Deliberative Mini-Publics

Drawing inspiration from normative theories of deliberative democracy, one category of democratic innovations is the deliberative mini-public (Curato Reference Curato2021; Smith and Setälä Reference Smith, Setälä, Bachtiger, Dryzek, Mansbridge and Warren2018). Examples include the British Columbia Citizens’ Assembly on Electoral Reform in Canada, the Oregon Citizens’ Review Initiative in the United States (Gastil Reference Gastil2020), the Irish Citizens’ Assembly (Farrell et al. Reference Farrell, Suiter and Harris2019; Muradova Reference Muradova2021), an assembly on constitutional revision in Iceland (Landemore Reference Landemore2020), and the French Citizens’ Convention for Climate (Galván Labrador and Zografos Reference Galván Labrador and Zografos2024).

Deliberative mini-publics share a set of core features. First, participants are selected through random sampling to produce groups that are representative of the broader citizenry, hence the term ‘mini-public’ (Curato Reference Curato2021; Smith Reference Smith2009). Second, participants discuss and make recommendations about policy issues in a context conducive to quality deliberation. In principle, legitimacy from mini-publics derives from these features: random sampling ensures representative groups and the deliberative process should generate more legitimate decisions or recommendations (Manin Reference Manin1987).Footnote ¹

Several recent studies investigate how the broader public evaluates the legitimacy of deliberative mini-publics. Már and Gastil (Reference Már and Gastil2023) find limited trust in mini-publics in the United States. By contrast, research in Germany (Goldberg and Bächtiger Reference Goldberg and Bächtiger2023) and Finland, Ireland, and the United States (Goldberg et al. Reference Goldberg, Lindell and Bächtiger2025) finds that citizens are generally favorable towards mini-publics, but mainly when they play an advisory role. Those dissatisfied with politics are more likely to support mini-publics with decision-making power (Goldberg and Bächtiger Reference Goldberg and Bächtiger2023), as are those who distrust politicians (Bedock and Pilet Reference Bedock and Pilet2021). Experience participating in mini-publics and greater trust in the decision-making abilities of other citizens also enhances support for mini-publics with decision-making authority (Goldberg et al. Reference Goldberg, Lindell and Bächtiger2025). Related to the findings on experience, learning about mini-publics also increases perceptions of legitimacy (Boulianne Reference Boulianne2018). Legitimacy is also enhanced when political leaders follow the mini-publics’ recommendations (Germann et al. Reference Germann, Marien and Muradova2024). While some research shows that citizens prefer random selection (Goldberg et al. Reference Goldberg, Lindell and Bächtiger2025), others find no evidence of a preference for selection by lot (Már and Gastil Reference Már and Gastil2023; Pow Reference Pow2023). Finally, people prefer mini-publics when they agree with their decisions (see, for example, Goldberg et al. Reference Goldberg, Lindell and Bächtiger2025; Werner and Marien Reference Werner and Marien2022).

While this recent research has generated important findings, open questions remain. First, much of this research studies mini-publics in isolation, leaving open the question of how the public evaluates mini-publics relative to other kinds of democratic innovations and to the status quo, the latter being important in terms of generating an appropriate baseline for comparison (Werner and Marien Reference Werner and Marien2022). Second, the knowledge base is focused on countries in North America and Europe, an additional gap that this paper seeks to fill.

Participatory Innovations

Innovations built upon a popular assembly model represent an additional category of democratic reform. The popular assembly or open town hall approach is distinct from the deliberative mini-public approach in that these processes are open to all citizens (Smith Reference Smith2009). In addition, these open citizen groups have decision-making power. In emphasizing reforms that open up decision-making processes to all citizens, these innovations are related to ideas put forth by advocates of participatory democracy (Barber Reference Barber1984; Pateman Reference Pateman1970 Reference Pateman2012).

