Skip to main content
Where I Live
A-Z Index

BBC News

BBC Election 2005

Watch the BBC Election News
  • Election news alerts
  • Email services
  • Mobiles/PDAs
  • News for your site
Last Updated: Monday, 9 May, 2005, 12:35 GMT 13:35 UK
Did the opinion polls get it right?
David Cowling
By David Cowling
Editor, BBC Political Research

How did the opinion polls fare in this general election?

Generally remarkably accurately, is the honest answer, especially compared to some of the concerns that have been expressed at previous elections.

The traditional test for the pollsters is their last published poll before election day.

Overall five of the six polls tracked by the BBC gave an accurate result within their margin of error.

Taking the average of all six, the share of the vote for Labour was 37.6% (actual share 36%), Conservatives 32% (33%) and Lib Dems 22.6% (23%).

And the BBC/ITV exit poll was able not only to predict accurately the share of the vote, but was spot-on in its predictions as to the size of Labour's reduced majority.

Share of the vote
Poll CON LAB LD Lab lead
over Con
Communicate Research/IoS 31% 39% 23% 8%
ICM/Guardian 32% 38% 22% 6%
NOP/Independent 33% 36% 23% 3%
Populus/Times 32% 38% 21% 6%
YouGov/Telegraph 32% 37% 24% 5%
MORI/FT 33% 38% 23% 5%
ELECTION RESULT 33% 36% 23% 3%

The 2005 Gold Medal goes to NOP whose poll for the Independent was spot on, while the least accurate was Communicate Research (which was carried out the previous week for the Independent on Sunday).

However, the election was an impressive overall performance by the polling companies and goes a long way to repair the damage their collective reputation has sustained in recent elections.

Reduced Labour bias

In the past few elections, the opinion polls have consistently over-stated the Labour vote.

See the pattern of the polls through the campaign

In the 1997 and 2001, the polls accurately predicted a Labour victory, but overstated its size.

For example, in 1997 the five leading opinion polls predicted a Labour share of the vote of 47%, and a Labour lead over the Conservatives of 16%. The actual result in the General Election was a 44% share and a 13% lead.

There has been much discussion of the reasons for such a bias, including sampling error (Conservatives less likely to answer pollster's questions) and the "shame factor" (people generally not wanting to admit they voted Conservative).

But whatever the causes, the pollsters this time seem to have devised methods to overcome it.

Campaign results

Not only did the opinion polls accurately predict the results in the final poll, they also produced much more stable figures throughout the campaign.

Some 51 voting intention polls were published throughout the 2005 campaign.

Most show a range of just of few percentage points in the share of the vote gained by Labour and the Conservatives over the four weeks of campaigning.

The biggest fluctuation appears to have been in the MORI polls (as it was in the 2001 campaign). Not only did they produce the only Conservative lead in a poll (5% in the Financial Times, 5 April) but their range for Labour was a full 10% (29% to 39%).

The stability of the polls was also a reflection of the fact that, as far as we know, they was no late swing of the kind that produced a very different result in 1992.

This in turn may be a reflection of the relatively cautious election campaign run by all the main parties, with the emphasis on not making any mistakes that could jeopardise their vote.

Exit poll success

This is the first year that the BBC and ITV have combined forces to do a joint exit poll, whose results were known just after the polls closed at 10PM.

The BBC/ITV News exit poll predicted a Labour majority of 66 seats which was exactly the final outcome (assuming that the Staffordshire constituency, where the vote was postponed because of the death of the LibDem candidate, returns a Conservative).

The polling companies NOP and MORI interviewed some 20,000 voters as they left 120 polling stations throughout Britain on 5 May.

It predicted 37% for Labour, 33% for the Conservatives, and 22% for the Liberal Democrats.

The voting figures they collected were sent to a group of academics (led by Professor John Curtice of Strathclyde University and Professor David Firth of Warwick University) who fed them through their model and on behalf of the BBC and ITV News came up with the final prediction of 66 seats.

Without these adjustments, the raw exit poll would have predicted a larger majority for Labour on a uniform swing.