TOWARD INFORMED OPINIONS BY THE MEANS OF DELIBERATION Kasper M. Hansen1 University of Southern Denmark
Today public opinion polls are omnipresent. An unexpected alliance between democratic theory and opinion research has introduced alternative methods to gain insight into public opinion. There are two general problems with the existing opinion polls. First, seldom do the citizens have salient political opinions, thus the responses are quite random. Secondly, social choice theory has taught us that the winning alternative when aggregating individuals’ choices depends on the methods of aggregation. The solution to both problems is to give the citizens the opportunity to take part in deliberation on the issues before the final opinions are expressed. Comparing opinions prior and after deliberation shows that post-deliberative opinions are more embedded and salient than pre-deliberative. Furthermore, the collective choice is not affected in the same degree by the choice of aggregation method. The largest effect of deliberation is on the individuals’ opinions, while the collective opinions are more robust to deliberation. Key words: Deliberative Poll, opinion consistency, opinion transitively, single-peakedness, cycles, deliberation, deliberative democracy, nonattitudes.
1. Introduction The commercial public opinion agencies are standing in line to measure our top-of-the-head political opinions and the media are the uncritical bullhorn of these numerous polls. Over time, the number of political public opinion polls has increased. Despite the increasing use of the public opinion polls, its use and validity are seldom questioned. Previous some researchers (e.g. Key 1961) pointed to the many problems of measuring public opinions as nothing else but what happen to be salient in our minds during dinner when telephone polls often are conducted. Today much of the research in public opinion is concerned with the extent to which the public opinion polls correctly predict the election results. Research has shown that there is variation in how the polls are conducted as well as there ability to predict the election result correctly (Thomsen 1998; 2002). The degree to which the public opinions polls predict the election results is a relatively objective comparison of the quality of the public opinion polls, but when the polls are concerned with political issues which are not up for election or election will not held for some time this comparison disappears. In addition, such a comparison would neglect that there are also measurement errors in relation to the individual answers the participants give in the polls. Maybe the participants have not thought about the issue before they are confronted with the question and therefore do not possess any real opinion on the issue. The 1
University of Southern Denmark, Department of Political Science, Campusvej 55, DK-5230 Odense M, Denmark, E-mail:
[email protected]: www.sam.sdu.dk/staff/kmh. Many thanks to Peter Kurrild-Klitgaard and Mickael Bech for discussions and comments on the paper.
1
measurement error in this case is in addition to the statistical error, which only relates to the random sampling because all answers are taken as an expression of a real opinion.2 Thus, two traditions are related to measuring public opinion. The first tradition is represented with Thompsen’s research. This research focuses on the tool for measuring public opinion – the public opinion polls, and the statistically error related to sampling and size of the sample. The second tradition focuses on the answers the participants give in the polls and discuss the extent to which these opinions are stable and express a real opinion. This paper focuses on the second tradition, that is to what extent do the opinions express real opinions, and how can the measurement error relation to the individual answers be reduced? In this case, the solution applied is to give the participants access to information on the issue and give them the opportunity to deliberate with each other, experts and politicians before giving their final answers. In the next section, some general arguments from the literature on opinion formation and public opinion polling are discussed. Especially the American research tradition is discussed. The following section presents some results from four Regional Hearings, which the Danish Regions3 conducted in the fall of 2003. The issue was the future models for the regional structure. These hearings gathered 168 citizens for a five-hour deliberation with each other, experts and politicians. At the recruitment and at the end of the hearings the participants were asked to rank three different models for the future regional structure in Denmark4. The Regional Hearings provided the opportunity to compare the citizens’ opinion before and after deliberation. The analyses of the pre-deliberative and post-deliberative opinions are analyzed by using two approaches – the effect on the individual opinions and the effect on the collective opinion. The paper ends with a short conclusion bringing the two approaches together. 