Almost all the pre-election polls, other than the final eve-of-poll predictions, are now in (results here). They vary widely, from the latest MORI/Financial Times poll which shows Labour with a 10 point lead over the Conservatives, to the YouGov/Sunday Times poll which estimates Labour has just a 3 percentage point lead. Whatever the result on Thursday, someone will have egg on their face.
To put those poll results in perspective, Labour won a majority of 167 seats in the 2001 election. The swing-o-meter at Electoral Calculus estimates that with no tactical voting effects, the MORI/FT figures would see a Labour majority of around 162 - almost no change. By contrast, if the YouGov poll is acurate that majority would be cut to 96 seats - still quite a workable majority.
However if the LibDem to Labour tactical voting advantage turns negative (from +2% to -2%), that majority falls to 62 seats. Figures around those levels would start to make PM Blair uneasy, and might prompt an early move to oust him by Chancellor Gordon Brown.
One solution to the divergent poll results is to pool the major surveys. The Financial Times have done this on their website, using UK Polling Report numbers:
It shows that the earlier Tory campaign gains have been pared back over the last two weeks. Several commentators have attributed this to backlash over the Conservative's negative campaigning and attacks on Blair, but it may also reflect the Tory's campaign peaking too early, or their narrow focus on populist issues such as immigration that galvanised their electoral base, but did little to attract swinging voters.
Anthony Wells, on his UK Polling Report Blog, gives reasons why he expect YouGov to be the most accurate pollster. I think he may be right, for several reasons:
1) There is widespread anecdotal evidence of large Labour protest vote. Many Labour supporters appear likely to either stay home or vote Liberal Democrat.
2) The Liberal Democrat tactical voting which steered about 2% of votes in Labour's favour in 2001 may well disappear, and could even turn negative.
3) Labour has a lot fewer rank and file supporters on the hustings than last time, which will make it harder for them to get out the vote on Thursday.
4) There is some evidence of "shy Tories" who are too embarrassed to admit to an interviewer that they support the Conservatives, understating support for the Conservatives (Wells discusses this effect).
5) The latest ICM poll suggest the Tory campaign focus on key marginals is bearing fruit. I expect the Conservatives to pick up more marginals than the overall national swing would predict.
6) The betting markets have for some time favoured a much lower Labour majority than the polls suggest. According to Monday's Independent:
Punters are betting that Labour's majority will be as low as 60 seats. Few punters are backing a majority of more than 80 seats.
A final variable to factor in is the far higher number of undecided voters this time around. The MORI/FT poll claims that 36% of voters may yet change their mind, compared with just 21% still undecided in the final week last time. Of course most could vote Labour's way, but if polls continue to indicate a large Labour win I expect quite a few to register a protest vote.
For what it's worth, I expect the popular vote to be very close - with a good chance that the Conservatives will get more votes than Labour in England - the so-called Al Gore scenario. There won't be a hung Parliament, but Labour's majority will be drastically trimmed - down from 167 to between 60 and 90 seats. The Liberal Democrats, meanwhile, are on track for their best result in 22 years.
UPDATE: Tony Blair has achieved an historic third term, but with his majority slashed - as most commentators, including myself, had expected. The final declared result saw a 5.4 percentage point swing against Labour, bringing their national vote down to just 35.3% - the lowest ever winning share of the vote. But most of those lost votes went to the Liberal Democrats (up 3.8 percentage points), not the Conservatives (up just 0.6 ppts). The populist Tory campaign may have consolidated its base, but it won few converts. Labour suffered from a big protest vote, not a swing to the right.
The BBC/ITN Exit Poll, conducted by MORI, predicted a Labour party majority of 66 seats. That was where we ended on election night. The final majority, with all seats declared, was 67. The Sunday Times rightly dubbed this a "triumph for the pollsters".
But the opinion polls and political betting also got it right. As Anthony Wells explains, the election-eve opinion polls were all very close to the actual voting shares, ranging from a very small 0.2% average error for the NOP/Independent poll to a still respectable 1.6% average error for the Populus/Times poll. These were all well within the usual 3% margin of error.
Finally, electoral betting sites had been predicting an outcome of this magnitude for weeks, even while earlier opinion polls were proclaiming a big Labour win. As Mike Smithson comments, "other spread markets seemed to have been good reflection of what actually happened."
Recent Comments