Abstract

Practical experiments with deliberative democracy, instituted with random samples of the public, have had success in many countries. But this approach has never before been tried in Sub-Saharan Africa. Reflecting on the first two applications in Uganda, we apply the same criteria for success commonly used for such projects in the most advanced countries. Can this approach work successfully with samples of a public low in literacy and education? Can it work on some of the critical policy choices faced by the public in rural Uganda? This essay reflects on quantitative and qualitative results from Uganda's first Deliberative Polls. We find that the projects were representative in both attitudes and demographics. They produced substantial opinion change supported by identifiable reasons. They avoided distortions from inequality and polarization. They produced actionable results that can be expected to influence policy on difficult choices.

The last two decades have seen a great rise in interest in deliberative democracy, in both theory and practice.1 In political theory, this “deliberative turn” has largely supplanted the previous enthusiasm for “participatory democracy,” a change sometimes decried by advocates of the latter. Participatory democracy generally relies on self-selected mass participation. In development contexts, an iconic form is the “participatory budgeting” practiced in Porto Alegre, Brazil.2 By contrast, the form of deliberative democracy that we will discuss here emphasizes designs that promote both the representativeness and the thoughtfulness of public participation. Instead of mobilizing as many people as possible, the idea is to foster thoughtful weighing of the arguments for and against policy alternatives by representative microcosms of the public.3 The numbers who participate may be smaller than in mass participatory institutions, but the conclusions offered can represent the public's considered judgments.

In practice, this kind of deliberative democracy has found applications in various parts of the world with designs that foster public input for actual policy-making. The designs vary, but they generally attempt to facilitate the discussion of competing reasons for policy alternatives in a context in which members of the public can become more informed about the issues in question. The more rigorous versions carefully select the participants by recruiting a microcosm or “mini-public” of the relevant population through random sampling.4 The basic idea is that if the sample is representative and the participants deliberate under good conditions for considering the issues, then the results should represent what the public would think were it to engage with the issues under similarly good conditions. This strategy makes deliberative democracy a practical and implementable theory, at least for the policy issues selected.

How widely can this approach be applied? There have been successful cases in postconflict situations and instances of ethnic division.5 There have also been successful cases across multiple linguistic barriers, as when a Europe-wide sample deliberated in Brussels in twenty-two languages with simultaneous interpretation.6 There have even been successful cases in authoritarian systems lacking electoral competition.7 But never before have there been applications of deliberative democracy with random samples of the public in Africa, where populations with low literacy and low levels of education often face extraordinary policy challenges. Is it applicable in such contexts? Or is deliberative democracy just an approach for advanced countries with highly educated populations? We reflect here on a pilot effort to apply deliberative democracy, through randomly selected microcosms, to produce public input for policy-making in Africa.

Billions of people around the world live in poverty and deprivation. Development efforts to assist them increasingly invoke the idea that the people should be consulted. Those who might be affected by policies should in some way have a voice about them. Some argue that policies the public can accept will be more effective. Others argue that long-term development will be more sustainable in open and inclusive societies in which people participate.8

But how is this to be accomplished? There are various approaches. Some efforts engage stakeholders or policy experts who speak on behalf of the people. Some take decisions to the people themselves in self-selected forums or meetings. Sometimes researchers employ focus groups and key informant interviews to get voices from the people and from those who might have relevant local knowledge.

However, stakeholders or policy experts may turn out to have different views from those of the people themselves. Self-selected forums are inevitably unrepresentative and usually dominated by those especially motivated to turn out. Further, self-selected forums to discuss the distribution of benefits are likely to foster mobilization for the benefits, rather than deliberation about the general good of the community. For example, in the famous “participatory budgeting” in Porto Alegre, a practice now spread around the world, self-selected groups mobilize for specific benefits but the broader population is not well represented.9 The question we explore here is whether the move from participatory to deliberative democracy, a move made prominent in democratic theory,10 can be retraced in the practice of public consultation in developing countries. More specifically, is it practical to consult populations in developing countries through deliberating microcosms selected through random sampling?

There have been a few efforts to incorporate elements of deliberative democracy in public consultations in Africa. In Benin, for example, politics scholar Leonard Wantchekon has reported on a field experiment looking at how the discussion of different kinds of platforms (clientelist or public-policy related) just before an election affected voting.11 In Príncipe and São Tomé, political scientist Macartan Humphreys and colleagues have found in a deliberative democracy experiment “robust evidence that the influence of leaders on the outcome of deliberation is extremely strong, with leadership effects accounting for a large share of the variation in views elicited across the country.”12 Their finding that discussion leaders effectively determine the outcome would obviously undermine the aspiration for using genuine deliberation by the public to influence public policy.

However, neither of these studies involved random samples of the public. Rather, they involved random assignment of self-selected participants to different treatments. And the project that seems to have been a self-conscious application of deliberative democratic theory (the São Tomé and Príncipe experiments) apparently gave moderators a great deal of flexibility and discretion to present their own opinions and advocate for them. Moderators appear to have used that discretion freely, resulting in the apparent distortion of the outcomes to conform to their views. As in other applications of deliberative democracy, the precise institutional design can be consequential. It remains to be seen what would happen with a deliberative design closer to the microcosms that have been applied thus far in developed countries.13 If deliberators were recruited through random sampling rather than self-selection and if the moderators were strictly constrained to avoid advocacy, rather than having free rein to promote their own views, might the efforts be more successful? Until now, the basic idea of deliberating microcosms chosen by random sampling had not been tested in Africa. We report on such an effort here.

