Deliberative democrats have a history of aversion against referendum and initiative processes (see, for example, Chambers 1998; Cohen 1989; Fishkin 1995; Gastil & Richards 2013; Gutmann & Thompson 2004; LeDuc 2015; Leib 2006; Mansbridge 1983; Offe, 2017; Parkinson 2001; Urbinati 2006). Various reasons underpin the rejection of these popular vote processes as institutions that could translate into practice an ideal of legitimate government as decision-making based on the exchange of reasons and on considered judgment.1

Referendums and initiatives would empower mass publics whose ‘deliberative competence […] is suspect’ because they are in no conditions ‘to think about how they are to exercise that power’ (Fishkin 1995: 21) to make decisions on laws and policies. Voters casting their secret ballot – on pre-selected, possibly unclear or even flawed policy alternatives – in solitary voting booths, without any risk to have to give reasons for their ballot decisions, would even be disincentivized to make the effort of getting informed and developing considered judgments. Moreover, the majoritarian logic of popular votes combined with the inequalities of resources present in contemporary democracies would undermine the opportunities for deliberative exchange in public debates and push strategic political elites to try and misinform and polarise voters. The prioritization of ‘votes’ over ‘voice’ (LeDuc 2015) promoted by referendums and initiatives would finally result in poor quality policies adopted through the aggregation of unconsidered individual opinions, or even limitations to the rights of minorities. On this view, conventional representative democracy in which elected representatives only decide on policy issues would be more appropriate to institutionalize deliberative democracy: ‘democratically elected and accountable representatives of citizens may be better deliberators’ than citizens (Gutmann & Thompson 2004: 31).

The scholarship on deliberative democracy has increasingly recognized the limitations of these diagnoses. Some researchers have argued that popular vote processes can be made more deliberative by being coupled with small-scale deliberative forums of citizens selected by lot, such as citizens’ assemblies and minipublics (see Chambers 2018b; Gastil & Knobloch 2020; Gastil & Richards 2013; Landemore 2018; Levy 2017). Adding these designs in ways that are visible to the broader public, in great part thanks to media coverage, is expected to guarantee that the decision-making moment of the popular vote is preceded by and connected to high quality deliberation – thus enhancing the provision of information to voters, ensuring that various perspectives and arguments are heard, creating incentives for higher quality debates, and inducing more reflexivity on the part of voters (Chambers 2018b; Landemore 2018; Parkinson 2020; Warren 2008). Different kinds of coupling have been proposed (Gastil & Richards 2013; Saward 2001), two of which have been deemed particularly successful: organizing citizens’ assemblies to participate in determining which policy options should be put to popular votes, as in the 2004 British Columbia Citizens’ Assembly and the Irish 2012 Constitutional Convention and 2016 Citizens’ Assembly experiences (Suiter & Reidy 2019; Warren & Pearse 2008); and empowering small minipublics to deliberate intensely on an issue put to a popular vote for four to five days to produce a one page information document destined to be distributed to all their fellow voters – namely, the Citizens’ Initiative Review process imagined by John Gastil and first implemented in 2010 Oregon (Gastil, 2000), which has since gained tremendous support and is being experienced in a variety of contexts, from other US states to Finland and Switzerland.2

This approach is only one way to reconcile popular vote processes with deliberative democracy. It corresponds to what André Bächtiger and John Parkinson name an ‘additive’ approach to deliberation: it focuses on ‘adding in deliberation at some point of a process’ as a way of enhancing the overall deliberative quality of popular vote processes (Bächtiger & Parkinson 2019: 44). Another approach is the ‘summative’ approach to deliberation, according to which deliberativeness is a quality of political systems that results from the interactions between its various parts, ‘no part of which need be fully deliberative (or fully democratic) on its own’ (Bächtiger & Parkinson 2019: 104; see also Mansbridge et al. 2012). It emphasizes that popular vote processes alone, even without being coupled to deliberative processes and even if they fail to meet all the criteria of good deliberation as ‘discrete entities’ (Felicetti, Niemeyer, & Curato 2016: 428), can play a valuable role in fostering deliberation at the system level. In this perspective, the notable idea has emerged that (certain kinds of) referendum and initiative processes have a specific and valuable role to play for deliberative systems in fostering actual, catalysed mass communicative exchange between various actors of democratic systems (Chambers 2009, 2018b; el-Wakil 2017; Lang & Warren 2012; Parkinson 2020; Setälä 2006).

