Arguably, the most striking aspect of mini-publics is the mode of selection.7 Random selection (or sortition) has a long democratic heritage: it was the preferred method for selecting positions of political authority in the Athenian polis and continued to play a part in republican thought and practice throughout the Middle Ages and the Renaissance and into the seventeenth and eighteenth century (Manin 1997). Given this democratic heritage, it is perhaps surprising that it has played little or no role in contemporary politics, where selection by competitive elections is generally perceived to be the democratic mechanism of choice. The most prominent exception to this rule is the randomly selected jury used in a number of legal systems in advanced industrial democracies. For Athenians, lot and rotation gave full expression to the principle of democratic citizenship by providing the occasion for citizens to rule and be ruled in turn. As Barbara Goodwin notes, the ‘choice of leaders by lot averted the danger that power would go to the rich or to those who desired it’ (Goodwin 2005: 46). The use of random selection acts as a defence against oligarchic tendencies and realizes the democratic principle that any citizen is capable of political judgement (Barber 1984: 293).
Random selection replaces the principle of equal opportunity to participate that is the hallmark of open or popular assemblies with an equal probability of being selected to participate (Saward 2000: 16; Brown 2006: 212-13). Advocates argue that this is a just and acceptable restriction: where the opportunity to participate is to be limited, then random selection is a fair mechanism to distribute an ineradicable inequality (Goodwin 2005: 45). It is a mechanism that ensures that no citizen or social group from the given population is systematically excluded from participation.
In actual practice, mini-publics rely on ‘near-random selection’ (Warren and Pearse 2008: 6). There are three reasons why pure random selection is not achieved. The first two are well-known sampling problems that affect recruitment. The first relates to the incomplete nature of any database from which the sample of citizens is taken. For example, the initial sample for the BCCA was drawn from the province’s voters’ list. Not all residents of the province will or can be registered.
A second sampling problem relates to the element of self-selection in the recruitment process. Since none of the innovations can require participation, those who are invited can choose not to participate. In the case of the BCCA, citizens from a large random sample drawn from the electoral register were asked twice whether they would be willing to participate:
The initial letter, mailed to 23,034 randomly chosen citizens, invited the recipients to decide if they wanted to participate in the Assembly process. Those who responded positively and then attended a selection meeting were again asked to confirm their willingness to commit to the project and accept the responsibilities of membership. The Assembly members were then chosen by lot from this group of attendees.
(Citizens’ Assembly on Electoral Reform 2004: 39)
From the initial 23,034 invitations, 1,715 citizens responded positively; 964 attended selection meetings; those who wished to participate were entered into a lottery.
There is a danger that such a low response rate will skew the sample and thus undermine the legitimacy of mini-publics. Organizers of the BCCA were confident that ‘the final membership of the Assembly generally reflected the distribution of the provincial population’ (Citizens’ Assembly on Electoral
Reform 2004: 40). Similarly, the Ontario Assembly appeared to reflect many of the socio-demographic variations across the province, including place of birth, languages and occupations (Ontario Citizens’ Assembly Secretariat 2007: 47). Experience from deliberative polls suggests that ‘[t]hose who decide to attend are usually somewhat more politically active and better educated than the initial sample’ (Fishkin and Farrar 2005: 74). However, differences between the final sample and the wider population should not be exaggerated: the use of random selection ensures that participants are drawn from a range of social groups and are not simply the self-selecting few who typically engage in the political process.
The third reason why pure random selection is not achieved in most minipublics is actually by design. The Citizens’ Assemblies, citizens’ juries and consensus conferences all use a form of stratified sampling. Quotas are assigned to particular social groups to ensure their presence and then participants from these groups are selected randomly. Pure random selection for a single mini-public (especially one small in size) is likely to lead to the absence of citizens from numerically small social groups and potential imbalances between other groups. While no systematic bias would have been in operation, the lack of presence of certain groups means that their perspectives may not be articulated or considered in deliberations, thus potentially affecting the perceived legitimacy of the body and its recommendations. The BCCA used three criteria - geographical district, gender and age - as the basis of quotas in its selection process. In comparison, citizens’ juries and consensus conferences often stratify the sample to include other characteristics, for example, ethnicity, social class and - on occasion - political or social attitudes. Over-sampling of particular social characteristics ensures that the panel reflects politically salient characteristics from within the wider population. Given their larger size, deliberative polls and planning cells tend not to use quotas.
The failure of the BCCA to select on the basis of ethnicity may have had ramifications on the proceedings and the selection of a preferred electoral system. As Michael James argues: ‘by stratifying for region and gender but not for race or ethnicity, citizen assemblies deliberating about electoral systems could potentially skew the agenda against the interests of racial or ethnic minorities and in favour of women and regional minorities’ (James 2008: 180). There was official recognition that the selection process had failed to recruit citizens from aboriginal communities - hence the Assembly chair requested that two additional members be included in the process; a request that was granted (Citizens’ Assembly on Electoral Reform 2004: 39). But there were no quotas for other potentially salient minority ethnic groups. Learning a lesson from British Columbia, the Ontario Assembly ensured that one self-defined Aboriginal citizen was selected, although, again, the presence of other minority ethnic groups was not considered significant.