An influential example is Participatory Budgeting in Porto Alegre, Brazil (see, for example, Baiocchi Reference Baiocchi2005). Participatory Budgeting involves a multistage and multilevel process beginning with localized, neighborhood-level popular assemblies, moving up to regional assemblies and then to a municipality-wide body (Smith Reference Smith2009). These popular assemblies have real decision-making power over the city budget. Participatory reforms have been adopted elsewhere in Latin America and beyond (Kapiszewski et al. Reference Kapiszewski, Levitsky and Yashar2021; Peruzzotti and Selee Reference Peruzzotti and Selee2009; Van Cott Reference Van Cott2009), though it must be noted that some ‘participatory’ programs are participatory in name only.

Most studies of participatory innovations study the processes themselves (Avritzer Reference Avritzer2009; Baiocchi Reference Baiocchi2005 Reference Baiocchi2011; Peruzzotti and Selee Reference Peruzzotti and Selee2009; Wampler Reference Wampler2007) or their impacts on participants (see, for example, Altschuler and Corrales Reference Altschuler and Corrales2012). While there is a broad literature focused generally on citizen demand for greater participation, there is only a limited amount of research that evaluates the extent to which citizens perceive the outcomes of participatory innovations as legitimate or prefer processes such as Participatory Budgeting (van der Does and Kantorowicz Reference van der Does and Kantorowicz2021) relative to other democratic alternatives and to the status quo.

Citizen-Elite Deliberation

A third category centers on opportunities for citizens and elites to deliberate and engage with one another about policy issues or political decisions. These forms of engagement are distinct from traditional town halls or elite-level deliberation in that they involve meaningful consultation between political leaders and citizen groups. While not always discussed in conversations about ‘democratic innovations’, advocates of these approaches are motivated by similar concerns regarding the state of democracy and dissatisfaction with it (Castradori Reference Castradori2021; Neblo et al. Reference Neblo, Esterling and Lazer2018). They also view them as an alternative to the status quo that more effectively engages citizens in politics and decision making. As such, they fit the general conceptualization of a democratic innovation.

Working in the United States, Neblo et al. (Reference Neblo, Esterling and Lazer2018) advocate for what they term ‘directly representative democracy’, or opportunities for citizens to engage in ongoing deliberation with elected representatives. Hibbing et al. (Reference Hibbing, Theiss-Morse, Hibbing and Fortunato2023) argue that the public desires ‘attentive democracy’ whereby ‘elites still make formal policy decisions but do so only after they have been in sustained, intimate contact with everyday people and only in the interests of those people’ (p. 4). This framework relates to (though is distinct from) the work of Paller (Reference Paller2019) on ‘dignified public expression’ in Ghana – accountability generated through regular, respectful interactions between representatives and constituents that lead citizens views to actively inform policy making or decision making. Paller (Reference Paller2019) emphasizes that, in Ghana, ‘democratic accountability requires that the leaders owe the people an account of what they have been doing. [Ghanaians] expect their leaders to do more than deliver the goods; they also expect them to listen to their ideas and hear their voices’ (p. 13).

The expansion of ongoing opportunities for citizen engagement with political leaders could enhance legitimacy for several interrelated reasons. With sufficiently high levels of participation and inclusion, these processes can ensure representation for a broad range of citizens (Neblo et al. Reference Neblo, Esterling and Lazer2018). Because they are ongoing and operate outside of election time, they can increase accountability and help to ensure that politicians take citizens’ perspectives into account (Neblo et al. Reference Neblo, Esterling and Lazer2018; Paller Reference Paller2019). By promoting personal connections between leaders and citizens, they have the potential to increase mutual respect, which should enhance legitimacy (Paller Reference Paller2019).

Research has examined citizen-elite deliberation in a number of contexts, with the focus mainly on participants. In Benin, citizen-elite deliberation reduced clientelism and improved informed, policy-centered voting (Fujiwara and Wantchekon Reference Fujiwara and Wantchekon2013; Wantchekon and Guardado Reference Wantchekon and Guardado2024). In the Philippines, it led to shifts in policy preferences (Wantchekon et al. Reference Wantchekon, Lopez-Moctezuma, Fujiwara, Lero and Rubenson2022). In the United States, it generated greater political interest and led to increases in political knowledge (Neblo et al. Reference Neblo, Esterling, Kennedy, Lazer and Sokhey2010; Neblo et al. Reference Neblo, Esterling and Lazer2018). As far as I am aware, there has not been research on how the broader public evaluates the democratic legitimacy of decisions involving citizen-elite engagements.

