Citizen participation in a deliberative poll: Factors predicting attitude change and political engagement

Abstract

In November, 2002, a randomly sampled body of citizens was brought together to participate in a deliberative poll on the issue of a bill of rights. Participation in deliberative polls is thought to cause attitude change and an increased sense of political engagement, but the underlying processes are little understood. We surveyed poll participants before and after the poll, proposing that either exposure to information, fair treatment or identification as a poll participant would predict attitude change and engagement. None of these affected attitude change but all three predicted political engagement. Results suggested that participation in the poll increases levels of political engagement through its capacities to (a) give people an opportunity to exchange views in a respectful atmosphere and (b) create or increase a sense of connection to a relevant social group.

A deliberative poll (Fishkin, 1991) is a process designed to give citizens the opportunity to access information about public policy issues and to consider these issues thoughtfully. A random sample of voters is surveyed then brought together to deliberate on an issue over a 2-day period. During this time, participants are presented with a range of competing arguments, which they consider in small group discussions chaired by an impartial facilitator. After the poll, attitudes are surveyed again with the expectation that attitudes would have shifted. Importantly, only attitude change of some kind is predicted, rather than change in a specified direction. This is because a deliberative poll aims to encourage critical thought and social or political engagement on the issue of concern rather than movement towards a particular view.

The experience of taking part in a deliberative poll is said to move participants from a position of ignorance and disengagement to a position of considered opinion and engagement. Fishkin, Luskin, and Jowell (2000) suggest that ordinary people start from a position of ignorance not through any deficiency of their own but as a by-product of the way they are usually treated in relation to issues of public interest. Fishkin et al. argue that contemporary democracies don't ordinarily provide citizens with the means to have their voices heard. People learn that their vote (on the occasions that they get to cast it) is just one of many and is of little consequence. They learn that being informed and engaged has no utility for them. Fishkin et al. (2000), after Downs, (1957) calls this a position of “rational ignorance”. Participation in a deliberative poll is said to move people away from a position of rational ignorance by providing them with the opportunity and motivation to think more about issues.

Ryan (2002) goes even further and suggests that pre-poll ignorance is not simply due to a lack of opportunity or motivation but is an artefact of the way the human cognitive system functions. This argument is based on research into human decision-making (e.g., Fiske & Neuberg, 1990; Messick & Mackie, 1989; Neale & Bazerman, 1991; Rothbart, 1993), which shows that when making decisions, humans often appear to take “mental shortcuts”. It is assumed that these serve the purpose of filtering and aggregating information, thereby speeding up the decision-making process. As examples of cognitive shortcuts, people may stereotype others, ignore arguments if they come from an unfamiliar source or make judgments by evaluating new facts in comparison to existing (and perhaps incorrect) beliefs. In the literature on cognitive information-processing, these behaviours are seen as errors that can be attributed to a lack of information processing capacity or opportunity. Ryan argues that deliberative polling gives people the time and information they need to develop more evidence-based, finely tuned and accurate opinions rather than relying on cognitive shortcuts.

In contrast, it has also been suggested that attitude change has very little to do with quantities of time or information (Oakes & Turner, 1990). This alternative is based on a tradition of research drawn from self-categorisation theory (Turner, Hogg, Oakes, Reicher, & Wetherell, 1987). At the heart of this approach is the argument that our attitudes (and our willingness to change them) are necessarily linked to the way we categorise ourselves in relation to the world. According to this approach, our attitudes are linked to our membership of social groups (e.g., as Australian, female, or an opinion group such as “human rights supporters”; see Bliuc, McGarty, Reynolds, & Hendres, in press). Groups that we mentally place ourselves into can be thought of as reference groups or ingroups (as opposed to outgroups, or groups that we do not identify with) (Tajfel & Turner, 1979). Importantly, research has shown that we are much more influenced by ingroup members than outgroup members (David & Turner, 2001; Turner, 1991). This means that, in order to be influenced towards a certain point of view we must see that point of view, as relevant or meaningful to a valued group of which we feel a part (e.g., Haslam, McGarty, & Turner, 1996; McGarty, Haslam, Hutchinson, & Turner, 1994). Consistent with this, one of the best predictors of political engagement is thought to be identification as a political activist (Simon et al., 1998). From this perspective, attitude and behaviour changes do not occur simply as the result of an individual's encounter with new information. Instead, they come about when people begin to take on a new identity that brings with it a new view.

