Support for direct and deliberative models of democracy in the UK: understanding the difference

The models of direct and deliberative democracy are broadly considered the major alternatives to representative democracy. So far, the two models have been merged under the broad umbrella of participatory democracy and thus little is known about why citizens support direct democracy and/or deliberation. They are distinct procedures, driven by di ﬀ erent logics and outcomes and this makes it likely that the preference for them rest on di ﬀ erent premises. This article ﬁ lls this gap in the literature and distinguishes between the models proposing two central arguments. First, we expect that several general determinants have a positive impact on the support for both direct democracy and deliberation because they are di ﬀ erent from representative democracy. Second, we test the e ﬀ ect of speci ﬁ c determinants that drive people towards supporting more one of the two alternative models of democracy. We use individual level data from an original survey conducted in December 2018 on a representative sample of 1094 respondents in the UK. The results indicate that the supporters of direct democracy di ﬀ er from those of deliberative democracy in several ways. The survey pre-test did not highlight any biases in the direction of direct democracy. This question included two forms of direct democracy and two forms of deliberative democracy. The forms of direct democracy were the referendum and citizens ’ initiatives. The latter are direct democratic instruments that do not necessarily end up in a vote on a policy. Nevertheless, they follow the aggregation logic of collecting signatures in order to put a topic on the political agenda. The forms of deliberative democracy were mini-publics and participatory budgeting. 3 The available answers indicated preference for each of these forms and were recorded on a four-point ordinal scale between not at all (1) and very much (4). The answers about referendum and citizens ’ initiatives were merged to form the support for direct democracy, while the answers about mini-publics and participatory budgeting were merged to form the support for deliberative democracy. Both variables are coded on a seven-point ordinal scale ranging from 2 (not at all for both forms merged) to 8 (very much for both forms). The values of the Cronbach ’ s Alpha for each dependent variable indicate high levels of internal consistency: 0.79 for direct democracy and 0.84 for deliberative democracy.


Introduction
The flaws of contemporary representative democracies have determined citizens and politicians to look for alternatives. An extensive body of literature acknowledges the various demands of citizens and focuses on their preferences for alternative models of political decision-making (Hibbing and Theiss-Morse 2001;Esaiasson, Gilljam, and Persson 2012;Bengtsson and Christensen 2016;Gherghina and Geissel 2017). This coincides with a momentum gained by participatory institutions and procedures in real-life politics. In the last three decades, an increasing number of countries introduced direct democratic procedures to enable citizens to make final political decisions (Scarrow 2001;Gherghina 2017). Other countries use deliberative procedures to enhance the quality of decisions, e.g. Iceland and Ireland with their constitutional conventions (Farrell, O'Malley, and Suiter 2013;Reuchamps and Suiter 2016). A recent study shows that many countries choose between these two forms of participatory institutionsdirect democracy or deliberationrather than using both (Geissel and Michels 2018).
So far, the functioning and consequences of these two forms of participatory democracy have been intensely documented (Bengtsson and Mattila 2009;Webb 2013;Coffé and Michels 2014;Font, Wojcieszak, and Navarro 2015;Caluwaerts et al. 2018). The preferences for different forms of participation have been explored and described in detail, with emphasis on trends, general meaning and relationship to political participation (Bengtsson and Christensen 2016; Gherghina and Geissel 2017;Landwehr and Steiner 2017). Several studies went one step further and sought to identify what triggers the demand for referendums as the most common form of direct democracy (Hibbing and Theiss-Morse 2002;Bowler, Donovan, and Karp 2007;Anderson and Goodyear-Grant 2010;Qvortrup 2014). More recently, citizens' initiatives were included in analyses next to referendums, thus providing a fuller understanding of citizens' preferences for direct democracy (Gherghina and Geissel 2019).
In spite of this broad coverage, little is known about why citizens support direct democracy and / or deliberation. They are distinct procedures, driven by different logics and outcomes, and this makes it likely that the preference for them rest on different premises. And yet many studies collate the two under the broader umbrella of participatory democracy, while others analyse them separately. One notable exception to this rule is the study of Landwehr and Faas (2015) who compare the two forms (see below). One explanation for this limited comparison is the scarcity of comparative data. For example, international surveys such as World Value Survey and the Eurobarometer have included questions on preferences for direct democracy, but they lack the deliberation component. To address this gap in the literature, this article aims to answer the following research question: Why do citizens' support direct democracy and/or deliberative procedures?
Our quest for an answer relies on individual level data from an original survey conducted in December 2018 on a representative sample of 1094 respondents in the UK. We formulate and test empirically two central arguments. First, we argue that several general determinants such as interest in politics, consumption of political news, internal efficacy and dissatisfaction with institutions of representative democracy have a positive impact on the support for both direct democracy and deliberation. Second, there are specific determinants that drive people towards supporting more one of the two alternative models of democracy. Apart from their common shared idea of involving citizens in the decision-making process, direct democracy rests on different principles than deliberative democracy and for this reason we expect support to be triggered by different variables.
