Is Labour’s margin really as big as it seems?

Now here’s a funny thing – in the three Guardian ICM polls for January-March 2007 the firm was reporting leads for Cameron’s party of 6% – 10%. Yet looking at the raw data FEWER people were telling the pollster then that they planned to vote Conservative than in ICM’s three published polls in July.

So for the 2007 Q1 surveys the average number of people saying they would vote Tory was 194. Compare that with the average of 205 people who said the same in the pollster’s three July surveys. The sample sizes were almost identical in each of the surveys.

Ah you will say – “Smithson’s fiddling the figures!” And you are right. I’m looking at the base responses before the weightings were applied and it is these which have played a part in determining Labour’s margin.

For two of the driving factors behind the move to Labour since PM Brown arrived are a greater determination by Labour supporters to vote and, surprisingly, an apparent change in the pattern of responses to the question of how respondents voted in 2005.

For in the 2007 Q1 surveys for every ten people who said they had voted Tory in May 2005 nineteen were saying they had voted Labour. Compare that with the three ICM July polls this proportion was down to fourteen declared 2005 Labour voters for every ten Tory ones.

In fact the final ICM poll of July had fewer people saying that they had voted Labour than any survey from the pollster that I have been able to find since the last general election.

This is critical in past vote weighting surveys like those from ICM. For the fewer people who say they voted for a party last time the greater value will be put on their opinions when working out the final party vote shares. So the quite sharp increase in declared 2005 Tory voters that has been experienced has been a major factor in the party’s reduced vote share.

I am not saying that ICM is wrong – in fact I’m a great fan of the pollster and its methodology. But I record this to increase the understanding of how polling works.

There is a positive consequence for the Tories and a negative one for Labour in all of this. For the past vote weightings that are used are determined by the 2005 election result itself averaging a series of recent findings on vote recall. If this trend continues then the weightings should move a bit away from Labour in future surveys.

UPDATE: Please see comment by Nick Sparrow of ICM at comment number 50.

Mike Smithson