While the idea of deliberative democracy has acquired many enthusiasts over the last two decades, it has also attracted criticisms. Some of those criticisms might plausibly define barriers to applying the approach in developing countries. Consider three. First, deliberative democracy is often criticized as an elite form of democratic practice. Even the term was coined in a discussion of James Madison's theory of representatives who would “refine and enlarge the public views by passing them through the medium of a chosen body of citizens.”14 The refinement derived both from the process of weighing arguments on the merits in the legislative assembly and from the selection process for selecting the “best” representatives. The term in its first coinage thus applied to deliberations among the highly educated and supposedly especially virtuous, who would choose in the interests of the public good for the rest of us. In the last two decades, the term has been adapted, at least for practical applications, to deliberations by the people themselves and especially, for our purposes, by random and representative samples. But the question remains whether ordinary citizens who vary widely in education and expertise can usefully weigh the competing arguments at issue in actual policy choices. Some critics even question whether ordinary citizens in developed countries such as the United States have the capacity to weigh competing policy arguments.15 From such a perspective, it would seem even less plausible that participants who lack education or even basic literacy could do so.

Second, arguments “against deliberation” have found a footing in normative theory, building on the jury literature.16 In particular, there is the worry that advantaged groups will dominate the deliberations and impose their views on everyone else. In some juries, the men dominate, or the more educated, or those who have higher social status. Juries are the most studied deliberative institution, and while one can make a case that juries do fairly well at not reflecting the power relationships found in society,17 any pattern of domination by the more advantaged raises questions about whether people are really deliberating on the merits, rather than deferring to those who may be seen to have more competence or authority.18 To the extent that the deliberations are not genuine decisions on the merits, the institution loses its claim to determine legitimate outcomes. Of course, juries are generally deciders of fact, not policy, but the worries about juries have inspired concerns about the broader use of citizen deliberation for policy questions. If deliberating microcosms or mini-publics are distorted by deference to the advantaged, or by the ability of the advantaged to impose their will on the other participants, then the ideal of deliberation, nicely captured in German philosopher Jürgen Habermas's famous phrase as the “unforced force of the better argument,” would be undermined.19 In the context of developing countries, this worry might well be exacerbated. Those few participants who are well educated, or who have high status for other reasons, might have a great advantage over the rest of the participants who lack basic education and preparation for the discussions. The less well educated might defer to the advantaged, thus distorting the process. Hence this critique, often applied generally to applications of deliberative democracy, would seem to pose a special challenge in developing countries.

A third critique has centered on what has come to be known as polarization, or the tendency of groups engaging in discussion to move toward extremes. Building on earlier work on the “risky shift,” Cass Sunstein and various colleagues hypothesized a “law of group polarization.”20 On an issue for which there is a midpoint, if most participants in the discussion are to the right of the midpoint, then there will be movement away from the midpoint to the right. But if most participants are to the left of the midpoint, then there will be movement to the left. This polarization occurs, the argument goes, because of two factors: an “imbalance in the argument pool” and “a social comparison effect.” If the group is mostly on one side, then more of the arguments voiced are likely to be on that side. Hence the tendency to move to a “more extreme” position away from the midpoint. Second, as people pick up on the conclusions of others, they will feel social pressure to conform to the dominant position.21 More recently, Sunstein has added a third argument. Those who feel “tentative” in their views may choose initial moderation out of uncertainty, but these “tentatives” are more easily swayed by the other two factors to conform to the apparently dominant arguments.22

The vulnerability of deliberative discussion to polarization is likely a matter of institutional design. While found in jury-like experiments conducted by Sunstein and his collaborators, this pattern has not applied universally to deliberating microcosms chosen by random sampling. For example, we have not found it in Deliberative Polls.23 If designs have elements of balance and confidentiality, those elements may well defeat the imbalance in the argument pool and the social comparison effect. The design of Deliberative Polls includes elements of balance, such as balanced briefing materials, balanced plenary session panels, and moderators who are trained to draw attention to the competing sides of the argument in the briefing materials. It also ensures confidentiality for the final considered judgments by collecting them in confidential questionnaires. The facilitators are trained to bring out minority opinion and to set a tone for respecting the opinion-givers equally. These elements appear to protect at least this kind of deliberative microcosm from the polarization pattern.24

But what would happen with far less educated and literate respondents? None of the cases thus far have involved samples mostly composed of less-literate respondents with low levels of education. It is an open question whether the avoidance of the “law of group polarization” might apply in such contexts. It is easy to see why this might be a problem. The balance achieved through written briefing documents and the ability to weigh arguments from competing experts on either side of the issues in the plenary sessions might be undermined for nonliterate and low education respondents. Further, the less literate might more easily defer to group pressures, allowing the social comparison effect to determine the outcome and moving the mean of the group to more extreme positions. Perhaps less literate respondents will tend to be more “tentative” and defer to those with considered judgments or with higher social status. Such speculations imply that in the development context, the “law of group polarization” may well prove to be more of a challenge.

What is at stake here? Why is polarization a challenge for applications of deliberative democracy? If there were a predictable pattern of group psychology moving arguments to more extreme positions, then it would be hard to argue that the results were really the product of participants weighing the issues carefully on the merits. But if roughly half the time the groups move further from the midpoint and roughly half the time they move toward the midpoint, then the potential dynamic toward group polarization would have been stopped. As noted, Deliberative Polls have produced this nonpolarizing result with samples of more-educated and -literate populations, indicating that the postulated “law of group polarization” is not an inexorable law, even though it appears in jury-like designs.25 The dynamic among largely nonliterate and uneducated samples has only just been tested in research.

