Guardian

Might the Labour bias corrections have gone too far?

The Guardian’s polling expert, Julian Glover, raises an interesting question this morning – could all the measures that have been made in recent years to correct the systemic pro-Labour polling bias have gone too far?

And if that’s the case then could this offer a glimmer of hope for Brown’s party as it sees the Tories staying in the 40s looking on course for victory?

For unlike any previous election we go into the coming 2010 battle with every single one of the national pollsters now utilising measures of one sort or another to deal with Labour overstatement.

Glover writes: “For the past 25 years telephone pollsters have been wrestling with a ­persistent tendency of the polls to overstate Labour’s share of the vote. Since 1983 the only final poll before election day to have proved to be too kind to the Conservatives was ICM’s in 1997. Pollsters have got used to voters treating the Conservatives as the underdogs. They use adjustments to allow for the fact that telephone samples contain more Labour voters than the electorate as a whole, and that Labour voters overstate their likelihood to vote.

They have also had to deal with the so-called spiral of silence – the fact that some voters whose party choice is unpopular will hide their party allegiance from pollsters. Plot election polls for the last 25 years on a chart, and most overstate the likely lead of the winning party. All this has had the effect of increasing figures for Tory support – and making the polls more accurate.

But what if, in the context of 2010, these assumptions are turned on their head? We already know that in the last 18 months or so the spiral of silence has helped Labour: pollsters now find themselves having to increase Labour’s share, to take account of people who say they have switched away from the party but may still turn out for it on polling day. Perhaps some shy Labour voters are even now evading the pollsters’ radar..”

I think that Glover is wrong to put so much emphasis on “shy Tories” and the spiral of silence adjustment because it gives the impression that this is the major measure that pollsters deploy. It’s not and in any case it is confined to ICM, Populus and ComRes. The main elements have been the measures to ensure politically balanced samples and to take into account certainty to vote.

Glover’s also not correct to state that the “spiral of silence” adjustment started helping Labour only eighteen months ago. This trend began well before the 2005 election.

Overall the evidence is still that the most accurate poll is the one that has Labour in the least favourable position as we saw at the 2008 London mayoral race, the 2009 EU election, and at Crewe & Nantwich and Glasgow East

As I wrote in June I had thought that my “polling golden rule” about the polls and Labour might not survive the EU elections. I was wrong. Populus and YouGov had Labour’s share at 16% just above the 15.7% that was actually achieved. Other firms had Labour higher with one of them at 22%.

In 2005 the firms did quite well – but even then the final polls from all of them but NOP over-stated Labour. That pollster got it precisely right and then got sacked by its commissioning paper, the Independent.

I think that Glover is right to raise the issue – but so far there’s no evidence to support the argument that the pollsters have gone too far.

Mike Smithson