Deliberative democracy is generally viewed as more valuable when a diverse array of voices is represented equitably in the discussion. Given the inequalities and lack of opportunities to express one's views in daily life for most individuals from relatively powerless groups, a forum to freely express and discuss views as equals should be appealing. For this reason, some scholars (e.g., Walsh Reference Walsh2007; Williams Reference Williams, Kymlicka and Norman2000) argue that deliberative democracy helps mitigate the gap between politically influential and marginalized groups because in deliberation one's arguments matter more than money or social status. However, other theorists argue that inequalities found in broader participatory democracy are merely transferred to the deliberation process (Sanders Reference Sanders1997) or perhaps even heightened in small groups (Young Reference Young2001). For example, in a group deliberation, women tend to speak less than and feel intimidated by men (Bryan 2004). Being excluded from the conversation in this way may cause participants to feel disrespected, like they do not belong to the group (Simon and Sturmer Reference Simon and Sturmer2003), and prevented from achieving their interactional goals (Smith-Lovin and Brody Reference Smith-Lovin and Brody1989, 485). These types of negative experiences may cause individuals to opt out of future deliberations, thereby depriving themselves of the opportunity to develop civic skills and work through complex political issues as well as depriving other people from hearing an alternative perspective that could inform their own views. This article explores whether members of traditionally less powerful demographic groups have, on balance, positive or negative views of small group deliberation and explores whether the inclusion of more participants sharing these characteristics improves the assessment of the deliberation.
The literature on deliberation focuses less on answering this question and instead examines whether deliberation leads to changes in political opinion or increases political knowledge (Myers and Mendelberg Reference Myers, Mendelberg, Huddy, Sears and Levy2013). Among the few works that measured participants’ evaluation of the deliberation experience (e.g., Price and Cappella Reference Price and Cappella2002; Stromer-Galley and Muhlberger 2009; Wojcieszak et al. Reference Wojcieszak, Baek and Delli Carpini2009), none explicitly compared the evaluation of participants from less powerful groups against their more advantaged counterparts. If critics of deliberative democracy are correct that deliberation simply transfers real-world inequality to another venue, then we should expect participants from relatively more disadvantaged groups to evaluate deliberation less positively. Alternatively, if deliberation is a “natural ally” of marginalized groups (Williams Reference Williams, Kymlicka and Norman2000), then we should expect the participants to have a more positive perception of the deliberation.
Satisfaction with the quality of deliberative experiences may be improved by the composition of the deliberative group. Existing studies suggest it matters for minority participants how many other discussants are from the same minority group. For instance, Hannagan and Larimer (Reference Hannagan and Larimer2010) find that higher proportions of women in groups results in group outcomes closer to the women's preference. In a series of studies, Mendelberg and Karpowitz (Karpowitz et al. Reference Karpowitz, Mendelberg and Shaker2012; Mendelberg et al. Reference Mendelberg, Karpowitz and Oliphant2014a; Mendelberg et al. Reference Mendelberg, Karpowitz and Goedert2014b) find that the greater the proportion of women in a group the larger the influence of women in the group when the decision rule favors the majority, but not under unanimity rules. If superiority in number amplifies the dominance of more politically powerful groups in a deliberation, as the existing studies suggest, then it is likely that deliberation would be an unpleasant experience for minority participants, which could deter participation in future deliberative settings.
We answer these two questions by analyzing 27 deliberative polls conducted in 2004. Participants were randomly assigned to small groups for discussion and this random assignment created natural variation in the composition of the groups. We first compare how members of four relatively less powerful groups (women, the young, ethnic minorities, and people without college degrees) evaluate the deliberative experience on three dimensions. We find that people from these groups generally have slightly more positive impressions of the deliberative groups than older, white, male college graduates. Taking advantage of the natural variation in group composition created by the random assignment, we find no evidence that group composition colored respondents’ impression of the deliberative experience. Presumably hegemony exhibited in the small groups would be unpleasant and cause subjects to react negatively to the deliberation, so our results should temper concerns about inequality in deliberative democracy.