The Deliberative Polling projects in Uganda took place in the Mount Elgon region in two districts, Bududa and Butaleja, both troubled by frequent environmental disasters (floods and rock slides). Each district has about two hundred thousand inhabitants, mostly working in subsistence farming. The areas are characterized by low levels of education and high population density. The average population density in Uganda is 195 persons per square kilometer, but in the Mount Elgon region, the average population density is 950 persons per square kilometer. This population density puts pressure on the subsistence farming and prevents investment in education. The low education levels are especially pronounced for women and girls, many of whom get no formal education at all. The common environmental disasters lead to periodic evacuations and issues of resettlement.

The Deliberative Polling efforts were led by a team from Makerere University, more specifically, the East Africa Lab in the Resilient Africa Network sponsored by USAID and housed in the Makerere School of Public Health. The Stanford Center for Deliberative Democracy, also with USAID support, provided assistance at each stage of the two projects.

The Lab ran focus groups and key informant interviews in the two communities to identify challenges faced by the two districts. The project also convened an advisory group, including academics, key government officials (both local and national), and NGOs to provide further input. The advisory group is listed in the report, available online.26

The advisory group developed an agenda with specific policy options in three broad areas. Because of the region's recurrent environmental disasters, the agenda focused on three related topics: resettlement management, land management, and population pressure. The committee identified policy options under each of these headings that might feasibly be implemented in Bududa and Butaleja. These options were the principal topics in the pre- and post-deliberation questionnaires.

Deliberative Polling assesses the representative opinions of a population, both before and after it has had a chance to think about an issue and discuss it in depth. The idea is to gather a representative sample and engage it in transparently favorable conditions for considering the pros and cons of competing policy options. Most citizens, most of the time, in most countries around the world, do not spend much effort considering public policy questions in depth. The premise is that when policy options are important for a community, then public consultations about them should be representative of the population and thoughtfully based on the best information available. Hence the need for recruiting a random sample and engaging it in good conditions for considering the issues and the arguments for and against various policy options.

The method offers certain advantages over other methods of public consultation. Self-selected town meetings are unlikely to be representative because they involve only those who feel strongly enough to attend. Focus groups cannot be used to represent opinion because they are too small to be statistically meaningful. Rather, they are useful for uncovering the way the public frames an issue as a step in facilitating more systematic research. Conventional polls, while potentially representative when done well, offer the public's impression of sound bites and headlines. They do not reflect what the public would think if it actually thought in depth about the issues. Deliberative Polling is a method that offers representative and informed opinion. It offers a road map to the policies the public would accept upon reflection, and for what reasons. It can also indicate those policies the public would have reservations about, and for what reasons.

How should we evaluate these first Deliberative Polls in Africa? First, is the sample representative? We can compare the participants (those who take the initial survey and attend the deliberations) with the nonparticipants (those who take the survey and do not attend the deliberations). The comparisons should include both demographics and attitudes. The idea is to recruit a microcosm of the viewpoints and interests of the community. Voter lists, census data, and random digit dialing have all been used in other countries to provide the sampling frame. What approach might work in rural Uganda, where the data for such approaches are flawed and where the cell phone coverage is unreliable?

Second, do the opinions change? Ultimately we are interested in the final considered judgments of the sample, regardless of whether they stay the same or move away from where they began. But if Deliberative Polls rarely yielded significant net change, then few consultations would go to the trouble of creating these balanced and informed discussions. It would be easier just to do conventional polling. Hence, statistically significant net change indicates that something is happening when citizens deliberate.

Third, are there identifiable reasons for the final judgments? Does the process produce considered judgments that people reach on the basis of having considered competing arguments?

Fourth, does the process avoid the distortions we have already identified as potentially undermining the deliberative process? Two distortions have been especially prominent in the literature: group polarization and domination by the more advantaged. These potential distortions pose a challenge to deliberation in that they would appear to offer explanations for the results independent from the merits of the arguments. Rather, they would render the results an artifact of group psychology or of the domination of the more advantaged or educated. As mentioned above, if these distortions pose a challenge in the most developed countries with highly educated populations, they are even more likely to occur in developing countries whose populations have low education levels.

The participants were recruited through a random selection of households and a random selection within the households. In Bududa, there were 210 initial interviews, with only eleven refusals. Of those 210 initial interviews, 201 completed the full two days of deliberation. Counting the eleven refusals in the total, the response rate for the actual event was about 91 percent, an extraordinarily high level for surveys by international standards and especially among processes requiring two days of discussion. In Butaleja, there were 232 initial interviews, again with only eleven refusals. Of those 232 who took the initial interview, 217 completed the full two days of deliberation. Counting the eleven refusals in the denominator, the response rate in Butaleja is 89 percent, also an extraordinary level of participation. Why this high level? The projects had strong buy-in from community leaders and local authorities. The topic was one of great interest to the communities. And an honorarium and transport costs helped make participation attractive.27

Ten percent of the Bududa participants had no education, and 58 percent had only primary education. For Butaleja, 8 percent had no education and 57 percent had only primary education. Eighty-seven percent of the Bududa participants and 86 percent of the Butaleja participants were farmers. As best we can judge, it was an excellent sample with one serious distortion: an overrepresentation of men.28 Yet, as we will see below, issues of gender and the interests of women were reasonably well represented in the discussions, despite the underrepresentation of women among the participants.

