As Mike noted at the time the latest round of 2014Euro polls came back with a pretty wide range of results, ranging from UKIP leading by 11 points with ComRes, down to ICM putting them in 3rd and 4 points off the top.
So I did a little digging. What I think is causing at least the main part of the disparity is how the different pollsters treat certainty to vote responses. ICM operate a weighting system based on a responders self-described certainty to vote alongside whether they voted in the last election. ComRes uses a filter system to only count those describing themselves as 10/10 certain to vote.
If we go into the certainty to vote data we see that UKIP consistently has a significantly higher proportion of its supporters reporting a 10/10 certainty to vote, with ComRes giving them the largest lead.
What I suspect has happened is that ComRes has randomly had a sample at the upper end of UKIP’s certainty lead (the data to see if they consistently produce higher certainty leads for UKIP isn’t available from their past poll tables as far as I can see) and that their method of turnout filtering has magnified that to create the disparity we see compared to other pollsters.
Equally ICM’s method of down weighting those who did not vote in the last election might be hurting UKIP if a larger proportion of their support is coming from former non-voters than is normal.
To add another layer of complexity to how this will translate into actual election results, those areas that also have local elections at the same time are likely to see higher turnout.
UKIP’s best scores are being driven by the greater passion their voters have for voting and the European elections in particular, and if they can convert that passion into higher turnout then they’re likely to come out on top.
Edited to add:
Anthony Wells has an interesting take on the same subject.
If you look at the tables, we can work out what the polls would have shown using different methods, letting us compare like-to-like. So, if all three pollsters who’ve reported in the last couple of days only took those respondents who said they were 10/10 certain to vote the figures would be:
ComRes – CON 20%, LAB 27%, LDEM 7%, UKIP 33%, GRN 6%
ICM – CON 21%, LAB 30%, LDEM 7%, UKIP 29%, OTH 14%
YouGov – CON 21%, LAB 26%, LDEM 8%, UKIP 29%, GRN 10%
But if all three respondents included the answers of all respondents giving an intention to vote the figures would be:
ComRes – CON 21%, LAB 30%, LDEM 8%, UKIP 27%, GRN 7%
ICM – CON 26%, LAB 29%, LDEM 7%, UKIP 24%, GRN 6%
YouGov – CON 23%, LAB 27%, LDEM 9%, UKIP 26%, GRN 9%
(Not the catchiest twitter handle, but it’s mine).