, revision of the constitution, as well as land use decisions. Next, practical hurdles play a role (H5). The expected duration of the assembly has a very clear effect: the longer an assembly, the lower citizens’ intent to participate. The timing is also important, with assemblies on weekends being less attractive than those scheduled on weekday evenings. The length of the journey to the assembly location plays a rather minor role, however: out of the various journey durations, only the 20-minute journey has a significant negative effect. Interestingly, the findings for individual incentives (H6) do not all go in the expected direction. Offering no such incentives does not have a significant effect on the likelihood to participate – a finding that is most likely explained by the fact that this is the current practice in the three municipalities under scrutiny. But while post-assembly drinks and nibbles significantly increase individuals’ likelihood to participate, gifts offered for assembly participation decrease intent to participate rather than increasing it as we had expected. The most plausible explanation for this rather counterintuitive finding is that assembly participation is seen as a moral civic duty by most respondents, and that they dislike having this moral duty associated with material benefits. With respect to group composition (H7), we find that being accompanied by family and friends make a difference, while being accompanied by neighbors is not any different than going there alone. Interestingly, having work colleagues as company deters participants. This suggests that participation in communal assemblies in the three places under scrutiny is also seen as a community event that one wishes to attend with others who one feels close to. Finally, the meeting style hypothesis (H8) is not buttressed by the results. We expected consensual meeting styles to attract and adversarial meeting styles to deter potential participants, but the results suggest otherwise. Indeed, the absence of a debate, or the existence of an initial consensus decrease likelihood to participate in an assembly as compared to more adversarial atmospheres. However, a controversial debate that leads to agreement is preferred to an open dispute. This suggests a curvilinear relationship between meeting style and intent to participate: very consensual meetings are not attractive for potential participants, but very adversarial ones aren’t either. It is the middle ground on the continuum between very harmonious and very adversarial meeting styles that are preferred: deliberative consensus is most favorable to assembly turnout.
Representation and assembly design features
Given that individual citizens have different propensities to participate, the question arises whether people who tend to belong to marginalized groups value assembly design characteristics differently than regular assembly-goers. Therefore, we used the above regression model to differentiate two groups of respondents. On the one hand, one has a group of respondents who have a high propensity to participate in communal assemblies (‘engaged citizens’) and are thus likely to be present in assemblies. On the other hand, one has the group of those whose participation propensity is low (‘disengaged citizens’) and who, therefore, are likely to be marginalized by the assembly system. More specifically, we used the full regression model (see Table 3) to calculate, for each respondent, the individual probability that he or she falls into the ordinal category of those who declared having never participated in a communal assembly in the five years prior to the survey. Respondents with a probability below 50% were categorized as ‘engaged’ citizens, those with a probability equal or above 50% were categorized as ‘disengaged’ citizens. This yielded two groups (n = 1359 ‘engaged;’ n = 686 ‘disengaged’).Footnote 7 A look at the distinctive features of the two groups (see Table A6 in the online supplemental material) confirms that the previously identified traits associated with self-reported assembly (non-)participation are accentuated between the two groups.
To analyze whether and how assembly design features play out among different kinds of individuals, we ran the conjoint analysis separately for the two groups, that is, the engaged and disengaged citizens (Fig. 2). Many of the assembly design features previously identified as influencing the willingness to participate in general have quite similar effects on engaged and disengaged citizens. Indeed, political importance, practical hurdles, individual incentives, group composition and meeting style are important determinants for assembly participation in both groups.
Effects of assembly design features on likelihood to participate: conjoint analysis on engaged vs. disengaged citizens.
Notes: The squares represent the marginal means of the two strata. The horizontal lines drawn through the squares are the 95% confidence intervals. The calculation is based on weighted data. Disengaged: NObservations = 2676, NRespondents = 686; Engaged: NObservations = 5282, NRespondents = 1359. The data is clustered. See Table A10 for full results.
Nevertheless, some distinct effects between the two groups can be identified. First, while political importance of the items on the assembly agenda is important for both groups, disengaged citizens seem less attentive to particular aspects of more important agenda items than engaged citizens – who seem to make a difference for assemblies devoted to land use planning revisions, which they find most attractive. Second, practical hurdles deter both groups of citizens, but engaged citizens are deterred a little less than disengaged citizens – as is shown by them being insensitive to the length of the journey to the assembly location. Third, with respect to amenities, both groups are deterred by individual gifts; but disengaged citizens have a higher willingness to attend an assembly when drinks and nibbles are offered, while engaged citizens are indifferent about drinks and nibbles and prefer assemblies where there are no amenities at all (thus the status quo). Fourth, regarding group composition, both engaged and disengaged citizens are sensitive to being accompanied by friends, while disengaged citizens are less willing to participate in assemblies when they have to go there on their own. Finally, there are interesting differences between the two groups with respect to the expected meeting styles. Engaged citizens are less interested in very consensual assemblies, and more attracted by open conflicts and, especially, by deliberative consensus. Disengaged citizens, in contrast, are less deterred by very consensual meeting styles, but their willingness to participate increases only when they expect a deliberative consensus.
