Full text of DOI:10.1093/pa/53.4.657

Deliberative Polling and Public Consultation
BY JAMES S. FISHKIN, ROBERT C. LUSKIN AND ROGER JOWELL
At the heart of any notion of democracy is some element of public input
into the policy-making process. Inputs can take the form of direct
democracy or parliamentary democracy or both. Parliamentary democ-
racy is a form of representative democracy, in which the electorate
determines who its representatives are; the majority party or coalition
forms the government; and the government decides public policies.
Historically, British democracy has been representative democracy.
Direct democracy is manifest in a referendum or an initiative, in which
electors vote directly for or against adopting a policy. Institutions of
direct democracy are prominent in Switzerland and in American states
such as California, but they have been very unusual in the past in
Britain. Tony Blair has made the promotion of more popular inputs in
public policy a leading feature of his approach to government, for
example endorsing the idea of national referendums on electoral systems
and on joining the European Monetary Union.
But the distribution of votes in elections and referendums is not
necessarily a good indicator of public preferences. One obvious source
of error is low turnout: voters and non-voters may differ significantly
in their preferences. A more fundamental difficulty is the amount of
knowledge and thought underlying the votes that are cast. Substantial
scholarly research has produced an overwhelming consensus on the
obvious: most people know very little and have thought very little about
most policy issues.1 While there is disagreement about how much the
lack of information and interest affects people’s views, it is possible that
voting preferences would be noticeably different if everyone was more
knowledgeable about, attentive to and reflective about the issues
involved. Even if every voter were fully informed and deliberated about
major public policies, there is the possibility that no party reflects all
their views. In a two-party system in which a red and a blue party
compete for control of government, a voter who prefers yellow may be
unrepresented. Furthermore, when several issues are important, a voter
may prefer the policy of different parties on different issues.
Public opinion polls have become a kind of advisory input to policy-
making between elections, but much the same objections can be made
to the results of public opinion polls highlighted in the media. There
can be difficulties in recruiting a representative sample because some
groups are hard to locate or unwilling to be interviewed, as in the
British general election of 1992 when the opinion polls unanimously
Ą Hansard Society for Parliamentary Government 2000
Parliamentary Affairs (2000), 53, 657–666

658
Parliamentary Affairs
predicted a Labour victory. At least a part of their error (and some
scholars say most) was attributable to unrepresentative samples.2 Some
opinions are difficult to measure, especially if people hold conflicting
views that lead them to vary their answers according to how a question
is worded. The fundamental problem with conventional public opinion
polls concerns the quality of the responses.
Respondents in public opinion polls are often asked to give off-the-
cuff responses to questions of public policy that are usually of limited
interest to most voters. Whether to oblige the interviewer or to avoid
appearing ignorant, they usually comply, even when they are offered
the chance to say that they ‘don’t know’ or ‘haven’t thought much
about the question’. Thus, large proportions of random samples of the
American public supported or opposed the non-existent Public Affairs
Act of 1975—and, more recently, the equally non-existent legislation
to repeal it. Responses with little or no prior information or commit-
ment are what Philip Converse famously termed ‘non-attitudes’.3
Responses manufactured on the spot are not necessarily what respon-
dents would say in answer to the same questions if they had had some
information and time to think or discuss with others what was involved.
The interaction between an interviewer and a respondent is not a
discussion, for interviewers are explicitly instructed not to provide
information or venture their own ideas or opinions when asking respon-
dents to give theirs. While this avoids the risk of interviewers influencing
respondents, the situation is politically unrealistic: just as policy-makers
only take decisions after endless discussions and exchanges of opinions,
so individuals arrive at firm opinions by engaging in conversation about
them with friends, family, people at work and so on.
On most issues, a relatively limited portion of the public has well-
formed policy preferences; most people do not pay enough attention to
current affairs to have solidly based opinions about most major issues.