Conjoint Experiment Design

The goal is to study Ghanaians’ preferences for – and views about the legitimacy of – alternative options for reforming democratic decision making, relative to the status quo. I focus on preferences about a process leading to the allocation of local public goods – public goods that benefit local areas such as school facilities and clean water projects – by local district governments. I focus on local public goods provision for two reasons. First, in Ghana and similar contexts, local public goods provision is of critical importance to voters (Brierley et al. Reference Brierley, Kramon and Ofosu2020; Harding Reference Harding2015; Weghorst and Lindberg Reference Weghorst and Lindberg2013). Political competition tends to be structured around valence issues such as the provision of public goods for development, rather than broad ideological divisions (Bleck and Walle Reference Bleck and van de2013). Studying the allocation of local public goods focuses us on political decisions that are most salient and important to Ghanaians. Second, in local public goods allocations there are clear ‘winners’ and ‘losers’ – some areas get the goods, which are highly valued, while others do not. This allows us to assess how preferences about processes change when citizens are on the winning or losing end of the decision, which is key to analyses of legitimacy (Gibson et al. Reference Gibson, Lodge and Woodson2014; Gibson Reference Gibson, Bornstein and Tomkins2015; Werner and Marien Reference Werner and Marien2022).

The conjoint experiment asks participants to consider a hypothetical situation in which the local district government is trying to decide where in the district to implement a project that would improve local public goods. The prompt notes that resources are not available to implement the project in all communities in the district and so we would like to know their views about how communities should be selected for the program.Footnote ²

We presented subjects with two profiles, each with three attributes, summarized in Table 2. Each profile varied on the decision-making process used, whether the decision received an endorsement from either of the main political parties (NPP, NDC, or no endorsement), and whether or not the participant’s community received the project (Yes or No). The descriptions of the decision-making processes are presented in Table 3.

Table 2. Conjoint experiment design

Table 3. Democratic process descriptions

Each participant evaluated two sets of profile pairings (that is, completed two conjoint tasks). In each task, they viewed and evaluated two profiles. For each profile, the levels of the three attributes – process, endorsement, and outcome – were independently and randomly assigned from their full set of possible values.Footnote ³

Randomization across profiles and tasks helps mitigate potential carryover effects (for example, earlier choices influencing later ones) and profile order effects (for example, profiles shown first being more/less likely to be selected). In this experiment, the attribute order within profiles was fixed: process, followed by endorsement, then outcome. While attribute order randomization is sometimes used to reduce ordering bias, in this case I chose a fixed order to avoid awkward or confusing profile text.

Appendix Section B presents results from balance tests. Demographic and other pretreatment variables are not significant predictors of assignment to any of the democratic process conditions (Table B1), the endorsement conditions (Table B2), or the project outcome conditions (Table B3).

Appendix Section C presents diagnostic checks for the conjoint experiment (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2014). I present no evidence of order or carryover effects, that is, that responses to the second set of choices are impacted by the first set of choices (Figure C.1). I also present no evidence of profile order effects, which indicates that being the first or second option presented in a choice set did not impact the probability of being selected (Figure C.2).

Outcome Measures

I gathered three outcome measures. The first measures relative preferences by capturing the respondents’ profile choice as a binary variable (chosen or not). This permits identification of how different levels of each attribute impact the probability that a profile will be selected. Because respondents must make a selection, the choice measure captures relative preferences rather than overall levels of support or legitimacy. I therefore also gather two additional measures that capture legitimacy assessments of each profile separately. These measures capture levels of support for each profile and allow for comparisons. I measure legitimacy with assessments of process fairness (Werner and Marien Reference Werner and Marien2022) and an evaluation of how democratic the process is.

• Choice: 0/1 binary measure indicating whether the option was selected as the preferred process (‘Which of these processes would you prefer?’)