The pre- and post-deliberation questionnaires were administered in individual interviews taking approximately thirty-five to forty minutes each. The use of oral interviews combined with video briefings allowed the nonliterate to respond and participate in the process. We will return to the questions of whether the participation was relatively equal and whether the groups were able to avoid the predictable distortions in group discussion.

In both communities there were thirty-six policy options posed for deliberation covering three topics: resettlement, land management, and population pressure. Upon first contact at home and at the end of the weekend, the participants were asked to rate the thirty-six options in importance on a scale from zero (extremely unimportant) to ten (extremely important), with five in the middle. Here we will report the percentage saying simply that an option was “important” (a rating above five); the means of the respondents' ratings appear in Tables 1 and 2.

Table 1

Bududa: Significant Policy Changes for Participants

Question/IssueT1T2T2-T1P-value
1. Rezone high-risk areas for no settlement. 0.766 0.840 0.075 0.001*** 
4. Give support to the host families for helping those who move. 0.685 0.759 0.073 0.003*** 
5. Strengthen the local disaster management committees. 0.760 0.827 0.068 0.002*** 
6. Raise funds to support the work of the local disaster management committees. 0.646 0.766 0.119 0.000*** 
8. Build peri-urban centers where people can resettle. 0.752 0.810 0.058 0.013*** 
9. Make sure new peri-urban centers are nearby so people can farm. 0.812 0.843 0.031 0.108* 
13. Ensure that the early warning system works with the local disaster committees. 0.715 0.756 0.041 0.076* 
16. Communities should manage the wetlands during the dry season. 0.602 0.671 0.069 0.015** 
17. Communities should create more rice schemes, but not in the wetlands. 0.529 0.631 0.102 0.003*** 
20. Communities should be responsible for desilting riverbeds. 0.585 0.674 0.089 0.002*** 
21. Government should assist communities in desilting riverbeds. 0.528 0.580 0.052 0.076* 
22. Communities should build sanitation drains for the reduction of malaria. 0.836 0.872 0.036 0.074* 
27. The government should raise narrow bridges. 0.811 0.858 0.047 0.015** 
29. Communities should build ladders in the highlands where there are not roads. 0.420 0.483 0.063 0.053** 
34. Families should consider their resources in planning the size of their families. 0.740 0.797 0.058 0.011*** 
Question/IssueT1T2T2-T1P-value
1. Rezone high-risk areas for no settlement. 0.766 0.840 0.075 0.001*** 
4. Give support to the host families for helping those who move. 0.685 0.759 0.073 0.003*** 
5. Strengthen the local disaster management committees. 0.760 0.827 0.068 0.002*** 
6. Raise funds to support the work of the local disaster management committees. 0.646 0.766 0.119 0.000*** 
8. Build peri-urban centers where people can resettle. 0.752 0.810 0.058 0.013*** 
9. Make sure new peri-urban centers are nearby so people can farm. 0.812 0.843 0.031 0.108* 
13. Ensure that the early warning system works with the local disaster committees. 0.715 0.756 0.041 0.076* 
16. Communities should manage the wetlands during the dry season. 0.602 0.671 0.069 0.015** 
17. Communities should create more rice schemes, but not in the wetlands. 0.529 0.631 0.102 0.003*** 
20. Communities should be responsible for desilting riverbeds. 0.585 0.674 0.089 0.002*** 
21. Government should assist communities in desilting riverbeds. 0.528 0.580 0.052 0.076* 
22. Communities should build sanitation drains for the reduction of malaria. 0.836 0.872 0.036 0.074* 
27. The government should raise narrow bridges. 0.811 0.858 0.047 0.015** 
29. Communities should build ladders in the highlands where there are not roads. 0.420 0.483 0.063 0.053** 
34. Families should consider their resources in planning the size of their families. 0.740 0.797 0.058 0.011*** 

Note: T1 denotes before deliberation; T2 denotes after deliberation; T2-T1 denotes after deliberation minus before deliberation; P-value denotes statistical significance. Proposals were rated on a scale of 0 to 10, with 0 being extremely unimportant, 10 being extremely important, and 5 being the midpoint. Data are the means of respondents' ratings.

In the significance column, * indicates a P-value of 0.10 or below, ** 0.05 or below, and *** 0.01 or below.