They are, in Anthony Downs’s well-known phrase, ‘rationally ignor-
ant’.4 In a country with tens of millions of electors, such as Britain or
the United States, no one vote is likely to affect the outcome of any
election or referendum. This is a strong disincentive to learn or think
much about politics and policy. Why spend a lot of time and effort
becoming more informed about the merits of competing policy
alternatives?
The deficit of citizens is not just a matter of factual information or of
time to think about issues. It is also a matter of deliberation, that is
discussing issues with others with different experiences, holding differ-
ent views and representing varied and sometimes conflicting interests.
Ordinary people simply do not discuss politics a great deal, and when
they do, it is chiefly with others like themselves and of a similar cast of
mind, for everyone moves in limited social circles and selects conversa-
tional partners partly on the grounds that their friends and companions
think as they do.

Deliberative Polls
659
How much does this matter? Evidence from various experiments and
statistical simulations with survey data indicates that increasing the
deliberation and information of citizens does have some influence on
policy preferences and votes.5 There is a gap between what citizens
initially say they would prefer and, given more information and the
opportunity to reflect, what they then think. This gap must trouble
anyone concerned with democracy. The spirit of democracy requires
that government reflects and responds to the public’s opinions—but
less so when these are ‘non-attitudes’ or unconsidered opinions instead
of statement by people who have become informed and thought seri-
ously about the issues. Until recently, politicians have had to rely on
snap opinion polls or anecdotal examples of often uninformed voters,
as this has been the only available direct expressions of the public’s
views. Not only has this strengthened the influence of elected represen-
tatives and parties in policy-making but also that of non-elected gate-
keepers who organise the discussion of public policy in television and
in the press.
Frustration with inadequate methods of assessing public opinion has
led researchers around the world to devise fresh means of eliciting
evidence of more informed and reflective public opinion. Focus Groups
are the longest-established method of discovering how voters discuss
issues and see politics from the perspective of very limited information
and reflection. The Choice Questionnaire developed by Peter Neijens,
Wilem Saris and their colleagues in the Netherlands and imported by
Hanspeter Kriesi and Danielle Bu¨tschi to Switzerland introduces rele-
vant information about policies while conducting a survey interview.6
The Televote developed by Ted Becker and Christa Daryl Slaton
interviews a random sample, tells the respondents they will be re-
interviewed and sends them briefing materials to be read between
interview and reinterview.7 Another approach is to gather small groups
of citizens together to discuss policy issues over periods of a few days.
The Citizen Juries pioneered by the Jefferson Center in Minnesota and
introduced in Britain by the Institute for Public Policy Research involves
small group deliberations and opportunities to question expert ‘wit-
nesses’.8 Planning Cells, developed by Pieter Dienel and his colleagues
in Germany, not only provide information but also ‘coax participants
toward processing it on rational lines’.9 Consensus Conferences, devel-
oped in Denmark and brought by the Science Museum to the UK, bring
small (and sometimes large) groups together to discuss and try to reach
a set of agreed conclusions about policy issues.10
We have developed Deliberative Polling to meet the simultaneous
requirements of providing voters with information about public poli-
cies, giving them also opportunities to discuss the information with
people of diverse views, and measuring the resultant change.11
A
deliberative poll begins by interviewing a random sample of the popu-
lation. Following the interview, respondents are invited to come to a

660
Parliamentary Affairs
conference centre for a weekend of face-to-face discussions and are sent
a balanced package of background information to encourage learning
and thinking about an issue. These materials are made available to the
general public and are often certified by an advisory board for balance
and accuracy. During the weekend, discussions are held in randomly
assigned small groups chaired by an impartial moderator, and they
alternate with question-and-answer plenary sessions with experts and
policy-makers on the issue. After a weekend of information and discus-
sion, in both small groups and plenary sessions, participants complete
the same questionnaire as when first contacted. In this way, a delibera-
tive poll can measure the extent to which opinions change as a result of
exposure to information and discussion. Changes in opinion have often
been substantial. Whereas the ‘before’ results are just like those from
any other random survey of public opinion, the ‘after’ figures represent
what public opinion would be like if everyone had an opportunity to
become informed by experts on all sides and to reflect on issues by
discussing them with people of diverse views.