2. The American public opinion tradition In 1936 George Gallup promised the Washington Post that he would be able to predict the presidential election better than the magazine Literary Digest, simply by asking a few thousand citizens compared to Literary Digest’s method by sending out around 10 millions questionnaires to people with telephone or car. If Gallup did not succeed, the deal was that he would pay for the survey himself. Gallup’s survey predicted that F.D. Roosevelt would win a landslide victory. Literary Digest predicted a landslide victory to Landon. Roosevelt was reelected with a convincing majority (Moon 1999; Moore 1992; Gallup & Rae 1940; Fishkin 1997). With Gallup, representative sampling was born and since then public opinion polls have been an increasing part of the daily news flow. Gallup believed that his public opinion polls were a method, which reflected already existing opinions among the public. A view essentially claiming that the opinions 2 3
4
Some polling companies are starting to use different filters such as political awareness and political interest in order to filter out nonattitudes from the samples. The Danish Regions is an interest organization representing the interests of all 14 Danish counties/regions and providing them with information and services. The Danish counties have the ability to set an income tax. The tax rate varies between the counties from 10.9% to 12%. The counties have directly elected councils and a mayor is elected within the council by the council. The council’s primary responsibilities are hospitals and the public health insurance (amount to 80% of the budget), but also some roads, public transport, regional development, environmental protection and high school education are responsibilities of the counties. The issue of the future regional structure in Denmark is most real and relevant. At the time of writing, the Danish government is deciding which of the models will be implemented. The three models for a new regional structure has been part of the ongoing debate in the media for some time.
2
existed and were salient in the minds of the citizens dominated the field until the beginning of the 1960’s when some critical argument to public opinion surfaced. V.O. Key (1961) argued that public opinion is nothing but an echo of the elites’ discourses communicated to the public in sound bits by the media – an interpretation which seems even more appealing today with the fast news flow. Another influential critique was the American political scientist Philip E. Converse (1964; 1970). Converse concludes on empirical findings that the public’s opinions are “extremely labile for individuals over time” (Converse 1964:241) and that “large portions of an electorate do not have meaningful beliefs, even on issues that have formed the basis for intense political controversy among elites for substantial periods of time” (Converse 1964:245). The conclusion was that large parts of the public do not have any real political opinions and their response is only ”non-attitudes”. That is, opinions which are not embedded in the mind of the respondent and are thus only random responses to the questions posed by the interviewer. Empirically, the concept of non-attitude suggests that the changes in the public’s opinions will largely fit a random pattern. (Converse 1964; 1970; Hill & Kriesi 2001; Hansen 2004:198). If large parts of the public only have nonattitudes, it has many democratic consequences. On the one hand, if you accept Converse’s conclusion why and how can decisions-makers be responsive toward the public’s wishes and why conduct public opinion polls altogether? On the other hand, the nonattitude interpretation of the instability has been challenged. Achen (1975) argues that what on the surface may seem to be nonattitude is in reality measurement errors such as vague wording, order of questions, interviewer bias, scaling errors, context in which the questions are asked, etc. and not vague opinions. Accordingly, measuring errors can also simply be that the participants misunderstand the questions or that the interviewer misunderstands the answers or codes the answers incorrectly. The approach blames so to speak the tool of opinion polls rather than the public. Various researchers give some empirical support to this interpretation (Achen 1975; Smith 1984; 1994; Zaller 1992; Inglehart 1990). Whether the instability is caused by the public’s lack of real opinions or the tools we use to measure public opinion have to some extend dichotomized the American public opinion research (Kinder & Palfrey 1993; Kinder 1998; Kinder & Sears 1985). Recently, the debate has been revitalized by yet another interpretation of the instability (Zaller, 1992). The approach suggests that instability is not due to citizens having any opinions, but that they have too many opinions and that their opinions are multidimensional and as such a more complex phenomenon. Accordingly, one opinion may lead to several different answers, which do not indicate instability, but rather that the opinions are complex. Citizens experience opinion-ambivalence rather than nonattitudes and the individual’s answers reflect what happens to be most salient at the time of the interview. What happens to be salient at a given point depends on the general level of the elites’ discourses, media coverage and