Table 2

Butaleja: Significant Policy Changes for Participants

Question/IssueT1T2T2-T1P-value
1. Rezone high-risk areas for no settlement. 0.553 0.670 0.116 0.000*** 
3. Resettle with host families in a low-risk area when there is a disaster. 0.563 0.626 0.063 0.017** 
11. Early warning system should use sirens. 0.761 0.821 0.061 0.008*** 
12. Early warning system should use text messages. 0.628 0.525 −0.103 0.000*** 
14. Plant trees to protect the river banks. 0.833 0.869 0.036 0.049** 
16. Communities should manage the wetlands during the dry season. 0.687 0.736 0.048 0.041** 
18. Communities should maintain water channels during the wet season. 0.749 0.657 −0.092 0.000*** 
20. Communities should be responsible for desilting riverbeds. 0.593 0.513 −0.080 0.006*** 
21. Government should assist communities in desilting the riverbeds. 0.846 0.874 0.028 0.091* 
33. The government should enforce the minimum age requirement for marriage of eighteen years. 0.840 0.881 0.041 0.032** 
37. Which option do you prefer? Spend money on more roads and fewer bridges, or spend money on more bridges and fewer roads (on a scale of 1 to 7, with 1 showing preference for more roads and fewer bridges and 7 more bridges and fewer roads). 0.612 0.462 −0.150 0.000*** 
Question/IssueT1T2T2-T1P-value
1. Rezone high-risk areas for no settlement. 0.553 0.670 0.116 0.000*** 
3. Resettle with host families in a low-risk area when there is a disaster. 0.563 0.626 0.063 0.017** 
11. Early warning system should use sirens. 0.761 0.821 0.061 0.008*** 
12. Early warning system should use text messages. 0.628 0.525 −0.103 0.000*** 
14. Plant trees to protect the river banks. 0.833 0.869 0.036 0.049** 
16. Communities should manage the wetlands during the dry season. 0.687 0.736 0.048 0.041** 
18. Communities should maintain water channels during the wet season. 0.749 0.657 −0.092 0.000*** 
20. Communities should be responsible for desilting riverbeds. 0.593 0.513 −0.080 0.006*** 
21. Government should assist communities in desilting the riverbeds. 0.846 0.874 0.028 0.091* 
33. The government should enforce the minimum age requirement for marriage of eighteen years. 0.840 0.881 0.041 0.032** 
37. Which option do you prefer? Spend money on more roads and fewer bridges, or spend money on more bridges and fewer roads (on a scale of 1 to 7, with 1 showing preference for more roads and fewer bridges and 7 more bridges and fewer roads). 0.612 0.462 −0.150 0.000*** 

Note: T1 denotes before deliberation; T2 denotes after deliberation; T2-T1 denotes after deliberation minus before deliberation; P-value denotes statistical significance. Proposals were rated on a scale of 0 to 10, with 0 being extremely unimportant, 10 being extremely important, and 5 being the midpoint. Data are the means of respondents' ratings. Question 37 posed a trade-off on a 1 to 7 scale.

In the significance column, * indicates a P-value of 0.10 or below, ** 0.05 or below, and *** 0.01 or below.

In Bududa, the rating of eleven of the thirty-six policy options changed significantly after deliberation; four other options had changes that were marginally significant (see Table 1). The changes were mostly in the direction of increased support for what became the most favored options.29 Some options started high and went significantly higher. Before deliberation, 76 percent of respondents viewed the rezoning of high-risk areas for no settlement as important; post-deliberation, 85 percent viewed it as important. Before deliberation, 67 percent of respondents viewed supporting host families to help those who move as important; post-deliberation, 78 percent viewed it as important. Some of the changes were large: the perceived importance of raising funds to support the work of the local disaster management committees jumped from 58 percent to 79 percent. There was also significantly increased importance given to proposals involving community action: to create more rice schemes (but not in the wetlands), to manage irrigation for cultivation (from 48 percent to 57 percent), for taking responsibility to desilt the riverbeds (from 52 percent to 64 percent), and for building sanitation drains to reduce malaria (from 87 percent to 94 percent). On the subject of family planning, there was a significant increase in support for the notion that families should consider their resources in planning the size of their families. The endorsement of this proposal increased from 76 percent to 87 percent.

The online appendix lists the top priorities after deliberation for Bududa and for Butaleja. In Bududa, after deliberation, the top priority of all thirty-six proposals was that the community should encourage girls as well as boys to go to school. This proposal, which began with very high support (96 percent) ended with virtually unanimous support (99 percent). The online appendix includes transcript excerpts exhibiting the reasoning in support of the top priorities.30

The Butaleja deliberation also produced significant changes on eleven policy attitudes. These changes are depicted in Table 2. Some of these changes show interesting reversals with deliberation. All are significant at the 0.05 level or better.

Rezoning high-risk areas for no settlement began with only 46 percent of respondents endorsing it as important before deliberation; but after deliberation, the level rose twenty points to 67 percent. Support for an early warning system using text messaging went down from 60 percent to 42 percent, while support for an early warning system using sirens went up from 79 percent to 92 percent. We think that the unreliability of electric power for charging and the unreliability of the cell connections moved people to support sirens as a more dependable system than text messaging. While there was an increase in support for communities to manage the wetlands during the dry season (from 70 percent to 82 percent), there was a drop in support for the idea that communities should maintain the water channels during the wet season (from 78 percent to 67 percent) and that communities should be responsible for desilting riverbeds (from 55 percent to 42 percent). Discussions revealed a growing awareness of the machinery and scale of work required to get these tasks done.

In the family planning area, there was an increase in support for the government enforcing the minimum age for marriage of eighteen years from the already-high level of 87 percent to 94 percent.

The online appendix shows the priorities after deliberation for Butaleja. Government assistance in drilling for clean water tops the list with 98.6 percent of participants endorsing its importance. The second highest priority post-deliberation was that the community should encourage girls to go to school as well as boys. As in Bududa, this moved from 97.4 percent before deliberation to about the same level as the top priority (98.6 percent) after.

Our third question, whether the final considered judgments seem to reflect reasoned deliberation, gets ample support from the transcript excerpts detailed in the online appendix. Consider the top priorities.

