A brief word on the importance of the low turnout and its effect on the polls. MORI's final poll projection for The Times was Conservative 30%, Labour 45%, Liberal Democrat 18%; the "poll of polls" (average of all the companies' polls conducted during the final week) was Conservative 31%, Labour 45% and Liberal Democrats 18%. Both close to the final result (32.7%:42.0%:18.8%), and within the standard 3% margin of error for all parties - though, naturally, we would like it to be even closer.
But readers of some newspapers will have noticed that it took less than a week before the "expert commentators" started criticising the polls again, noting that all the final polls overestimated the Labour lead, and rehearsing again their theories about "shy Tories" and biased samples. As we have pointed out many times, you should watch the shares, and the turnouts, not the gap. Looking simplistically at the lead risks completely missing the point.
MORI's final poll found 63% said that they were certain to vote. In fact, only 59% voted. As we said on numerous occasions before the election was likely to happen, "old Labour" voters in safe seats in particular realised that they didn't need to vote after all - so Labour lost last-minute votes, but not seats, and won the majority tha should have gone with a 13-point lead even though in actual votes cast they had only a 9-point lead.
The trouble with percentages is that they depend on the base - if lots of people tell us they will vote Labour and then don't vote at all, that not only reduces Labour's percentage share but it increases the percentage shares of the other parties, because their unchanged numbers of voters are now part of a smaller total. The difference between a 63% turnout and a 59% turnout is big enough to have a significant effect on the party shares, if the extra stay-at-homes are disproportionately from one party.
So instead of thinking in percentages, let us think in millions of real voters. On a 63% turnout, a 30% Conservative share translates to 8.2 million votes, 45% for Labour equals 12.2 million and 18% for the Lib Dems is 4.9 million. And the actual votes? The Tories got 8.36 million, Labour 10.7 million and the Lib Dems 4.8 million.
In other words, that final poll almost exactly predicted the numbers of Conservative and Lib Dem voters. But it overestimated the number who would turn out for Labour instead of staying at home, and as a result got the percentages - slightly! - wrong. Interestingly, exactly the same was true of MORI's last poll before the 1999 European elections: the Conservative vote was correctly projected to the nearest tenth of a million, but Labour's turnout was overestimated.
So the likeliest explanation if the small error of the polls - which was present in all of them to differing degrees, despite the wide range of sampling methods used - is that, far from there being anything wrong with the samples, the samples were pretty good; the difficulty is identifying all of those who though they would support Labour if they voted at all will decide in the end not to bother in a low turnout election.
The so-called newspaper "experts" are wrong in their diagnosis! It's not shy Tories at all, it's stay-at-home Labour, and probably old Labour.