The datasets for last night's sensational BMG poll appeared late this afternoon. There's not a huge amount in them, because it's just a basic two-question poll, and BMG's datasets are in certain respects more limited than those produced by other firms. However, there are a few important nuggets of information -
* As I suspected, the Herald misled us badly by claiming that there is a 56%-44% majority against holding a second indyref before Britain leaves the EU. The question actually asked was : "In 2014 there was an independence referendum in Scotland. In your opinion, should there be another independence referendum held prior to Brexit negotiations being concluded between the UK and the EU?" Logically, anyone who thinks that an independence referendum should be held after the end of negotiations but before the actual date of Brexit would answer "no" to that question. That's not some sort of technical objection, because the recent Panelbase poll found that almost as many people thought there should be a referendum soon after negotiations are concluded (23%) as thought there should be one while negotiations are ongoing (27%). Therefore, the BMG poll doesn't tell us one way or another whether there is a majority for a referendum before Brexit, because that quite simply wasn't the question asked.
Indeed, on the face of it, the 44% in the BMG poll in favour of a referendum before the end of negotiations is much higher than the equivalent 27% in the Panelbase poll - although the Yes/No format of the BMG question may have effectively forced some supporters of a referendum to plump for "yes".
* As in the previous BMG poll, it looks as if the Yes vote may have been significantly downweighted due to 2014 vote recall, because virtually as many people in the unweighted sample recall voting Yes in the first indyref as recall voting No. Rolfe suggested on the previous thread that this is encouraging, because it perhaps means that some people who decided to vote No at the last minute and regretted it may be falsely claiming that they voted Yes. There's no hard evidence of that happening, but if by any chance it has, it would screw up the weightings and potentially lead to the Yes vote being slightly underestimated.
* The message of the recent YouGov aggregate figures is repeated - people who voted Yes in 2014 and then Leave in 2016 are potentially a problem for the next Yes campaign, because 19% of them have now switched to No (after Don't Knows are excluded). In proportionate terms, although not in terms of absolute numbers, that's much higher than the 8% of No/Remain voters who have switched to Yes. But, of course, this is an opportunity in disguise - if Yes can bring some of the straying voters back home (while holding on to what they currently have), it would be enough in itself to nudge them into the lead.
* Yes/Leave voters look very much like the 'swing respondents' on the question of when a referendum should be held, because they split practically 50/50 on whether there should be an indyref before Brexit negotiations are concluded. By contrast, Yes/Remain voters favour a very quick indyref by a whopping 5-1 margin.
* People who consider themselves to be left-wing break 2-1 for Yes, while right-wingers break 3-1 for No. That's not actually an advantage for No, because obviously - this being Scotland - there are far more left-wingers than right-wingers. However, the single biggest grouping is made up of self-defined centrists, who favour No by a razor-thin 53-47 margin.
* Yes stay in the lead among men (they were slightly ahead even last month), and have cut the gap among women to eight points. They are also ahead in every age category up to 44. The relatively small Yes deficits in the 45-54 and 55-64 categories mean it is certain that there is a significant Yes lead among under-65s in general. The killer, as ever, is the over-65 category, where No are ahead by an almost 3-1 margin. That problem is magnified by the fact that older people are considerably more likely to turn out to vote.
* * *
If you enjoy my writing, you can follow me on Twitter here, like the Facebook page here, or make a donation here.