In this article, I focus on this second approach to reconsider the deliberative potential of popular vote processes. I propose two contributions. First, I build on recent literature to argue that popular vote processes can serve deliberative systems by promoting what Jane Mansbridge recently coined ‘recursive representation’ (Mansbridge 2019). This dialogue-like form of exchange between citizens and their various representatives supports deliberative systems by promoting more discursive interactions in represented-representative relationships – enhancing the representation of citizens’ actual concerns in collective decision-making processes and the provision of persuasive justifications by representatives for their actions and decisions. I argue that the incentives for representatives to behave recursively can be explained by the changes that popular vote processes introduce in the formal opportunity structure of these actors. Highlighting these structural changes opens a promising path to reconsider their value for deliberative systems. My second contribution is to lay some ground for this future research. I demonstrate why it would benefit from going beyond the existing literature, which has often narrowly focused on popular votes in use, to account for the ways in which popular vote processes modify the formal opportunity structure of the actors of the political systems in which they are included even when no popular vote is organized. I close by offering possible theoretical and empirical research questions to explore whether and when political systems including popular vote processes can serve deliberative systems by promoting processes of recursive representation on the long term.

To be sure, this article should not be understood as a comprehensive defence of referendums and initiative processes. Such a defence would not only require more space but also going beyond the deliberative framework (see el-Wakil 2020; Lacey 2017; McKay 2019; Parkinson 2020; Saward, 2003; Warren 2017). Rather, my objective is to lay out one general way of reconsidering the potential of popular vote processes in order to expand our understanding of their role and value in deliberative systems.

In various ways, the literature has provided insight on how the moments of mass voting and mass campaign on policy issues triggered by popular vote processes provide unique opportunities for actual, mass communicative exchange that can serve deliberative systems. For instance, John Uhr takes popular vote processes to constitute moments ‘where political elites and voters can come together to deliberate over quite fundamental political issues’ (Uhr 2000: 209, emphasis added). Simone Chambers similarly proposes to understand the 2014 Scottish referendum as an example of how referendums can promote ‘a fruitful two-way conversation and practical cooperation between civil society and government representatives’ and activate the public in ways that ‘a purely government-initiated information campaign’ could not achieve (Chambers 2018a: 153, emphasis added). And Parkinson argues that popular vote processes can play a valuable role in deliberative systems by ‘connecting political actors with everyday political talk’ (Parkinson 2020: 3). More specifically:

That deliberative systems need to provide space for actual dialogue between represented and representatives is a relatively recent concern in theories of deliberative democracy. It follows a general shift in the field from an ideal of legitimate collective decision-making based only on the considered opinions of deliberative actors and justified by reasons that the represented would hypothetically accept (see e.g., Cohen 1997) towards an ideal of collective decision-making that accounts for and answers to the actual concerns of and contestations of members of mass publics (see Chambers 2009; Neblo, Esterling, & Lazer 2018; Williams 1998; Young 2002). On this view, legitimate deliberative systems need to enable citizens to see how their actual preferences relate to collective will formation and decision-making processes (Mansbridge 2019: 305; Parkinson 2020: 3; Warren 2017: 44).

Representation plays a constitutive role in creating these connections at the mass level by enabling and structuring communicative processes. It makes it possible to mobilize citizens, to develop their preferences and opinions, and to include these preferences and opinions in empowered spheres of decision-making (Disch 2011; Saward 2010). This is especially the case when representation is ‘recursive’, namely when it promotes more deliberative, interactive exchange between citizens and their representatives, in which both can iteratively ‘take in what the other is saying, update, revise, and respond on the basis of their own experience, and then listen to the others’ response’ (Mansbridge 2019: 307). Representatives, in particular, ought to be incentivized ‘to hear, to respond, to explain legislative actions, and to act on citizens’ responses to those explanations’ (Mansbridge 2019: 307). These interactive communicative processes should aim at implementing ‘the criteria of good deliberation’ such as inclusion, mutual respect, and limiting power asymmetries and attempts at manipulation (Mansbridge 2019: 307). Establishing this kind of exchange as a guiding ideal for mass processes of representation in deliberative systems may already enhance the normative and perceived legitimacy of deliberative systems.

