The breaking of the yellow tide

If you discussed the election with me just before polling day, you will be able to verify my claim; if not, you will just have to take my word for it. I told friends and colleagues that my biggest fear as a pollster was that we would overstate support for the Liberal Democrats. I suspected that some people said they would vote Lib Dem and genuinely meant it, but that, faced with a ballot paper, they would recoil from their intended act of protest and vote for either a Labour or a Conservative candidate.

And that, indeed, is what seems to have happened. Every poll published on 6 May put Lib Dem support too high. In contrast, the exit poll, conducted by NOP and MORI, was remarkably accurate. This suggests that voters told the truth and changed their mind. As a result, YouGov correctly predicted that there would be a 5 per cent swing from Labour to Conservative, and that the swing would be higher in Labour's marginal seats. However, like everyone else, we wrongly predicted significant gains for the Lib Dems.

This point is not only relevant to pollsters and the National Union of Political Number Crunchers; it tells part of the story about the most extraordinary feature of the general election campaign - the impact on voters of the first TV debate by the three party leaders.

Two tribes

Since the day after polling, the conventional wisdom is that this impact made no difference in the end, as the Lib Dems ended up losing seats, not gaining them. The conventional wisdom is wrong. Before the first debate, Lib Dem support averaged about 19 per cent. Afterwards it averaged 29 per cent. On election day it slipped back to 24 per cent. So the party retained about half the gains that Nick Clegg secured by his performance in that debate. I estimate that without those gains, the Lib Dems would have won 20 fewer seats: 14 more would have gone to the Tories and six to Labour.

The political effect of those 20 "extra" Lib Dem seats has been huge. Without them, the new House of Commons would have contained 321 Conservative MPs, 264 Labour and 37 Lib Dems. David Cameron would be so close to an overall majority that he would have been prime minister by the evening of Friday 7 May. In other words, it is because of the extra votes that Clegg secured and retained following the first televised debate that the Tories have fewer seats than Labour and the Lib Dems combined, and that the prospect of an anti-Tory government could be entertained at all.

So the debates did make a difference. But is it now possible to add Labour's 29 per cent share of the vote to the Lib Dems' 23 per cent and claim a progressive majority among Britain's voters? The answer is not clear-cut.

During election week, YouGov asked people where they placed themselves on the left, centre or right. We offered three variations of left and right - "very", "fairly" and "slightly". At first sight, Labour and Lib Dem supporters look fairly similar, and very different from Conservative supporters. Of those who expressed a preference, Labour supporters divide 54 per cent left, 23 per cent centre and 5 per cent right, while Lib Dem voters divide 43 per cent, 29 per cent and 9 per cent respectively. Contrast those figures with the Tories' 5 per cent, 21 per cent and 57 per cent.

If that were the only piece of evidence we had, then the conclusion would be irresistible: most British voters belong to one of the two tribes on the left bank of British politics, while only a minority belong to the one significant tribe on the right bank. But that is not the only evidence. When we asked the same Lib Dem voters which they would prefer if they had to choose, 51 per cent plumped for a Labour government led by Gordon Brown, while 36 per cent opted for a Conservative government led by David Cameron. So the advantage resides with Labour, but it is far from overwhelming - and very different from 2005, when Lib Dem voters opted by 3:1 for Labour/Blair over Conservative/Howard.

Among the electorate as a whole, the division is 47 per cent Conservative/Cameron, 43 per cent Labour/Brown. This tells a different story from the one derived by totting up the votes and saying that progressive Lab-Lib Dem voters outnumber Conservative voters by 52 per cent to 36 per cent. A Cameron-led government has greater public appeal than the idea of a Brown-led government and the assumption that a Lab-Lib coalition would command majority support under any circumstances is simply wrong. All of which helps explain why Brown agreed to signal his own departure in a dramatic statement on Downing Street on 10 May.

Despite the numbers, a progressive coalition could become popular. Voters change their minds (as roughly one million of them did on 6 May, deciding in the end not to vote Lib Dem). Indeed, the sheer variety of movements from day to day (especially after the first TV debate) and seat to seat - far greater than in any election in modern times - testifies to the erosion of tribal loyalties. Voters are increasingly footloose. For those who remember their teenage physics, we are less and less like hard steel (tough to magnetise, but also tough to demagnetise) and more and more like iron (easy to magnetise and easy to demagnetise).

Trimmed fringe

Two of the subplots in this election confirm this. Labour did especially badly in 2005 in constituencies with large numbers of students (because of tuition fees) and those with large numbers of Muslim voters (because of the Iraq war). This time, Labour support fell by only half the national average in university seats - and actually increased by 4 points in seats with the highest share of Muslim voters.

One final point. Despite Caroline Lucas's stunning victory in Brighton Pavilion, this was not generally a good election for the smaller parties. The Greens secured almost exactly the same number of votes (285,000) as in 2005. The British National Party won almost three times as many votes - but stood in almost three times as many seats; and the party suffered heavy defeats in Barking, where Nick Griffin failed to lay a glove on Margaret Hodge and his party lost all its councillors.

Only Ukip made any advances. Despite failing to unseat the Speaker, John Bercow, in Buckingham, its vote climbed to 917,000. Next time it could become the first distinct, new party to pass the million mark at a general election since Labour did so almost a century ago.

Peter Kellner is president of YouGov.