• Fairness: ‘Thinking of process A [B], how fair would you say it is?’ (1–7)

• Democratic: ‘Thinking of process A [B], how democratic would you say it is?’ (1–7)

Conjoint Experiment Results

Figure 2 presents the main results from the conjoint experiment with a focus on the binary choice measure. The points in the plot capture the change in the probability that a profile with a given level is selected relative to the baseline category (indicated with the point at 0), averaged across all other combinations of attributes. Hainmueller et al. (Reference Hainmueller, Hopkins and Yamamoto2014) refer to this as the Average Marginal Component Effect (AMCE). The bars represent 95 percent confidence intervals calculated using standard errors clustered on participant (in all analyses, I cluster standard errors by participant). Numerical regression results are presented in Appendix Table D.1.

Figure 2. Full conjoint experiment results: average marginal component effects on the binary choice outcome measure.

Ghanaians prefer all three alternative democratic options to the status quo, which was selected only 37 percent of the time. The differences are large (and all statistically significant): the citizen-elite process is 23 percentage points more likely to be selected, while the participatory and mini-public options are 15 and 20 percentage points more likely to be selected, respectively. The citizen-elite process receives the highest level of support, including relative to the participatory and mini-public options. The citizen-elite process is about 7 percentage points more likely to be selected than the participatory process (p = 0.005) and 3 percentage points more likely to be selected than the mini-public (p = 0.15).Footnote ⁴

Figure 2 also illustrates that outcome favorability has a large effect on process preference. Participants are about 26 percentage points more likely to prefer the process if their community receives the project, consistent with prior research (for example, Goldberg et al. Reference Goldberg, Lindell and Bächtiger2025; Werner and Marien Reference Werner and Marien2022). On average, there is no impact of the party endorsements. In Appendix Section D.2, I display the results for those who express a partisan attachment to the NPP and the NDC separately. Unsurprisingly, party supporters are less favorable to the process when the other party endorses the outcome and more favorable when their own party makes the endorsement (Figure D.1, Figure D.2, and Table D.1).

Figure 3 turns to legitimacy evaluations of each process (how fair and how democratic). Unlike the choice outcome measure, which captures relative preferences, these analyses allow us to make comparisons and examine overall levels of legitimacy for each process. In the figure, triangles represent perceptions of process fairness, while circles represent evaluations of how democratic the process is.

Figure 3. Legitimacy evaluations of democratic processes.

The average view of the status quo process is below the neutral midpoint (4) of the 1–7 scale, indicating that on average subjects do not believe the status quo is fair or democratic. Each of the three alternatives are rated as substantially more fair and more democratic than the status quo. The differences are large in magnitude, each representing almost a full standard deviation increase in the legitimacy assessments.

The citizen-elite process receives the highest legitimacy ratings, with an average above 5 on the fairness and democratic scales, although both other alternative processes do receive positive legitimacy assessments. For example, on the democratic scale, the citizen-elite process scores on average 0.30 points higher than the participatory process (p = 0.00) and 0.241 points higher than the mini-public (p = 0.002) (see regression results in Table D.2).

How the Outcome Impacts Legitimacy Assessments

I now turn to preferences and evaluations of each process, depending on whether the participant received their preferred outcome (getting the project). This is important as a key test of process legitimacy centers on how people react when they are on the losing side of decisions (Germann et al. Reference Germann, Marien and Muradova2024; Gibson Reference Gibson, Bornstein and Tomkins2015; Werner and Marien Reference Werner and Marien2022).

I present choices and evaluations separated by process and outcome.Footnote ⁵ Figure 4 focuses on the binary choice variable. Regardless of process, subjects prefer processes in which they achieve their preferred outcome. This is consistent with prior research (for example, Goldberg et al. Reference Goldberg, Lindell and Bächtiger2025). The status quo process where the participant is on the losing end is by far the least favorable. People are more than twice as likely to select the status quo option if they are on the winning end of the decision.

Figure 4. Process preferences by preferred outcome.

Importantly, among the losers of the decision (triangles), the alternative processes substantially improve the likelihood that the process will be selected. In each case, the alternative processes more than double the likelihood that those who do not prefer the outcome will select the decision-making procedure. This is most pronounced for the citizen-elite process. In fact, the citizen-elite process without getting the project is equally likely to be selected as the status quo process while getting project. That is, losers of the citizen-elite process are equally likely to select the process as are winners from the status quo.

