What is the true state of public opinion? Were Tony Blair to call a snap election this autumn, who would win? Every opinion poll this year has pointed to an even bigger Labour landslide than in 1997. Every real election contest, however, has pointed to Conservative gains – and, in the case of June’s European elections, outright victory for William Hague. Never in modern times have the political weathervanes pointed so many different ways. How do we explain this mystery?
Midway through the last parliament, Labour was miles ahead, whichever way you looked: polls, by-elections, local elections, European elections – all produced huge swings to the opposition. People debate the precise magnitude of the swing. But the overarching fact – that John Major’s government was floundering in a terrible mid-term slump – was beyond question.
Two common responses to the present confusion are: first, that the polls are making a hash of their job and second, that the data is so inconsistent that nothing useful can be said about party loyalties today. Both statements are wrong. The polls are telling a truth of sorts, though some doubts still remain about their accuracy; and if the election results are studied carefully enough, they start to tell a plausible story – and one, moreover, that can be reconciled with a sensible interpretation of the polls.
The table on page 32 sets out the basic figures. At the last election, Labour enjoyed a 13-point lead over the Tories in the popular vote. If we average the polls published this year (and indeed last year), we find a 5 per cent swing to Labour – enough to push its majority up to 275, and to depose Kenneth Clarke, John Gummer and Sir Edward Heath. Nobody expects a real election in two years’ time to produce such a result – not even the pollsters.
Yet this does not necessarily mean the polls are “wrong”. They do not predict the results of elections two years or even two months away. They give a snapshot of loyalties at the time they are conducted, in the form of answers to a hypothetical question: “How would you vote if a general election were held tomorrow?” Time and reality are liable to change the figures.
The first doubt about the polls concerns sampling technique. Most pollsters now accept that in 1992 they interviewed too many working-class and too few middle-class electors. Since 1992, however, this fault seems to have been corrected. If every single adult in Britain were tracked down and asked which party he or she supports, the chances are that, of those who gave a firm answer, around half would say Labour and only three in ten Conservative.
This conclusion is supported by two extra pieces of evidence. Last year, the academic group Crest, which runs the British Election Panel Study, reinterviewed more than 2,000 people it originally contacted just after the 1997 election. Comparing their 1998 loyalties with their 1997 vote, Crest found a 5 per cent swing to Labour. ICM conducted the same exercise with 1,000 people in April last year. It, too, found a 5 per cent swing. Movements of this scale suggest that the post-election shift to Labour has been real and substantial.
There remains, however, a trickier problem to do with the political interpretation of polling data. Are the figures warped by a “spiral of silence” – Tory supporters who say “don’t know” rather than admit the terrible, unfashionable truth?
ICM’s Nick Sparrow (who also conducts the Tories’ private polls) has long believed that there is a spiral of silence and he adjusts his raw data accordingly. That is why ICM’s polls in the Guardian produce consistently lower Labour leads than MORI’s in the Times and Gallup’s in the Daily Telegraph.
I have doubts about ICM’s method of adjustment, but I am sure Sparrow is right in principle. Every unadjusted eve-of-election poll in 1997 overstated Labour’s support, on average by 4 percentage points. Unadjusted polls in Scotland in the run-up to the May elections overstated Labour’s support by around 8 points.
My own guesstimate of Labour’s true Britain-wide support is 47-50 per cent, compared with 29-31 per cent for the Tories and 14-16 for the Lib Dems. That, however, still leaves a chasm between poll-based figures and those generated by real election results. How come?
Turnout is a huge factor. By no means everyone who expresses a party preference to a pollster actually votes in real elections. Some stay at home. In almost all elections, turnout tends to be higher in middle-class than working-class areas. In 1997 turnout in a typical inner-city seat (around 60 per cent) was four-fifths of that in a plush suburb (75 per cent). In the European elections inner-city turnout was often only half that of the most middle-class constituencies.
These figures chime with Nick Sparrow’s analysis of ICM’s data for the European elections. He finds that very few people shifted from Labour to Tory, or vice versa. The big difference was that Tories were far more determined to vote than Labour folk.
But that is not the end of the story. If we trim the unadjusted poll figures, Labour’s lead is around 18 points; if we tweak the European result to allow for turnout, the lead is around 2 points. What can explain this remaining large gap?
Look again at the table below. Each election tells a different story: Labour support fluctuating between 28 and 45 per cent, the Lib Dems between 13 and 24, minor parties collectively between three and 23. Only the Tories, on 34-36 per cent, remain steady. What is happening is that voters are increasingly taking a horses-for-courses approach to elections. Thus the Lib Dems score well in many local government contests but less well in national elections; the Greens and the UK Independence Party pick up votes when the voting system is proportional, but not in first-past-the-post elections; and voters are keener to send nationalist representatives to Cardiff and Edinburgh than to Westminster.
That explains why Labour’s share of the vote held up at the Eddisbury by-election: it was not in Scotland or Wales (so no nationalist surge), it was not a local council election (so hard luck for the Lib Dems) and not a PR election (so no help to the Greens or UKIP). Labour might even have gained the seat had the turnout not fallen to 51 per cent, from 76 per cent in 1997: a MORI poll in the constituency put Labour ahead among all electors, but the Tories ahead among those “certain to vote”.
So a pattern does emerge. Labour is doing best in the elections that matter most and worst in those that matter least. For the moment, voters will turn out to keep Tony Blair in Downing Street – but many will stay at home or vote for a minor party when he is not directly involved, especially if a fringe candidate stands a real chance of getting elected.
There is no evidence to suggest that in a high-turnout election the Tories could win much above 30 per cent. Even if we look solely at real elections and ignore the polls, William Hague’s Tories are vying with Michael Foot’s Labour Party in 1981 for the record of least popular mid-term opposition since the war.
Were a general election held this autumn, Labour would almost certainly win another landslide victory. But whether the party has enough solid, consistent and enthusiastic support to build a lasting hegemony – that is an altogether different and more doubtful matter.