Authors concerned with the lack of recursive representation in conventional representative systems have generally not considered popular vote processes as a possible solution, favouring adaptions in electoral systems or deliberative designs (see e.g., Ferejohn 2008; Mansbridge 2019; Neblo et al. 2018). Yet, the literature mentioned above suggests that initiatives and referendums have a particular advantage in incentivizing mass conversation-like exchanges about policy issues between ordinary citizens and their representatives.

Of course, this may not always be the case; popular vote campaigns have been widely criticized as a privileged space for strategic and resourceful representative actors to try and manipulate the public by introducing divisive messages and misinformation using mass scale technology (Ferejohn 2008; Neblo et al. 2018; Wooley & Howard 2019). These are important issues that should inform the institutional design of popular vote processes in ways that I cannot fully consider here. What I wish to highlight, rather, is that popular vote processes also share a feature that justifies the expectation that they foster recursive representation: compared to conventional representative systems, systems that include referendums and/or initiatives redistribute the formal empowerments of citizens in ways that can motivate representatives to listen and reach out to them (see also Cheneval & el-Wakil 2018; el-Wakil 2017, 2020; Lacey 2017; Parkinson 2020; Setälä 2006). More specifically, while the right to vote of ordinary citizens in conventional representative systems is limited to casting ballots to participate in selecting candidates in elections, it is formally extended by popular vote processes to include casting ballots on policy issues.

It entails that, just as representatives have increased incentives to engage in recursive communication with voters to win elections (Mansbridge 2019), they have increased incentives to foster recursive exchange with citizens to see their favoured policy adopted in popular votes. To take a trivial example, let us imagine an elected representative engaged in protecting the interests of bicycle drivers, and who participated in developing a new policy to allocate some of the funding for roads to build bicycle paths. If this policy is put to a popular vote, this representative has a strong interest in the result of the vote being a ‘yes’ in support of their policy. This creates two kinds of incentives. First, trying to mobilize citizens and convince them that they should approve the proposal, which involves listening to and understanding their possible concerns as well as providing them publicly with acceptable answers in order win the support of a majority of them. Second, taking seriously their opponents in the campaign, such as the representatives of car drivers who bring the concerns of their constituencies into public debates, and answer their arguments in ways that are accessible and understandable to citizens and oriented towards convincing them to support their favoured policy option. These incentives for recursive representation might be stronger when voters are empowered to cast ballots in binding popular votes, since the decisions of the majority of voters cannot be overturned; but we can expect them to be present in advisory popular votes, too, which are politically difficult to overturn (see Chambers 2018b).

Furthermore, two kinds of referendum and initiative processes also formally empower nonelected actors to bring the actual concerns of a part of the constituency into public debates by demanding popular votes on issues of their choice (el-Wakil 2020, Chapter 5; Guillaume-Hofnung 1994; Höglinger 2008; Parkinson 2020). If they successfully collect the required number of signatures within the specified time, groups can trigger a facultative referendum to challenge a policy recently adopted by elected representatives to a popular vote, or a popular initiative to put a new proposal of their choosing to a popular vote. In both cases, these groups can gain a privileged position in the campaigns preceding the popular vote, thus increasing the chances of seeing these concerns included in public debates and answered by representatives. To build on the previous example: if an association for car-drivers triggers a facultative referendum on the newly adopted policy to fund bicycle paths with some of the budget previously allocated to roads, they gain a platform to raise their grievances and arguments against this policy in public debates. The elected representatives who passed the new legislation in the first place have strong incentives to take these arguments seriously and to answer them publicly, in ways that aim at convincing citizens to support their proposed policy change.

