What happens if you don’t measure “likelihood to vote”?
One of the most important questions asked by four the five pollsters that regularly carry out national voting intention surveys in the UK – ICM, Populus, Ipsos-Mori and Communicate Research – is how likely it is that respondents will actually vote.
Those surveyed are asked to rate, usually on a scale of 1-10, how likely it is that they will turn out and these answers play a huge part in determining the headline numbers.
To look at the impact of this compare the February and May ICM Guardian polls. In the former, only 76% of Labour support counted for the final figure against 86% for the Tories. In the latter, that came out on Thursday, you find that the Labour turnout rating had moved up to 84% while the Tories were still rated at 86% likely.
Not only does this have the affect of giving a big boost to the Labour share but it also dilutes the Tory share because the overall number of responses on which the final figures are calculated is greater.
So an immediate impact of Blair’s departure and Brown impending arrival is that many supporters feel better about the party and are telling pollsters that they would be more likely to vote.
Which brings us to YouGov which does not normally apply a turnout filter to its final figures.
If the change in the desire of Labour supporters to turn out is having such an impact with the other firms is there a danger that YouGov might not be picking up the scale of any Brown-induced move back to the party?
Certainly the internet pollster has not recorded the Labour lows on the same scale this year as the other four – maybe it won’t record the “highs” as well.
If my analysis is correct then expect to see bigger moves to Labour in the Communicate Research and Ipsos-Mori May surveys which are still to come out.
Ipsos-Mori, of course, only counts those scoring 10/10 in its headline figures and the big swings that it often records can be put down to an increased or a decreased desire by Labour supporters to vote.