Thursday, May 06, 2010

Polls apart

Polls apart

OK.  Here we go.  Polling day.

It’s been a remarkably odd campaign – characterised by a Lib Dem surge that may or may not have petered out, a Labour campaign that staggered from farce to farce right up until the last few days and a Tory campaign that never knew whether it wanted to go positive or negative – and ended up doing sort of both and sort of neither.

But what’s going to happen tonight?  Well the first thing to say is that the polls have more or less converged on a pattern of 36/28/28.  But, as Fraser Nelson says, the polls also say that lots of people haven’t made up their minds.  I’ve been tracking the opinion polls more assiduously that has been good for my state of mind – and I’m not alone, just watch the comments on politicalbetting trend from hope to despair and back again every night.  But a nagging doubt has surfaced.

Back when the Sun started its daily YouGov poll there was a host of complaints and criticisms raised about its new methodology.  The introduction of ‘Labour Disloyals’, the heavy down-weighting applied to Tory respondents – raw numbers were being put through the wringer with such a heavy touch that Tory leads of up to 13% were being reported as leads of 6%.  Now, I’m not suggesting any impropriety in this of course, but what it made me realise was that the raw numbers that pollsters take in frequently bear little relation to the managed numbers they put out.

There are good reasons for this – pretty much every poll since 1983 has understated the Tories and overstated Labour.  No-one’s quite sure why, but that’s the way it’s always been.  A host of mathematical jiggery-pokery has therefore been applied to try and reverse this – and this has often been successful, YouGov was pretty much spot on with the mayoral elections.

But will it work this time?  One of the aims of the methodologies is to prevent ‘silly’ results where, because of a lack of Labour respondents, the Labour number looks wrong.  But Angus Reid, a new Canadian pollster, has been finding Labour shares of 23-24% for months.  Could they be right?  Do the pollsters formulae account for a Lib Dem surge?  Could the Tories be understated (again) and really be closer to 40 than 35?  Are we, in other words, about to witness 1992 mark II for the pollsters?

Well, maybe.  I don’t know.  And nor does anyone else, but it does add a certain amount of spice to tonight don’t you think?


Post a comment

Subscribe to Post Comments [Atom]

<< Home