The question of who occupies Downing Street on Friday might appear a foregone conclusion, but after a bruising election campaign the battle of the opinion polls will not be decided until all the votes have been counted.

“This election is as much a test for the pollsters as it is for the parties,” Philip Cowley, a reader in politics at the University of Nottingham, said on Friday.

This test of the pollsters should help to settle a long-running argument over rival methodologies. The three leading and longest established pollsters ICM, Mori and NOP are at odds with two newer firms, YouGov and BPIX, who conduct surveys over the internet rather than using phone or face-to-face interviews.

Some of the messages from the polls are clear. For one thing, Labour's popularity is well down on 2001. Not one of the nearly 50 opinion polls carried out since the election was called giving Labour a 35 to 42 per cent share of the vote shows it exceeding its 42 per cent result in the 2001 election.

In contrast, the Conservative share is ranging between 29 and 37 per cent in the polls, compared with a 33 per cent share of the vote in 2001. The Lib Dems' share has firmed a little, ranging in the polls from 18 to 24 per cent compared with its 19 per cent result in 2001.

Apart from these trends, the outstanding feature of the polls is the clear gap that has persisted between the YouGov and BPIX results and the rest. The internet polls increase the Tory vote and reduce Labour's, giving an average gap of 2 percentage points between Labour and the Conservatives, compared with the 7-point gap in the other polls.

The 8 to 10-point leads reported recently by some of the traditional pollsters should mean another landslide win for Tony Blair.

If claims by both main parties that the fight is much closer than these headline figures suggest prove to be correct, the traditional firms will have some explaining to do. According to Mr Cowley: “If, come election night, Labour are a point ahead or even-stevens, it's a disaster for the pollsters.”

It would not be the first time the pollsters have proved to be a fallible guide to election results.

Most opinion polls in the past three general elections overstated Labour's lead. In 2001, the leading three traditional pollsters got the Labour lead over the Conservatives wrong by an average of more than 5 percentage points.

The pollsters have tried to correct for the tendency to over-egg Labour's lead in this campaign. The principal problem is that Labour supporters are less likely to vote than Tory supporters a differential turnout that can be exacerbated by the polls themselves.

Surveys suggesting that the election is a foregone conclusion may induce some Labour voters to stay at home, through apathy or the desire to curb the size of Tony Blair's majority.

“It's been widely suggested polls can have an impact on turnout. The fact that in 2001 people were being told it was all over bar the shouting may have helped to depress the turnout,” said John Curtice, professor of politics at Strathclyde University.

Polling organisations are trying to prevent a repetition this time by changing how they crunch the numbers. Roger Mortimore, senior political analyst at Mori, said: “The big reason why we were overstating Labour in 2001 is we weren't trying to adjust for turnout.”

Mori now sieves the 1,000 or 2,000 people it has questioned to exclude people who say they are not certain to vote from the headline figures. Many other leading pollsters also sieve or weight their sample by likelihood of voting.

This is not the only statistical juggling that affects the final figures. Samples are weighted by age, gender, social class and region to match the country's demographic profile. YouGov also weights by newspaper readership. Some pollsters further tweak the numbers according to how interviewees say they voted last time.

So, divergent headline figures may in part be a result of not comparing like with like. There is also the statistical fact that all the surveys have a margin of error, typically plus or minus 2 to 4 points. The odds suggest that one in 20 surveys will be a “rogue poll” with unrepresentative findings.

Prior to Thursday night's slew of polls, the leading firms showed a Labour lead of between 2 and 13 percentage points and they cannot all be right. But the risks of attaching too much significance to absolute survey results does not make them worthless. The trend of the polls is usually right. What is more, the underlying data for this campaign support the headline figures.

With hindsight, many 1992 polls' prediction of a Labour victory did not tally with all the data. The surveys suggested voters preferred John Major to Neil Kinnock as a leader and liked the Tory policies. This time, Tony Blair is consistently seen as the best prime minister and Labour is rated as best on all the big issues that matter to voters, bar immigration.

The polls may have swung around a fair bit during the campaign but none of them suggests Mr Blair will be kicked out of Downing Street.

Copyright The Financial Times Limited 2018. All rights reserved.

Comments have not been enabled for this article.