I mentioned in my post this morning that my decision to introduce a Poll of Polls on this blog was triggered by the absurd misreporting of one or two individual polls that showed improbably high leads for the No campaign - and, yes, I was mainly talking about polls from Progressive Scottish Opinion, a non-BPC pollster with appalling standards of transparency and a track record of wild volatility in their Holyrood polling. In a way I was hoping we'd finally seen the last of them, but perhaps we should be glad we haven't, because their first poll in several months has shown a huge swing towards the pro-independence campaign.
Do you believe Scotland should become an independent country?
Yes 34% (+7)
No 54% (-2)
When Don't Knows are removed, it works out as...
Yes 39% (+6)
No 61% (-6)
As you can see, the real referendum question was not posed in this poll, which is rather peculiar, because when I asked about this last autumn I was told that PSO do use the real question. It appears that they've since had a good hard think about how they can possibly improve their dubious methodology, and have come to the conclusion that their question was just far too close to the thing they are actually trying to measure. However, I can't see anything in the new wording that would benefit the Yes camp - if anything the opposite might be true, because the replacement of the word "be" with "become" subtly draws attention to the position of No as the status quo option. So there's no reason to doubt the extent of the swing to Yes in these numbers - unless of course it was simply caused by PSO's notorious volatility as a pollster.
And the swing is what matters here. John Curtice has today gone through the motions of effectively saying "yes, PSO are one of the most No-friendly pollsters, but that doesn't necessarily mean that they are wrong, and if by any chance they're right this poll is a reminder that Yes would still have a lot of ground to make up". But a "reminder" is all it is - that's not new information, any more than it would be new information to point out that if ICM and Panelbase happen to be correct, Yes do NOT have a lot of ground to make up, and are already almost neck-and-neck. We won't find out which pollsters are right and which ones are wrong until polling day, so for now all we can do is look at the trend shown by each firm and see how it fits into a broader pattern. With Don't Knows excluded, PSO are today suggesting a swing of 8% to Yes since September, which is at the upper end of the range of swings shown by other pollsters - ICM are also showing a swing of 8% over the same timescale (although in their case that has taken Yes from 40% to the apparent brink of victory at 48%), while at the other end of the scale Panelbase are suggesting only a 3% swing since September. So the new information we have today slightly increases the chances that the movement to Yes over the last few months has been bigger than some of the more conservative estimates have implied. (I emphasise the word 'slightly' - I'd be more confident if we were getting these numbers from a pollster other than PSO.)
I said earlier that Progressive Scottish Opinion aren't exactly the most transparent of pollsters, and the best illustration of that point is the fact that I can't actually work out something as basic as the data collection method for today's poll - was it face-to-face, was it online, or was it telephone? Neither the Sunday Mail article nor John Curtice's analysis seems to bother telling us (unless I skimmed over the crucial bit). PSO's website bills the company as both an online and face-to-face pollster, and reveals that fieldwork for their online polls is conducted by their "partners" YouGov (let joy be unbounded). Given that this poll took place over just two days, presumably the balance of probability is that the fieldwork was online and used the YouGov panel, which we can reasonably suspect to be one of the more No-friendly online panels. However, the weighting procedures will have been decided by PSO rather than YouGov, which clearly has the effect of producing a somewhat bigger No lead than even YouGov do. This may be because PSO no longer weight by past vote recall.
Nevertheless, PSO are no longer the outright most No-friendly pollster - with these new numbers they've allowed themselves to be 'overtaken' by Ipsos-Mori. The latter haven't published a poll for over two months (although we know they've been conducting unpublished polls since then), and the trend in that early March poll was somewhat out of line with what other pollsters were showing. So it'll be fascinating to see if the next Ipsos-Mori poll to be conducted for public consumption shows any convergence with the norm.
Meanwhile, our dear friends over in the official anti-independence campaign continue with their relentless quest to shed any last vestiges of self-respect. When the PSO poll was published a few hours ago, they posted a graphic on Twitter comparing it to the YouGov poll from ten days ago, rather than to the last poll conducted by PSO. By sheer coincidence, this piece of jiggery-pokery gives the grossly misleading impression that there has in fact been a big swing to the No campaign, rather than the other way around. But don't worry - you can rest assured that they'll maintain consistency at all times, and when a new Panelbase poll appears in the near future, they'll be comparing it with this PSO poll, rather than the last Panelbase poll.
Oh do stop sniggering - these guys have integrity.
In case you're wondering, PSO's reappearance doesn't affect the Poll of Polls because they don't adhere to British Polling Council rules. The mean average for all BPC-affiliated pollsters remains at Yes 35.8%, No 47.0%, and with Don't Knows excluded it's Yes 43.2%, No 56.8%.