We see deliberative polling as offering various significant advantages
when compared to other forms of public consultation. First of all, it
provides opportunities for extended discussion and deliberation. It is
difficult to process new information about complex issues during the
course of an interview, as in the choice questionnaire. Without discus-
sion, it is also difficult to gain a full appreciation of the competing
arguments or the circumstances and interests behind them. In asking
respondents to read the materials and discuss the issues with friends
and family, the televote, seeks to generate some deliberation, but there
is no assurance that they will do so and even if they do, the conversa-
tions will tend to be with people with similar backgrounds and views.
In contrast, in deliberative polling each participant spends a great deal
of time talking to others of diverse views. Not only do moderators
ensure consideration of all the major arguments for and against major
policy options, but random assignment to small groups produces discus-
sion among people who think and vote differently and would not
normally be exposed to one another.
The representativeness of those involved in deliberative polling is
another important consideration. Without a random sample of hun-
dreds of people there can be no assurance that the results of asking
people to discuss questions bear any resemblance to what would be
obtained if the whole population could be given the same experience.
Other methods of encouraging deliberation, such as focus groups,
citizen juries, consensus conferences and planning cells fall particularly
short on this point. Citizen juries are usually not randomly selected and
in any case normally consist of a dozen or so people, far too few to
reflect a random cross-section of the population. Whatever their num-
ber, the participants in consensus conferences are substantially self-
selected, typically from advertisements placed in newspapers. Planning

Deliberative Polls
661
cells employ random samples from a number of localities, but pooling
the groups does not create a national sample. Moreover, variations in
activities from cell to cell raise questions about whether those involved
are being exposed to the same variety of views.
In deliberative polling, samples are relatively large and random. The
initial survey normally interviews about a thousand people and the
participants in the deliberative weekend number several hundred. The
overall response rate in the initial interview is similar to that in the best
public opinion surveys and the samples are representative of the popu-
lation as a whole. Although there are occasional statistically significant
differences between the characteristics of the initial interviewees and
those participating in subsequent deliberative discussions, these have
been relatively few and modest. The participant samples have repre-
sented both the initial survey sample (the ‘before’ measure) and the
general population remarkably well.
The purpose of deliberative polls is also distinct. consensus confer-
ences, as the name implies, citizen juries and planning cells all tend to
seek a collective, consensual decision among participants. However,
induced consensus not only masks differences of opinion, which are
normal in policy-orientated discussions; it also opens the door to
excessive influence by ‘opinion leaders’. In contrast, deliberative polling
seeks neither. The moderators who conduct the small group discussions
are trained to make sure that nobody feels under pressure to agree with
anyone else and the groups are under no pressure to reach a consensus.
The aim is for each participant to become better informed and refine
their individual views, whether in opposition to or in support of
preponderant opinion. Our interest is not in reaching an agreed verdict
but in measuring opinions—and opinion change—by pre- and post-
deliberation confidential questionnaires. The distinctive result of a
deliberative poll is the evidence it reveals of the scale and direction of
change in opinions before and after exposure to information and
alternative opinions. The poll also shows who changes, in what ways
and by how much. Deliberative polling is not like a jury or committee
meeting, seeking agreement about a decision; it is more like an idealised
version of an election campaign where opinions and behaviour at the
end of a campaign are likely to be more informed and, as a result, to
differ from what they were at the start.