In Bududa, the top priority after deliberation was that the community should encourage girls to go to school as well as boys. Education can reduce the outcome of girls getting pregnant and married at too young an age. With schooling, they may be able to go to technical schools and find jobs. The second top priority was creating more Health Center 2s (local clinics) in small villages. Currently, the distance to health centers is too great for many community members to receive treatment for emergency illnesses. The local clinics could provide a first response even if they do not offer all the equipment and services found in the larger hospitals. Moreover, in other discussions, community health centers were cited as offering support for family planning, meaning that some women were reluctant to pursue family planning assistance because medical help was too far away. The third top priority in Bududa was creating one-classroom schools for elementary education in remote areas. The distances to school now are too far for children to walk, so many receive no elementary education at all. Concern that girls may be sexually assaulted if they have to walk long distances to school also motivates parents to keep their children at home.

In Butaleja, the top priority was government assistance in drilling for clean water. Participants were very concerned about the disease risks of dirty water, but they needed help with drilling. The second top priority was encouraging girls as well as boys to go to school, for reasons very much like those in Bududa. The third priority was the government building roads in remote areas so residents could bring their produce to market. The local communities did not have the resources to build the roads themselves.

Another way to explore the reasons supporting the final ratings of the policy options is to run regressions to isolate the levers of opinion change. The questionnaire, covering thirty-six policy options plus demographics and other questions, did not have much room for explanatory variables. However, it did probe some basic values, allowing us to make connections in the regressions between those values and the policy options. The online appendix illustrates some of these connections. The floods and rock slides, for example, periodically threaten the basic well-being of the communities. The disasters threaten the order and security required for people to prosper economically and maintain their access to basic necessities. Hence, support for the early warning system in Butaleja is associated in the regression analysis with basic values such as making sure everyone has clean air and water and promoting economic growth. These two values plus the importance of education are associated in Bududa with support for new infrastructure, such as building roads in remote areas, building bridges, and raising narrow bridges. The link between valuing education and supporting travel infrastructure may reflect that the difficulties of travel pose a major impediment to education. The values of economic growth and clean air and water are also associated in the regression analysis with the policy of offering more education for family planning, probably because large family size in these communities impedes economic growth and better access to the necessities of life.

Both the transcripts and the regressions provide evidence that the final considered judgments were supported by the reasoning of participants grappling with trade-offs and priorities.

Turning to our fourth major issue, did the process avoid the distortions that have plagued some other group discussions? The two we focused on are polarization and domination by the more advantaged.

Polarization, as we discussed earlier, is the idea that on a given issue, if a group starts out to the left of the midpoint, it will move further to the left. If it starts out to the right of the midpoint, it will move further to the right. If this were a consistent pattern for the issues, then it would undermine the claim that participants were deliberating on the merits. For example, in a study of polarization in group discussions in two locations in Colorado, researchers found 80 percent of the group issue combinations (the movements of small groups on a given issue) polarizing in this way in Boulder (moving left politically) and 93 percent in Colorado Springs (moving right).31

In the Uganda projects, there were fourteen groups in Bududa and fifteen groups in Butaleja. Both projects used the same questionnaire with thirty-six policy options. Hence, there were 504 group issue combinations (potential group movements on the specified issues) in Bududa and 540 group issue combinations in Butaleja. In contrast to the results predicted by polarization theory, only 54 percent of the group issue combinations polarized in Bududa and 51 percent in Butaleja.32 That is, about half the time the groups moved away from the midpoint in the hypothesized direction and about half the time they moved toward it in the opposite direction. We see no evidence of a “law of group polarization” distorting these deliberations.

The second distortion is domination by the more advantaged: do they impose their views on the others? In rural Uganda, where many of the respondents lack education, this risk was of special concern. Using the group issue combinations, we began by looking at the starting points of the more privileged. If they were dominating the discussions and imposing their views on everyone else, then the group issue combinations should move in the direction of the views held by the privileged or advantaged. We looked at three possible patterns of domination: males imposing their views on females, older participants imposing their views on younger ones, and the more educated imposing their views on the less educated. No such patterns arose. In Bududa, the group issue combinations moved in the direction of the males only 21 percent of the time, in the direction of the older respondents only 47 percent of the time, and in the direction of the more educated only 24 percent of the time. In Butaleja, the group issue combinations moved in the direction of the males only 25 percent of the time, in the direction of the older respondents only 53 percent of the time, and in the direction of the more educated only 42 percent of the time.33 In these contexts, then, we see no evidence that the advantaged are imposing their views on others. Instead, all sectors seem to be learning from each other, sharing arguments, and coming to conclusions about what should be done.

The results of our studies in Bududa and Butaleja, Uganda, reveal two cases of representative and thoughtful deliberation expressing the considered views of the communities. Rather than self-selected group meetings or stakeholder consultations, these two projects show a way for the public to provide input directly on what they consider to be the most urgent issues. The results have already been remarked upon by local and national officials in Uganda and in the donor communities that hope to see many of the policies implemented. Consider two examples of useful input: the policies on schools and health care centers. District officials had previously been closing secondary health centers (or failing to rebuild them after natural disasters) in order to consolidate health care provision in bigger and better centers. But these larger facilities are fewer in number, requiring many people to travel much farther. Officials had also been consolidating the schools to make them bigger and better, again increasing travel time for those who attend. The idea of small one-room schools in more remote villages to provide elementary education, particularly to girls, was not on the agenda. Yet these deliberations highlight the merits of placing both the local health clinics and the schools as close to the communities as possible. Although the trade-off between distance and quality for schools and for health care generates arguments on both sides, the district officials found that, deliberating together, the people from these communities had reached a different decision from the one they had made. The officials were, however, receptive to the idea that unless the health centers were located close to the villages, many people would not get critical health care, including family planning, and unless there were school facilities close to the villages, many people would not get elementary education. These burdens would fall especially on the women, for family planning, and on the girls, for education.

