Why Did the Polls Get it Wrong in 1992?

Opinion polls play a major role in politics, they can be used by the Government
to decide when to call and election, and, among other things, how their pre-
election campaigns are run. Throughout the history of opinion polling, from the
time when polling began to be widely used before an election, in 1945, until
1987, the last general election before 1992, the polls have on average been
correct to within 1.3% of the vote share between the three leading parties, and
the 'other' category (Crewe, 1992, p. 478). This puts all the previous opinion
polls well within the +/-3% margin of error. Because of the past accuracy of
opinion polling, the system has had great credibility and has always been
trusted, both by the public, and political parties. The way polling forecasts
can affect the way people vote is very dramatic, this is because they can be a
'self fulfilling prophecy', in that some voters like to back the 'winning team',
and others only vote for a party they feel has a real chance. This was
demonstrated in 1983, when the Alliance, frustrated with the media concentrating
only on their position in the polls, leaked their own private polls to the press,
resulting in a late surge of support (Crewe, 1992, p.478).

Britain generally has a much greater number of opinion polls carried out than in
other countries, this is due to the large number of national newspapers, and the
amount of current affairs programming on television. The period prior to the
1992 general election saw a much greater intensity of opinion polling than ever
before. During the 29 days between the date of the announcement of the actual
election date, 11th March, and the election date itself, 9th April, there were
a total of no less than 57 national opinion polls.

The 1992 election will always be remembered as the one the pollsters got wrong,
during the lead up to the election, they almost all showed Labour ahead of the
Tories. Of the four polls carried out in the two days prior to the actual
election date, all of them pointed to a hung parliament; one put the
Conservatives 0.5% ahead, one put Labour and the Tories neck and neck, the other
two showed Labour ahead by a narrow margin (Crewe, 1992, p. 8). On the actual
day of the election, exit polls carried out by the BBC and ITN both showed there
would be a hung parliament, although both of them had the Conservatives slightly
ahead. They were both not far from the actual Conservative 43%, and Labour 35%,
and if they had predicted using a uniform swing assumption, they would have been
very close to the real result. But they adjusted the figures as they were
suspicious of the results being so far out of line with the mornings polls.

The polls were not up to their normally high closeness to the actual results for
one, or both, of two very broad reasons. Firstly there must have been a late
swing of undecided voters to Conservative, or secondly, that the polls that were
carried out were all inaccurate, obviously for the same or similar reasons.

Looking at the first explanation, the theory that there was a late swing of
'undecided' voters in the favour of the Tories, this would have meant that the
polling companies had all been correct at the time. But this, in itself, could
not possibly have accounted for the incorrectness of the polls. The swing would
have had to be in the order of 4%, which is unbelievably high. Although there
were an exceptional number of 'undecideds' on the eve of the election, and it
was evident from the post election recall surveys that there was a late swing
towards the Tories (Crewe, 1992, p. 485).

Before we can look at the second explanation, that the polls were simply wrong,
we should look at where the 1992 polls differed from the past, remarkably
accurate polls. Polling practices had not changed much from previous years, nor
had the style of the polling, the questions, samples, etc. One reason that has
been put forward is that the polls didn't check that people were eligible to
vote or not, this may have caused major discrepancies in the outcome of the
polls. The reason this may have caused such a big problem is that a lot of
people may have taken part in opinion polls when they were not registered to
vote, this is because they were avoiding having to pay poll tax. In general the
people avoiding the poll tax in this way were Labour voters, which could explain
why the forecast polls showed Labour