According to this perspective, attitude change and a sense of engagement on poll issues should come about due to the development of a new willingness to see the self as the type of person who does and should think about politically relevant issues, that is, as a result of the development of a new sense of oneself as a politically engaged person (Kelly & Breinlinger, 1996). How may such a new sense of self develop as a result of participation in a process such as deliberative polling? A mechanism may be provided by the group value model (Tyler, 1989), which suggests that we are more likely to develop a social identity or a sense of ourselves as being part of a social group when we receive positive or fair treatment from others within that group. Such treatments include the provision of respect and of a voice, or of the opportunity to have a say. According to the group value model, the provision of respectful treatment and an opportunity for voice is likely to increase the sense of connection that people feel to the relevant social group and to the norms and values it embodies. Consistent with this, Simon and Stürmer (2003) have found that respectful treatment by people in one's group serves to increase one's sense of commitment to that group.

Past research work on deliberative polls shows that a large proportion of individuals do in fact shift their attitudes to some degree as a result of participation in a poll (e.g., Luskin, Fishkin, & Jowell, 2002). Similarly, anecdotal evidence suggests that participation in a poll can make people feel more politically engaged (Smith, 1999). But how and why do these shifts occur? Is it a matter of utility and information or does a process of developing a sense of social connection to others mediate the effects of deliberative polling? The purpose of the present research is to provide an investigation based on this question. The poll we studied was conducted in the Australian Capital Territory (ACT) in November 2002 and centred on the issue of whether or not the ACT should introduce a bill of rights. The poll process was organised and managed by Issues Deliberation Australia (IDA), a not-for-profit organisation. We surveyed participants immediately before and after the poll. We had several research aims. First, we sought to establish whether or not participants' attitude towards a bill of rights shifted over time. We also aimed to investigate the relative contributions of several factors that have been hypothesised to contribute to attitude shift and political engagement on an issue (here operationalised as participants' level of interest in the rights issue): (a) exposure to information, (b) being treated with respect and being offered a voice (“procedural fairness”) and (c) identification as a community representative. We focused on identification as a community representative because this approximates the self-perception that poll organisers encouraged among the group. That is, participants were asked to think of themselves as people who had been chosen to deliberate on the ACT bill of rights issue as representative citizens.

Based on the above review and past research, we suggest that attitude change of some kind is likely following participation in a deliberative poll. We suggest two alternatives to explain this shift. First, based on the cognitive information processing approach, we suggest that perceived exposure to a wide range of opinions should play the biggest role in producing attitude change. Alternatively, based on the self-categorisation approach, we suggest that a tendency to identify as a representative citizen should play the biggest role. In relation to participants' sense of engagement, we make similar predictions. Based on the cognitive approach, we suggest that perceived exposure to a wide range of opinions should predict engagement on the rights issue. Alternatively, based on the self-categorisation approach, we suggest that a tendency to identify as a representative citizen should predict engagement. Finally, according to the group value model, we expect that perceptions of fair treatment should be related to the development of participants' sense of identification as representative citizens.

Method

Participants and design

Two hundred and three citizens of the Australian Capital Territory (ACT) attended a deliberative poll in November 2003 in Canberra. A total of 195 of these poll participants completed both pre- and post-poll surveys.