The remainder of this article proceeds as follows. The first section reviews the literature explaining citizens' preferences for participatory procedures in general and formulates four testable hypotheses. Next, we review the studies discussing citizens' preferences for direct democratic versus deliberative procedures formulate two further testable hypotheses for each of these two major forms of participation. The third section describes our research design, followed by a section on empirical analysis and interpretation of results. The conclusion summarizes the main findings and discusses broader implications for the study of citizens' attitudes towards participatory decisionmaking. 1 Citizens' support for direct and deliberative procedures This section reviews the literature seeking to explain the preferences of citizens for participatory procedures as alternatives to the purely representative form of democracy. The debate on preferences for alternative models of democracy has been shaped by two major analytical approaches. The first approach rests on the question, if the dissatisfied and less educated citizens favour direct democratic decision-making more than those citizens who are politically interested and educated (Donovan and Karp 2006). The second approach builds on the dichotomy between enraged (dissatisfied) and engaged citizens to understand the preference for participatory democracy (Bowler, Donovan, and Karp 2007). In essence, these two approaches point in the direction of the following key determinants: interest in politics, consumption of political news and satisfaction with institutions of representative democracy. We add internal efficacy due to its crucial importance for participation, which can intuitively also influence preferences for participatory democracy. This intuition is supported by existing research in the fields of political attitudes and political participation.
We build an integrative theoretical framework that proposes testable hypotheses with four crucial determinants: interest in politics, the consumption of political news, internal efficacy and dissatisfaction with the institutions of representative democracy. We test for the effects of these variables on preferences for both direct and deliberative procedures, because they refer to general participatory preferences, i.e. for non-representative forms of government. We use these predictors because they are applicable to both direct and deliberative democracy.
Interest in politics is a crucial predictor for political participation (Verba, Schlozman, and Brady 1995;Norris 2000;Christensen 2017) and also for participatory preferences. The logic behind this relationship is that citizen, who care about what happens in politics, do actively engage in it. Citizens with political interest are more willing to have a greater say in the political will-formation and decision-making process. This mechanism most likely translates to the relationship between interest and preferences: 'citizens with greater interest in politics may … desire greater participatory democracy' (Donovan and Karp 2006, 872). Several studies show that people with high interest in politics are more interested in having more say in politics, wish to have an active role in the world of politics and embrace participatory democracy (Donovan and Karp 2006;Bowler, Donovan, and Karp 2007). In this context, the literature also refers to the increase of post-materialist values in advanced societies, which includes not only enhanced political interest but also goes hand in hand with an increased 'desire for a more participatory role in politics' (Bowler, Donovan, and Karp 2007, 354). Participatory procedures can be expected to be valued by citizens with post-materialist values, who at the same time are also more interested in politics than their fellow citizens. Some authors nuanced this perspective and argued that disinterested citizens are more inclined to support direct democracy, because they do not want to spend too much time with politics. Highly interested citizens may prefer deliberation because it allows them to understand a political issue in more detail (Bengtsson and Mattila 2009).
Internal efficacy has been also considered a potential factor. Its relationship with political behaviour is widely acknowledged (Finkel 1985;Rudolph, Gangl, and Stevens 2000;Valentino, Gregorowicz, and Groenendyk 2009;Jung, Kim, and de Zúñiga 2011). For example, highly efficacious citizens are more likely to vote than others. The logic behind this relationship is that citizens' confidence in their ability to participate provides people incentives to get involved. Similarly, it is reasonable to argue that citizens with strong feelings of internal efficacy are likely to be more supportive towards participatory procedures that open up new possibilities for involvement in political will-formation and decisionmaking. Surprisingly, there are only few studies examining the effects of internal efficacy on participatory preferences. Some studies examine whether direct democratic options and procedures enhance internal efficacy (Dyck and Lascher 2009), but only few works look at the effects of internal efficacy on participatory preferences. One of the studies on participatory preferences indicate that German citizens with high internal efficacy show more support for both forms of participatory democracy (Landwehr and Faas 2015, 14).
The dissatisfaction with institutions of representative democracy has been for a long time one of the usual suspects when speaking about preferences for alternative participatory models of democracy. The flaws of representative democracy have determined an increasing number of citizens to voice their discontent. Over time, citizens have become more critical and more demanding towards the system of political representation (Norris 2011). The inability of the system to solve its problems and to cope with citizens' interests and demands have triggered a desire to look for alternatives within the democratic realm rather than contesting the system overall. In line with this reasoning, research shows that citizens who are dissatisfied with the institutions of representative democracy are more likely to prefer participatory options (Dalton, Burklin, and Drummond 2001;Bengtsson and Mattila 2009;Christensen 2017;Gherghina and Geissel 2019). There are some contradictory findings according to which citizens with trust in institutions of representative democracy are stronger supporters of participatory procedures (Bowler, Donovan, and Karp 2007). However, the mixed evidence appears to have more methodological roots rather than theoretical or substantial reasons: the studies refer to different institutions and people, mixing for example trust in government, parliament, parties and politicians, which are quite different and can have separate effects (Gherghina and Geissel 2019). Overall, it is unanimously accepted that the level of (dis-)satisfaction with various institutions of representative democracy has an effect on participatory preferences.
