Saturday, September 8, 2018

New YouGov subsample offers reassurance for SNP

So as you'll probably remember, the first polling straw in the wind after the Alex Salmond story broke was a Scottish subsample from a GB-wide YouGov poll, and although it had the SNP in the lead, the party's share of the vote was down to 34% - breaking a long, long sequence of YouGov subsamples that had the SNP in the high 30s or low 40s.  Although no individual subsample can be regarded as reliable, YouGov's Scottish subsamples are unusual in that they appear to be separately structured and weighted - which probably explains the relative stability of the results over time.  So the drop to 34% might have been a coincidental and meaningless blip caused by normal sampling variation - but it might just have been a warning sign that the Salmond story had caused some damage.

As Marcia pointed out last night, a new YouGov subsample is now out which appears to show that normal service has been resumed...

SNP 40%, Conservatives 23%, Labour 21%, Liberal Democrats 9%, UKIP 4%, Greens 3%

If it does turn out that we're back to normal, and that's a big if, it'll be impossible to know whether the dip was real but transitory, or didn't happen at all.  But the local by-election in Fife on Thursday certainly didn't show any sign of a catastrophic drop in SNP support.

Less encouraging are three subsamples from other firms - two from Survation that have the SNP in second place, and one from BMG that have them just about in the lead but in a virtual three-way tie.  But those are based on very small samples, and probably aren't separately weighted in the way that YouGov subsamples are.  For now the YouGov figures are of most interest.

*  *  *

Friday, September 7, 2018

Horror show for Labour in Fife by-election

Tonight we have the result of the first local government by-election in Scotland for several months...

Inverkeithing and Dalgety Bay by-election result (first preferences):

Conservatives 37.3% (+0.7)
SNP 28.1% (-2.7)
Labour 12.0% (-4.8)
Liberal Democrats 9.1% (+4.1)
Independent - Collins 8.4% (n/a)
Greens 4.2% (+0.7)
Independent - Macintyre 0.6% (n/a)
Scottish Libertarian 0.2% (n/a)


I know the SNP were talking up their chances of outright victory, but for my money this is a very creditable result in difficult circumstances. In spite of the endless stream of negative headlines about Alex Salmond over the last couple of weeks, the SNP are only down a smidgeon on an election last year in which they won the national popular vote by a relatively comfortable margin. And although the Tories usually have an inbuilt advantage in local by-elections due to the greater motivation of their supporters to make it to the polling stations, the modest swing from the SNP they achieved would not be enough to overhaul the SNP's national lead if extrapolated to the whole country.

Technically this has to be reported as a "Tory gain from Labour", but that's just one of those wildly misleading quirks of the STV voting system. The Tories won the popular vote in the ward last year, and Labour were a distant third, so the chances of Labour 'holding' the seat were always remote (although arguably not totally non-existent, because Labour did finish a strong second in the ward back in 2012). Nevertheless, Richard Leonard ought to be horrified to see Labour's vote slip back more than the SNP's. We all know that any real threat to the SNP's predominance in Scottish politics would have to come from Labour, because there is a natural ceiling on Tory support. So the fact that the SNP have somehow improved their position relative to Labour in this ward is extremely heartening in the current climate. OK, Labour would probably argue that this was a classic third-party squeeze, with Labour voters lending their support to either the SNP or the Tories depending on whether they happen to be unionist diehards or not. But if it's quite as simple as that, why did the Lib Dems and the Greens both increase their vote share in fourth and sixth place respectively?  Is Labour's lack of clarity on Brexit costing them?

*  *  *

Thursday, September 6, 2018

Sometimes being ruthless just means losing Ruth

Phantom Power's acclaimed Journey to Yes series is back - and the latest film is perhaps the most remarkable so far. It features Ashley Graczyk, who was elected as a Scottish Conservative councillor in Edinburgh only last year, but who has now left her party and become a passionate supporter of Scottish independence. Watch her story below.



*  *  *

Tuesday, September 4, 2018

The case for an early independence referendum has already been made - it can't and mustn't depend on transitory opinion poll results

I'm more than a little troubled by Jason Michael's latest post on Random Public Journal.  Obviously I agree with him that now is the time for the SNP to go on the offensive against the British government, but we really must be careful not to put all our eggs in the basket of what is frankly a ropey interpretation of a single opinion poll.  Jason says that the Deltapoll survey conducted for the Best For Britain campaign shows that "52 per cent of Scots think self-determination is the best option for our nation’s future...this represents a seven-point swing in favour of independence since the 2014 referendum". That simply isn't true.  As I pointed out in my previous post, the main question on independence actually shows a 51-49 split in favour of No, although admittedly it was only the tiniest of tiny fractions away from being rounded up to a 50-50 dead heat.  So in fact the swing to Yes since 2014 implied by the poll is actually around 4 or 5 points.  That's a fabulous showing for Yes by most recent standards, but it's not quite an outright lead and it's not a 7 point swing.

