ComRes show swing back to the Conservatives
ComRes have released what is probably the final poll of the year (though I'm conscious I was rather premature saying that last year when a final YouGov poll emerged after Christmas!). The topline figures, with changes from ComRes's last poll, are CON 39%(+2), LAB 34%(-2), LDEM 16%(+2).
I'm always a bit wary of weekend polls conducted this close to Christmas - there is the potential for the unusual shopping patterns to produce an unusual sample. Still, if we take it at face value then, like YouGov last week it shows a shift back towards the Conservatives, though given that ComRes were showing the smallest Conservative leads this only brings them back into line with the other companies. Until we see the polls in January I still think it's a bit early to be declaring - as the Independent are doing - that the poll definitely shows the second Brown bounce is over.
Still, with all the polls now showing Conservative leads between 4 and 7 points, we can at least end the year with a good idea of where the parties stand, even if we can't be quite so sure which way the trend is going.
UPDATE: The ComRes poll also including questions about how people would vote under various scenarios. Sadly, I don't think they tell us very much.
Firstly, they are not comparable to the standard voting questions. A voting intention question in a poll is normally quite an involved business - when ComRes do it involves three questions - how likely people are to vote, how they would vote, and how they would vote if forced to by law (plus the past vote and party ID questions used for weighting and reallocation of don't knows). When alternative voting intention questions are used there is a tendency to skip bits and just do it with one question, rather than go through that whole rigamarole again, therefore making them non-comparable. In this case, likelihood to vote is assumed to be constant throughout (and, in fact, the extra voting intentions aren't weighted by turnout), so if the Conservatives or Labour promising something energised or alienated their base and made them more or less likely to vote it wouldn't show up in these questions.
More importantly, they don't really get at what they are trying to. The argument that people may vote differently in or after a recession is largely based on psychological factors and loss aversion. If that argument is true then these would be psychological biases that people are not necessarily consciously aware of - we wouldn't expect them to show up in a survey like this. Secondly, people are rubbish predictors of how they will react to future events anyway. It was easily predictable that there would be a big boost in Labour support after Gordon Brown became leader, yet people consistently said they would be less likely to vote Labour with him there - they failed to forsee that they would want to give him a chance, or would be caught up and won over by his honeymoon in office. We can have no confidence that they will or will not be able to predict how they would react to a government calling an election mid-recession, or how they will look at politics after one.
The other questions, about how people would vote if the Conservatives did X, don't show us much either. If you give a prompt that only mentions one party and says something positive they would do, then miraculously it greatly increases the proportion of people who say they would vote for them. My favourite example is this one from MORI back in 2004, commissioned by UKIP. It found UKIP support in the European elections at 2%. They then asked "At the European Parliament elections the UK Independence Party will be campaigning nationwide for Britain to leave the European Union and put an end to unlimited EU immigration. Assuming the UK Independence Party were the only moderate party campaigning for this, which party would you vote for?" and found support for UKIP at 35%. Magic!
Going back to ComRes, the fact that putting the question from a Conservative angle (the Conservatives will spend less and not increases taxes) and a Labour angle (Labour will spend more, but will increase taxes) results in almost identical answers would be fascinating... if the questions were put independently to different people using a split sample. As far as I can tell, they were asked one after the other to the same people. Asking different groups of people the same question using slightly different wording can give you fascinating results. Asking the same people the same question using slightly different wording normally gives you the same answer.
