Saturday, February 22, 2014

New Panelbase poll shows the pro-independence campaign closing the gap

We know that there will be at least two referendum polls tonight, and the headline figures from the first one have now been released. Panelbase of course bucked what seemed to be a universal trend a couple of weeks ago by showing a slight increase in the No lead, but that has been reversed tonight, no doubt courtesy of the Yes campaign's greatest asset, Mr George Gideon Oliver Osborne.

Should Scotland be an independent country?

Yes 37% (n/c)
No 47% (-2)

(UPDATE, 10.40am : Panelbase's Ivor Knox has let me know that this poll used a different preamble to the firm's previous referendum polls.  The new wording is very much in line with ICM, and strikes me as being impeccably neutral.  It reads : "There will be a referendum on an independent Scotland on the 18th of September.  How do you intend to vote in response to the question : Should Scotland be an independent country?"  So assuming this wording is maintained in future polls, we no longer have to worry about the results being slightly distorted.

UPDATE II : OK, this is getting very complicated - I now realise that the new poll was commissioned by the SNP, which means that the neutral preamble was probably used at their request, and that there is a big danger of the subtly biased wording being reinstated in future polls for the Sunday Times.)

With undecideds excluded, the figures work out as -

Yes 44% (+1)
No 56% (-1)

We'll have to wait and see whether what I'm about to say still holds true when the ICM poll emerges later tonight, but as things stand Panelbase remain the only pollster to show that the No lead is slightly higher than in their final pre-White Paper poll.  Whether we should worry about that is highly doubtful given the clear evidence from several other pollsters (especially TNS-BMRB) that Yes have closed the gap since the autumn.

In trying to make sense of this poll, we also have to take account of Panelbase's relatively recent methodological change, which boosted the weighting of older voters in the sample.  The last time I mentioned this issue, I suggested that it might be responsible for artificially boosting the No lead by up to 4% - that was based on my recollection that John Curtice had stated that the change might be expected to decrease the Yes vote "by a point or two", which presumably meant that it also ought to have increased the No vote by a similar amount.  I later checked and discovered that I had got the quote wrong - Curtice had actually said that the Yes vote ought to have fallen by "a point or so".  But it's still reasonable to assume that this means the No vote would be similarly affected, which implies that we have to adjust the overall No lead by 2% in order to make a fair comparison with earlier Panelbase polls.  The company's No leads during the campaign so far (excluding the SNP-commissioned poll) have ranged from 8% to 13% - which if adjusted to take account of the new methodology works out as 10% to 15%.  In other words, the No lead in this new poll falls within Panelbase's "normal" range - but is at the absolute bottom end of it.

And now, while we await the ICM figures, let's have a little bit of light relief, straight from the celebrated "who needs satire when you have the London media?" collection.  A couple of hours ago, the Sunday Express published yet another of those virtually pointless pan-UK polls about the referendum.  A picture of Alex Salmond was given this caption: "The exclusive research is a resounding rejection of Alex Salmond's independence bid." But just below the picture, a 'minor detail' from the poll's results was revealed -

"The poll results among Scots only showed that of the 188 questioned, 52 per cent indicated they supported independence – 34 per cent 'strongly', and 18 per cent 'somewhat'.

Only 24 per cent supported the idea of the rest of the union having a say."

Translation : To the very limited extent that the poll interviewed people who actually have a vote in the referendum, it found that such people supported Alex Salmond's position (as opposed to "resoundingly rejecting" it), thus driving a coach and horses through the whole premise of the article! I think in your shoes I might have been tempted to keep that part quiet, guys...

* * *

UPDATE : The new ICM poll is now out, and it shows a No lead that is exactly equidistant between the wildly diverging previous two ICM polls from September and January respectively. This almost certainly means that there were sampling issues in the extraordinary January poll, leading to an overstatement of the swing to Yes since the early autumn, which has nevertheless clearly been substantial. Once undecideds are taken out of the equation, the Yes campaign have actually retained the bulk of the gains that they appeared to make in January. In other words, you can basically discount pretty much every syllable of the Scotland on Sunday spin, which attempts to transform the mundanity of normal sampling variation into a fabulously implausible narrative of a mega-slump for No in January being followed by a surge in February.  That version of events is simply not borne out by any other pollster.

Should Scotland be an independent country?

Yes 37% (n/c)
No 49% (+5)

With undecideds excluded, the figures come out as...