In this article, we differentiate between the legislature and the executive because they have specific roles and the dissatisfaction of citizens can get specific meanings. In a representative democracy the parliament is the arena for debating and deciding upon legislation. The theory of political representation explains how elected politicians are supposed to get together and produce the most appropriate laws for the societal development. In reality, there are many instances in which the laws are delayed, contested, the result of compromise or inappropriate for certain segments of society. When citizens are not convinced about the ability of this institution to make laws, they could look for alternatives. Direct democracy may be one of the alternatives preferred since it involves citizens directly in the decision-making process, bypasses parliamentary debates and provides a non-mediated translation of preferences into policies (Setala and Schiller 2012;Altman 2018). The deliberative procedures can also be an important alternative because they develop new ideas, bring information into the will-formation process, or provide suggestions for broadly accepted solutions (Gutmann and Thompson 2004;Warren and Pearse 2008;Reuchamps and Suiter 2016). Both alternatives provide a relevant voice to citizens and diminish the influence of an institutionthe parliamentin the decision-making process.
The government has the executive functions of initiating legislation and the implementation of policies and decisions. While citizens may be content with parliamentary law making and decision-making processes, dissatisfaction could arise from the ways in which these decisions are put into practice. People may have an issue with the pace and the form of the implementation process, government priorities and so on. For the reasons outlined above, both direct and deliberative democracy provide useful avenues to address the problem that citizens have with the government. For example, citizens who wish to have a new policy do not have to wait for the government to schedule its initiation and implementation. Instead, a citizens' initiative or a referendum could place that policy on the agenda at a much faster pace and without all the trouble associated to the executive process.
Following all these arguments we expect that the support for participatory democracy, i.e. both direct and deliberative procedures, is driven by the following variables: H1: Political interest drives support for participatory democracy.
H2: Internal efficacy support for participatory democracy.
H3: Doubts on Parliament as best arena for decision-making support for participatory democracy.
H4: Dissatisfaction with the government support for participatory democracy.

Citizens' support for direct vs deliberative procedures
We turn now to the differences between the preference for direct vs deliberative procedures. The two are different institutions. Direct democracy can be defined as 'a publicly recognized, institutionalized process by which citizens of a region or country register their choice or opinion on specific issues through a ballot with universal and secret suffrage' (Altman 2018, 6). It rests on the idea that citizens have an equal input into the collective decision-making (Saward 1998). This input shifts the control from politicians to citizens and provides the possibility for larger acceptance of decisions in society. Any changes occurring through direct democracy are often easier accepted if debates and decisions involve the citizenry rather than being limited to the political elites (Gherghina 2017). In essence, direct democracy is about the direct impact of ordinary citizens on political decisions, which can take place either through voting on policies or through the support of citizens' initiatives (Setala and Schiller 2012). The referendums and citizens' initiatives are the most common forms of direct democracy. The referendums follow the logic of 'aggregating votes' as the main mechanism of decision-making and create winners. They can create important divisions in society especially when the issue is salient in society (e.g. Brexit).
Deliberative democracy is based on the idea that the individual and collective exchange of reasoned arguments and public debates precede decisions. It marks the shift from a vote-centric to a talk-centric conception of democracy (Chambers 2003). According to this understanding, 'the public deliberation of free and equal citizens is the core of legitimate political decision making and self-government' (Bohman 1998, 401). Deliberation involves key components such as respectful communication and conversation, confrontation with different opinions and experiences with the aim of finding the best argument (Fishkin 2011). The key concept behind deliberation is that citizens' interactions in deliberative procedures may improve both the epistemic and the ethical quality of the choices made (Habermas 1996;Dryzek 2000). The quality of decisions and individual preferences can change when people provide considered judgments and engage in communication (Smith 2009). The literature refers to two major types of deliberation.
There is a traditional Habermasian perspective that focuses on consensus. It emphasizes the exchange of rational arguments and the search for the common good. The alternative is a type of deliberation that is open to various modes of communication (e.g. stories, testimonies) in order to adapt the ideal of deliberation to the real world (Bächtiger et al. 2010). The two major forms of deliberation are the mini-publics and participatory budgeting. The latter has a strong component of public deliberation in which citizens are involved at the regular meetings about the budget (Sintomer, Röcke, and Herzberg 2016). The deliberation is aimed both at informing participants about the process and available resources and at debating about specific projects (Baiocchi and Ganuza 2014).
