Journey into the bizarre as Herald describes an increase in support for independence as "the reversal of a surge"

A new YouGov poll on independence has been released, and in line with the other three credible polls to be published since the EU referendum, it shows an increase in support for Yes...

Should Scotland be an independent country?

Yes 47% (+1)
No 53% (-1)

I regret to say that the Herald's reporting of this poll warrants yet another caution from the psephology police.  They've made a determined attempt to compare apples with oranges by claiming that the surge in support for independence in the immediate aftermath of the Brexit vote has been "reversed".  To state the bleedin' obvious, the polls that put Yes in the outright lead in late June were Panelbase and Survation polls - and tonight's poll was not conducted by Panelbase or Survation, but by a different firm with its own distinct methods.

It is quite possible that a Panelbase or Survation poll conducted now would show the Yes vote holding steady in the low 50s.  It is also quite possible that a YouGov poll conducted in late June would have produced an identical result to tonight's.  There is absolutely no hard evidence in this poll of a surge being reversed - and that is not an opinion, it is a fact.  The new numbers can only be compared with the last YouGov independence poll from well before the EU referendum - since when there has been a small 1% swing towards Yes. 

UPDATE : The datasets are now out, and the big question that immediately forms in my mind is whether the results have been weighted by EU referendum vote.  At first glance it looks like they have been, because 553 Remain voters in the unweighted sample have been reduced to 500.  But the note at the bottom of the datasets doesn't mention EU referendum vote as one of the factors that have been weighted for. 

The reason this matters is that there have been signs in other polls that some Leave voters falsely say they voted Remain, possibly because of embarrassment - meaning that there will generally appear to be too many Remain voters in unweighted samples.  As there is a strong correlation between voting Remain and supporting independence, any misguided downweighting of Remain voters could potentially lead to support for independence being underestimated.

I would also strongly urge people to pay only limited heed to supplementary questions that appear to show slightly worse results for Yes.  Respondents are already factoring in Brexit when they answer the headline question on independence, so that's the result that matters.  Questions with more exotic wordings will always produce slightly different results.

Nevertheless, as has been pointed out in the comments section below, it's striking that a significantly bigger minority of SNP voters seem to oppose independence when the question is posed as a straight choice between the UK and the EU.  That illustrates the potential danger of making retention of EU membership the lead campaign pitch in Indyref 2 - because it may alienate Yes voters from 2014 who have since voted Leave.  My own view is that we have to draw a distinction here - the issue of being dragged out of the EU against our will matters enormously, because it's the casus belli for calling a second referendum.  But once that referendum is actually underway, we'll need a more inclusive campaign that appeals to both Remain and Leave voters.

UPDATE II : I'll just quickly run through some of the other concerns that have been raised about the poll in the comments section.  Yes, absolutely, it's extremely reprehensible that YouGov haven't interviewed 16 and 17 year olds, given that it seems almost inevitable that the voting age for Indyref 2 will be 16 rather than 18.  The Yes lead among 18-24 year olds is a whopping 26%, so if you assume that 16 and 17 year olds would break in a similar way, it's perfectly possible that the headline result of this poll should really be Yes 48%, No 52%.  That's just speculation, obviously, but it's scarcely unwarranted.

In theory, there's no reason why the poll should be missing a pro-Yes surge among citizens of other EU countries.  YouGov weight by country of birth, and the 'born outside the UK' group is weighted to make up roughly 9% of the sample.  However, there's no way of knowing whether that group contains the correct proportion of EU and non-EU citizens.

It's true that the recalled EU referendum vote is slightly skewed in favour of Leave in the weighted sample - it's Remain 59%, Leave 41%, whereas it should be Remain 62%, Leave 38%.  For the reasons given above, I'm not sure that weighting by EU referendum vote is a good idea - but all the same, you'd expect to see too many Remain voters in the sample (because of the problem of false recall), not too few.  So there may be a small issue there - and if so, the poll may be slightly underestimating the Yes vote.