But, while most pollsters are confident that the trend shown by ICM was due to statistical error, they are all nervous about the levels of party support.
Tom Simpson, managing director of Harris, is working on estimates of what would happen if there were a late swing to the Conservatives of the same order as in 1992. On average, the final polls were nine points adrift then, with about one-third of the error down to out-of-date information about the make-up of the electorate. So if voters have the same last- minute change of heart as five years ago, Labour's lead may end up being six points lower than in the opinion polls.
"Now the same factors may not be at work, but it would be a legitimate exercise to do alongside our final poll next week," Mr Simpson said. But on Harris's numbers today it would only cut Labour's lead to 12 points, and leave Tony Blair with a Commons majority of just over 100 seats.
The unusual feature of Harris's polls in the run-up to this election is that the number of don't knows has risen during the campaign, instead of falling as expected. From 12 per cent five weeks ago, 16 per cent are now saying they do not know how they will vote.
This feature, reflected in other polls, has been seized on by the Tories as evidence that there is "all to play for". But many of these are already counted by the pollsters as "shy Tories" on the basis of how they say they voted last time, and who they think the best prime minister would be. Only MORI refuses to make this adjustment.
Today's Independent/Harris survey also continues to suggest the Labour vote is firmer than the Tories'. Of Labour voters, 80 per cent say they are "certain to vote Labour", while only 60 per cent of Tories say they are "certain to vote Conservative". Twice as many Tories (29 per cent) as Labour supporters (15 per cent) say "there is a chance that I may change my mind before I vote".
The Liberal Democrats are the most popular second choice among potential switchers (28 per cent of waverers name the Lib-Dems, 17 per cent Labour, 14 per cent Tories).
There remains the puzzle of why ICM consistently put Labour lower, and the Tories and Liberal Democrats higher, than the other polls. Nick Sparrow, ICM managing director, attributes the difference to random telephone interviewing.
But Gallup, which uses the same method, produces results more in line with the other companies whose interviewers speak to people face-to-face. Andy Brown, head of research at Gallup, claims his methods are superior to ICM's because telephone numbers are generated at random by computer, while ICM take numbers at random from telephone directories and then add another random number to the last digit. "ICM's method means they get fewer ex-directory numbers, because we know that ex-directory numbers tend to cluster together," Mr Brown said.
He added that, surprisingly, the one-third of the population which is ex-directory is more pro-Labour, reflecting the fact that they tend to be younger, unmarried, in rented housing and less likely to be in professional occupations.
Meanwhile, analysis of the last local council by-elections before the general election by Colin Rallings and Michael Thrasher of Plymouth University, suggests a Labour lead of 13 points. This is lower than the opinion polls, but what is heartening for the Tories is that the same analysis just before the general election five years ago put the two main parties neck and neck. If that relationship holds good this time, Labour would win the general election by the same sort of margin at the five point lead in this week's ICM poll, pointing to a Labour majority of around 40 seats.
Harris Research interviewed 1,177 adults face-to-face in their homes between 18 and 21 April.Reuse content