What we need to know is how people will vote?

As someone who uses polls to try to predict elections and win bets I’ve always been very wary of polling questions which ask respondents for their opinion – not for an indication of what they would do in an election.

This based on the last two general elections only about 60% of people actually vote so the results you get from the opinion questions are distorted by the views of non-voters. Secondly the voting intention questions from the major pollsters are subject to a lot of controls that often don’t apply to the findings that are about people’s views.

Then there is the challenge for the pollster of devising a question that is useful. Just look at the example above from the weekend’s News of the World YouGov survey and it’s hard to draw any conclusions.

What’s needed is the form of polling that we saw during the Tory leadership election three years ago and was a regular feature of Populus and ICM polls until May 2007 – the month before Gordon got his coronation – “How would you vote if Brown/Miliband/Johnson was Labour leader, Cameron was Tory and XXXX was Lib Dem leader.” These were subject to the same political weighting and turnout filters as the standard polls from the firm and have proved themselves with what they suggested about Brown.

I am hoping that within the next few weeks, certainly ahead of the Labour conference, that we’ll have the findings of polling questions like this in relation to possible other Labour leaders. How better/worse would Labour be doing with Miliband/Johnson at the helm rather than Mr. Brown. My guess is that there will not be that much variation.

The other polling material really does not contribute a great deal to our understanding and I tend not to feature it much here. I do not, for instance, pay much attention to the latest PH5000 data on David Miliband – which seems to have got the Guardian a bit excited.

Mike Smithson