The late eighteenth century American Constitution starts with the
words ‘We, the people’, and the political values underlying many
different methods of direct democracy flourished first in the United
States. Contemporary American campaign practices are often drawn on
for lessons in conducting campaigns in Britain and in other established
democracies. Deliberative polling has much more cosmopolitan origins,
for it was heavily influenced by discussions between James Fishkin, an
American, and Peter Laslett, a Cambridge political theorist, and the
first poll was undertaken in Britain in 1994. Of the total of 16 polls,

662
Parliamentary Affairs
five national Polls have been undertaken in Britain funded by Channel
Four Television, with the help of Granada Television and the National
Centre for Social Research, and each has been nationally televised. The
topics have been how to reduce and deal with crime (1994), Britain’s
future in Europe (1995), the monarchy (1996), the economic issues in
the General Election (1997), and the future of the NHS (1998). On the
eve of the United States presidential nomination campaigns of 1996, we
conducted a nationally televised poll about family policy, the future of
the national economy and America’s role in the world. The Poll,
conducted with the support of the Public Broadcasting Service (PBS),
the National Opinion Research Center (NORC), and the University of
Texas, involved televised dialogues with a panel of Republican presiden-
tial candidates, as well as with Vice-President Gore. In addition, we
have conducted nine local or regional deliberative polls, eight of them
on electric utility matters in Texas and neighbouring states. Lastly, we
collaborated in a nationally televised poll in Australia before the 1999
referendum on whether or not Australia should cease to have a monarch
and become a republic. Conducted with Issues Deliberation Australia,
The Australian Broadcasting Corporation, the Nine Network’s Sixty
Minutes programme, the Australian newspaper and the Australian
National University, it reached a large portion of the Australian public,
providing considered judgments of an informed microcosm before
people voted on a complex constitutional issue.
Does information and deliberation make much difference? Every
deliberative poll to date has yielded sizeable and statistically significant
changes of opinion on a large proportion of the policy items covered.
The greatest opportunity for deliberation arises when laws require that
public consultation be undertaken before a binding government decision
is taken. In Texas, a statutory requirement for public consultation was
introduced for energy resource decisions by regulated electric utilities.
When Converse wrote about ‘non-attitudes’ in the early 1960s, one of
his key examples of pseudo-opinions concerned attitudes toward the
government’s role in electric power. Now, nearly forty years later, we
have been able to discover not only what more informed opinions on
such questions would be like, but to insert them into the policy-making
process, having conducted eight polls on energy issues in Texas and
neighbouring states.
Because most people have little information about the relationship
between conservation issues and energy use, the response to a one-shot
survey of public opinion is likely to be heavily influenced by the questions
asked. If questions focus on environmental protection, a non-economic
good, then a majority of people may be in favour. If they focus on the
desirability of keeping energy prices down, this may produce results
unfavourable to environmentalists. A deliberative poll gives people
information about the costs and benefits of both environmental protec-
tion and cheap energy—and exposes individuals to arguments for giving

Deliberative Polls
663
each priority. The results of polling in Texas in 1996 show the major
changes in opinion of Texans before participation in a weekend deliber-
ation on the issues. Before becoming informed, when people were asked
if they were prepared to pay at least $1 more a month for more
environmentally friendly renewable energy resources, in three polls an
average of 55% said they would do so. After participation in a weekend
discussion, an average of 88% said they would. Moreover, the propor-
tion giving first priority to energy conservation rose an average of 31%,
focusing attention on reducing energy demands rather than on choices
between alternative sources of energy. These polls registered far greater
public interest in and sensitivity towards environmental concerns than
policymakers had expected, and have led to the largest-ever investments
by Texas in renewable energy (such as wind and solar power), conserva-
tion subsidies for low income consumers, and, most recently, lower
prices for customers buying renewable energy.
The Australian referendum on the monarchy posed a single question,
asking people whether or not they wished to replace a Governor
General representing the Queen with a President chosen by Parliament.
