Survation have won the race to conduct the first poll since Alex Salmond humbled Alistair Darling in Monday night's BBC debate...
Should Scotland be an independent country?
Yes 47% (+4)
No 53% (-4)
With Don't Knows taken into account, the figures are...
Yes 42% (+5)
No 48% (-2)
The dramatic changes in this poll completely wipe out the supposed gains that the No campaign made in the last Survation poll, which caused a flurry of excitement across the anti-independence media, given that it was the first poll to be conducted since Alistair Darling's largely fictional "victory" in the first TV debate. Yes are now back to the all-time high with Survation of 47% that they first reached in June, and held onto for the next two polls from the firm. I expect one or two commentators will try to make the case that Salmond's success on Monday ought to have generated at least the same temporary honeymoon effect that No enjoyed after the first debate - in other words Yes shouldn't simply have returned to the status quo ante, but should instead have gone well beyond 47%. The problem with that claim is that, as you know, I don't think there's the slightest evidence that No ever had a honeymoon after the first debate, even a fleeting one - the most likely explanation for the bounce they secured in that one individual poll is that it was an illusion caused by Survation's ongoing problems with having to upweight certain groups of respondents by an extreme amount.
Just as importantly, those problems may mean that the progress made by Yes in this poll is, in any case, even more significant than a cursory glance would suggest. On each and every one of the three previous occasions when Yes reached 47% with Survation, there was always something to be found in the datasets that left a big question mark over whether things were really as good as they appeared. In the first two of those polls, Yes were heavily reliant on an unusually good showing among the tiny sample of under-25s, who had been upweighted by roughly three-fold, thus hugely magnifying the effect of any potential error caused by random sampling variation. And in the third poll, Yes had a thoroughly implausible lead in the sample from the South of Scotland electoral region, which was upweighted two-fold. Tonight's poll is the very first time that Yes have hit 47% with Survation without there being any obvious alibi in the datasets for No. As usual, young people and respondents from the south have been upweighted sharply, but there's nothing out of the ordinary in the results from either of those samples. Under-25s are in fact the third most No-friendly age group in this poll out of six, while the south is the second most No-friendly region out of eight (after only the Lothians). So there's a case to be made that the dubious results from previous Survation results are masking the fact that Yes are now at a higher level of support than ever before, and that there has indeed been at least a modest post-debate bounce that has gone beyond simply returning us to the status quo ante. If so, we'll probably find out for sure at the weekend, because there seem to be Panelbase and YouGov polls in the works.
At an absolute minimum, we're certainly entitled to conclude that Yes are now at a higher level of support than they were during the first half of this year, because Survation consistently had them at either 44% or 45% for several months in the late winter and early spring.
As always, we have to put a health warning on Survation's headline figures, because they have once again failed to join the new orthodoxy of weighting by country of birth. There's no way of knowing for sure whether they have too many English-born people in their sample, but as that's a problem that has been observed in the results of all of the other three online polling firms, you'd think the balance of probability is that it applies to Survation as well. If so, the No lead in this poll should probably be a smidgeon lower.
Respondents in the poll were asked who they thought won Monday night's debate - as with the equivalent poll after the first debate, the results are virtually meaningless, because they'll have been tainted by people's exposure to media reporting of the debate, which in turn will have been influenced by the instant poll from ICM. Once again, the person who the media reported as the winner is given a wider margin of victory in the Survation poll than in the instant poll, which is exactly what you'd expect. However, one finding that can't be so easily dismissed is that 35% of respondents who are currently undecided say the debate has made them more likely to vote Yes, compared to just 8.5% of undecideds who say that it has made them more likely to vote No.
In the comments section below, Colin has raised an issue that I intended to cover a day or two ago (before I got sidetracked). The gist of what he said is : "It's good that Yes are making progress, but as people are already voting by post, isn't it a problem that Yes are still behind?" Well, that depends on our reason for being confident that Yes can win - it might be that we think they're behind but can close the gap, or it might be that we think the polls are understating them slightly and that they're already ahead or level. Under the latter scenario, the fact that voting is underway obviously poses no problem.
But if Yes really are a few points behind right now? It's important to keep a sense of perspective - in spite of the huge number of postal vote applications that have been made, we're still only talking about 1 in 6 of the registered electorate. The turnout will probably be higher among postal vote applicants, which might take the figure up to 1 in 5. For the sake of argument, let's take a very conservative estimate of where the Yes vote might be at the moment - perhaps 43%. If the postal votes being cast right now reflect that state of play, then Yes would need the equivalent of 52% support by polling day to overturn that small disadvantage. It's important to stress, though, that they wouldn't need 52% support in the polls, which from now on will (or at least should) ask people to give either their voting intention or how they have already voted. So a Yes lead in any poll will still mean what it says - it's just that it'll be a touch harder to achieve the swing necessary to get there, given that a significant minority of votes are already cast in stone, and can't be changed even if some people end up regretting what they've done.
That said, it may well be that postal vote applicants are disproportionately likely to be committed voters, and that people who are undecided or open to persuasion are more likely to be planning to vote on polling day - in which case the impact of postal voting on the Yes campaign's chances of closing the gap will be very limited.
* * *
Swing required for 1 out of 6 pollsters to show Yes ahead or level : 2.0%
Swing required for 2 out of 6 pollsters to show Yes ahead or level : 3.0%
Swing required for 3 out of 6 pollsters to show Yes ahead or level : 4.5%
Swing required for 5 out of 6 pollsters to show Yes ahead or level : 6.5%
* * *
SCOT GOES POP POLL OF POLLS
There have been a spate of very good polls for Yes recently, and it's only really been the one bad poll from Survation a few weeks ago that has been suppressing the Yes vote in the Poll of Polls. With that factor finally removed, it's no surprise to see the No lead slump to a new all-time low. With Don't Knows taken into account, it's fallen to below 10% for only the second time, and at 9.8% is 0.1% lower than the previous record low.
MEAN AVERAGE (excluding Don't Knows) :
Yes 44.4% (+0.8)
No 55.6% (-0.8)
MEAN AVERAGE (not excluding Don't Knows) :
Yes 38.7% (+0.9)
No 48.5% (-0.3)
MEDIAN AVERAGE (excluding Don't Knows) :
Yes 43.7% (+1.1)
No 56.3% (-1.1)
(The Poll of Polls is based on a rolling average of the most recent poll from each of the pollsters that have been active in the referendum campaign since September 2013, and that adhere to British Polling Council rules. At present, there are six - YouGov, TNS-BMRB, Survation, Panelbase, Ipsos-Mori and ICM. Whenever a new poll is published, it replaces the last poll from the same company in the sample. Changes in the Poll of Polls are generally glacial in nature due to the fact that only a small portion of the sample is updated each time.)