Friday, September 16, 2016

It's a statistical tie as "remarkable" Ipsos-Mori TELEPHONE poll puts support for independence at heady 48%

Should Scotland be an independent country?

Yes 48%
No 52%

More details to follow eventually.  I'm physically blogging this from the Davis Cup semi-final in Glasgow, so it's not straightforward...

Tuesday, September 13, 2016

Drama as TNS poll reveals that Brexit has increased support for independence by more than expected

Back in May, in the run-up to the EU referendum, the polling firm TNS asked Scottish voters how they would vote in a new independence referendum if Brexit occurred.  The results were a tad surprising, given that other firms had reported that a Leave vote would significantly boost support for independence.

MAY 2016 POLL :

It is possible that Scotland may vote to remain in the EU, but the UK as a whole votes to leave. In those circumstances, if there was a new referendum on Scottish Independence how would you vote in that independence referendum?

Yes 44%
No 56%


As you can see, TNS were suggesting that Brexit would for some reason lead to the Yes vote being very slightly lower than in the first independence referendum in September 2014. But, of course, respondents to polls are not always terribly wonderful at answering hypothetical questions. And sure enough, now that we have the 'before' and 'after' photos to compare, it appears that the reality of Brexit is less palatable than people had anticipated, and that the result in June has indeed increased rather than decreased support for independence. A new TNS poll hot off the press has found that the Yes vote is actually 2% higher than it was in the 2014 vote.

NEW POLL :

Should Scotland be an independent country?

Yes 47%
No 53%


Other polls have been better still for Yes. There have now been six credible polls on independence since the shock of 23rd June, and HALF of them have shown an outright Yes lead. ALL of them have shown a higher Yes vote than in September 2014 - and bear in mind that most firms now weight by recalled referendum vote, so there's no good reason to doubt that the increase is real. What we would have given for these numbers when the last indyref campaign got underway...

*  *  *

UPDATE : There wasn't much more I could say about the poll last night, because only the headline results had been made known.  But the datasets are now out, so here are a few more miscellaneous points...

*  TNS have, as expected, continued with their excellent practice of including 16 and 17 year olds in their sample, in line with the voting age of 16 that has now become the norm for Scotland-only elections and referendums.  This again puts to shame YouGov's laziness in excluding 16 and 17 year olds from their Scottish polls, and highlights the fact that YouGov's recent figures of Yes 46%, No 54% cannot be regarded as reliable.  It's unlikely that they under-reported the Yes vote by more than 1% (and over-reported the No vote by more than 1%), but when the race is so tight, such small errors make a big psychological difference.  YouGov should be called out on this basic flaw in their methodology until they put it right.  No more excuses.  If they innocently claim to be a politically neutral organisation (and they do), they simply can't get away with knowingly excluding a Yes-friendly part of the electorate from their polls.

* John Curtice had a go at winding up Yes supporters this morning by claiming that the TNS poll shows a "6% swing to No".  That's technically a justifiable claim, but he's measuring that swing from an ancient poll conducted a year ago - which strictly speaking was the last time TNS asked a voting intention question on independence, and which reported figures of Yes 53%, No 47%.  The problem for Curtice is that we also have the much more recent TNS poll from May that I mentioned above, which showed that in the event of Brexit the Yes vote would be only 44%.  For some reason TNS didn't bother taking the obvious step of simultaneously asking a straight voting intention question on independence for the purposes of comparison, but we know from polls conducted by other firms that, if they had done so, it's very unlikely that the 'real' Yes vote would have been higher than the hypothetical 'Brexit' Yes vote.  It's more probable that it would have been somewhat lower (ie. lower than 44%).  So it seems clear enough that today's Yes figure represents an increase since the spring - perhaps an increase within the margin of error, but certainly enough of an increase to open up the possibility that a genuine change has occurred.

The poll from last September looks like an oddity in retrospect.  It may well have been a rogue poll, but as TNS have polled on independence so infrequently since the indyref, it's impossible to say that for sure.  The alternative possibility is that they may have picked up a spike in support for Yes that occurred last year, but if that is the case, it certainly seems to have long since faded by the time of the May poll.  It doesn't change the fact that today's poll appears to represent an improvement for Yes since the spring.

* There have been no fewer than NINETEEN voting intention polls on independence from TNS since the start of 2013 - and today's is the THIRD-BEST for Yes. The only better ones were last September's, and the one immediately before polling day in 2014. At one point, in August 2013, No had an almost two-to-one lead, so we've come a long, long way since then.

* There's a gender divide in the new poll - but it's not the one we're used to.  TNS are reporting that women are now significantly more favourable towards independence than men are.  Among men, No have a double-digit lead, but among women the race is a statistical tie.  When you see a finding as unexpected as that, it does make you wonder whether there's something strange about the sample.

* The age divide is the familiar one, with younger people breaking more for Yes - but it's particularly extreme in this poll.  Among under-65s, Yes have a healthy enough lead of 55% to 45%.  But that's overturned by over-65s breaking for No by a whopping 80% to 20% margin.  Quite honestly, I think it's unlikely that No have a four-to-one advantage in that age group (or any other age group), but of course you can't necessarily cherry-pick one part of a sample and say that only that bit is implausible.