Symbolically, the issue was whether or not Australia should become a
republic, an important issue in a land where the population increasingly
has no British ancestors or only very distant British antecedents. The
method of choosing the President became an issue during the campaign,
because appointment by Parliament appeared ‘undemocratic’ in that it
denied popular choice of the head of state. A representative sample of
347 Australian voters participated in a deliberative poll in Canberra
shortly before the 1999 referendum. During the weekend they were
exposed to the rationale of Parliament appointing a President by a two-
thirds majority (so that the president would not overshadow the legit-
imate authority of a Prime Minister accountable to Parliament) and the
desirability of a head of state who would be above the political fray
and not to have to stand for office. In consequence, the number
favouring a directly elected President plummeted from 50% to 19%.
After listening to advocates of each side, the percentage of participants
favouring a republic with a president chosen by the proposed model
increased from 53% to 73%.
The deliberative poll before the 1997 British general election exposed
participants in the weekend discussion to Liberal Democrat policies
alongside those of the two main parties—in a context where there was
time to reflect and discuss the pros and cons of voting for a party that
receives little media attention during the life of a Parliament. In conse-
quence, the proportion expressing an intention of voting Liberal Dem-
ocrat increased by 22 percentage points, drawn almost equally from the
ranks of previous Conservative and Labour supporters and ‘don’t
knows’. When we checked after the election how the participants had
actually voted in the election, we found that they did indeed vote
according to their new-found preferences.12

664
Parliamentary Affairs
In the deliberative poll we conducted before the start of the American
Presidential primary process in 1996, the preliminary survey found that
most people wanted to get rid of foreign aid. They also had the
impression that foreign aid consumed a major part of the federal
budget, a view that was wildly inaccurate. Participants in the weekend
deliberations about issues were given documentary evidence that by any
measure foreign aid was about one per cent of federal spending. This
was a revelation to many participants, and it increased the proportion
in the sample thinking the current level of spending on foreign aid is
about right from 26% to 41%. Equally important, presidential candi-
dates who were cross-questioned by our sample could not get away
with any of the usual platitudes about such a complicated subject.
Televising substantial parts of all the national deliberative polls at
peak times increases the impact of such polling. Television multiplies
the audience for deliberation, so that millions of viewers acquire more
knowledge about a contentious issue and see that there is more than
one point of view. They are thus stimulated to re-evaluate their previ-
ously held opinions, whether shallow or firm. Because the central figures
in polling are a cross-section of ordinary voters, viewers can identify
with actual participants in the deliberation; they are not simply passive
spectators watching a public argument between candidates or an aggres-
sive interview.
Deliberative polls are intended to complement the familiar institu-
tions of representative democracy rather than supplant them, for there
is always a need for deliberation on major issues by small groups,
whether in Cabinet committees, party caucuses or Parliament. The real
issue is not to deflect policy-makers from being influenced in their
deliberations by what the public thinks but to expose them to what a
more informed state of public opinion would be like.
Today, only a minority of elected office-holders will defend the
traditional idea that they are elected to form their own judgments about
public policy; most proclaim that they are also there to give voice to the
wishes of the people. But the voices that claim to speak for the people
are themselves unrepresentative, such as journalists, editors and propri-
etors of the popular and elite media. Most media-sponsored opinion
polls provide little information about popular attitudes, focusing instead
on the horse race aspect of politics—which party is in the lead?—or
tabulating the current state of “non-attitudes” and snap judgements
about current issues Similarly, reports of public opinion based on chats
with taxi drivers or speedily convened focus groups are, at best,
reflections of uninformed opinions, and probably unrepresentative ones
as well. Although it is axiomatic that snap judgements generally make
for poor policies, that is precisely what opinion polls are designed to
measure.
Democratic governance ought also to reflect what the public would
think if it had the opportunity to reflect on the issues of the moment.

Deliberative Polls
665
The merit of deliberative polling is that it employs social science
methods to produce evidence about informedpublic opinion. Every
aspect of the process is designed to give participants a chance to
consider a topic in a balanced way, discussing information about all
sides of a case in small groups and with balanced panels of experts,
politicians and decision-makers with different points of view, and then
to register their opinions in questionnaires that insulate individuals
from social pressures to conform.