These features indicate clear differences in the aims, functioning and consequences of direct and deliberative democracy. Direct democracy often aims at producing a change in society and is oriented towards reaching decisions. Deliberative procedures are set for a variety of different purposes and are highly heterogenous (Esau, Fleuß, and Nienhaus 2020). While they might fulfil different functions also for democratic-decision-making, they usually do not have the power to make final political decisions. In addition, they require different citizens' competence: direct democratic procedures are characterized by limited citizen involvement, while 'deliberative practices attract knowledgeable and motivated citizens willing to invest time in assemblies or town meetings' (Font, Wojcieszak, and Navarro 2015, 155). In spite of these differences, there are some instances in which participatory procedures may combine deliberative and direct democratic elements. For example, in some countries, the participatory budgeting procedures based on some kind of deliberative citizens' assemblies, can unfold decisive aspects. Another example is the participatory process that includes a combination of deliberation and referendum, in this order (e.g. Ireland).
In general, direct democratic and deliberative procedures follow different logics, require different levels and types of citizens' engagement and focus either on decisionmaking (direct democracy) or political will-formation (deliberation). These differences indicate that there are good reasons to expect that these two forms of participatory procedures are favoured by different types of citizens. So far, very few empirical studies tried to disentangle citizens' support for direct versus deliberative procedures. The following lines build arguments, explained below in detail, about how (1) doubts on politicians, as decision-making actors, and experience with direct democracy may favour preferences for direct democracy, and (2) civicness, which most likely entails an interest in discussion and interaction, and experience with deliberative democracy may favour preferences for deliberative democracy.
As previously explained, one of the main differences between direct democratic and deliberative procedures is about who decides. Whereas deliberative procedures serve as means for political will-formation with politicians remaining the decision-makers, direct democratic procedures shift the decision-making power to the citizens. As such, we may expect the negative perceptions of politicians to influence the preferences for direct democracy but to have little or no impact on citizens' preference for deliberative democracy. More precisely, if citizens do not trust politicians to act in the interest of people and pursue own interests instead, they are likely to prefer direct democracy. This puts the decision-making directly into the hands of the citizens themselves and takes it rightly out of the hands of the politicians who are perceived as corrupt and incompetent. Simply put, 'people disenchanted with the performance of traditional representative institutions may … be expected to be more enthusiastic about direct democracy' (Bowler, Donovan, and Karp 2007, 253). Citizens, who doubt that politicians make decisions in the interest of 'the people', are likely to favour direct democracy because they are more inclined to take (major) decisions out of the hands of politicians. In contrast, supporters for deliberative democracy may have more faith in politicians as 'good' decision-makers. The latter are probably more willing to foster the deliberative process of politicians via citizen deliberation, leaving the final decision to politicians.
The level of civic engagement is a well proven indicator for political participation (Rosenstone and Hansen 1993;Verba, Schlozman, and Brady 1995;Putnam 2000) and we can expect a differentiated impact on direct vs deliberative democracy. Citizens involved in (political or non-political) organizations usually have an intrinsic interest in collective discussion and social interactions, which might involve a greater interest in deliberative procedures compared to not engaged fellow citizens. Through civic engagement citizens get a sense of collective will formation within their communities which might strengthen their preferences for deliberation (Leighley 1996). Many deliberative procedures are oriented towards addressing community problems and civically engaged citizens might have experienced that these problems can be solved via communication and discussion, i.e. deliberation. We argue that civically engaged citizens are more inclined towards deliberative procedures because these create a broader sense of community involvement and deliberation instead of majoritarian decision-making. The entire idea of deliberation lies on the promotion of dialogue and exchange of ideas, with formats of mini-publics aiming to reach agreement and to find solutions agreeable to the participants.
Equally important, deliberative procedures are more demanding in terms of resources to be allocated. Civically engaged people often know how to gain access to resources, how to use them and how important they are in the entire process of generating changes. They are likely to develop a positive attitude towards another community-oriented process, i.e. deliberative procedures. In contrast, non-engaged citizens may prefer direct democratic procedures because casting a ballot is less demanding. Direct democracy requires less information and lower level of resources. Citizens, who are engaged in civil activities, might also prefer to work together towards a common solution in case of divisive issues. In contrast, non-engaged citizens might be less inclined to deliberate on controversial topics, but to 'get things decided' by majority vote.
In addition to these, we expect experience with direct and deliberative procedures to have a positive impact on citizens' preferences for respective procedures. We might speculate that citizens, who have already taken part in a direct democratic or deliberative procedure, are more inclined to favour the respective options. Empirical studies provide a more complex picture. Font and Navarro (2013) show that Spanish citizens with experience with local participatory procedures are less favourable for these options. Another study illustrates how skepticism towards participatory procedures can occur after the frustration generated by personal experience in participation (Fernández-Martínez, García-Espín, and Jiménez-Sánchez 2019). However, this skepticism is often due to frustrating experiences with really existing processes and citizens are able to abstract from concrete experience and still favour participatory procedures over purely representative democracy (Levine and Nierras 2007). All in all, we can assume that experiences with direct democratic and deliberative procedures most likely influence preferences, although the direction if influence is not obvious.
Based on these considerations we expect the support for direct democracy to be rooted in the following variables: H5: Doubts on politicians' public interests drives support for direct democracy.