Figure 5 presents the same analysis but focused on the fairness and democratic measures. Across the board people believe that processes are more fair and more democratic when they like the outcome of the process. Qualitatively, this has the biggest impact with the status quo process. Those on the losing end of the decision are very negative about the status quo process, believing it to not be democratic or fair, while those on the winning end are on average somewhat positive.

Figure 5. Legitimacy evaluations by process outcome.

Crucially, losers from the three alternative processes still believe that the processes were on average fair and democratic (average values above the neutral midpoint of 4). While they prefer the processes when they get the project, they are still on average favorable toward the processes even when they lose. To the extent that democratic legitimacy – and ultimately compliance with decisions and policies – requires that those on the losing end still see the process as fair and democratic, this is an important result. These dynamics are most pronounced for the citizen-elite deliberation process. Losers from the citizen-elite decision-making process believe that the procedure is more fair and more democratic than do winners from status quo decision-making.

I also conducted exploratory subgroup analyses to assess whether the patterns presented above differed by different groups in the population – for example, those with less education (Dijk et al. Reference Dijk L, Turkenburg and Pow2023). There are generally no substantive differences except that more educated respondents were more favorable toward participatory processes relative to the status quo (see Appendix Section D.3).

Reforms to Address Partisan Tensions

Experiment 2 turns to a related but distinct question: Which reforms do Ghanaians believe would most effectively address the challenge of inter-partisan tension in the country? Because existing research highlights how affective polarization and partisan animus present challenges for democracy (McCoy et al. Reference McCoy, Rahman and Somer2018), democratic reforms to address democratic crises (Dryzek et al. Reference Dryzek, Bächtiger, Chambers, Cohen, Druckman, Felicetti, Fishkin, Farrell, Fung and Gutmann2019) should be equipped to handle this challenge.

Experiment 2 is a conjoint experiment with only one attribute, the process, and four levels. As such, it is essentially a vignette experiment, an approach that has been used in other recent studies in this area (Werner and Muradova Reference Werner, Muradova, Ercan, Asenbaum, Curato and Mendonça2022). The questions ask participants to consider options for democratic reforms to reduce partisan tensions in Ghana.Footnote ⁶

There are four proposals, presented in Table 4. Each respondent is randomly presented with two options from which to select. The first, competitive elections, can be considered a proposal to improve the status quo operation of democracy. The other three are distinct alternatives. Each participant considered one set of choices. Appendix Table B.4 presents balance tests for this second experiment. Appendix Figure C.3 provides evidence that the profile order had no impact on the probability of being selected.

Table 4. Democratic reform descriptions

I gathered three outcome measures.

• Most Effective: Which one of these two options do you think would be most effective in reducing tension and hostility between NPP and NDC supporters? [binary]

• Most Democratic: Which of these two options do you believe is the most democratic way to make policy decisions? [binary]

• How Effective: Thinking of option A [B], how effective do you think it would be in reducing tension and hostility? [scale 1–7]

Experiment 2 Results

Figure 6 presents the main results from experiment 2, focused on the two binary measures (effective and democratic). The patterns are very similar for the 1–7 effectiveness scale and so I present those results in Appendix Figure E.1.

Figure 6. Preferences on reforms to address partisan tensions.

The main finding is that the citizen-elite process is perceived as the most effective and democratic mechanism for addressing the challenge of partisan tension. Respondents were about 11 percentage points more likely to select the citizen-elite process than the proposal to increase the competitiveness of elections (p = 0.004) and the proposal to introduce direct democracy (p = 0.008). They were 16 percentage points more likely to select the citizen-elite process compared to the proposal on mini-publics (p = 0.000).

The citizen-elite process is also the only proposal that is considered more effective and democratic than the proposal to make elections more competitive. Direct democracy is viewed very similarly to the elections proposal. By contrast, the deliberative mini-public proposal is actually less favored relative to the elections reform. This may be because, in the abstract, people do not necessarily conceive of random selection for mini-publics as being democratic (Bedock and Pilet Reference Bedock and Pilet2021) or because they envision partisan tensions emerging in the deliberations.Footnote ⁷