Government officials and other policy-makers can weigh these trade-offs by themselves if they so decide. But if they want policies that are sustainable because the people can buy into them, then they need to hear from the people. To date, the only practical method for getting that public input in both a representative and informed way is through the kind of deliberative process outlined here.

The challenges to applying deliberative democracy in Africa have previously been thought overwhelming. Yet the Uganda projects have already helped inspire subsequent Deliberative Polls in Ghana (in Tamale), in Senegal (in an area near Dakar), and in Tanzania (on a national level). All of these deliberative projects have featured high participation rates and intense deliberation. The complex story of the policy impacts of these projects will require separate analysis. In the meantime, these first projects stand as demonstrations that it is entirely feasible to consult populations in Africa in a representative and thoughtful way about the policies affecting their communities. We need not leave it only to stakeholders and elites to speak for the people. With the right design, the people can speak for themselves.

ENDNOTES

Authors' Note: This essay was made possible by support from the United States Agency for International Development (usaid). It was prepared jointly by the Center for Deliberative Democracy (CDD) at Stanford University and the Resilient Africa Network Eastern Africa Resilience Innovation Lab at Makerere University. The contents are the responsibility of the authors and do not necessarily reflect the views of USAID or the United States Government. We also acknowledge the contribution of the advisory group members, especially those from the local governments of Bududa and Butaleja. Data and materials from the projects discussed here are available from the Center for Deliberative Democracy at Stanford University and from the East Africa Lab in the Resilient Africa Network at Makerere University. The authors would like to thank Kathleen Giles of the CDD for her tremendous contributions to this project.

1

For some collections that gather the most influential of these discussions, from various perspectives, regarding deliberative democracy, see James Bohman and William Rehg, eds., Deliberative Democracy: Essays on Reason and Politics (Cambridge, Mass.: The MIT Press, 2007); Jon Elster, ed., Deliberative Democracy (Cambridge: Cambridge University Press, 2008); and James S. Fishkin and Peter Laslett, eds., Debating Deliberative Democracy (Oxford: Blackwell Publishing, 2003). The phrase “deliberative turn” was coined by John Dryzek in his Deliberative Democracy and Beyond: Liberals, Critics, Contestations (Oxford: Oxford University Press, 2003).

2

Carole Pateman, “Participatory Democracy Revisited” Perspectives on Politics 10 (1) (March 2012): 7–19.

3

If and when it is possible to engage the entire public on an issue, that also exemplifies deliberative democracy by the people. For a proposed method of doing so, see Bruce Ackerman and James Fishkin, Deliberation Day (New Haven, Conn.: Yale University Press, 2004). Most of the focus of recent empirical work on public deliberation has, however, been on the microcosmic or mini-public strategy. Other work has addressed the quality of deliberation in democratic legislatures; see Jürg Steiner, André Bächtiger, Markus Spörndli, and Marco R. Steenbergen, Deliberative Politics in Action: Analyzing Parliamentary Discourse (Cambridge: Cambridge University Press, 2004).

4

For an overview, see Kimmo Grönlund, André Bächtiger, and Maija Setälä, eds., Deliberative Mini-Publics: Involving Citizens in the Democratic Process (Wivenhoe Park, United Kingdom: ECPR Press, 2014).

5

Robert C. Luskin, Ian O'Flynn, James S. Fishkin, and David Russell, “Deliberating across Deep Divides,” Political Studies 62 (1) (2014): 116–135. For the case of public deliberation about the condition of the Roma, see Center for Deliberative Democracy, “Deliberative Polling on Policies toward the Roma in Bulgaria,” http://cdd.stanford.edu/dp-locations/europe/bulgaria/.

6

James S. Fishkin, Robert C. Luskin, and Alice Siu, “Europolis and the European Wide Public Sphere: Empirical Applications of a Counter-Factual Ideal,” European Union Politics 15 (3) (2014): 328–351.

7

See James S. Fishkin, Baogang He, Robert C. Luskin, and Alice Siu, “Deliberative Democracy in an Unlikely Place: Deliberative Polling in China,” British Journal of Political Science 40 (2) (2010): 435–448. For the more general issue, see Baogang He and Mark Warren, “Authoritarian Deliberation: The Deliberative Turn in Chinese Political Development,” Perspectives on Politics 9 (2) (2011): 269–289.

8

Joseph E. Stiglitz, “Participation and Development: Perspectives from the Comprehensive Development Paradigm,” Review of Development Economics 6 (6) (2002): 163–182. See also Amartya Sen, Development as Freedom (New York: Alfred Knopf, 1999).

9

See Toward a More Inclusive and Effective Participatory Budget in Porto Alegre, World Bank Report No. 40144-BR (Washington, D.C.: World Bank, 2008). For the critique of representativeness, see volume 1 of the report, 27–30.

10

For a critique of the move from participatory to deliberative democracy, see Pateman, “Participatory Democracy Revisited.”

11

Leonard Wantchekon, “Clientelism and Voting Behavior: Evidence from a Field Experiment in Benin,” World Politics 55 (3) (2003): 399–422.