Materials and procedure

The poll was focused on the issue of whether or not the ACT should adopt a bill of rights. Once at the poll venue, participants completed an initial survey that questioned them about their attitudes towards a bill of rights. They were randomly assigned by the poll organisers to subgroups (approximately 15 members each). All groups then took part in three facilitated group discussions interspersed with three plenary sessions. During discussion sessions, groups developed questions to ask experts at each plenary.

Immediately following the close of the last group session, participants completed a post-poll survey that again asked them about (a) their attitudes as well as (b) their commitment to this opinion as well as a range of other measures including (c) perceived exposure to all sides of the issue, (d) fair treatment (two-item scale: “I was able to express my own views feely” and “all views treated with respect”, Pearson's r = .61), (e) identification as a community representative (three-item scale: confidence in own ability to represent the community, feeling of pride in participation, likelihood of telling others in the community that you took part, Cronbach's α = .77) and (f) political engagement, operationalised as interest in the issue (three-item scale: belief that it is important to develop a view on the issue, the hope to play an active role in the development of others' views, and intention to discuss the issue with others, Cronbach's α = .75). All items were presented in the form of statements. Participants indicated their level of endorsement of each statement by responding on nine-point Likert scales where 1 = strongly agree and 9 = strongly disagree.

Results

Attitude change over time

Initially, we tested the prediction that the attitudes of poll participants would change following participation in the deliberative poll. We found that, on average, people were less in favour of a bill of rights after participation compared to their positions before the poll (Ms pre-poll = 4.5, post-poll = 5.11, where lower numbers indicate more support for a bill of rights), t(141) = −3.54, p < .001. It is interesting to note that there was a lot of variance around these means, both before (SD = 2.24) and after (SD = 2.69) the poll, meaning that poll participants were widely distributed in their attitudes towards a bill of rights before the poll and slightly even more so afterwards.

Relationships between variables

From Table I we can see that none of the three independent variables, exposure to information, identification and fair treatment, significantly related to attitude change or commitment. This means that we did not find support for the prediction that either (a) exposure to information (based on the cognitive information processing) or (b) social identification as a participant (based on the self-categorisation approach) would predict attitude change or contribute to participants' feelings of commitment to their positions. Therefore, we did not proceed with further analysis to investigate attitude change per se. However, some interesting patterns did emerge for political engagement (a variable also central to deliberative polling). All three independent variables were significantly related to political engagement. The independent variables were also correlated with each other to some degree. There were significant relationships between exposure and identification, between exposure and fair treatment and between identification and fair treatment.

Table I. Correlations between variables

	Identification	Fair treatment	Attitude change	Commitment	Political engagement
Exposure	.389*	.537*	−.168	.141	.312*
Identification		.633*	−.045	.199	.759*
Fair treatment			−.069	.188	.533*
Attitude change				.007	.048
Commitment					.170*
Note. *Correlation is significant at the .01 level (two-tailed).

Factors that influence political engagement

Initially, we aimed to investigate the relative contribution to political engagement made by the three predictor variables exposure to information, identification as a participant, and fair treatment. A linear regression was performed on the data, with all three predictor variables entered at the same time. As can be seen in Table II, identity emerged as the only significant predictor (unstandardised regression weight b_identity = .724, p < .05).

Table II. Regression testing model 1 (relationship between exposure to information, identification, fair treatment and political engagement)

IV	DV	ß	Standardised β	R^2
Identification	Political engagement	.724	.721	.579
Fair treatment		.070	.071
Exposure		.020	−.022

Taken together with the correlations seen between the predictor variables and political engagement, this pattern of findings supports a model that places identification as a mediator between (a) exposure and fair treatment and (b) political engagement. This model is displayed in Figure 1. The model was tested using hierarchical regression (see Table III). If identification as a participant were acting as a mediator variable we would expect to see (a) a significant relationship between identity (the mediator) and engagement (the dependent measure), (b) the relationship between identity and engagement to remain significant in the context of non-significant relationships between the independent variables exposure and fair treatment, and the dependent measure engagement, and (c) significant relationships between the independent variables exposure and fair treatment and the mediator variable identification as a participant. Hierarchical regression analysis showed that relationships (a) b_identity = .761, p < .05, and (b) b_identity = .724, p < .05; b_{fair treatment} = .07, p > .05; b_exposure = −.02, p > .05 held but that (c) and (b) b_{fair treatment} = .573, p < .05; b_exposure = .05, p > .05 did not. These findings indicated that the relationship between fair treatment and political engagement was mediated by identification. However, identification did not appear to act as a mediator in the relationship between exposure to information and political engagement.