H6: Experience with direct democracy drives support for direct democracy.
The support for deliberative democracy is likely to be driven by the following variables: H7: Civic engagement drives support for deliberative democracy.
H8: Experience with deliberative democracy drives support for deliberative democracy.

Control variables
The analysis includes three control variables with potential impact on preferences for both direct and deliberative democracy: consumption of political news, education and age. 2 Starting with the consumption of political news, citizens following public affairs are more involved than citizens not consuming respective information (McLeod, Scheufele, and Moy 1999). News consumption can be a potential driver for preference formation (Donovan, Tolbert, and Smith 2009;Schuck and de Vreese 2011;Gabriel 2013). Citizens who actively search for political information are more inclined to participate and, in turn, might be more in favour of direct and deliberative procedures that expand the realm of participation.
Education is a key variable in the field of political attitudes. Empirical findings are mixed and mostly refer to preferences for direct democratic procedures. Studies on Finland and the Netherlands show that citizens with lower education are more likely to favour direct democracy (Bengtsson and Mattila 2009;Bovens and Wille 2010;Coffé and Michels 2014). In the US there is no correlation between education and support for citizens as decision-makers (Bowler, Donovan, and Karp 2007). Considering differences between preferences for direct and deliberative procedures, Landwehr and Faas (2015) do not identify an effect of education among German respondents. Age can also play a role: Several studies show that younger people favour participatory procedures and institutions more strongly than older citizens (Dalton, Burklin, and Drummond 2001;Jeydel and Steel 2002;Donovan and Karp 2006).

Research design
To test our hypotheses, we use individual level data from an original survey conducted online in the UK in December 2018. The survey uses a sample of 1094 respondents (complete answers), which is representative at national level on several key socio-demographic variables. This ensures proportionality in the sample according to the most recent census in the UK: area of residence (region), age, education and gender. The high level of Internet usage in the UK (estimated around 90% in the year of the survey) reduces to a large extent the likelihood of self-selection bias. The survey draws on an online panel that includes a very large pool of respondents, which allows for a sampling strategy that matches the general features of the British population. The survey was pretested in October and November 2018. The questionnaire was designed by the authors of this article and included specific questions about preferences for direct and deliberative democracy, general political attitudes and political behaviour. The 'do not know' and 'no answer' options were treated as missing values and excluded from analysis.
The UK is the appropriate context to investigate preferences for direct democracy and deliberative procedures due to their availability and use at different levels of government. Two referendums were organized at national level in the last decade: 2011 on an alternative voting system and 2016 on Brexit. At the regional level several referendums have taken place, mainly related to devolution and independencein Northern Ireland, Scotland and Wales and in parts of England (in the North East and London). At the local level there are many referendums and citizens' initiatives. Deliberative procedures took place in the UK, mainly at regional and local level. In the fields of local transport or health public deliberation plays a crucial role. Another example, in Scotland the government has recently revealed plans for a Citizens' Assembly, while Edinburgh runs a participatory budgeting for several years.

Variable operationalization
The two dependent variables of this study are the support for direct or deliberative procedures. They are measured as the answers provided to the following question 'To what extent would you like to see adopted and / or implemented at national level the following options, which give citizens more direct say in the UK politics?'. The wording 'give citizens more say' does not include a certain bias towards direct democracy because deliberative practices also allow citizens to express preferences. The survey pre-test did not highlight any biases in the direction of direct democracy. This question included two forms of direct democracy and two forms of deliberative democracy. The forms of direct democracy were the referendum and citizens' initiatives. The latter are direct democratic instruments that do not necessarily end up in a vote on a policy. Nevertheless, they follow the aggregation logic of collecting signatures in order to put a topic on the political agenda. The forms of deliberative democracy were mini-publics and participatory budgeting. 3 The available answers indicated preference for each of these forms and were recorded on a four-point ordinal scale between not at all (1) and very much (4). The answers about referendum and citizens' initiatives were merged to form the support for direct democracy, while the answers about mini-publics and participatory budgeting were merged to form the support for deliberative democracy. Both variables are coded on a seven-point ordinal scale ranging from 2 (not at all for both forms merged) to 8 (very much for both forms). The values of the Cronbach's Alpha for each dependent variable indicate high levels of internal consistency: 0.79 for direct democracy and 0.84 for deliberative democracy.
Interest in politics (H1) is an ordinal variable measured on a five-point scale as the answers provided to the question about how much interest respondents have in politics: 1 = not at all and 5 = very interested. Internal efficacy (H2) is also ordinal, operationalized as the answer to the question about how confident respondents are in their ability to participate in politics. The available answers range from not at all (1) to completely (5). Doubts on parliament as the best arena for decision-making (H3) is measured on a six-point ordinal scale (1 = totally disagree, 6 = totally agree) as the answer provided to the following statement 'Parliament is the best institution to decide on laws'. The satisfaction with the government (H4) is measured through the following question: 'Thinking about the UK government, how satisfied are you with the way it is doing its job?'. The available answers ranged between extremely dissatisfied (0) and extremely satisfied (10).