on the individual level of political awareness, exposure and access to information, political knowledge, and political interest (Zaller 1992; 1994; Hansen 2004:198). Zaller’s (1992; 1994) conclusions create new dynamics in the field of public opinion research. On the one hand, political behaviorists and political psychologists conducted new empirical analyses such as split-sample design and more advanced panel design in their attempt to gain insight into why the public opinion seems so instable over time (e.g., Sniderman 2001; Johnston et al. 1992). On the other side, the dynamics came from a rather unexpected research tradition – democratic theory. Some political theories argued that there was a strong need to combine empirical research and political theory. Only through a combination of the two, it would be possible to move beyond the somewhat entrenched discussion between liberal and republican interpretation of democracy (Held 1996; Rothstein 2002; Smith 2001; Hansen
3
2004:140). With a focus on deliberative democracy, an attempt was made to bridge the gab between the liberal and the republican interpretation of democracy The focus of deliberative democracy is deliberation, which can be defined as an unconstrained exchange of arguments that involves practical reasoning and potentially leads to a transformation of preferences (Hansen 2004:98). The concept of deliberation is thus that the citizen must reflect upon his/her own and other opinions. Such a process broadens your mind and potentially causes opinions to change. In this way, deliberation includes an informative and educational element as well as a potentially transformative element in relation to your opinions (Hansen 2004). James S. Fishkin (1991; 1997) is one of the democratic thinkers, who took the abstract words of finding a symbiosis between democratic theory and empirical research serious. He argued for combining a liberal democratic ideal, with focus on individual rights, with a republican democratic ideal, with focus on citizenship and democratic participation and deliberation. Supported by the fast growing debate on deliberative democracy (see Hansen 2004) and the need to develop the tool of public opinion polls, Fishkin developed the idea of Deliberative Polling. The idea got some attention among democratic theorists as well as public opinion researchers to gain insight into the citizens’ reflective and informed opinion. The idea behind the Deliberative Poll is to create an arena for political participation and deliberation, which tries to create an atmosphere, which stimulates a deliberative process. The Deliberative Poll’s fundamental idea is to invite a representative sample of the citizens together and let them deliberate with each other, experts and politicians on the basis of balanced information. Prior to and after the event the participants are asked a number of identical questions. These questions provide the opportunity to compare the citizens top-ofthe-head opinions polled before the event with the opinions polled after the participants have had the opportunity to deliberate with each other, experts and politicians and in this way have had the opportunity to form an informed opinion. The pre-deliberative opinion is a picture of what traditional representative public opinion polls provide, whereas the post-deliberative opinion is a picture of what the citizens’ opinions would have been if the public had had the same opportunity to engage in the same kind of deliberative process before they had to take a position on the questions (Fishkin 1997; Hansen 2004). The empirical material in this paper – the Regional Hearings – resembles a Deliberative Poll, where the citizens’ opinions are polled before and after the deliberation. The hearings have multiple purposes depending on the view. On the one hand, it had a normative democratic purpose. That is creating an arena for public involvement and political deliberation on an important political issue. Another purpose was to provide the media and decisions-makers with insight into what the opinion of the citizens would be if they were provided with the opportunity to engage in deliberation before deciding the possible path for regional development in Denmark. On the other hand, the event had an experimental purpose in order to create insight into the effect of deliberation on a representative sample of the public. It is the latter which is in focus in this paper. 3. The Regional Hearings In the fall of 2003, a Danish interest organization composed of all 14 Danish counties – Danish Regions - conducted four regional hearings in Denmark – the Regional Hearings – on the future of the regional/county structure. The aim of the Regional Hearings was to bring together a representative sample of citizens for five hours of intensive deliberation with each other, experts, and politicians. The participants’ opinions were measured before and at the end of the five hours of debate with the purpose of creating insight into the top-of-the-head opinions as well as the
4
