Standing before an audience in the Institute of Government overlooking The Mall near Buckingham Palace on Thursday morning, Dr Stephen Fisher from Trinity College, Oxford produced a pie chart of near-bewildering complexity.
In all, Fisher outlined a dozen possible results that could emerge from the May 7th election, among them majorities for Labour and the Conservatives (highly unlikely); a Conservative/ Liberal Democrats/DUP alliance (a 14 per cent chance); and a Labour/DUP pact (a 1 per cent possibility).
Everyone agrees that the House of Commons will be “seriously hung” afterwards, said Fisher, with both Labour and the Conservatives not only failing to reach the 326-seat finishing line, but unlikely to get over 300.
Election 2015 is the most polled electoral contest in the UK’s history. Hundreds of surveys have been taken. Yet, with just three weeks to go to voting, frequency has not brought clarity. If anything, it has added to the shadows.
Indeed, the changes in British society, illustrated by the emerging differences between the numbers collected from opinion polls gathered over the telephone and those gathered from internet panels, has attracted increasing attention.
Earlier in the week, a Guardian/ICM survey (one taken by telephone) put the Conservatives ahead by an extraordinary six points. Meanwhile, a YouGov survey from an internet panel (published in the Sun) put Labour three points ahead.
The two results are not outliers. A 10-day average “poll of telephone polls” puts the Conservatives in the lead with 35.5 per cent of the vote, compared with Labour on 33.5 per cent, the Liberal Democrats on 8.5 per cent, Ukip on 11 per cent and the Greens on 6 per cent.
Meanwhile, a similar exercise for online polls goes the other way. There, Labour is in the lead, with 34 per cent of the vote and the Conservatives trailing by two points. The Liberal Democrats record the same as in the phone poll but, curiously, Ukip scores 15.5 points.
Longtime addresses
The differences are explained by a number of factors, including the fact that more older people have lived at their addresses for a long time, have landline telephones and are Conservative supporters.
Most significantly, said Dr Rob Johns of the University of Essex, internet polling panels are made up of volunteers, “who are disproportionally more likely to be politically active and interested”.
This was graphically illustrated during last year’s Scottish independence referendum, when polling company Survation eventually closed off its polling panel to new members because of an influx of independence- supporting applicants.
Internet polls tend to have a disproportionate number of people living in rented accommodation who may have recently moved house and may not have registered to vote at their new abode.
Drop in voters’ list
With just three days left for registration, the voters’ list has fallen by 800,000 – partly on the back of new rules requiring annual registration, plus a clean-up of a list that still included the recently deceased.
“I wonder if we are not reaching the end for [landline] telephone polls,” said Professor Paul Whiteley of the University of Essex. “They may have reached the end of their sell-by date because of the growth of mobiles, call- blocking and whatever”.
Polling via the internet is now the most common way of conducting polls, said Professor John Curtice of the University of Strathclyde, though pollsters have worked hard to “overcome potential shortcomings”.
Results are weighted in order to identify past voting loyalties, if any; the likelihood that the person polled will actually vote; and to make predictions about the likely behaviour of those who classify themselves as “Don’t Knows”.
Underestimating Tories
In the past, opinion polls tended to exaggerate Labour’s popularity and underestimated the Conservatives’ strength. This was especially true in 1992, when Tory leader
John Major
came back from the dead to win.
Curtice offered caution, however: “We may question how far that experience is an adequate guide to the accuracy of polls being conducted now, especially as the polls actually underestimated Labour support in 2010.”
The lack of clarity exists even though there are more numbers than ever before.
However, Election 2015 is unique: the fixed-term parliament legislation means that the date was set in stone, rather than left up to David Cameron to choose to his own best advantage.
Most significantly of all, of course, there is Scotland, where the Scottish National Party juggernaut shows no sign of slowing down in its bid to smash Labour north of the border.
If the polls fail to offer guidance, perhaps the bookmakers can. On Thursday Ladbrokes cut the odds on a hung parliament to 7/1 on. The odds on Labour's Ed Miliband being in charge by mid-summer, however, have been cut: from 10/1 in March to just 4/1 now.