New TNS-BMRB poll gives the SNP a 10% lead - but the fieldwork dates overlap with the polls that gave the SNP much bigger leads, so there's no "narrowing of the gap"

A new full-scale Scottish poll is out tonight from TNS-BMRB...

Scottish voting intentions for the May 2015 general election (TNS-BMRB, 14th January - 2nd February) :

SNP 41%
Labour 31%
Conservatives 16%
Greens 6%
Liberal Democrats 4%
UKIP 2%

Although this poll shows a smaller SNP lead than recent polls from other companies, it's important to stress that this categorically does NOT constitute "a narrowing of the gap" (as a Labour troll is already trying to have us believe on Twitter).  We haven't previously had a full-scale Scottish poll from TNS-BMRB since the referendum, so we don't have any baseline numbers to work from.  It's quite possible that a marginally better showing for Labour is simply an in-built "house effect" of TNS-BMRB's unusual methodology, in much the same way that they were consistently one of the most No-friendly firms during the referendum (until the Great Convergence at the very end).

The only poll this isn't different from is the last Panelbase poll, which in fact was completely identical in as much as it had the SNP on 41% and Labour on 31%.  But the afterlife of that result ought to teach us an important lesson, because we later found out that Panelbase had used a dodgy methodology - they had asked a leading question immediately before the main voting intention question.  Could something similar have happened here?  It's unlikely that TNS would use an unusual question sequence, but there's always the perennial issue of past vote weighting to consider.  For their later referendum polls, they did use the more reliable 2011 weighting, but not until they were coaxed into it after one of their polls produced an obviously skewed result.  Is there a chance they will have foolishly thought to themselves "this is a Westminster poll, so we need to revert to Westminster weightings"?  Let's hope not, but it shouldn't be forgotten that Panelbase did make that mistake in October, and we therefore can't know for sure until we see the datasets.

[UPDATE, 11.45am : Shockingly, it turns out that TNS have indeed used 2010 weighting.  Although they diluted the effect by using 2011 weighting as well, this may partly explain them showing a lower SNP lead than other firms.  More details in a fresh post HERE.]

Assuming that there isn't any issue with the weighting procedures, the most likely explanation is simply the fact that TNS are the only firm that still use the old-fashioned face-to-face approach.  As Calum Findlay points out in the comments section below, this method has produced unusually favourable results for Labour in recent Britain-wide polls, and it could be that the same is happening in Scotland - albeit that's still only enough to put Labour a "mere" 10% behind in Scotland, as opposed to the outright leads they've enjoyed in the Britain-wide TNS polls.  To put it in perspective, if tonight's numbers were accurate, the Electoral Calculus predictor suggests that the SNP would win 38 of the 59 Scottish seats at Westminster - an enormous gain of 32 seats.  Labour would be cut in half from 41 to 20, and the Liberal Democrats would be completely wiped out (albeit we know in reality they're likely to hold Orkney & Shetland due to the special political environment in that part of the world).

UPDATE : Now this is crucial - I've just seen a claim on Twitter that the fieldwork dates for the poll ran from mid-January through to 2nd February.  That's not unusual for TNS, who as a result of their face-to-face methodology often produce polls that are already well out-of-date by the time of publication.  It means the poll partly overlaps with the fieldwork for the Ipsos-Mori poll which gave the SNP an enormous 28% lead, and it wholly overlaps with the YouGov poll that gave the SNP a 21% lead.  (In fact, the YouGov poll should really be considered more recent than TNS, because it was entirely conducted towards the tail-end of the long TNS fieldwork.)  There's also considerable overlap with the sensational Ashcroft constituency polling that put the SNP in such a commanding position.  All of this makes it literally impossible that we're looking at a "narrowing of the gap" (as David Maddox of the Scotsman has now, to his eternal shame, joined the Labour trolls in claiming).  The lower SNP lead must, as stated above, be a house effect of the TNS methodology - unless of course this is an out-and-out rogue poll, but that's much less likely.

As we always used to say during the referendum campaign, when different firms are contradicting each other by light-years, there's no way of knowing for sure who is closest to the truth.  So to a limited extent Labour might be justified in taking some comfort from this poll, in spite of the fact that there has been no narrowing.  If the TNS methodology is exactly right, a 10-point deficit with three months to go would leave them with a fighting chance of winning more seats than the SNP, because the electoral system works in their favour if they are only slightly behind on the popular vote.  However, they would still require a very significant swing to whittle the SNP's lead down sufficiently, so the challenge would still be formidable.  And more to the point, even if we assume that the more SNP-friendly pollsters are getting something terribly wrong, it's very hard to believe that both YouGov and Ipsos-Mori would be showing SNP leads of over 20 points if the real gap was only 10.  So, in truth, all tonight's poll does is slightly increase the chances that the real position is somewhere in between the two extremes - which would suggest Labour have a huge deficit in the mid-to-high teens.  In normal circumstances, that gap would be considered utterly insurmountable, and I'm struggling to see why we shouldn't reach that conclusion in this case.

And isn't it curious that I've just referred to YouGov and Ipsos-Mori as "SNP-friendly pollsters"?  There seems to be no real correlation at all between which firms were favourable to the Yes campaign, and which are favourable to the SNP now.

Yes-friendly pollsters in the referendum campaign :

Survation
Panelbase 
ICM

SNP-friendly pollsters after the referendum :

Ipsos-Mori
YouGov
Survation

No-friendly pollsters in the referendum campaign :

Ipsos-Mori
YouGov
TNS-BMRB

Labour-friendly pollsters after the referendum :

TNS-BMRB
Panelbase

Although whether Panelbase should really be in the latter list is open to question, because they've produced three post-referendum polls, all using different methodologies, and all with very different results.  ICM have only conducted one post-referendum poll with a reasonably middling result, so it's too early to say where they're going to slot in.

*  *  *

SCOT GOES POP POLL OF POLLS

It feels very peculiar to receive the results of a new full-scale Scottish poll, and yet to exclude it from the subsequent Poll of Polls update.  But the rules I laid down in the autumn for the Poll of Polls are very clear - full-scale polls are included if the fieldwork was at least partly conducted within the last seven days.  I'm writing this in the early hours of the 10th, and both the TNS and YouGov polls were concluded on the 2nd.  So they're both out, and instead we revert to an update based on the Scottish subsamples from GB-wide polls conducted entirely within the last seven days - three from YouGov, two from Populus and one from Ashcroft.

Scottish voting intentions for the May 2015 UK general election :

SNP 42.2% (-3.4)
Labour 26.0% (-0.5)
Conservatives 19.2% (+3.0)
Liberal Democrats 5.5% (+0.6)
Greens 3.0% (-0.3)
UKIP 3.0% (-0.8)

(The Poll of Polls uses the Scottish subsamples from all GB-wide polls that have been conducted entirely within the last seven days and for which datasets have been provided, and also all full-scale Scottish polls that have been conducted at least partly within the last seven days. Full-scale polls are given ten times the weighting of subsamples.)