The idea that popular vote processes can contribute to deliberative systems by providing representatives with increased incentives to engage in practices of recursive representation offers a basis to reconsider previous debates about the deliberative value of referendums and initiatives. In particular, it puts into perspective the relevance of concerns about the deliberative quality of voters’ decisions, which have been at the centre of much normative and empirical research on referendums and initiatives in deliberative democracy (see Colombo 2018; Leib 2006; Levy 2017). To be sure, the general assumption that ‘deliberativeness’ in popular vote processes should be found among voters themselves is not an unreasonable one. Majorities of voters are a decisive actor in popular vote processes and, following deliberative standards, we should want them to be in conditions where they can develop some form of considered and informed opinion (see McKay 2019) or to benefit from the quality deliberations of minipublics ahead of popular votes, as proposed by the innovative coupling designs mentioned in the introduction. Yet, this focus on whether the opinions of citizens casting their ballots have been ‘filtered’ through forms of deliberation (Lafont 2015) has somewhat obscured questions related to the ways in which referendums and initiative impact representative processes at the system level. For instance, which kinds and institutional designs of referendums or initiatives can best incentivize recursive representation? What contexts and designs can limit the temptations and options for representatives to try and manipulate or misinform citizens in popular vote campaigns? Which actors should be enabled to trigger bottom-up processes and bring citizens’ concerns into public debates? Should they also have to offer responses to these concerns, or are elected representatives the main actors who ought to listen and respond? And which and when do representatives engage in recursive communication – and under what conditions can this be done in more or less successful ways? Representation processes are central to deliberative – and democratic – systems; better conceptualizing and understanding of how referendums and initiatives impact these processes is essential to fruitfully reconsider their deliberative potential.

I have so far focused on how popular vote processes can impact the behaviour of representatives in valuable ways when they are used. This follows the existing deliberative literature, which has generally focused on the deliberations (or absence thereof) that surround popular votes that take place. To take, again, some examples, Hélène Landemore insists that deliberative democracy encourages us to focus on the campaigns preceding popular votes (Landemore 2018: 322). Parkinson conceives popular votes as ‘the end point of longer, multi-site, multi-method processes that include at least one focal, well-publicized deliberative forum’ (Parkinson 2020: 23). And Chambers insists that ‘a deliberative democratic approach to referendums focuses on the structure and content of debate’ preceding votes (Chambers 2018b: 307). However, this focus tells only part of the story about the impact of popular vote processes on collective processes of will formation and decision-making. In this section, I suggest that a broader outlook that accounts for the ways in which referendums and initiatives institutionally reshape opportunities for deliberation throughout the system is needed to answer questions such as the ones listed above – and to raise new relevant paths for research.

In addition to highlighting the relevance of studying the representative processes incentivized by popular vote processes, another insight from the argument of the previous section is indeed that including referendums and initiatives in democratic systems modifies the formal opportunity structure in which various actors are embedded. This, in turn, impacts these actors’ incentive structures and behaviour (see Smith 2009: 198–199; Warren 2017: 43) – for instance, encouraging representatives to adopt more recursive behaviours in popular vote campaigns, as argued above.

The impact of these changes in the formal opportunity structure does not stop at the close of the popular vote. The availability of top-down referendums gives elected majorities the option to determine whether and when to organize popular votes on questions of their choosing at any given time. In contrast, mandatory referendums keep limiting their competence to change certain kinds of legislation without the support of majorities of voters. And bottom-up referendums and initiatives leave the option to nonelected groups to demand popular votes on specific issues at any time, thus making the competence of elected representatives to have the last word on these issues conditional on the trigger of such processes (see el-Wakil 2020). Such distributions of formal empowerments and constraints have generally remained outside of the main concerns of deliberative democrats (Warren 2017: 40). Yet, without acknowledging them, we cannot explain why initiatives and referendums have been shown to impact processes of collective will formation and decision-making even when no such process is triggered (Boehmke, 2005; Gerber, 1999; Leemann, 2015; Magleby & Patterson, 1998; Papadopoulos, 2001; Sciarini & Trechsel, 1996). This calls for broadening our guiding question from ‘how can the use of popular vote processes contribute to deliberative systems?’ to ‘how can political systems including popular vote processes promote deliberation?’ and ‘can they do so in ways that other political systems cannot?’

