BMG were supposed to put out their final EU poll yesterday, but it was put back for a day because of the murder of Jo Cox and eventually emerged this morning. BMG carried out parallel telephone and online polls, and unlike ICM who no longer find any difference at at in their dual-mode experiments, BMG continue to find a big gulf:
- In their online poll BMG found topline figures of REMAIN 41%, LEAVE 51%, Don’t know 9%. Like other recent polls this reflected a big shift towards leave, with remain down by 2 points, Leave up by 6 points
- In their telephone poll BMG found a small lead for Remain – REMAIN 46%, LEAVE 43%, Don’t know 11%
BMG also reallocated don’t knows based upon how they answered other EU questions in their phone survey, this produced final topline figures of Remain 53%, Leave 47%. Note that the fieldwork for the poll was conducted between the 10th-15th June, so wholly before the death of Jo Cox. Full details of the polls are here.
Secondly today we have a poll from Surveymonkey. Many readers will recognise Surveymonkey as a software platform for conducting surveys – the poll was conducted by randomly picking some of the people taking other Surveymonkey surveys and then directing them an EU survey – so very different from panel-based online surveys. Surveymonkey did the same at the general election with somewhat mixed results: their poll had the Conservatives six points ahead of Labour, so in that sense was far more accurate than other polls… but the reason was because they significantly underestimated both Conservative and Labour support, so actually had a larger average error than some other polls. Anyway, on the EU referendum they found topline figures of REMAIN 48%, LEAVE 48%, No answer 4%. Make of that what you will – full details are here. Fieldwork was between the 8th and 15th June, so again, before Jo Cox’s murder.
Third is a poll from a company called qriously, whom I have never previously heard of. As far as I can tell the poll was conducted by embedding survey questions in adverts on smartphone apps. The data is weighted by age, gender, region, past vote and education so is making an effort to produce representative results – the question is to what degree, if at all, the sampling method is capable of producing a representative sample, which we cannot really tell. Their poll between the 13th and 16th June found topline figures of REMAIN 40%, LEAVE 52%, Don’t know 9% – so more favourable towards Leave than any other polling. They also released figures for people interviewed on Friday morning after Jo Cox’s murder, which were REMAIN 32%, LEAVE 52%, Don’t know 16% – a significant movement from Remain to don’t know. I would treat these Friday figures with a lot of caution, it’s a method that is unproven in political polling, the shift from remain to don’t know doesn’t make much intuitive sense as a reaction to the murder, and most importantly, the fieldwork was only conducted on a weekday morning, which may itself skew the make up of the sample. I would strongly suggest waiting to see what other polls conducted after the murder show. Details of the polling are here.
Tonight we should get new figures from at least YouGov in the Sunday Times and Opinium in the Observer, possibly others.