Log in

Eagle Eye

Previous Entry | Next Entry

John Rentoul: Pollsters don't believe own polls, contd.

Posted by Eagle Eye
  • Sunday, 11 April 2010 at 10:26 pm
Alex Barker adds two good points to the debate about pollsters' predictions. One, a consensus in any group is a warning. Two, although most of the pollsters predict a Conservative majority, most academic commentators (about whom I commented here) seem to expect a hung parliament.

Indeed, Professor John Curtice of Strathclyde University, for whom we did not have space in The Independent on Sunday, predicts the Conservatives will win 318 seats, eight seats short of a majority. (Although, as president of the British Polling Council, he is as much a pollster as an academic.) 

Meanwhile, in my new role as one of the IoS's team of Randall and Rentoul (right), I have provided this guide to opinion polls:

You can't predict elections by asking just 1,000 people. Yes, you can: if they are a good enough sample, their answers can be adjusted – or weighted – so that the sample matches the total electorate by sex, age, class and, in YouGov's case, newspaper readership. The findings should be accurate to within two percentage points 95 per cent of the time. The polls had problems in the past with Labour supporters being more willing to take part and being less likely actually to vote, but in 2005 they were pretty accurate and in the most recent big Labour-Tory test YouGov got Boris Johnson's margin over Ken Livingstone right in London 2008.

The polls are all over the place/public opinion is unusually volatile. No, they aren't, and no, it isn't. You would expect the figure for the support for any party to vary by four points (two up, two down) in 19 polls out of 20 carried out using the same method. Hence the difference between two parties' share of the vote – the figure for the Tory lead – could vary by up to eight points even if people's views remain unchanged. Most of the time, polls from the same company tend not to vary that much.

But the Conservatives are spending a fortune in the marginal seats where the election will be decided, so the national polls are useless. Money can make a difference, but not that much. An Ipsos-Mori poll in the marginals last week suggested, like previous surveys, that the Conservatives are doing better in these seats than nationally. This could have the same effect as adding one or two points to the Tory lead in national polls.

So the seven-point lead in your Poll of Polls means David Cameron is sitting pretty; why is everyone talking about a hung parliament? Because a seven-point lead, or even a nine-point lead, may not be enough for the Conservatives to win a majority in the Commons. Two reasons: even after new boundaries, Labour constituencies tend to be smaller than Tory ones and so need fewer votes to win; plus Tory voters behave differently, tending to come out to vote in safe seats where Labour supporters tend not to bother. That is why it is possible for Labour to win more seats than the Tories with fewer votes.

Who is going to win, then? Whichever party wins most seats is the short answer. That will be the Conservatives if their share of the vote is five percentage points ahead of Labour's. The problem is that their average seven-point lead in the polls could conceal a systematic bias one way or the other. If so, it is unlikely to be more than a few points either way – but, this time, a small difference could make a big difference to the result. Different polls suggest anything from Labour being the largest party in a hung parliament to a small majority for Cameron. Meanwhile, there are the "events" of the campaign itself.

PS. Thank you to fox_c for your kind words. I recommend following this blog on Google Reader.


General Election is in 26 days time
quietzapple wrote:
Sunday, 11 April 2010 at 09:40 pm (UTC)
It is reasonable to anticipate that changes in intentions between polls now and the actual election will follow reason.

Those who imagine that the balance of the switches will be to the Buggins' turn "Tiem for a Change" may well believe that the Tories will obtain an overall majority.

Those who have noted the changeability of the Tory message and relative steadiness of Labour - reflected in the switch to Darling as most preferred Chancellor in todays ICM poll - may expect these trends to continue, and a Labour majority of 50 seats overall.

Some "Events, dear boy, Events" are almost predestined. Tory contradictions and dissemblances generate events like piranha. People don't like that.


RSS Atom

Report Comment

To report an offensive comment for review, please send a Personal Message and provide a link to the comment. The moderators will review it and take action if necessary.
Powered by LiveJournal.com
Designed by chasethestars