DATA
Twenty-seven deliberative polls were conducted during January and October of 2004. Respondents chosen by a random sample constructed by the Survey Research Center at the University of California at Berkeley were contacted by phone and given a short pre-survey. A randomly selected portion of those contacted during the survey were offered to attend a day-long deliberation in a pre-designated site.Footnote 1 The January wave was held in 10 sites across the country with 721 participants, whereas the October wave took place in 17 sites with 1,661 participants. Both waves followed the same procedure in the recruitment of subjects, structure of the deliberative polls, instruction to group moderators, and survey questions answered by subjects (Farrar et al. Reference Farrar, Green, Green, Nickerson and Shewfelt2009; Fishkin and Luskin Reference Fishkin and Luskin2005). Participants were randomly assigned to 10 discussion groups at each site that met twice. Discussions in the small groups were moderated and concerned designated topics (the Iraq war and free trade in January and the war on terror and jobs in October). A general panel with policy advocates and experts followed the two small group sessions, after which participants completed a survey about issue positions and attitudes toward the deliberation experience.
Dependent Variables
Three questions about the small group deliberation portion of the deliberative poll are our outcome of interest. The first question (small-group experience), “How valuable in helping you clarify your positions on the issues was participating in the small group discussions?” (scale 1–10), gauged the subject's assessment of the utility of the small group exercise. The second question (feeling respected), “To what extent would you say the other discussion participants in your group respected your point of view on the issues?” (scale 1–4), measured the degree to which the respondent felt s/he had been respected. The final question (efficacy) assessed how efficacious the respondent felt in the small group discussion by asking “How much impact do you think your participation in the small group discussions had on the other members of your group?” (scale 1–4). All three questions measure different dimensions of deliberation in small groups and are coded so higher numbers indicate a positive experience.
Independent Variables
Our analysis focuses on social groups that are at times politically marginalized: women; youth; ethnic non-whites; and people without a college degree. Whether or not a subject is a member of one of these descriptive groups is not randomly assigned. However, since small groups in each site are randomly assigned, the composition of the small group is subject to random variation.Footnote 2 That is, some subjects will be assigned to groups with more or fewer people (perhaps even none) sharing their descriptive characteristic.Footnote 3 Thus, we calculated for each of the four descriptive categories the proportion of people in the group other than the respondent falling into the category. For example, in a group of one female and four males, for the four males proportion of female would be 0.25 and for the female it would be 0.
There were no reported problems in the randomization procedure. We checked for covariate balance, which can be diagnostic of problems in randomization, and present the results in the online supplemental appendix (Tables A3 and A13). If pre-treatment covariates predicted treatment assignment (as measured by the F-test), it would suggest an error in randomization or in the data pipeline (or very bad luck). We find no evidence of problems. Lastly, our analysis controls for group size and the site of the deliberative poll. The online supplemental appendix presents a table of all variables’ descriptive statistics.
ANALYSIS
Since our two research questions consist of one descriptive claim (i.e., members of relatively less powerful groups will evaluate the sessions differently) and another claim tested by randomly assigned data (i.e., descriptive representation in the group matters), our analysis is relatively straightforward. To gauge how particular subgroups evaluate the deliberative process relative to their converses (i.e., women vs. men; young vs. not young; non-whites vs. whites; no college degree vs. college degree), we regress each of our three outcome variables on a dummy for sub-group membership, the proportion of the discussion group who are members of the sub-group, group size, and a dummy variable for each polling site. Positive coefficients indicate that members of each subgroup rated the small group deliberation more highly than their relatively more advantaged counterpart.
Figure 1 presents the results of these regressions for each group of interest on each measure of the small group deliberation process (see online supplemental appendix for the full regression results). For example, women ranked the groups on average 0.37 points higher on the 10-point scale than men, but evaluated their effectiveness on the group discussion an average of −0.13 points lower on the efficacy scale. With only two exceptions (females on feeling efficacious and the less educated on feeling respected), Figure 1 provides evidence that participants from the politically less influential groups evaluated the deliberation experience more favorably than their more influential counterparts and argues against the notion that deliberation will recreate unequal—and presumably unpleasant—experiences for members of these subgroups.Footnote 4
Figure 1 Mean Difference of the Groups.
While several are statistically significant, these differences are not substantively large. For instance, the question regarding small group experience has a standard deviation of 1.88, but the detected average difference between women and men is only 0.37. That is, the difference between men and women constitutes only 20% of the standard deviation in evaluating the small group experience. The proportions for other subgroups are even smaller. As a proportion of overall variance the biggest—still relatively small—difference on feeling efficacious is by race. Non-whites on average felt 0.20 point more efficacious than whites did, but this is only 28% of the standard deviation of the efficacy variable (0.71). So while the differences in evaluation between the relatively less powerful social groups and their complements are measurable and statistically significant, they are not large in absolute terms. Thus, any hegemony manifested during the deliberative process does not cause participants from traditionally less powerful groups to react negatively to the experience.
