I've finally had a chance to look at the datasets for this week's Survation poll, and as I suspected might be the case, it turns out that the introduction of a new weighting procedure (weighting by recalled referendum vote) has had quite a significant effect - without it, the pro-independence vote would have had a slight lead. That's not to say that the reported result is necessarily wrong or misleading, but the important point is that at no stage during the referendum campaign did Survation ever have the Yes vote higher than 48% - so the fact that this poll would have had Yes above 50% if the old methodology had been used suggests that there are now more independence supporters out there than ever before. That's entirely in line with the findings of YouGov and Panelbase, both of which did have Yes in the lead.
I was tempted to use "Daily Record caught fibbing" as the title of this blogpost, because you might recall that they claimed that the independence numbers in the poll were Yes 47%, No 53%. John Curtice says that the Yes vote is actually 48%, and sure enough, when I looked at the raw numbers in the datasets, they worked out as Yes 47.52%, No 52.48%, which on the face of it ought to be rounded to Yes 48%, No 52%. However, there is a potentially innocent explanation here - even the "raw numbers" of weighted respondents in polling datasets are themselves rounded up or down, so they can sometimes give a misleading impression when things are very finely-balanced. It could be that Survation privately relayed to the Record that the correct Yes figure was 47%, without putting that in the datasets and without alerting Curtice. But I'm not sure whether we should take that on trust.
A really interesting set of figures tells us whether people who voted for each party in 2011 are now more or less likely to vote in a given way as a result of the referendum campaign. Obviously there is the usual meaningless effect of people who were never likely to vote for Party X in the first place claiming they are now even less likely to vote for Party X, so the most useful results relate to whether people are less likely than before to stick with their actual 2011 party.
Labour voters less likely to vote Labour : 21.9%
Liberal Democrat voters less likely to vote Liberal Democrat : 15.2%
SNP voters less likely to vote SNP : 8.1%
Conservative voters less likely to vote Conservative : 3.1%
It's easy to look at these numbers in a superficial way and think "great news for the Tories", but you have to bear in mind that we're starting from a baseline of an overwhelming SNP landslide in 2011. If the SNP lose only 8.1% of their voters from that contest and don't gain a single extra vote from elsewhere, it would still leave them with an overall vote share of 41.7% - comfortably a winning position (albeit probably not enough to retain an absolute majority in 2016). This should probably set our minds to rest about there being any danger of throwing the baby out with the bathwater - ie. the SNP aren't going to lose their No-voting heartlands as they chase Yes-voting Labour seats, or at least that won't happen unless the small percentage of disgruntled SNP voters identified in this poll are extremely heavily concentrated in certain geographical areas.
Murdo Fraser said the other day that he was looking forward to taking to the doorsteps of Perthshire with the message of "vote SNP, get Miliband". Well, Murdo, absolutely nothing you guys have tried over the last NINETEEN YEARS has succeeded in dislodging the SNP in Perthshire, so if even this wheeze doesn't work, maybe it'll be time to give up the ghost completely?