Regulating
election polls
Olof Petersson
Research Director,
Democratic Audit of Sweden, Centre for Business and Policy Studies,
Stockholm
August 2003
The fear of undue
influence on voters is the reason why many democracies have discussed and
quite a few have decided to regulate the publication of election polls. The
controversy around election polling shows that several difficult questions
arise. Do polls really influence voting behaviour? Is a total or partial ban
on election polls compatible with basic democratic principles such as the
freedom of the press? Could a national embargo on the publication of polls
really be effective in a world of Internet and global media?
The impact of
opinion polls: theory and evidence
Early in the
history of mass surveys both scholars and politicians started to worry about
their possible effects. This might be the social science equivalent to the
Heisenberg uncertainty principle: measuring public opinion would change
public opinion. A pioneer team of election researchers found evidence for a
bandwagon effect: electoral behaviour was obviously influenced by
perceptions of the likely winner (Lazarsfeld et al. 1944).
The theoretical
problem was solved by Herbert Simon, who came up with a mathematical
demonstration of how self-correcting polls could be designed (Simon 1954).
As long as the voter reaction function is known the published figures might
be adjusted so that the subsequent voter opinion would match the initially
observed values. But since the exact voter reaction is never known, Simon’s
proof remains elegant but useless.
The measurement of
polling effects is problematic since there are several theoretical ways in
which the publication of a poll can change a voter’s electoral choice. The
famous underdog effect assumes that a political party or candidate gains by
a positive polling trend. But there are other hypothetical effects than this
simple version of the spiral of silence theory (Noelle-Neumann 1982). One
should add that there might also be important indirect effects, linking
polling results with voter reaction via the impact upon party strategies,
media bias and other channels.
This variety of
hypothetical effects, and an abundance of experimental as well as
non-experimental empirical data, have lead some observers to completely
discard the problem of possible effects: ‘As a whole, the effects remain
first of all minimal and secondly they can be seen as completely harmless’ (Donsbach
2001).
Obviously many
different factors determine why and how people vote, and no one would argue
that opinions polls are a major cause. But there is strong evidence that
opinion polls under certain circumstances might in fact influence election
results (Holmberg and Petersson 1980; Petersson and Holmberg 1998). One
important example is proportional election systems with a threshold limit,
such as Germany, Poland, Denmark, and Sweden. Surveys show that tactical
voting, that is whether or not to support a party close to the barrier for
parliamentary representation, is partially based on media reporting of poll
data.
Although there is some empirical support for
polls influencing elections the normative conclusion is not obvious. In an
open society with freedom of speech there are many examples showing how
citizens are influenced by different kinds of information. The publication
of economic forecasts sometimes have a tremendous impact on consumers and
investors. Political opinion data might help voters who want to use their
vote for tactical purposes (Särlvik 1971).
Regulatory policies
A study from 1997, commissioned by
international polling organizations, found that 30 of 78 countries had some
kind of embargo concerning publication on or prior to election days. In 9
cases the embargo applied to the election day only (Røhme 1997).
One of the most restrictive regulation of
polls in a democracy was introduced in France in 1977. The law made it
illegal to ‘publish, disseminate or comment’ on opinion polls during the
week preceding an election. The law did not prohibit actual polling, but
tried to shield voters from knowing the results. The law also contained
rules on the publication of polls between elections. Any publication of a
poll must include information about the identity of the polling
organization, sample size and time of fieldwork.
Over the years the law became more and more
inefficient. Polling before the election continued, but the results were
restricted to political insiders. The general public was not given access to
the polls, except via foreign media and the Internet. The constitutionality
of the law was also questioned. Finally, the law was changed (Law 2002-214).
The embargo was reduced from one week to two days. Polls can now be
published freely, except on election day and the day preceding the election.
The modification of the French legislation
indicates a growing consensus in Europe. Election polls are more frequent
than ever and concerns about their detrimental effects are often voiced.
However, legislation is not seen as the main instrument. Many countries
primarily rely on the self-regulation of media institution and continuing
public debate to raise general awareness about the limitation of opinion
polls.
The Council of Europe, with its 45 member
states, has discussed the need to harmonize national legislation about
election polls, but refrained from doing so. Instead in 1999 The Council of
Europe in issued a recommendation concerning media coverage of election
campaigns (R (99) 15).
The Council of Europe recommendation
underlines both the independence of media and the responsibility of media
professionals. Opinion polls are mentioned in section III.2:
Regulatory or self-regulatory frameworks
should ensure that the media, when disseminating the results of opinion
polls, provide the public with sufficient information to make a judgement on
the value of the polls. Such information could, in particular:
- name the political party or other
organisation or person which commissioned and paid for the poll;
- identify the organisation conducting the
poll and the methodology employed;
- indicate the sample and margin of error
of the poll;
- indicate the date and/or period when the
poll was conducted.
All other matters concerning the way in which
the media present the results of opinion polls should be decided by the
media themselves.
Any restriction by member States forbidding
the publication/broadcasting of opinion polls (on voting intentions) on
voting day or a number of days before the election should comply with
Article 10 of the European Convention on Human Rights,
as interpreted by the European Court of Human Rights.
Similarly, in respect of exit polls, member
States may consider prohibiting reporting by the media on the results of
such polls until all polling stations in the country have closed.
These non-binding rules could be considered as
a set of generally accepted standards in today’s Europe. Legislative
regulation might be accepted to protect the election day itself, but should
otherwise be kept at a minimum.
References
Council of Europe, Recommendation R (99) 15,
http://cm.coe.int/ta/rec/1999/99r15.htm
Donsbach, W., 2001. Election Polls? Normative
and Empirical Arguments for the Freedom of Pre-Election Surveys, The
Foundation for Information, ESOMAR.
Holmberg, S., and Petersson, O., 1980. Inom
felmarginalen. En bok om politiska opinionsundersökningar, Publica,
Stockholm.
Lazarsfeld, P. F., Berelson, B., and Gaudet,
H., 1944. The People's Choice. How the Voter Makes up his Mind in a
Presidential Campaign, Duell, Sloan and Pearce, New York.
Noelle-Neumann, E., 1982. Die Schweigespirale.
Öffentliche Meinung: unsere soziale Haut, Ullstein, Frankfurt a.M. [1984,
The Spiral of Silence: Public Opinion, Our Social Skin, University of
Chicago Press, Chicago.]
Petersson, O., and Holmberg, S., 1998.
Opinionsmätningarna och demokratin, SNS Förlag, Stockholm.