There are also health warnings that need to be put on the poll.  Even the main question had a non-standard wording.  (It's not a wording that in any way contained a pro-independence slant, but nevertheless the question should really have been 'Should Scotland be an independent country?')  It appears from the datasets that the only political weighting in the poll was by recalled EU referendum vote.  It's not unheard of for pollsters to eschew political weighting, but it's certainly unusual - the majority of firms would have weighted by recalled indyref vote and recalled general election vote.  This is also Deltapoll's first foray into independence polling, which means there is no baseline to judge from.  In other words, there is no hard evidence in this poll of a recent surge for Yes.  The unusually good result may just be a 'house effect' generated by how the poll was conducted.  That doesn't necessarily mean the figures are wrong - as Scottish Skier pointed out in the previous thread, it's perfectly possible that weighting by recalled indyref vote is making other firms' results less accurate, not more so.  The fact that Ipsos-Mori (who don't weight politically) were a No-friendly firm before the indyref but are now more towards the Yes-friendly end of the spectrum might conceivably lend some weight to that theory.

But the bottom line is that if No-friendly firms like YouGov and Panelbase were to publish independence polls tomorrow, there's no reason to automatically think they'd show anything other than the type of No leads they usually show.  So if we get too wedded to a narrative of "a referendum is coming because the polls now show a Yes lead", we're just setting ourselves up for a fall in very short order.  The case for a referendum should not depend on whether Yes are currently at 43% or 47% or 49% or 52% - all of those are potential platforms for victory (or indeed for defeat).  I'm particularly aghast to see Jason unintentionally echoing the rhetoric of those who want to see the referendum kicked into the long grass by stating that Yes now need to use their current momentum (which may not even exist) to kick on and reach an utterly fantastical and unattainable target figure of 60% by the date of Brexit, which is less than seven months away. 

If we're serious about wanting an early referendum, we simply cannot afford to run away with ourselves and set unrealistic expectations of what the polls will show over the coming weeks and months.  If a significant swing to Yes occurs, it's much more likely to be after the referendum is called, and not before.

*  *  *

Sunday, September 2, 2018

Deltapoll delirium as new survey finds Scotland split down the middle on independence

There's a lot of excitement tonight because of an independence poll featured on the frontpage of The National which sort-of-shows Yes support at 52%.  The problem is that Deltapoll asked a non-standard question based on the hypothesis that Britain will leave the EU.  Although this is obviously a hypothesis that is extremely likely to come to pass, some respondents may have felt that they were 'supposed' to adjust their current view on independence when the imagined future was taken into account, which could make the results unreliable.  People are notoriously bad at answering hypothetical questions - 'bad' in the sense that their answers don't tally up with how they actually react when the event comes to pass.

But never fear, because there is something very interesting and encouraging about this poll.  Before the hypothetical question was posed, respondents were first asked a more neutral question about independence.  Unfortunately it's still not the standard independence question (ie. 'Should Scotland be an independent country?' Yes/No), so it can't be directly compared with other independence polls.  But there's absolutely nothing about the question that would artificially steer respondents towards a pro-independence position - if anything, the opposite is true.  The results are startling.

In a referendum on independence for Scotland held tomorrow, how would you vote?

For Scotland to remain as part of the United Kingdom: 50.6%
For Scotland to become an independent country: 49.4%

That's as good a result for Yes as we've seen in a long time - essentially a dead heat.  I haven't been able to track down the fieldwork dates yet, so if anyone knows what they are I'd be very interested to hear.  That could obviously be a point of great significance given recent events.

*  *  *

UPDATE: Stuart Campbell kindly sent me the methodological note from the poll last night, which states that the parallel Northern Ireland sampling took place between the 27th and 30th of August, but irritatingly doesn't specify the fieldwork dates for Scotland.  The National article states that polling took place between the 24th and 29th.  Either way it appears to be entirely after the Alex Salmond story broke, which is very encouraging.

*  *  *

First Scottish subsample since the Alex Salmond story broke has the SNP still in the lead

When news of the complaints against Alex Salmond was irresponsibly broken by the Daily Record, I said that we'd just have to wait for opinion poll evidence to see whether there would be much (if any) negative effect on the SNP.  Frustratingly, the end of August has seen an unusual dearth of polls even at Britain-wide level.  (It goes without saying there was never much chance of a full-scale Scottish poll when you really wanted one.)  However, we do at last have a Britain-wide YouGov poll which was conducted on the 28th and 29th of August, and the Scottish subsample shows the following:

SNP 34%, Conservatives 27%, Labour 24%, Liberal Democrats 8%, Greens 3%, UKIP 3%

Now, to state the obvious, any individual subsample cannot be regarded as statistically reliable, and I normally wouldn't dream of looking at one in isolation.  However, when a major event has occurred and you're waiting for any clues at all about the impact, obviously you're going to be interested in the first straw in the wind, no matter how many health warnings have to be attached to it.  The good news is that the above figures are not at all untypical for Scottish subsamples since last year's general election, but the bad news is that they're less impressive for the SNP than the most recent batch of subsamples from YouGov.  The previous eight had the SNP on 38%, 39%, 42%, 40%, 40%, 42%, 38% and 42%.

The dip to 34% might easily be explained as the meaningless effect of random sampling variation, but obviously it could also be a sign that the media hysteria of recent days has had a genuine effect on public opinion.  If it proves to be the latter, I think we should take heart from the fact that there's no sign yet of the SNP losing their lead outright, and that it's perfectly possible that any setback will be quickly reversed once the media obsession eventually blows itself out.

*  *  *