Of course, citizens do not normally have the opportunity to form
their opinions after careful deliberation. Nor are they obliged to do so.
Politicians are elected to do that job for them. But to the extent that
politicians want to take account of the opinions of ordinary citizens in
policy-making, they would be wise to treat ordinary poll data about
most specific policy issues with caution. Snap judgments can quickly
change when an issue becomes more widely debated. And there is little
doubt that democracy would work better if politicians could somehow
be better informed about the likely direction of such changes.
The aim of deliberative polling is to provide inputs into current policy
deliberations. In the absence of such inputs, politicians come up with
decisions that they hope are also consistent with the views of their
constituents, but which may fail to reflect what those constituents
would want if they had a modicum information and the time to make
considered judgments. While organising such a poll costs a substantial
sum of money, the mistakes of government cost even more.
Deliberative polling gives insights into how electors can produce
better-reasoned preferences grounded in evidence about the complexi-
ties of controversial public issues, offering a reliable and valid quanti-
tative representation of majority (and also minority) opinions about
major issues—and how these opinions change after due deliberation.
The method is not suitable for every problem facing government. Crisis
measures require instant decisions. It is especially suitable on issues
about which the public is likely to have little information or when the
choice of a public policy depends on trade-offs between competing
goods. It can help the public articulate considered opinions. As such, it
is both a social science experiment and a form of public education in
the broadest sense.
The Deliberative Polling project has benefited from the efforts of many collaborators. The term itself is a
registered trade mark, and the Center for Deliberative Polling at the University of Texas uses fees from the
trade mark to support research. An earlier version of this article appeared in the UK Ceed Bulletin.
1
See R.C. Luskin, ‘Measuring Political Sophistication’, American Journal of Political Science, November
1987; M.X. Delli Carpini and S. Keeter, What Americans Know about Politics and Why It Matters,
Yale UP, 1996.
2
3
See R. Jowell et al, ‘The 1992 British Election: the Failure of the Polls’, Public Opinion Quarterly 57,
Summer 1993.
See P.E. Converse, ‘Attitudes and Non-Attitudes: Continuation of a Dialogue’ in E.R. Tufte, (ed.), The
Quantitative Analysis of Social Problems, Addison-Wesley, 1970.

666
Parliamentary Affairs
4
5
A. Downs, An Economic Theory of Democracy, Harper and Row, 1956.
In the way of statistical simulation, see L. Bartels, ‘Uninformed Votes: Information Effects in
Presidential Elections’, American Journal of Political Science, February 1996, and Delli Carpini and
Keeter,op. cit.
6
7
8
9
See P. Neijens, The Choice Questionnaire, Free University Press, 1987.
C.D. Slaton, Televote: Expanding Citizen Participation in the Quantum Age, Praeger, 1992.
See A. Coote and J. Lenaghan, Citizens’ Juries: Theory into Practice, IPPR, 1997.
See P.C. Dienel and O. Renn, ‘Planning Cells: A Gate to Fractal Mediation’ in O. Renn, T. Webler and
P. Wiedemann (eds), Fairness and Competence in Citizen Participation: Evaluating Models for
Environmental Discourse, Kluwer, 1995.
10 See ‘Consensus Conferences’ in R.E. Sclove, Democracy and Technology, Guilford Press, 1995.
11 For an account of Deliberative Polling and its rationale see J.S. Fishkin, The Voice of the People, Yale
UP, 1997.
12 See J.S. Fishkin and R.C. Luskin, ‘Bringing Deliberation to the Democratic Dialogue’ in M. McCombs
(ed.), A Poll with a Human Face: The National Issues Convention Experiment in Political Communica-
tion, Lawrence Erlbaum, 1999.