Yes 43% (-3)
No 57% (+3)

John Curtice has in fact just revealed that ICM have made a modest (but perhaps significant) methodological change since their last poll, suggesting that a direct comparison with this one may not even be meaningful at all. The bottom line is that ICM - regarded by many as the "gold standard" of UK pollsters - are continuing to show a significantly lower No lead than the likes of YouGov and Ipsos-Mori, and are far more in line with Yes-friendly pollsters such as Panelbase (and indeed Survation under their new methodology, who are at least for now the most Yes-friendly of the lot).

* * *


Although the two polls tonight show diverging trends, one curiosity is that they both show a static Yes vote on the headline figures, meaning that the pro-independence campaign remain at their highest level of support since the Poll of Polls started in early December.

MEAN AVERAGE (not excluding Don't Knows) :

Yes 34.7% (n/c)
No 48.7% (+0.4)

MEAN AVERAGE (excluding Don't Knows) :

Yes 41.6% (-0.2)
No 58.4% (+0.2)

MEDIAN AVERAGE (excluding Don't Knows) :

Yes 42.0% (n/c)
No 58.0% (n/c)

(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, and that adhere to British Polling Council rules. At present, there are seven - Angus Reid, 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.)

To put the trend over recent weeks into perspective, the No lead has fallen from 17.5% to 14% on the headline figures since the Poll of Polls started. That may be a slight underestimate of the true decline in the lead, given that Angus Reid are still represented in the sample by an old poll from last summer. (At some point, I may have to stop taking Angus Reid into account, but hopefully they'll return to the fray soon.)

The reason why the median average is unchanged is that ICM and Panelbase both remain on the Yes-friendly end of the spectrum. The current ranking of the pollsters in order of their favourability to Yes is as follows -

1) Survation
2) Panelbase
3) ICM
4) Angus Reid
6) YouGov
7) Ipsos-Mori

Only the bottom two are not showing a Yes vote of above 40% when undecideds are excluded.


  1. I read John Curtice's article. He now sees gains to Yes from last year across all age groups. I suspect the Yes campaign being fought at grassroots level rather in the media has been the antidote to the negative press/media spin. The No side must be getting rather frustrated in what they thought would be an easy thing to do by frightening the public but that is now wearing bit thin.


  2. Public poll results always seem to vary from internal private polling. Given that movement to the Yes camp is probably greater than stated in the results I would hazard a guess that one more month of this will lead to Darlings "sick leave". The removal of a "leader" is the most public view of how the land really lies.

  3. Marcia : "The No side must be getting rather frustrated in what they thought would be an easy thing to do by frightening the public but that is now wearing bit thin."

    Yes, and that's been the value of a long campaign - if it had been shorter the scare tactics would probably have worked much better. No wonder the favourite refrain of anti-independence people in 2011 and 2012 was "oh this is so BORING, for the love of God let's just get it OVER with".

    In case anyone misses it, I've made an important update at the top of the post to make clear that Panelbase have introduced a new and much more neutral preamble. The fact that the lead has only dropped by two points probably means that the preamble wasn't having as much impact as we feared (and may well not have been having any impact at all), but I'm still hugely relieved - it's important that we can all have confidence in the methodology, and that there is full transparency.

  4. CH

    There are two versions of that question being asked by Yougov. One saying Labour would win and the other that the Tories would win. Yougov should question themselves as to why they are sending out conflicting questions.

  5. I spoke too soon - I've had to update the post yet again, because it turns out that the Panelbase poll was commissioned by the SNP. That means it may well have been the SNP who insisted on the use of a neutral preamble, and that there is a significant danger of the subtly biased one being reinstated for future Panelbase polls.

  6. "Vision Critical" are "Angus Reid"!

    I presume (although I haven't checked) that they are still BPC members.

  7. An interesting point is made at Wings Over Scotland, that all of the polls from different companies are converging.

    The lastest polls from companies that have conducted polls this year shown the no lead between 9%-18%. Comparing that to last year where we were having different polls saying Yes was 30% behind and 1% infront in the same week!

    I'd say this is down to refining methodology, so now all the companies (except Ipsos) weight by 2011 past vote. Also, ICM and Panelbase have fixed their sample errors as well.

  8. Oldnat : I'm a bit unsure about whether a poll under the Vision Critical name can be regarded as exactly the same thing as an Angus Reid poll, but it's an academic point in this particular case because it wasn't a full-scale poll of Scottish respondents (and didn't ask the referendum question directly).