Some warnings about hypothetical leader polls

The monthly online ComRes poll for the Independent on Sunday and Sunday Mirror has topline figures, with changes from their last online poll in early December, of CON 38%(+2), LAB 38%(-2), LDEM 11%(+1), Other 13%(-1). A neck-and-neck tie between Labour and the Conservatives is very much in line with the YouGov polling we've seen so far this month.

In the rest of the ComRes poll today they asked voting intention questions with various alternate Labour leaders. Questions like this are incredibly popular with the media when a party leader starts to teeter, but are extremely dicey methodologically. Firstly, if you want to compare them to current voting intention questions with the incumbent leader you need to treat them in the exactly the same way as a normal voting intention question - if you measure normal voting intention using three questions (likelihood to vote, voting intention and a squeeze question) then you need to do the same for other candidates. If, for example, you just do one likelihood to vote question and use it for all questions then you'd miss out on voters who might sit on their hands if X was leader, but come out if Y was leader.

There's also the fact that normal voting intention questions don't mention the party leaders, while by definition these questions do. There is a risk of skewing things if you mention only the leader of the party in question but if you get round that by mentioning the other leaders too, how do you know any difference is not down to mentioning them?

In short, the caveats on comparing hypothetical leader questions to current voting intention questions are significant.

More interesting are comparing the voting intentions with one alternative leader to another, though again, be careful. They are to a great extent defined by a lack of public awareness of most politicians, if you recall a few months ago only around 60% of respondents were able to recognise a picture of Ed Balls...if 40% of people don't even know what Ed Balls looks like, they are not likely to be a particularly sterling judges of how they would respond to an Ed Balls leadership.

Hypothetical questions about how people would vote with leaders who are very little known are virtually useless, as voters of that party naturally say don't know, while other parties' support holds up, giving the illusion of the party performing hopelessly under their leadership. For example, in this ComRes poll 50% and 45% of people said they didn't know how they'd vote with Chuka Umunna and Yvette Cooper as Labour leader, creating the false impression of Conservative leads of 26 and 20 points under their leaderships.

That's not to say these polls are always useless - they can give a good steer on how people might respond to a leader who is already very well known to the public. For example, the hypothetical polls in early 2007 showing Labour doing worse under Gordon Brown, a figure very well known by the public, were saying something worth listening to (though even then, were a rubbish predictor of the Brown bounce).

With the caveats out of the way, the ComRes poll found people saying that with David Miliband as Labour leader the party would have a 3 point lead, compared to a 6 point deficit with Ed Balls (or Tony Blair!), a 13 point deficit with Alistair Darling and a 15 point deficit with Harriet Harman. As far as I can see, the hypothetical voting intention questions were just done as a grid, without the likelihood to vote weighting or squeeze question that ComRes normally use, meaning they are not comparable to normal ComRes voting intention polls.