If certain mini-public designs are not aiming to achieve as pure a random selection of the population as is practicable, precisely how much presence is required for different social groups? One or two participants from a particular group is unlikely to be adequate - for example, the two Aboriginal members who were recruited as an afterthought to the BCCA, or the single self-defined
Aboriginal member required in the Ontario Assembly. James argues that the presence of a ‘critical mass’ from minority social groups may be necessary for a number of reasons. First, to ensure that there are sufficient numbers to communicate effectively - one voice can become isolated in a large assembly. Second, to provide support and bolster the confidence of speakers who may be offering a perspective that is uncomfortable for other participants. Third, to ensure that the perspective is heard in the different locations within the body - for example, the various break-out groups in the BCCA. Finally, in recognition that there is likely to be a plurality of perspectives from within social groups - they are not homogenous and closed communities (James 2008: 120-23). Achieving critical mass may require significant over-representation of small minority social groups, something that is difficult to achieve (if not impossible if there are a number of groups) in smaller mini-publics such as citizens’ juries and consensus conferences.
We are left with a series of difficult conundrums. How far should mini-publics move away from the principle of equal probability of selection (pure random selection) and towards stratified samples that ensure the presence and/or overrecruitment of particular social groups? Which social groups should be considered for quotas? And at what point do we achieve a critical mass for particular minority groups? These are highly pertinent theoretical and practical questions that demand further attention (and correspond to dilemmas faced by proponents of electoral quotas).
Once citizens agree to participate in mini-publics, evidence suggests that they typically do attend. Even the BCCA - which ran for eleven months and was thus more demanding than other designs - suffered only one withdrawal. Commenting on the recruitment process, the Assembly’s final report suggests: ‘This process appeared to create a sense of “buy-in” for the Assembly members that contributed significantly to their commitment to the process ... the fact that only one member withdrew in the course of 11 intensive months, suggests that this process of recruitment deserves further examination’ (Citizens’ Assembly on Electoral Reform 2004: 39).
The use of random selection might be a significant factor in encouraging engagement, particularly amongst citizens from politically marginalized social groups. The formal invitation to participate in a selective forum, combined with a modest honorarium, appears to play a crucial role in motivating citizens’ participation and support for the process. Citizens who typically do not participate in other forms of political activity are willing to engage in what are more intensive forms of political engagement. Citizens perceive they are being offered a rare opportunity to participate in a politically significant process. This was certainly the perception of participants in the NICE Citizen Council, where citizens reported that their motivation to participate was based on a variety of factors: on the belief that institutions should be more open to the public voice; that it is a public duty to make a contribution; and for reasons of personal growth and fulfilment. Citizens ‘also frequently referred to a sense of being “privileged” to have been selected’ (Davies et al. 2006: 80-1).
Random selection generates a variety of claims about the representativeness of mini-publics, although there are relatively few studies dedicated to exploring this issue in much theoretical depth (Brown 2006; Warren 2008). It is not uncommon to find practitioners and theorists claiming that mini-publics achieve statistical, descriptive or microcosm representation. This claim needs to be tempered on several accounts. First, the effect of the sampling problems involved in recruitment and the use of stratified sampling in many designs render such representative claims problematic (Parkinson 2006: 76). Second, descriptive representation is an unhelpful categorization because it often assumes that participants are somehow representing ‘people like them’ in a strong sense (Smith 2003: 91). There is a danger here of creating false essentialisms that fails to recognize that there are multiple perspectives within social groups. Jeffrey Abramson draws out the potential tension between representing a social group and achieving inclusive deliberation in relation to legal juries:
We do not want to encourage jurors to see themselves as irreconcilably divided by race, selected only to fill a particular racial or gender slot on the jury. Yet we do want to encourage jurors to draw upon and combine their individual experiences and group backgrounds in the joint search for the most reliable and accurate verdict. The difference is subtle but real.
(Abramson 1994: 11)
Abramson is pointing towards a different way in which mini-publics can be said to represent the broader public. Here, we can return to Fishkin’s much repeated comment: ‘A deliberative poll attempts to model what the public would think, had it a better opportunity to consider the question at issue’ (Fishkin 1997: 162). Arguably this is a different sort of claim: a claim that mini-publics offer a representation of the public’s considered judgement (Brown 2006: 216).
The use of random selection challenges our prejudices of what legitimate representation entails, but it does ensure that mini-publics engage a broad cross section of citizens with a diversity of social perspectives. The recruitment process generates a panel of highly motivated citizens who differ markedly from the highly skewed characteristics of citizens who routinely engage in consultation processes. For this reason alone, random selection deserves much more attention within theoretical work on institutional design (see also Geissel in this volume).