ICM/Guardian - CON 28, LAB 34, LD 11, UKIP 18

The Guardian have released this month's ICM poll, the first telephone poll conducted since the local elections. Topline voting intention figures with changes from a month ago are CON 28%(-4), LAB 34%(-4), LDEM 11%(-4), UKIP 18%(+9!).

The 18% for UKIP is the highest that ICM have shown and, more strikingly, the highest any company have shown. I will advise my usual caution about polls showing extremes and records, more often than not they tend to be outliers. ICM's methodology tends to produce some of UKIP's lower figures (ICM reallocate some don't knows to the party they voted for last time, weight down people who didn't vote last time and interview by phone... all things that tend to produce lower UKIP scores) so it is particularly surprising to see ICM with a record breaking UKIP score. While the scale of the UKIP increase may well be a bit of a blip though, the broader trend is the same as other companies - they are all showing a surge in support for UKIP.

It does make one ponder about how high we might see UKIP in some other companies' polls. If YouGov have them as high as 16% or 17%, if ICM have them at 18%, what might we see from companies like Opinium or ComRes who tend to show higher levels of UKIP support?

More unexpectedly the ICM poll also found a jump in support for the BNP, up to 4%, the highest any poll has had then at for years. This is strange. The BNP have certainly not had any great publicity boost, at the local elections they seemed essentially moribund. It may just be an odd sample, or perhaps as Tom Clark suggests it is just a case of confusion amongst respondents, with some people getting the names of the BNP and UKIP mixed up.

ICM also asked about voting intention in an EU referendum, finding voting intention fairly evenly balanced - 40% would vote to stay in (22% definitely, 18% probably), 43% would vote to leave (32% definitely, 11% probably).

UPDATE: ICM tabs are up here. Topline figures without reallocation of don’t knows would have been CON 27%, LAB 35%, LDEM 9%, UKIP 19%, BNP 5%.

That strange boost of support for the BNP is almost wholly amongst women, almost wholly amongst C2s, almost wholly amongst over 65s and almost wholly in Wales. The unweighted number of 2010 BNP voters in the sample was 1, increased to 18 by weighting. What that strongly suggests to me is that there was one little old C2 BNP-voting Welsh lady who got a very high weighting factor, and probably makes up almost all of that 4%! Such things happen sometimes, but it means the BNP blip is probably just a data artifact that can be ignored.