The part of tomorrow’s ComRes poll that will no doubt get the most attention is a series of questions asking how people would vote with a different leader. These appear to show that Labour would do better with any alternate leader, and that most of them would produce a hung Parliament instead of a Conservative victory. Looking more carefully though, the figures don’t appear to be comparable, and shouldn’t be compared to standard voting intentions or taken as voting predictions.

Generally speaking the “how would you vote if X was leader” sort of question is very popular with the press, but there are a bucketload of problems in carrying them out and interpreting them. Firstly, there is how to ask them. A normal voting intention question does not mention the leaders of the party by name, and adding their names does have an effect (Populus tested it when it was all the rage to ask questions about how people would vote with Brown as leader). Therefore one cannot ask “how would you vote if X was Labour leader” and then compare it to the standard voting intention question, since that did not include Gordon Brown’s name in the question, and would have given a different answer if it had.

On top of that, just including the name of the putative Labour party leader in the question itself skews the question, because it is not mentioning the names of David Cameron or Nick Clegg. Mention one, you should mention them all (and that too makes a difference, so shouldn’t be taken to seriously either.)

Thirdly, there is whether the questions are asked on an equal footing. For most of the polling companies voting intention is actually quite a cumbersome question, they ask how likely people are to vote, how they would vote, and in some cases a squeeze question too. If you want to ask a series of truly comparable “alternative leader” questions you’d need to ask again how likely people are to vote with each leader, then who they would vote for. Understandably this is sometimes skipped and it assumed that likelihood to vote remains the same, but that’s probably a false assumption.

Even if you do all this, it’s questionable how meaningful the answers are. People probably have a fair idea of what Jack Straw or Harriet Harman are like, they’ve both been in the public eye a long time… but Ed Miliband? Equally, while they may have an idea what sort of chap Jack Straw is, few have any real idea of what he, or any other alternative leader, might actually do were to to become Prime Minister or, to be honest, how they would react to it.

So, with all that in mind, let’s turn to the actual questions in the ComRes poll. ComRes asked “If there were a general election tomorrow, which party would you vote for if each of the following were leader of the Labour Party?”, so the question prompted by the name of the Labour party leader, but not the leaders of the other two parties. Most importantly their standard question is filtered and weighted by turnout, however as far as I can tell the questions for alternative leaders weren’t.

In other words, the questions aren’t comparable. Filtering by likelihood to vote almost always helps the Conservatives and hurts Labour, so it’s no great surprise that questions that aren’t filtered by likelihood to vote show Labour doing better (and that’s before the effect of prompting by only the Labour leader’s name is taken into account). We cannot conclude from these questions that the alternative leaders would do better than Brown.

What the poll is useful for is comparing the alternate Labour party leaders to one another. It’s not that useful of course, since several of the people asked about aren’t very well known to the public and the public are not necessarily very good at predicting what they would do or how they would perform if they did become Prime Minister. For the record though, Jack Straw did the best, followed by David Miliband, closely followed by his brother Ed. Behind that was Ed Balls and Harriet Harman, then Alan Johnson, then Peter Mandleson. The least impressive figures were for Jon Cruddas, probably because he is the least well known.

There’s a new ComRes poll in tomorrow’s Indy. The topline figures, with changes from their last poll, are CON 38%(-2), LAB 23%(-1), LDEM 23%(+2). It’s actually a smaller boost for the Lib Dems that the polls we saw from ICM and YouGov, but ComRes already had them at a high rating anyway, resulting in the unusual result of having Labour and the Lib Dems equal.

The survey also had a series of questions on how people would vote with different Labour party leaders, which is likely to get rather more attention. I’ll have a proper look at that and update later.


For anyone interested in the German elections, here are the most recent opinion polls in Germany. There is a surprising lack of polls carried out in the last few days – presumably there isn’t any law against them, since Forsa have managed to put one out with fieldwork carried out midweek.

Date CDU/CSU FPD Green SPD The left
Forsa 24/09/09 33 14 10 25 12
YouGov 24/09/09 34 11 10 27 11
Allensbach 21/09/09 35 13.5 11 24 11.5
GMS 17/09/09 36 13 11 25 11
Forschungruppe 17/09/09 36 13 10 25 11
Infratest 17/09/09 35 14 10 26 11
Emnid 16/09/09 33 14 10 25 12
INFO GmbH 15/09/09 34 12 10 27 12

At the last German elections in 2005 the polls performed very poorly – showing CDU leads between 7 and 9 points, when they ended up leading by only 1 point. Let’s see if they’ve done any better this time round.

UPDATE: Prior to the election the German electoral commission made a great fuss about the risk of people leaking exit polls on Twitter, which frankly is asking to become a self-fulfilling prophesy. Amusingly though, with twenty minutes to go Twitter is now flooded with lots of people posting contradictory fake exit polls.

There is also a new BPIX poll in the Mail on Sunday. The topline figures are CON 40%, LAB 25%, LDEM 22% so largely inline with the other pollsters, showing the Lib Dems up to the low twenties following their conference. Last month’s BPIX had very odd figures that looked as if they hadn’t been repercentaged to exclude don’t knows, and without access to the tables it was impossible to work out what was going on, but these look more normal.

The poll also asked if people agreed with Vince Cable’s “mansion tax”, 57% of people said yes. Despite a generally negative verdict in the papers, as I said yesterday my guess is that only thing most of the public would have picked up from the Lib Dem conference was a reminder that they and Nick Clegg existed, and that they were going to put a new tax on people richer than them – something which polls almost always show as being popular.

This weekend has also produced two new polls for the Irish referendum on the Lisbon Treaty.

A poll from TNS mrbi for the Irish Times has referendum voting intentions of YES 48%, NO 33%, so a slight narrowing of the YES lead since their previous poll which showed 46% to 29%.

A second poll in the Sunday Business Post, this time from Red C, had figures of YES 55% and NO 27%, so that has the YES lead up one since the last Red C poll.

There’s a week to go to the referendum. In last year’s referendum at this stage TNS/mrbi were showing a 5 point lead for NO, Red C a 3 point lead for YES. The final result was a 7 point victory for NO. The polls would have to be a lot more wrong than they were last time (or public opinion would have to shift to a much greater extent in the final days) for this to be anything but a YES victory.