opinions after the issue had been going through intense deliberation.5 The five hours of debate was conducted in small groups of 6-12 participants lead by a neutral moderator and in plenary sessions. The debate was very intense and most of the participants took active part in the deliberation in the small groups as well as during plenary sessions. The participants were introduced to three different models for a future regional structure: the Municipality model, the State model, and the Regional model. These models were briefly described in the questionnaires. The municipality model was described as: ”The counties responsibilities will be transferred to joined municipal companies administered by politicians elected at the municipality level”. The state model was described as: ”The counties responsibilities will be transferred to the Danish parliament and the state”. The regional model was described as: ”The counties responsibilities will be transferred to few and larger regions than today”. These three models were the focus of the debate during the event, not to mention the debate in the media before and after the event. The goal was to bring together 100 representative citizens to each of the four Regional Hearings, but only the half showed up. In this way the citizens who showed up was not representative of the population. Women and citizens between 18 and 30 years old were underrepresented. The lack of representativeness is the main reason why the Regional Hearings cannot be defined as a Deliberative Poll, because representativeness is needed in that case (Hansen 2004). The process of the recruitment seems to a large extends to be based on self-selection, which has increased the bias between the citizens and the public at large. The relative few participants and their bias to the general public do not allow generalizing the findings of the Regional Hearings to the public at large. Nevertheless, one important consequence of the large self-selection is that the citizens participating in the Regional Hearings are more political aware, more interested in the issue and thus have a more embedded opinion than a more representative sample would have had. It is likely that a more representative group would have had an even larger level of opinion change and weaker opinion consistency than has been observed among the participants at the Regional Hearings. Nevertheless, the opinion representativeness is easier achieved than sociodemographics representativeness. Even though before the deliberation the participants were more positive about the regional model than non-participants6, and non-participants were relatively more positive about the municipality model, the Regional Hearings succeed in bringing a group of citizens together to an intensive deliberation where a broad spectrum of opinions where present and express in an open debate. The focus of this paper is on the effect of the five hours of deliberation on the citizens’ opinions about the future regional structure in Denmark. Through telephone interviews and questionnaires pre- and post-deliberation, the participants were asked to rank the three models (Municipality, State, and Regional model) on four different dimensions.7 These dimensions are in the first column of table 1.
5
6
7
Hansen (2004) discusses in details the general problem with control of the stimulus in this kind of quasiexperiments. General speaking the stimulus is defined as everything after recruitment and to the end of the event. Thus, it is likely that the participants will start to prepare themselves to engage in deliberation before the event, e.g. following the news careful and discussing the issue with friends and family. Non-participants are those respondents who took part in the recruitment by phone and thus answered the questions of the first round of questions but either declined to participate or simply did not show up at the Regional Hearings. See Hansen, 2003 for a detailed analysis on the biases between the participants and the public at large. Ranking questions are seldom used in opinion polls because the respondents often have some difficulties conduction a complete ranking. In addition, it takes longer interview time than the traditional Likert-scale questions. This despite the fact that ranking questions provide more information than traditional Likert-scaled question as ranking forces the respondent to prioritize the alternatives.
5
4. The quality of pre- and post-deliberative opinion From a deliberative democratic perspective, it would be expected that deliberation would improve the quality of the individual citizen’s opinion by increasing the citizen’s ability to make an informed choice. Table 1 confirms this expectation, as more participants are able to choose between the three models after deliberation compared to prior the deliberation. Table 1: Share of participants with a first priority and complete ranking between the three models. Pre- and post-deliberative opinion (in percent) Have a first priority Complete ranking Dimension of the three models PrePostPrePostdeliberation deliberation deliberation deliberation 93 96 86 72** • Best to take decisions • Best opportunity for the 88 98** 80 78 citizens to influence the political decisions 84 93** 77 80 • Best to secure citizens’ rights • Best to prioritize the 86 96** 82 83 hospitals Note: N=168, ** significant difference between pre- and post-deliberation p