12

Macartan Humphreys, William A. Masters, and Martin E. Sandbu, “The Role of Leaders in Democratic Deliberations: Results from a Field Experiment in São Tomé and Príncipe,” World Politics 58 (4) (2006): 583–622, esp. 604.

13

The Ugandan projects reported on here did not collect separate survey data from the moderators. But in a more recent Deliberative Poll in Tanzania with the Center for Global Development, we did collect such data and intend to report on it separately. The Deliberative Poll process trains moderators not to offer their own substantive views during deliberations.

14

The quotation is from James Madison in Federalist No. 10. The term “deliberative democracy” was coined by Joseph Bessette in discussing Madison's theory of deliberation by representatives (not deliberation by the people themselves). See Joseph M. Bessette, The Mild Voice of Reason: Deliberative Democracy and American National Government (Chicago: University of Chicago Press, 1994).

15

Russell Hardin asks whether “deliberative democracy is the ‘democracy’ of elite intellectuals.” Russell Hardin, “Deliberation: Method, Not Theory,” in Deliberative Politics: Essays on Democracy and Disagreement, ed. Stephen Macedo (New York: Oxford, 1999), 103–119. Shawn W. Rosenberg argues that deliberation even among the educated in the United States is “problematic,” concluding from his case study in California that “even in what might be considered the favorable case of highly educated, empowered adults raised in a liberal democratic polity, the deliberation very rarely was rational, reasonable, or critically reflective in the ways assumed by deliberative democrats.” See Shawn W. Rosenberg, ed., Can the People Govern (New York: Palgrave, 2007), 157.

16

Lynn Sanders, “Against Deliberation,” Political Theory 25 (3) (1997): 347–364; and Iris Marion Young, Inclusion and Democracy (Oxford: Oxford University Press, 2000).

17

In the landmark study comparing views of judges with actual jury verdicts, Harry Kalven and Hans Zeisel found considerable agreement between the judges and juries. See Harry Kalven and Hans Zeisel, The American Jury (Boston: Little Brown, 1966).

18

For classic accounts of gender and class domination, see Fred L. Strodtbeck, Rita M. James, and Charles Hawkins, “Social Status and Jury Deliberations,” American Sociological Review 22 (1957): 713–719; and Reid Hastie and Steven D. Penrod, Inside the Jury (Cambridge, Mass.: Harvard University Press, 1983).

19

Jürgen Habermas, Between Facts and Norms, trans. William Rehg (Cambridge, Mass.: The MIT Press, 1998), 306.

20

Cass R. Sunstein, “The Law of Group Polarization,” in Debating Deliberative Democracy, ed. James S. Fishkin and Peter Laslett (Oxford: Blackwell Publishing, 2003), 80–101.

21

For an overview of these arguments and an extensive bibliography documenting their occurrence, see Cass R. Sunstein, Going to Extremes (Oxford: Oxford University Press, 2009).

22

Cass R. Sunstein and Reid Hastie, Wiser: Getting Beyond Groupthink to Make Groups Smarter (Boston: Harvard Business Review Press, 2015), 84.

23

See, for example, Robert C. Luskin, James Fishkin, and Roger Jowell, “Considered Opinions: Deliberative Polling in Britain,” British Journal of Political Science 32 (3) (2002): 455–487; Fishkin et al., “Deliberative Democracy in an Unlikely Place”; and Alice Siu, “Look Who's Talking: Examining Social Influence, Opinion Change, and Argument Quality in Deliberation” (Ph.D. diss., Stanford University, 2008).

24

For more on this, see Alice Siu, “Deliberation & the Challenge of Inequality,” Dædalus 146 (3) (Summer 2017).

25

See David Schkade, Cass R. Sunstein, and Reid Hastie, “When Deliberation Produces Extremism,” Critical Review: A Journal of Politics and Society 22 (2–3) (2010): 227–252.

26

See Deliberative Polling in Uganda: A Report on the Projects in Bududa and Butaleja Districts, Uganda (Stanford, Calif., and Kampala, Uganda: Center for Deliberative Democracy at Stanford University and Resilient Africa Network Eastern Africa Resilience Innovation Lab at Makerere University, 2014), Appendix Table 1, “Members of the Advisory Group,” 30, https://cdd.stanford.edu/2014/a-report-on-the-projects-in-bududa-and-butale-districts-uganda/.

27

Participants were paid the equivalent of about seven dollars a day as an honorarium.

28

Both the Bududa and Butaleja samples were overrepresented by men. Participants in Bududa were 58.7 percent male and in Butaleja were 66 percent male. The interviewers randomly selected the participant within each household, but they could only do so based on the list of inhabitants provided by the person first contacted. Despite this limitation, the interests of women seem to have been well represented in the discussions, as should be clear from the priorities after deliberation on the education of women and on population pressure.

29

We will discuss these results in percentages; the means can be seen in Table 1 and Table 2.

30

The online appendix for this article is available at Center for Deliberative Democracy, “Deliberative Polling in the Bududa and Butaleja Districts of Uganda,” https://cdd.stanford.edu/2014/deliberative-polling-in-the-bududa-and-butaleja-districts-of-uganda/. For the top-ten priorities after deliberation in Bududa and Butaleja, see Tables H and I. The transcript excerpts are included in Tables J and K.

31

Shkade, Sunstein, and Hastie, “When Deliberation Produces Extremism,” 232.

33

See Tables F and G in ibid.