Update about the YouGov poll : a big methodological change has been made

It turns out that the results of last night's YouGov poll are not directly comparable with previous YouGov polls, because Peter Kellner's firm has now fallen into line with Survation and Panelbase by weighting its figures to recalled referendum vote.  As you'll remember, it was the SNP's opponents who were unhappy about the failure to introduce that procedure until now, because it meant (on the face of it) that there were too many Yes voters in the YouGov sample, thus potentially leading to the SNP's vote, and the Yes vote in a hypothetical second independence referendum, being overestimated.  Curiously, though, the change in methodology has not actually led to a downweighting for the SNP in the new poll, because all of a sudden there are (slightly) too few Yes voters in the unweighted sample, and (slightly) too many No voters.

I find it hard to believe that something as unexpected as that could have happened by chance.  It may have done, but I think a more likely possibility is that YouGov have also made a methodological change at the point of data collection, and are now ensuring that they invite fewer Yes voters to take part in their polls, in an attempt to minimise the amount of weighting that is required.  If that is indeed the case, it means that it's literally impossible to tell what the results of the new poll would have been if the old methodology had still been in operation.  The only way we can make a comparison between this poll and the last one is by putting our faith in the new doctrine of weighting by referendum vote, and making a rough calculation of what the results of the last poll would have been if that weighting had been applied.  If we do that, even the trivial 2% drop in the SNP's lead is effectively scrubbed out, because of course the SNP's vote would have been a touch lower in the last poll.

The voting intention figures in the new poll for a hypothetical second independence referendum are Yes 49%, No 51% - which is identical to what they probably would have been in the last poll if the same weighting had been applied.  Under the old methodology, YouGov had been reporting a slight Yes lead, but at least now there are no more technical excuses for our opponents - this poll has recorded a hefty 4% swing in favour of Yes since the referendum, and that's undoubtedly a real swing because respondents are now weighted strictly in line with how they voted in September.  All polling firms that have asked the independence question over the last few months are united in showing a pro-Yes swing of a few percentage points, bringing the race into "statistical tie" territory - ie. where it's not possible to tell whether Yes or No are ahead, due to the standard 3% margin of error.  That, incidentally, makes an utter nonsense of yesterday's bizarre claim from the New Statesman's Jason Cowley that "Scots are not suffering from buyer's remorse" and that "a majority still wish to remain part of the UK".  Evidently we should stop listening to scientifically-conducted opinion polls, and start listening to Mr Cowley's gut feelings about what must be true.

For the first time, there's a crossbreak in the YouGov datasets specifically covering the portion of the sample that voted Labour in 2010, and Yes in the referendum.  (Hopefully that isn't the latest mutation of the Kellner Correction, and is simply there for information purposes.)  There are no great surprises - we always knew that the SNP wouldn't have been able to build up this huge lead unless ex-Labour Yes voters were now mostly in their column, and sure enough 81% of those people have made the journey across.  Just 18% have remained faithful to Labour.  Perhaps most importantly, 99% of them would vote Yes again if given the chance.

An enormous 88% of these "Labour-Yes switchers" think that Ed Miliband should leave open the possibility of a post-election deal with the SNP.  So perhaps he ought to be cautious about listening to the siren voices in the London commentariat who would have him believe that he can somehow regain popularity in Scotland by doing the complete opposite.

I mentioned Jim Murphy's dismal leadership ratings last night, but the true humiliation for him is that slightly more people (29%) think David Cameron is doing well as Prime Minister than think Murphy is doing well as Scottish Labour leader (26%).  Admittedly, Murphy's net disapproval rating isn't quite as bad as Cameron's, because there are more people who think that Cameron is performing poorly.

There's an old saying that "divided parties don't win elections", and that's where this poll has truly devastating news for Labour.  59% of voters think Labour is divided, compared to just 10% who say the same about the SNP.

If the unionist parties are hoping that they can turn things around in the formal campaign period, they'll be dismayed to learn that a formidable 69% of people planning to vote SNP say there is "no chance at all" that they will change their minds.  The equivalent figure for the much smaller group of people currently planning to vote Labour is 59%.

Incredibly, more respondents (27%) say that the SNP is the party best able to keep Britain in the European Union than say the same about any other party.  What a shocking indictment of the state of the pro-European lobby at Westminster (is there one anymore?)