Survation, UKIP and prompting for minor parties

There is a Survation poll in the Daily Star on Sunday with topline figures of CON 32%, LAB 34%, LD 12%, Others 22% (including UKIP 11%, Green 4%, BNP 2%). Clearly 22% is a very high total for Other parties, and while I don't have a spreadsheet of past UKIP polling figures, their highest rating in a GB Westminster voting intention poll that I'm aware of is this 10% from Harris in 2009, so this would be their highest.

The reason for these high figures is fairly straightforward: Survation prompted for the minor parties in their voting intention question. In other words, instead of asking something along the lines of "If there was a general election tomorrow, would you vote Con, Lab, Lib Dem or another party" they asked something like "If there was a general election tomorrow, would you vote Con, Lab, Lib Dem, UKIP, Green, BNP or another party" (in fact, they randomised the order of Con, Lab, LD & UKIP).

These days all the main pollsters include the name of the main three parties in their voting intention question, but none of them list the smaller parties. This makes a clear difference to the answers you get - if people are reminded of the existence of minor parties, then more people say they would vote for them. This Survation poll itself is good evidence of the effect - when they ran a poll without prompting earlier this month they found UKIP on 4% (though some of the difference could also have been between telephone and online methods). There is also the case of YouGov's Scottish election polling in 2007, where the prompt was changed half way through the campaign to include minor parties - the effect was to increase support for "others" from 11% to 19%, and to change support for the Scottish Greens from 4% to 9% (in the event, the Scottish Greens got 4%).

Now, I sometimes see supporters of minor parties complaining about pollsters not including their parties in the prompt and saying it is unfair. I suppose in many ways it isn't, and if one was arguing from first principles one might very well think that, given all the parties are on the ballot paper, they should all be in the prompt.

The reason other pollsters don't include minor parties in the prompt is, however, because in practice not including them produces accurate results. There is no particular logic to it, it is just what has worked in the past. At the last election no pollster included minor parties in their main voting intention prompt, and the polls were pretty accurate in their predictions of support for minor parties:

Including minor parties in the prompt would lead to significantly higher levels of support in polls, yet when compared to actual election results polls do not appear to be significantly underestimating support for minor parties. Certainly in the case of UKIP, most polling companies were pretty much spot on at the last election.