All in all, this suggests that assembly design features have similar impacts across the two groups. Nevertheless, some of the features found to matter for the participation of engaged citizens appear to be even more important for the disengaged citizens. Compared to engaged citizens, disengaged citizens’ willingness to participate in communal assemblies is even more pronounced when these are akin to community events: where one is accompanied by friends, where debates are controversial but agreements are reached, and when there also is an opportunity to have a drink together. Engaged citizens, of course, are not deterred by these community event characteristics of assemblies, but their willingness to participate is less decreased when they expect assemblies to become conflictual. Drawing on Jane Mansbridge’s (Reference Mansbridge1983) early conceptualization, we can say that disengaged citizens are more attracted by assemblies with a rationale of unitary democracy, while engaged citizens are prepared to persevere even in an atmosphere of adversary democracy.
Discussion and conclusion
The existing studies on assemblies in deliberative and/or participatory democracy have focused on their role in the wider institutional environment, on deliberative or decision-making processes within them, or on the educational effects on the participants. However, analyses of potential participants’ actual attendance or their motivation to attend such assemblies remain rare. This is surprising, given the fact that most individuals recruited or entitled to take part in such assemblies actually decline the opportunity. The aim of the present study was to better understand this abstention by identifying the factors associated with self-reported and intended (non-)participation in communal assemblies in Switzerland. Similarly to New England town meetings (Pateman Reference Pateman2012: 16), these assembly-based democratic legacy institutions tend, very unfortunately, to be neglected by political scientists (for an exception see Stadelmann-Steffen and Dermont Reference Stadelmann-Steffen and Dermont2016). But, as we argue in this article, a closer examination yields valuable insights on participation in face-to-face events that are also relevant for the debate on deliberative and participatory democracy more generally.
First, the analysis of our survey results confirms the importance of individual predispositions, with hypotheses H1 and H2 corroborated, and H3 partly corroborated (Table 4). Civic resources, engagement and recruitment, known to influence political participation in general, are also important predictors of participation in democratic assemblies. Individuals with more resources, higher levels of political engagement as well as stronger integration in recruitment networks were found to participate more frequently (H1). Psychological factors were also found relevant (H2): public meeting avoidance, as well as conflict aversion, were associated with less frequent participation. Perceptions of democracy also turned out to be important (H3). Frequency of participation in communal assemblies was associated with external political efficacy, as well as positive assessments of communal assemblies as an institution. However, trust in local government was not significantly associated with assembly participation: abstention does thus not equal mistrust of government.
Summary of hypotheses and results
Second, the conjoint experiment allowed us to identify features of assembly design that have an impact on citizens’ intent to participate. On the one hand, our results confirm previous evidence regarding the assembly agenda – intent to participate is stronger when there are politically important items on the assembly’s agenda (H4) – as well as regarding the practical hurdles to assembly attendance – duration, scheduling as well as assembly location have an impact on intent to participate (H5). On the other hand, our study yielded results that add to or differentiate findings made by other scholars. Based on the literature, we hypothesized that the intent to participate in an assembly is positively associated with individual incentives for such participation (H6). We found this positive effect only for community-related incentives, that is, providing drinks and nibbles after the assembly. Handing out individual gifts actually had a negative effect on intent to participate. While this result is contrary to what other scholars found (see Neblo et al. Reference Neblo2010: 574), it ties in with the results of the landmark study of deliberative democracy in the USA by Jacobs et al. (Reference Jacobs, Lomax Cook and Delli Carpini2009: 73 ff.), finding that, besides instrumental considerations, a sense of ‘duty’ as a citizen or member of the community is a very strong motivation to attend deliberative forums. It echoes more general arguments about the detrimental effect of gifts on civic behavior: external rewards can undermine the intrinsic motivation to engage in an activity, especially when this activity is viewed as an altruistic or a civic duty (Frey and Jegen Reference Frey and Jegen2001).
