YouGov/Sun - CON 31, LAB 34, LD 7, UKIP 14, GRN 8

YouGov's first poll of the year is out tonight, with topline figures of CON 31%, LAB 34%, LDEM 7%, UKIP 14%, GRN 8%.

YouGov have made a couple of methodological changes to start the election year. The first and most interesting is to include UKIP in the main prompt for voting intention. Prompting is something I've written about here many times before, most recently here. It's tricky because it really can make a difference, yet there is no real way of knowing when it is appropriate and when it isn't. There are instances when pollsters have overestimated the level of support for minor parties because they prompted when they probably shouldn't have (YouGov & UKIP in the 2004 European election, and the Greens in the 2007 Scottish election), but go back to the 1980s and polls that failed to prompt for the Liberals & SDP tended to underestimate their support. It's clear from history that you can both get it wrong by prompting when you shouldn't, and get it wrong by failing to prompt when you should.

I've written before about the difficulties of making the judgement call on this. There is no obvious way of drawing the line - whenever I write about it in the comments section people make helpful comments saying "why not if the party is in third place" or "if the party is over x%" or whatever... but all these are utterly arbitrary - perfectly reasonable in themselves, but no help in getting it right. It's not even clear what we should be looking for - it it a certain level of support, or a level of public awareness and familiarity, or a level of media coverage?

In the event however the difficult decision pretty much made itself. It's something YouGov have been quietly testing on and off over the years, and it seems to be making less of a difference. Testing at the end of last year showed about the same level of support for UKIP in prompted polls as in unprompted polls. Presumably UKIP are established enough in the public mind for prompting not to make a difference, at which point the decision became a simple one. Note that the fairly low UKIP score in today's YouGov poll - 14% - is not a result of the change, in our testing last year we were showing UKIP at around 17% when prompted, at at time when they were around 16%-17% in unprompted polls.

With YouGov shifting over it means three companies (YouGov, ComRes and Survation) now include UKIP in the main prompt, Populus, Opinium, Ipsos MORI, ICM and Lord Ashcroft polls do not. When ComRes made the switch they too seemed to show very little difference in levels of UKIP support (though their online polls seem to have made some changes to turnout weighting at the same time) but just because prompting doesn't make much difference to YouGov polls, it does not follow that it won't make a difference anywhere else - in particular, the impact of prompting may be very different in an internet survey with two pages to click through than in a telephone survey with a human interviewer. What is the correct approach for one sort of polling will not necessarily be the correct approach for another company's polls.

The other change in YouGov's methods is much smaller, a tidying up of the sample spec to try and reduce some of the oversampling and reducing the amount of weighting needed. The overall quota targets are the same and the weighting remains exactly the same so there should be no difference at all in the published voting intention figures. The only difference anyone might notice is in crossbreaks: YouGov have started to include ethnicity in the sample quotas for London, which may have an impact in the London crossbreak.

UPDATE: I somehow managed to miss the first Populus poll of this year this morning - figures there were CON 34%, LAB 36%, LDEM 9%, UKIP 12%. Tabs are here.