Figure 1. Model 2: proposed relationship between exposure to information, identification, fair treatment and political engagement

Table III. Model 2: Regression testing the relationship between exposure to information, identification, fair treatment and political engagement

		IV	DV	ß	Standardised β	R^2
Regression 1	Step 1	Identification	Political engagement	.761	.759	.576
	Step 2	Identification	Political engagement	.724	.721	.579
		Fair treatment		.070	.071
		Exposure		−.02	−.022
Regression 2		Fair treatment	Identification	.573	.596	.405
		Exposure		.050	.069

A third model was suggested (Figure 2). In this model, exposure acts as a predictor of fair treatment, which in turn predicts identification, which in turn predicts political engagement. If this model holds true we would expect to find (a) a significant relationship between identification and engagement, (b) a significant relationship between identification and engagement in the context of a non-significant relationship between fair treatment and engagement, (c) a significant relationship between identification and engagement in the context of non-significant relationships between fair treatment and engagement and exposure and engagement, (d) a significant relationship between fair treatment and identification in the context of a non-significant relationship between exposure and identification, and (e) a significant relationship between exposure and fair treatment. As can be seen in Table IV, analysis using hierarchical regression showed that all these relationships held as predicted, such that (a) identification predicted engagement (b_identity = .761, p < .05), (b) identification predicted engagement (b_identity = .721, p < .05), in the context of a non-significant relationship between fair treatment and engagement (b_{fair treatment} = .06, p > .05), (c) identification predicted engagement (b_identity = .724, p < .05), in the context of non-significant relationships between fair treatment and engagement (b_{fair treatment} = .07, p > .05), and exposure and engagement (b_exposure = −.02, p > .05), (d) a significant relationship was found to exist between fair treatment and identification (b_{fair treatment} = .573, p < .05), in the context of a non-significant relationship between exposure and identification (b_exposure = .05, p > .05), and (e) a significant relationship was found to exist between exposure and fair treatment (b_exposure = .413, p < .05). This pattern of findings suggests that Model 3 is the best one to describe the relationship between the three predictor variables and the dependent measure political engagement.

Figure 2. Model 3: proposed relationship between exposure to information, identification, fair treatment and political engagement

Table IV. Model 3: Regression testing the relationship between exposure to information, identification, fair treatment and political engagement

		IV	DV	ß	Standardised β	R^2
Regression 1	Step 1	Identification	Political engagement	.761	.759	.576
	Step 2	Identification	Political engagement	.721	.719	.578
		Fair treatment		.060	.061
	Step 3	Identification	Political engagement	.724	.721	.579
		Fair treatment		.070	−.022
		Exposure		−.02	−.022
Regression 2		Fair treatment	Identification	.573	.596	.405
		Exposure		.050	.069
Regression 3		Exposure	Fair treatment	.413	.537	.289

Discussion

First, we hypothesised that participation in the deliberative poll would be associated with a shift in attitudes. Data analysis showed that participants' attitudes towards a bill of rights did change following participation in the poll. On average, those who participated in the deliberative poll shifted their positions to become less in favour of a bill of rights following participation. The next step was to investigate the processes underlying the development of particular attitudinal positions in relation to the bill of rights. We suggested two hypotheses to account for attitude shift during participation and commitment to beliefs following participation. Based on the cognitive information processing approach to deliberative polls, (Ryan, 2002), we suggested that mere perceived exposure to a wide range of opinions could play a role in producing attitude change. Alternatively, based on the self-categorisation approach (Turner et al., 1987), we suggested that a tendency to identify as a participant in the poll could play a role in producing attitude change. However, neither exposure to information nor identification as a participant predicted attitude change or commitment to one's position. This means that the predictions of both the information processing approach and the self-categorisation approach were not supported for these measures (Footnote: in terms of attitude change and social influence identification with the “information sources” may have been a more precise measure). Similarly, we did not find a role for perceptions of fair treatment as a predictor of opinions about the bill of rights.