Regarding the impact of the group composition on the intent to participate in an assembly, our results also differ from what we expected (H7). We did find that group composition affected the intent to participate, but it was the prospect of going there together with like-minded people one feels emotionally close to – that is, friends or family – that increased individuals’ intent to participate. Finally, our results also corroborate the importance of the meeting styles for intent to participate. But while we expected (drawing on Collins Reference Collins2021) that people would feel more attracted to consensual rather than adversary meetings (H8), we actually found a curvilinear effect. Very consensual and very adversary meetings both decrease intent to participate. Instead, people feel most attracted to assemblies in which they expect a lively debate that eventually leads to an agreement – a ‘deliberative consensus’ as we have called it. This result not only ties in with the literature on deliberative democracy – emphasizing that deliberative quality depends on discursive diversity (Dryzek and Niemeyer Reference Dryzek and Niemeyer2008) – but also underscores the importance of the community dimension for participation in democratic assemblies. Indeed, the qualifications for H6, H7, and H8 all point in the same direction: individuals’ intent to participate increases with cues pointing to a stronger communitarian character of an assembly. The relevance of this aspect is further emphasized in light of the results on the interaction between individual predispositions and assembly design features (H9). The overall pattern of effects of participatory-enhancing features of assembly design on individual intent to participate plays out as expected across engaged and disengaged citizens. Thus, people tend to participate (or not) in assemblies for the same reasons, but compared with engaged citizens, disengaged citizens are even more sensitive to assembly design features that emphasize their communitarian character.
Limitations of the study are, as always with population surveys, the risk of uncontrolled sampling bias. Given the political topic of the questionnaire, it is likely that citizens with a higher interest in politics are overrepresented. In reality, the group of disengaged citizens is therefore likely to be larger than in our sample. In addition, the setup of the conjoint experiment invited respondents to express their participation preferences as a pairwise choice between two hypothetical assemblies. One could argue that this does not match real-world situations, in which citizens choose to either go to an assembly or not to go. One likely effect of our experimental setup could therefore be that it overestimates the intent to participate in assemblies at all. Finally, the real-world setting of the three communities under scrutiny restricted the range of attribute levels that could plausibly be used in the conjoint experiment. In particular, our operationalization of the composition of the participating group makes it difficult to directly address questions of assembly-member diversity as it is usually addressed in the literature, namely as a variation in cross-cutting views (Mutz Reference Mutz2006), discourses (Dryzek and Niemeyer Reference Dryzek and Niemeyer2008), worldviews and perspectives (Landemore Reference Landemore2013), or social characteristics (Beauvais and Warren Reference Beauvais and Warren2019).
But in spite of these limitations, our findings emphasize, more generally, the close relationship between democratic participation and community vibrance (Putnam Reference Putnam1993). The mechanisms at work in traditional democratic assemblies are obviously very similar to those of deliberative democratic innovations, where group-building is also an important condition for the quality of deliberation (Chambers Reference Chambers2018; Niemeyer et al. Reference Niemeyer2023). Our findings therefore have implications for assemblies both as democratic innovations and as traditional participatory institutions, as an existing sense of community is conditional to participation and deliberation in both instances. While this seems a challenging endeavor in the face of ongoing individualization in many contemporary societies, accounts of successful community building and democratic revitalization (Warren Reference Warren2001) forcefully demonstrate that the cause is by no means lost. Future research should establish how the organization and practice of democratic assemblies can be accompanied by efforts to actively strengthen or build community ties among those who are entitled or invited to participate and deliberate in such assemblies.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1755773924000341
Data availability statement
Anonymised data as well as R scripts used to produce the analysis are available at the FORS replication service using the following link: https://doi.org/10.25597/t450-zw11.
Acknowledgments
We acknowledge research assistance by Robin Huber, and we would like to thank Michael Schüepp from the Glarus cantonal Chancellery, as well as the communities of Glarus, Glarus North and Glarus South for their support and cooperation in the survey implementation. Previous versions of this paper were presented at the Annual Meeting of the American Political Science Association in Seattle in 2021, at the 2021 Online General Conference of the European Consortium of Political Research, as well as at the Publication Seminar of the Department of Political Science of the University of Zurich in 2022. We would like to thank participants at these events for their helpful comments, and acknowledge, in particular, the contributions of Sergiu Gerghina, Clayton M. Nall, Ashley E. Nickels, Marco Steenbergen, Francesco Veri and Hannah Werner. Thanks also to John Bendix for his help with English editing, as well as to the two anonymous reviewers for their insightful comments.
Funding statement
Research for this study was funded by the State Chancellery of the canton of Glarus (Switzerland), the Centre for Democracy Studies at the University of Zurich, as well as the University of Zurich’s COVID-19 emergency fund.
Competing interests
The authors state that they have no competing interests to declare.
Ethical statement
See section A5 in the supplemental material.
Open access