Here is not the place to develop an answer to these general questions. But it is worth mentioning that existing literature gives provisional support to extending the claim proposed in the previous section about popular vote processes in use – namely, to the idea that systems including popular vote processes, especially bottom-up ones, can set institutional incentives that promote deliberation by fostering mass processes of recursive representation over the long term. For instance, Maija Setälä argues that bottom-up referendums set institutional incentives for elected representatives ‘to be more reflective about the pros and cons of [all the decisions they make], and to consider how their positions could be justified in terms acceptable to voters’ in case a popular vote would be demanded (Setälä 2006: 716). Adopting a democratic systems framework, I similarly suggest with Francis Cheneval that ‘the availability of bottom-up referendum processes indirectly enhances electoral representation by bringing uncertainty to elected representatives, who have additional incentives to anticipate and stay in dialogue with larger parts of the citizenry’ (Cheneval & el-Wakil, 2018: 299). Joseph Lacey also argues that referendums and initiatives modify the structure of voting spaces and ‘act as a supplement to elections by providing additional incentives for representatives to attune their policy programmes with the values and interests of citizens’ – but also to foster more inclusive forms of representation of various interests in the society in the long term (Lacey 2017: 45). Furthermore, Amy Lang and Mark Warren evoke the idea that the anticipation of a popular vote might also encourage actors other than elected authorities to engage with the wider public. Knowing that a referendum on their selected propositions would take place would have pushed the participants in the British Columbia Citizens’ Assembly to design policies with the following guiding question in mind: ‘what policy will the public accept?’ (Lang & Warren 2012: 306).

These theoretical accounts highlight that the inclusion of popular vote processes in the opportunity structure of democratic systems adds formal constraints on representatives. Compared to representatives in conventional representative systems, they do not only need the support of citizens in elections, but also the support of majorities of citizens to some of their policies (or against policies proposed in initiatives that are detrimental to their agenda). Their incentives to anticipate these decisive moments over the long term by engaging with citizens, taking their actual views into account, and giving them reasons to favour their position in case popular votes would take place are thus higher (see also el-Wakil 2020). Empirical research, mostly developed on the Swiss case, has so far provided support to this hypothesis. In particular, it has shown that the availability of bottom-up popular vote processes goes hand in hand with the development of ongoing processes of consultation during legislative processes in government and parliament (see e.g., Neidhart 1970; Sciarini & Trechsel 1996; Papadopoulos 2001; Hug 2009). For instance, a study comparing the inclusion of citizen groups in pre-parliamentary phases of decision-making in Switzerland and in Denmark suggests that the presence of facultative referendums in Switzerland explains, in part, why more Swiss citizen groups than Danish ones are included (Christiansen, Mach, and Varone 2018, 541). Swiss elected representatives have increased interests in including the nonelected representatives of a variety of citizens, such as interest groups and associations, and to take their views seriously: ensuring that new policies can be supported not only by a majority of their fellow elected representatives, but also by the wider public enables them to either avoid the trigger of facultative referendums or to increase their chances to win popular votes.

Further theoretical and empirical research is needed in order to support the claim that political systems including referendums and/or initiatives foster recursive representation even when these processes are not used – and whether, and when, they do so in ways that serve deliberative systems. A number of questions ought to be considered. Some of them are conceptual and crucial to understand the extent to which popular vote processes modify the rules of the game and redistribute empowerments, as well as the dynamics they create in the political systems in which they take place: how do the inclusion of popular vote processes modify the opportunity and incentive structures of democratic systems? Which of these changes can be expected to promote recursive representation in better or worse ways? And which designs of popular vote processes are more or less conducive to recursivity on the long term? Top-down referendums might for instance foster continuous interactions between elected representatives and citizens less than other kinds of popular vote processes whose trigger is outside of their control. Bottom-up referendums in particular might set better incentive structures for recursive representation, as suggested in the literature cited above, since they introduce uncertainty for elected representatives as to which laws or policies might need the approval of majorities of voters to pass. The ways in which these opportunity structures interact with other ones, and how this affects recursive representation, should also be uncovered (see el-Wakil 2020; McKay 2019). Do the incentives for representative actors change when referendums and initiatives are combined with other political processes, such as proportional or majoritarian elections? And how can we expect broader contextual variable, such as media systems (Chadwick 2013), to impact the incentives for and the success of recursive representation?