The doubts about politicians' public interests (H5) is measured through the answer provided by respondents to the statement 'In their actions UK politicians pursue mainly citizens' interests'. Answer options range between totally disagree (1) and totally agree (6). Civicness (H7) is a six-point cumulative index of five dummy variables that measure the involvement (coded as 1) of respondents into social, economic, cultural, political and charity organizations. This item refers mainly traditional, hierarchical organizations but it does not exclude newly forms of involvement. Respondents could decide whether looser networks to which they belong fall under one of these categories. The question asks them to explain whether they were members of groups, associations or organizations in those areas. Experience with direct (H5) and deliberative democracy (H8) is measured on a six-point ordinal scale as the answer provided to the question 'How often did you undertake the following activities (referendums, participatory budgeting or other kinds of political consultations) in your life?'. The available answers range from never (coded 1) to very often (coded 6).
The consumption of political news is measured as the answer to the question about how often respondents follow political news on the web, TV, radio or newspapers. It is a five-point ordinal scale that ranges from 1 (never) to 5 (several times a day). Education is a four-point ordinal variable that ranges from basic/primary (1) studies to postgraduate degree (4). Age has been recoded into four categories with the respondents up to 30 years old in the first group, those between 31 and 45 in the second, those aged 46-60 in the third and above 61 in the last one (for descriptive statistics, see Appendix 1). 4 Figure 1 depicts the distribution of support for direct democratic and deliberative procedures among the respondents in the UK. The percentages show that a very limited share of respondents rejects the idea of direct or deliberative democracy. Only 10%, i.e. the sum of the first two bars in the graph, answered that they would not want either referendums or citizens' initiatives adopted or implemented further at national level. Less than 15% have a similar attitude towards mini-publics or participatory budgeting adopted at the national level. Overall, the British are quite favourable to direct and deliberative democracy. This observation may appear slightly counter-intuitive at least for direct democracy in the context of a highly divisive Brexit in 2016, which was ongoing at the time of the survey. 5 The positive attitude of many British towards direct democracy indicates that they distinguish between the procedures of direct democracy and the outcome that sometimes is not favourable to them. In other words, few oppose direct democracy even if they were not on the winning camp in the most recent popular vote.

Analysis and results
The aggregate distribution of support for both direct and deliberative democracy is quite similar. The correlation coefficient between the two preferences is 0.67, statistically significant at the 0.01 level, which strengthens the idea that people have similar attitudes towards the two types of participatory democracy. However, the individual level analysis of the two preferences reveals that an important share of respondents have different attitudes towards direct and deliberative democracy. In essence, we do not have the same individuals in the same categories. For example, only one-quarter of those respondents who want very little direct democracy (score 3) have exactly the same attitude towards deliberative democracy. Similarly, only 30% of those who indicate that they strive for more direct democracy (score 5) also strive for more deliberative democracy. Table 1 presents the results of the bivariate correlation analysis between each of the independent variables (plus controls) and the support for direct and deliberative democracy. There is empirical evidence for all the hypothesized effects and with one exception the relationships are statistically significant at the 0.01 level. The first argument of our article is that some features (H1-H4) favour support for both direct and deliberative democracy. The support for hypotheses goes in the expected direction but the strength of relationships differs. Interest in politics correlates highly with both support for both participatory procedures. Dissatisfaction with the government correlates the highest with support for direct democracy and internal efficacy with support for deliberative democracy. The doubts that the parliament is the best arena for decision-making correlates weakly (H3) with supporters of deliberative democracy and for direct democracy it lacks statistical significance. Among the controls, people who are more informed are likely to support both direct and deliberative democracy, while younger people support only the latter. Our second argument is that, due to their different logic and principles, the support for direct and deliberative democracy is driven by different factors. The evidence confirms our theoretical argument and indicates that doubts raised about parliament as arena for decision-making and experience with direct democracy correlate with support for direct democracy. None of these two variables correlates with support for deliberative democracy. Corollary, civicness and experience with deliberative democracy are associated with higher support for deliberative democracy. Both are very close to statistical independence relative to direct democracy.
Next, we run multivariate statistical models to test for the effects of these variables. We use Ordinary Least Squares (OLS) regression with robust standard errors. There are three models for each dependent variable: one with the four general variables, the second with the four general and the four specific variables, and the third with all variables plus the controls. In general, the values of the regression coefficients in all three models confirm the correlations but provide new valuable insights and nuances to the bigger picture.
To begin with the support for both direct and deliberative democracy (H1-H4), there is empirical evidence for all the hypothesized relationships. Figure 2 includes the average marginal effects for model 2; all statistical models are presented in Appendix 2. The regression analysis indicates that the supporters of both direct and deliberative democracy have interest in politics and have doubts that the parliaments are the best institution to produce legislation. These are the two strongest predictors, see Appendix 2 for the coefficients. Higher internal efficacy and lower satisfaction with the activity of the government also lead to more support for both direct and deliberative democracy but to a lower extent compared to the previous two variables. The effects are fairly similar and they always go in the same direction.
