It's been ages since I last calculated a polling average on independence, so I thought it might be time to give it a spin. For those who don't remember, the Scot Goes Pop Poll of Polls takes into account only the most recent poll from each firm that has asked the independence question at some point in the last six months. Back during the indyref, I had an exchange with a pollster (I think it was someone from Ipsos-Mori) who was insistent that my method didn't make sense, and that a Poll of Polls should take into account all of the most recent handful of polls regardless of which firms had conducted them. I pointed out to him that the 'house effects' in indyref polls were so extreme that his preferred method would generate crazily misleading trends - if you went from one average that was mostly based on polls from No-friendly firms (such as YouGov) to another average that was mostly based on polls from Yes-friendly firms (such as Panelbase), you'd get the firm impression there had been a sharp swing to Yes even if no such thing had actually happened.
The gap between pollsters is no longer as extreme these days, but there are still differences. Panelbase has moved to the other end of the spectrum and is now a No-friendly pollster, usually reporting a Yes vote that is a little lower than one or two other firms such as Survation. Needless to say YouGov remain a firmly No-friendly outfit, and as you know I've always been a bit cynical about them. It's hard to escape the conclusion sometimes that they start from the assumption that the Yes vote should be on the low side and work backwards to find a methodology that will produce that outcome. During the indyref, when they were still under the control of Labour supporter Peter Kellner, they used the notorious "Kellner Correction" to split SNP voters into two distinct categories and weight them separately, which magically produced figures that were much more No-friendly than other online firms - until the closing weeks of the campaign, when the small SNP group that they were artificially upweighting showed an enormous swing to Yes. That was why Damian Lyons-Lowe of Survation argued on the evening of September 18th that the campaign had not had much impact on voting intentions, while Peter Kellner standing right next to him was equally insistent that the swing during the campaign had been dramatic. It's impossible to know who was right, although I do suspect that if YouGov had been picking up a pro-Yes swing several months before polling day rather than a couple of weeks before, they might have changed their methodology again to make that swing look less significant. Not because they were consciously trying to 'rig' anything, but because Kellner was bringing unionist preconceptions to the table and was much more likely to search for reasons why Yes was being overestimated, rather than the reverse.
Anyway, for today's update of the Poll of Polls, four polls are taken into account: very recent polls from Panelbase and Survation, a poll from YouGov that was conducted in early June, and a poll from Ipsos-Mori that was conducted in early March. Obviously March is quite a while ago now, but that probably doesn't make too much difference - polling numbers on independence have been relatively stable of late.
SCOT GOES POP POLL OF POLLS
Should Scotland be an independent country?
* * *
If you haven't seen it yet, here is the second edition of Phantom Power's groundbreaking Nation documentary series starring Lesley Riddoch. This time the focus is on Iceland. Incidentally, did you know that more people speak Welsh than Icelandic? And yet try telling the people of Iceland that their national language is "useless" and that they should just get on with speaking English like normal people do...