Our second area of inquiry is whether the demographic composition of small discussion groups affected people's assessment of the deliberative experience. Since subjects were assigned to small groups randomly, our model can be spare with few control variables. For each subgroup, we regress each dependent variable on a dummy variable indicating membership in the group, the proportion of the small discussion group sharing this characteristic, and the interaction between the two variables. If a member of a group feels better about the small group discussion when the group contains more people sharing the descriptive characteristic, then the coefficient for the interaction term should be positive. This analytical strategy minimizes collinearity, avoids false negatives, and represents a maximally sensitive test for detecting effects from representation in the group.
Table 1 presents the results from the regression analysis. In none of the 12 models run does the coefficient for the interaction term approach conventional thresholds for statistical significance. In fact, the coefficient in seven of the models has the wrong sign, which would mean that subjects had a lower opinion of the small group deliberation when other participants shared a descriptive trait. We take these results to mean that these 27 deliberative polls provide no evidence that group composition moderates participants’ evaluation of the deliberation experience. Thus, the argument that power relationships outside of the deliberative process mar the deliberation experience receives little support here.
Table 1 Regression of Deliberation Experience with Interactions
Robust standard errors in parentheses. Dummies for sites are always included.
***p < 0.01, **p < 0.05, *p < 0.10.
DISCUSSION
Why would less politically powerful groups rate the deliberative process more positively than their more advantaged counter-parts and what does this difference mean for the arguments advanced by critics of deliberative democracy? Two mechanisms suggest themselves as potential explanations for this finding. The first is that deliberative discussion is generally a positive experience and the inequality may be less pronounced than in broader society. Members of relatively disadvantaged subgroups may give the small group deliberative sessions high marks because discussing policy on equal footing is enjoyable. By the same token, people not falling into one of the subgroups may find the level playing field less enjoyable than discussions occurring in the real world. This mechanism suggests that deliberation is a natural ally of less powerful groups (Williams Reference Williams, Kymlicka and Norman2000) where the opportunity to express opinions leads participants from the subgroups to have a more positive evaluation of the process.
Second, the moderators may have ameliorated the inequalities some theorists predicted. Moderators in the small groups were instructed to ensure that all group members were attended to when they were speaking. When participants from the less influential subgroups rated their experience, they might compare it to the un-moderated, free-flowing discussion in the real world where people from less powerful groups are less likely to be paid attention to. If moderators were the key to ensuring more equal speech, then many of the critiques of deliberative democracy may still apply. Deliberation without formal moderation may be less inclusive and control over moderator instructions could define the nature of the discussion.
We tried to gauge the extent of which the results are attributable to the moderators by running three additional analyses (see online supplemental appendix). The first and second analyses regress participants’ evaluations of the moderators and of unmoderated sessions on the participants’ characteristics, whereas the third replicates the original analysis but controlling for evaluations of the moderators. Our results suggest that the relatively marginalized participants (females, young people, non-whites, and those without college degree) evaluated the unmoderated sessions positively and moderators relatively negatively. Furthermore, the results of the original analysis hold up even after controlling for evaluations of the moderators. This analysis provides evidence that our results are not driven by the presence of moderators.
Our null finding regarding the effect of group composition does not necessarily contradict previous works on group composition (e.g., Mendelberg et al. Reference Mendelberg, Karpowitz and Goedert2014b; Hannagan and Larimer Reference Hannagan and Larimer2010) for three reasons. First, and most obviously, we are studying sentiment about the process rather than opinion formation. Second, the issues being discussed in our study (foreign policy, the economy, and free trade) were not directly related to social inequality. Talking about foreign policy with people from different races is arguably less uncomfortable than talking about affirmative action. Third, participants in our study were not required to arrive at a conclusion, which might create a less argumentative deliberation process resulting in a more positive experience.
Future work could address this gap by systematically varying the goal of the deliberation. While most deliberative settings (e.g., friendly conversation) do not require a decision to be reached, many politically important settings (e.g., juries) do. It is possible that the groups studied here have a less positive assessment of more conflictual settings where a group decision is expected. Such studies would also benefit from incorporating more detailed measures of the actual deliberation itself, such as speaking time and frequency of interruption, which are not measured here. While our study offers strong evidence that deliberation is welcomed by relatively less powerful groups, we view it as the start of a research agenda rather than a final answer.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/XPS.2016.6.