Two immediate conclusions can be drawn. On the one hand, the support for direct and deliberative democracy rests on similar grounds. The four variables have an effect in the same direction and they are statistically significant most of the times. Our results contradict earlier claims according to which disinterested citizens are more inclined to support direct democracy, while the highly interested citizens may prefer deliberation (Bengtsson and Mattila 2009). We show that political interest drives preferences for both direct and deliberative democracy. Among the controls (Model 3) age is the strongest predictoralso statistically significantwith younger citizens supporting both direct and deliberative democracy. The prevalence of age in explaining support for the two alternative models of democracy says something about their potential popularity in the near future, but also downplays the negative effect of education on support for direct and deliberative democracy. Since younger generations are more educated than the older generations, it is not surprising that we do not find, contrary to earlier studies, a strong and significant negative effect of education. As such, our findings confirm that participatory preferences are more than populist tendencies of low educated citizens (Webb 2013; Bowler et al. 2017).
On the other hand, the comparison of the two strongest predictors for each model of democracy indicates that the support for them lies on different grounds. That is stronger evidence for the second argument that we make in this article. The variable about doubts on parliament (H3) has a stronger effect on the support for deliberative democracy. Originally our argument was the other way around, we expected more doubt among the supporters of direct democracy. However, the empirical evidence points into a different direction. We interpret this finding as an indicator that UK citizens desire political decisions to be deliberated and not just decided. They do not consider the Parliament as the bestor the onlyarena and prefer the inclusion of citizens' perspectives and arguments. This also goes hand in hand with the finding, that supporters of deliberative democracy have doubts on politicians' public interest (see below). Very likely, the supporters of deliberative democracy want all arguments and interests to be taken into account in the will-formation processand they do not necessarily trust that parliament and politicians act accordingly.
The level of internal efficacy has stronger effects on preference for deliberation rather than on those for direct democracy. In other words, internal efficacy plays a bigger role in the support for deliberative and less in that for direct democracy. For deliberative democracy, the internal efficacy and doubts about parliament as arena for decisionmaking are the strongest explanations. This finding goes hand in hand with our expectation, which we can elaborate now. We assume that citizens with high level of internal efficacy not only trust themselves to deliberate due to their feeling of strong communicative and political competence. They also might have the impression that generally deliberation is a necessary step of political decision-making. For them politics is more about finding the best solution. Being highly efficacious also most likely encourages the wish and the demand to influence political will-formation and to make sure that citizens' competences find their way into and contribute to policies.
Moving on to the specific drivers of support for direct (H5-H6) and deliberative democracy (H7-H8), the empirical evidence illustrates that experience has an important impact. Individuals who participated in referendums are more willing to have direct democracy procedures in place. Similarly, once they have experience with some deliberative settings people are more likely to be supportive towards having them in place. These effects are presented in Figure 3, which contrasts the support for the two types of democracy. It is important to note that the British citizens who have experience with one type of democracy do not show support for the other type. For example, respondents who have more experience with direct democracy support deliberative democracy to the same extent as those respondents who had no experience.
The evidence for the remaining two specific variables provides a nuanced understanding of people's attitudes. The doubts that politicians pursue public interests do not appear to have an effect on the support for direct democracy. However, those people who believe politicians pursue citizens' interests support deliberative democracy. This is in line with earlier findings and with practical arrangements according to which deliberation is mainly a complementary of politicians' work. Many outcomes of deliberation depend on politicians to be implemented and thus a positive attitude towards politicians leads to more support for deliberation. Supporters of deliberative democracy desire thoughtful debates and meaningful input, and the involvement of politicians who pursue the public interest could take the discussion to the level of decision-making. The latter would be able to include deliberation that reflects citizens' interests. Although supporters of deliberative democracy do not consider parliaments as the best avenue for decision-making, they see the politicians oriented towards citizens' interests as an important feature of deliberation.
Civicness has a similar positive effect on support for both direct and deliberative democracy. The theoretical expectation was a strong effect for deliberative democracy. We had expected that citizens involved in community activities are more inclined to debate in order to find best solutions. However, the effect is fairly small and similar to what we observe for direct democracy. In itself civic engagement is not a crucial factor. Engaged citizens do not necessary prefer deliberation. Several explanations are possible for this finding. Citizens might be engaged in activities that are not dialogue-oriented, e.g. some charity organizations, and accordingly deliberation might not sound too attractive for them. And even if they are engaged in talk-centred activities, they might have little faith that deliberation has any effect. This would also explain, why civil engagement has in the statistical regression the same effect on preference for deliberative as well as on direct democratic procedures. Engaged citizens want to have more say, and they want more say as well in political-will formation (deliberation) as well as in decision-making (direct democracy).