Second, we drew hypotheses relating to the development of feelings of political engagement following participation. On the one hand, the cognitive information processing approach would predict that mere exposure to a wide range of opinions (“becoming informed”) should be most strongly related to political engagement following participation in the poll. Alternatively, the self-categorisation approach would predict that a tendency to self-categorise as a member of a relevant group or community (identification as a representative citizen) should be most strongly related to political engagement. In addition, consistent with the group value model (Tyler, 1989), we hypothesised that perceptions of fair treatment should be related to the development of a sense of identification. The results showed that exposure to information, identification and perceptions of fair treatment did play a role in predicting people's levels of engagement after the poll. Several predictive models were tested. We found that the best model was one in which (a) exposure to information predicted perceptions of fair treatment, (b) fair treatment predicted identification and (c) identification predicted political engagement. In this model perceptions of fair treatment mediated the relationship between exposure to information and identification, and identification mediated the relationship between fair treatment and political engagement. This means that exposure to information made people more likely to identify as a representative citizen to the extent that exposure was seen as part of an environment of fair treatment (in which participants were treated with respect and had the chance to have their say). It also means that fair treatment made people more likely to feel politically engaged to the extent that they also came to identify as representative citizens.

Taken together, these results suggest that people are not more likely to change their minds when they have the experience of being exposed to both sides of an argument (in fact, the correlation between exposure to information and attitude change was negative, suggesting that people who felt exposed to more views were more likely to stick with their original opinion). However, exposure to information was a factor in affecting people's sense of political engagement. Therefore, we found that the experience of being exposed to multiple sides of an argument can influence responses to participation in a deliberative poll, but not in the way that the cognitive information processing approach predicts.

A purely cognitive approach to influence suggests that exposure to information works due to its capacity to fill in informational blanks and allow deeper thought processing. However, our results suggest that exposure to information does not work its influence directly in this fashion but instead appears to do it as a function of its capacity to make people feel included and treated with respect. Consistent with the group value model (Tyler, 1989), this experience of fair treatment, in turn, can increase the likelihood that people will feel a stronger sense of belongingness to the social group involved. A key feature of this process is the opportunity that poll participation gives for people to see themselves in a new way, as part of a new collective (of politically engaged people). Therefore, to the degree that participation in a poll increases levels of political engagement, it appears to do so principally through its capacities to (a) give people an opportunity to listen to and give voice to their views and (b) create social ties with an ingroup for which “political engagement” is a normative position. In short, we found that the poll's climate of information exchange and respect helped to increase people's sense of themselves as part of a new social entity: the representative group of “bill of rights” poll participants. To the extent that people felt this new sense of self, they also became more interested in thinking and talking about the rights issue.

Therefore, this investigation showed that exposure to information, fair treatment and social identification can all play a role in making people feel engaged following participation in a deliberative poll. These factors seem to operate together in a way that is most consistent with a self-categorisation theory and group value model approach. Findings suggest that people can come to feel more politically engaged when they are treated with respect and given opportunities to discuss issues, ask questions and to air their views in collaboration with other members of a relevant community. A key factor in the success of the deliberative poll process as a tool is its capacity to provide this type of self-categorisation through the social environment it creates.

Acknowledgements

We would like to thank Pam Ryan for her input into this project, Issues deliberation Australia, and all those who participated in this research.

Notes

^*Accepted under the previous Editorial Board.