Other questions are normative. Part of this research agenda requires clarifying the value of recursive representation at the mass level for deliberative systems, especially as it is promoted by popular vote processes. One puzzle shows the relevance of clarifying how processes of recursive representation should be reflected in policy decisions – both in governments or parliaments and in popular votes. Indeed, researchers on popular vote processes tend to agree that one of their long term impact is to bring policies closer to the preferences of citizens (see e.g., Le Bihan 2018; Leemann & Wasserfallen 2016; Matsusaka 2004). It remains to be determined whether this is a result of the kind of recursive representation that popular votes might foster, and whether such findings could reflect that citizens’ preferences also come closer to those of representatives thanks to this kind of interactive exchange. But, at the normative level, and considering that it might entail that certain progressive policies are more difficult to implement (see e.g., Papadopoulos 2001), is this a possible long-term consequence of taking the actual preferences of citizens seriously for the substance of legislation that deliberative democrats should support? And if so, under what conditions should such legislation be considered legitimate? Relatedly, should there be filters to prevent that some kinds of concerns, for instance anti-democratic ones, are brought to public debates by nonelected groups and impact policy-making? If so, how to justify these filters and what form should they take? Other questions centre on how recursive representation should take place. Whose concerns ought to be listened to and answered by elected representatives as a matter of priority? Can there be a distribution of ‘recursive labour’, and what should it look like? Should all representatives engage in recursive communication to the same extent, or is it mainly about elected or empowered representatives? What role should various transmission belts (Kuyper 2016), such as the media, play in supporting such forms of communication between representatives and represented? And what constraints should be in place to avoid abuses on the part of recursive representatives – for instance, in terms of giving citizens control over their elected and nonelected representatives? Questions related to how these normative standards ought to inform our assessments of institutional designs and practices of popular vote processes should also be considered, such as: which designs of popular vote processes should we favour to foster recursive representation at the mass level? How should they be coupled with minipublics or other devices in order to create better opportunity and incentive structures to favour quality recursive exchange between over the long term? Answers to these normative questions are essential to clarify whether, and how, systems including popular vote processes have the potential to foster deliberative democracy at the mass level – or whether they fall outside its frontiers.

Finally, considering popular vote processes as possible mechanisms that can serve deliberative systems by fostering recursive processes of representation highlights paths for empirical research. Some have been mentioned above, such as testing whether processes of recursive representation are taking place in systems that include popular vote processes, and whether they are more prominent in these systems than in conventional representative ones (see Hug 2009). One way of observing the effect of popular votes in use on recursive representation could consist in observing whether elected representatives develop new arguments – arguments not previously mentioned in parliamentary debates – in popular vote campaigns, in response to concerns raised by their challengers (for instance, the groups triggering a facultative referendum). Studying such behaviours over the long term could also give us insights into the impact of the availability of popular vote processes in political systems. For instance, we could expect that, with time, fewer new arguments would appear in popular vote campaigns because elected representatives would have learned to engage in recursive exchange more continuously to anticipate the views of possible challengers ahead of popular vote campaigns. More generally, comparing different contexts to learn about the institutional conditions, formal regulations, and contextual framework that impacts the deliberative quality of such mass processes of communication between representatives and represented in terms of inclusion, respect, and limited attempts at manipulation would be very valuable in informing institutional design recommendations.

At a time when deliberative democracy and democracy in general face contestation that is increasingly considered to be due to empowered representatives’ failure to listen to and convincingly answer the actual concerns of citizens (Landemore 2017 10; Mansbridge 2019 326), I have proposed that the role that referendum and initiative processes could play in creating and sustaining bridges between empowered actors and citizens to support deliberative democracy at the mass level and over the long term might be larger than previously thought. Building on recent contributions, I have argued that popular vote processes can be particularly valuable in fostering forms of recursive representation. This is because introducing these processes in democratic systems changes the formal opportunity structure of representatives. This redistribution of empowerments can incentivize them to engage in conversation-like mass exchanges with the represented during popular vote campaigns and, possibly, in standard policy-making processes in the long term. This proposition calls for further study. But I hope that this new outlook on popular vote processes can demonstrate the relevance and potential of complementing important contributions studying how coupling popular vote processes with deliberative forums can enhance their deliberative quality with inquiries about whether, and how, popular vote processes can make political systems more deliberative in the summative sense.