Discussion and conclusion
This article sought to explain why citizens' support direct democracy and deliberative procedures, using attitudinal data on British citizens. We advanced two arguments for which we brought empirical evidence: (1) several general determinants have a positive impact on the support for both direct democracy and deliberation and (2) specific determinants could drive people towards supporting one of the two alternative models of democracy. We have three sets of findings. First, there is broad support for both direct and deliberative democracy among the UK respondents. Second, the support for both models of democracy is driven in the same direction by the general variables: interest in politics, efficacy, contestation of parliament and dissatisfaction with government. Third, the supporters of direct democracy differ from those of deliberative democracy in several ways. They are animated by greater political interest, dissatisfaction with the government and experience with direct democracy. The supporters of deliberative democracy rest more on contestation of parliament, internal efficacy and experience with deliberative democracy.
The implications of this analysis reach beyond the single-case study. The empirical findings bring contributions to the broader field of democratic preferences. The results about the support for participatory decision-making processes are in line with most studies in the field. Political interest and efficacy combined with political dissatisfaction are crucial determinants for participatory preferences (Gherghina and Geissel 2019). However, this is only half of the story since very few studies differentiate between support for deliberative and direct democratic procedures. This is where the main contribution of our work lies. This is one of the first analyses differentiating between these two types of support and reveals that citizens who prefer direct democracy show partly different features than those preferring deliberative democracy. Experience with one or the other form supports the preference for the familiar type of participatory involvement. Citizens, who support deliberative forms show more internal efficacy, which is intuitively convincing, since they have more faith in their own abilities to discuss political issues. They are probably more convinced of their skills to elaborate their policy preferences, political interests, and values. They also have less trust in the abilities of the parliament to find the best solution. In contrast, they seem to be more convinced that public deliberation would be a better way of preparing political decisions. One of the important methodological implications of our study is the use of specific survey questions on participatory preferences. Our study shows that it is analytically fruitful to differentiate questions on preferences for direct democratic versus deliberative procedures. British citizens show preferences for different participatory procedures and we have no reasons to believe that such an attitude is country specific. As a result, the use of specific batteries of questions in cross-country surveys could allow the capture of cross-country differences.
These results may also bear importance for policy-makers. To ensure congruence between citizens' participatory demands and the participatory developments of a democracy, it is crucial to know what the citizenry prefers. Policy-makers, who want to react responsively to citizens' procedural preferences need information about the interests and wishes of the latter. In this sense, they have to know if citizens are satisfied with the decision-making procedures purely based on elite decisions, if they prefer public deliberation while leaving the final decision to their representatives, or if they are more inclined to decide themselves.
Our study is limited to citizens within the UK and future research will show whether our findings can be generalized. A variety of cross-country surveys reveal that citizens around the world demand more possibilities to engage in decision-making (Geissel 2016); this indicates that the broad support for direct and deliberative processes is no British peculiarity. Yet, little is known whether preferences of citizens for direct democracy and/versus deliberative democracy vary in cross-country perspective and whether respective preferences can be explained with the same variables across several countries.
Future research may also go in a different direction and classify the respondents into four types according to their preferences for direct and deliberative democracy: preference for none, for one of them or for both. Such a typology could help coupling difference preferences with behaviours. Along similar lines, the cross-national differences can be analysed by using both individual and macro-level variables. For example, citizens in consensus-oriented countries might opt more strongly for deliberative procedures because they consider debate and consensus as strong values. In contrast, citizens living in majoritarian countries might prefer more say in direct decision-making, i.e. direct democracy.

Notes
1. For simplicity we use the terms 'participatory', 'direct' and 'deliberative' democracy to systems with strong participatory features, which can be either direct democratic, deliberative or both. These democracies are still organized in a representative setting and are not purely 'participatory', 'direct' or 'deliberative' but provide comprehensive respective procedures. 2. We have tested for other potential controls identified in the literature as driving support for direct democracy, for deliberative democracy or for both: ideology, income, medium of residence, media exposure. For example, ideology has been identified as a potential source of attitudes towards deliberative practices (Christensen and von Schoultz 2019). We tested for ideology both as left-right placement and party vote in the 2017 UK elections. None of these appear to have an effect on citizens' preferences. For reasons related to model parsimony, we report only the most often encountered controls although some of these also fail to produce an effect. 3. For citizens' initiatives, mini publics and participatory budgeting respondents had an explanation in brackets, in case they did not know what they stand for. The cognitive pre-test conducted prior to the survey and the answers to other questions in the survey about direct and deliberative democracy indicate that respondents understood what they mean even when they did not use them. 4. We tested for a series of other socio-economic characteristics such as income, medium of residence or gender. Their effects were neither strong nor statistically significant and thus not included in the analysis. 5. Nevertheless, it is important to note that the support for these two forms is high in the country even after the highly divisive referendum in 2016 about the exit from the European Union.

Disclosure statement
No potential conflict of interest was reported by the author(s). Appendix 2. Ordinary least squares regression to explain support for direct and deliberative democracy Note: Coefficients are standardized with robust standard errors in parentheses. **p < 0.01. *p < 0.05.