Can standard voting intention responses mislead?
Look at what happens when a more refined approach is used
Away from expenses and sleaze for a moment let’s look at a polling issue that’s been churning over in my head. It is possible that pollsters might be being mis-led when they ask standard voting intention questions? We know the sort of thing – “If there was a general election tomorrow which party would you vote for – the Conservative, Labour, the Liberal Democrats or some other party..”
For when these are being put in the standard format there’s good evidence that respondees often think nationally and name the party they most support – which isn’t always the same as the party that they would vote for where they happen to live. For here there might be a personal vote for an incumbent or tactical voting decisions which might make a difference.
For a dramatic example of this working in practice let’s go back to the massive 34,000 sample PoliticsHome survey that was carried out in July-August 2008 and was published the following month. There’s a full document that’s available for downloading which I recommend that everybody considering having a general election flutter should read.
For as is explained in the comprehensive report the poll adopted a staged approach to asking the voting intention question.
“Respondents were first asked a normal voting intention question. They were then asked whether they were likely to vote for their first choice of party, or vote tactically for a different party. Finally they were asked “And thinking specifically about your own constituency and the candidates who are likely to stand there which party’s candidate do you think you will vote for in your own constituency at the next general election?â€
Just look at the chart above to see the impact with both Labour and Tory voters in key LD>CON marginals. The standard polling approach just gives you what’s described as the “national vote”. Compare that with the “constituency vote” that the staged approach uses.
I’ve used the LD>CON marginals example simply because it shows more dramatically this working in action – but there’s evidence that this can operate in favour of all parties depending on the circumstances in each seat. My assumption for the coming general election, for instance, is that in LAB>CON marginals the the Lib Dem vote will be squeezed quite drastically and, again, this might not be being shown in standard polling.
One reason why I think you have to dig into the methodologies of the different pollsters is to find out how near their question format is to the second range. Of the existing firms ICM has the most constituency-based question which is why, sometimes, its figures can be out of line. I believe that the other pollsters could learn from ICM.
As I reported last week my understanding is that PoliticsHome will be updating its marginals survey ahead of the party conference season in September.