Predictions of Voting Behaviour from a Pre-Election Survey
| Date | 01 December 1985 |
| Published date | 01 December 1985 |
| DOI | 10.1111/j.1467-9248.1985.tb01586.x |
| Author | Catherine Marsh |
| Subject Matter | Article |
Polilical
Studies
(1985),
XXXIII,
642-648
Predictions of Voting Behaviour from
a
Pre-election Survey
CATHERINE
MARSH*
University
of
Cam bridge
Introduction
Forecasting the results
of
an election on the basis of
a
pre-election poll of
opinion is commonplace. Such polls account for a tiny fraction
of
all surveys
conducted, but are publicized and discussed widely; when they get the result
wrong, this too is highly visible. This research note examines the validity
of
traditional voting indicators. Surveys often overestimate the turnout rate,
either because the questions used to predict turnout are poor, or because of a
defect in sampling. To discover which explanation is more plausible, and to see
if
this could account for failure in predicting correct proportions of party
support,
a
check was conducted on whether respondents and non-respondents
in a single constituency pre-election survey actually voted in the election.
The
Survey
In the June 1983 General Election, staff and students at the Social and Political
Sciences Committee of the University of Cambridge carried out
a
survey
of
electors in the Cambridge constituency. The fieldwork, conducted with local
sixth-form pupils as interviewers, took place on the Sunday prior to the
election, and the resuits were published in the
Cambridge Evening News
on the
Tuesday before the election on Thursday
9
June 1983.
The original sample was
a
systematic random sample drawn from the
electoral roll and covered all wards of the constituency.
As
it was difficult to
know exactly how many volunteer interviewers would turn up at
9
a.m. on
Sunday,
a
much larger sample was issued than was needed. The organizers then
selected one in two polling districts at random to issue to interviewers, thus
introducing a small element
of
clustering. Interviewers were instructed that
once they had first attempted to contact
a
named respondent, they had to call
back up to four times;
if
it
was clear when the questionnaire was returned to the
office that no attempt had been made to contact anybody, the questionnaire
was discarded.
*
The research reported in this paper was conducted with the help
of
students at the Social and
Political Sciences Committee at the University of Cambridge and local sixth-form students
recruited by the
Cambridge Evening News,
all
of
whom took part on a voluntary basis; thanks are
due
to
them. Christopher Castleton, Colin Fraser and Les Goldman also helped substantially with
the project. lvor Crewe, Nick Moon and Norman Webb gave helpful comments on an earlier draft.
0032-3217/85/04/0642-07/$03.00
0
1985
Political
Studies
To continue reading
Request your trial