Friday, 5 September 2014

What's not happening

Since the publication of last week's yougov poll a state somewhat approaching hysteria seems to have descended on the Nationalist camp.

Partisans who have been maintaining that "yes will definitely win" since Eck finally realised that he had no alternative but to have a referendum now suddenly claim to have only truly believed this within the last week. Apparently they were whistling in the dark before but, trust them, they are not whistling in the dark now. And what is also rather inclined to be overlooked is that even this poll still showed the No side well ahead and much of the actual apparent narrowing over the last three Yougov polls being attributable to a change in methodology between the third and second last of these.

But, as I have argued before there is a self interest shared by the Yes campaign, the No campaign and the media in maintaining it is close. For the first it keeps up the spirits and efforts of their activists; for the second it serves to guard against complacency and for the third it sells newspapers or gets people watching television programmes.

So when Yes Scotland claims to have registered thousands of last minute (implicitly solidly Yes) voters nobody really has an interest in finding out if that is actually true. If it scares little old Tory ladies in the shires to make an extra effort to get to the polling station, Ruth Davidson might even have cause to thank the Radical Independence Campaign (sic). If it cheers up the Nats to believe their own propaganda, it might avoid too many early internal questions about why, with all these enthusiastic foot soldiers, even in this poll they are still losing. And for the media last minute voter registration is on any view a better story than simply repeating the sound bites that increasingly form the output of both the official campaigns.

But it suits nobody to inquire why, if all these trots and greens really have been swarming over, and indeed being received with acclamation within, the most deprived areas of Glasgow, large numbers (?) of potential voters only got round to thinking about registering to vote within the last 48 hours available to them.

And then we have the "look" of Scotland. There are on any view many more Yes posters in windows than No posters. And far more Yes badges too. But equally there are far more windows with posters from neither side and far more unbadged citizens of any sort. If this really was a popular insurrection about to throw over one of the world's most stable democracies shouldn't the whole thing just be considerably more visible?

And then we have the utter financial calm. Sure, the Yougov poll caused a (very) minor blip in the value of Sterling and sure Eck says we will keep the Pound and many are foolish enough to believe him but many others see through that pledge but yet seem in no hurry to get their money out of Scottish Financial institutions. Myself included. If there was any real anticipation of a Yes vote things would be very different. Sure, many nationalists would regard it as their duty to stand by the state they had created but anybody who thinks "rich Tory unionist capitalist bastards" would feel patriotically obliged to keep their money here and see what happens has a somewhat naive view of that clan. And, sure, optimists in Scotland would believe something could be worked out on currency while optimists in England might believe Scotland's departure would be of no great fiscal consequence. But would that be the view of the international exchange markets who, unlike we common herd, had alternatives to crossing their Zurich based fingers? Most certainly not. And yet the interesting thing about the financial markets is that they are so uninteresting.

And then we come back to the issue of polls. There is a recognised psephological phenomenon about polls in this sort of situation. They cluster. I have to start off by briefly explaining how polling is conducted. The standard sample is usually just over a thousand. But you don't just ask a random thousand people their view. You try to get a "representative" sample. 52% women being the most obvious example but you are also looking for the correct percentage of over 65s or under 30s; the correct percentage of blue collar and white collar workers; of economically active and inactive. And indeed various other determinants.

Now you can try to do this in your initial sampling but without polling much larger numbers and discarding many results you can only achieve a certain degree of precision. So you introduce "weighting". And that then involves value judgements by the pollsters themselves. If, for example, to weight up the under 30s means you increase the number of those economically inactive, the pollster has to judge which which is likely to be the more important determinant  of voting intention. There is nothing sinister about any of this, it is the way things have always been done.

BUT. In this campaign many of the polling companies decided early on that an important determinant was how people had voted in the 2011 Scottish Parliament elections. Thus, if crude sampling produced 30% of your sample saying they would vote SNP then in the published poll this was "weighted up" to reflect the 45% the Nats actually got in that election. Fair enough on one view except that obviously that 30% intending to vote SNP were understandably much more likely to vote Yes. This is all very well in a vacuum but of course it ignores three things. Firstly, it ignores the fact that in lower turnout Scottish Parliament elections the Nats always do better (and the Tories worse) than in other elections. No-one expects the referendum to be a low turnout affair. Secondly it ignores the fact that the 2011 Nationalist triumph was against the most inept campaign fought by a principal opposition Party in just about any election ever fought in any country at any time. And thirdly it ignores real elections that have taken place in Scotland since 2011, in almost all of which the SNP vote has gone down while both  Labour AND Tory votes have increased. The very absurdity of this method might best have been demonstrated by Survation who polled on both Independence and voting intention in the European elections in the run up to the latter.Ten days before polling they produced a poll, having applied "2011" weighting that produced a 56/44 No/Yes result. But using the same method they also predicted a 39% SNP vote share in the European elections. Ten days later the SNP got 29% in the actual election.

So maybe, just maybe, 2011 50% turnout voting intention is not the most accurate guide to  party affiliation now, the 80% turnout distribution of that affiliation or, most importantly, the knock on referendum vote based on that premise being accurate.

That certainly was the belief of Yougov who regarded it as of little importance and until recently produced much more robust leads for No. But of course that has changed with the most recent polls. So however has Yougov's weighting. This change has not been intended to bring Yougov into line with actual voting intention (for no-one knows what that is). It is however intended to bring Yougov more in line with other pollsters, albeit without adopting their precise method. Precisely the phenomenon I have described above as clustering. For no polling company wants to be an outlier. Ideally they all want to be right but there is less reputational risk in, at worst, all being wrong together.

Now, finally, if that clustering was indeed taking place then it would affect both sides, of course. If pollsters providing favourable results "to us" start aligning with pollsters providing favourable results "to them" then that cuts both ways.

Panelbase, the pollster who have provided consistently the most favourable results "to them" carried out a poll last week for the Yes campaign. It was a private poll but Yes Scotland have a, perfectly legitimate, track record of publishing favourable private polls. And of course where better to publish that poll than in today's Daily Record under the one off editorial control of Mr Alex Salmond.

But no such poll result appeared and I think we can now assume it is not going to appear. The polls are clustering. And that then leads me to observe this. Actually, although I use the phrase myself,  there have not been polls "more favourable" to Yes Scotland. There have only been polls less unfavourable. Never a single properly conducted poll putting them ahead from start to finish.

Now it might be in 13 days time all the polls are wrong. It might be that this will be just about the only referendum ever where the polls underestimated the vote for change and/or there was not a last minute swing back to the status quo. That might all be what happens.

But for less level heads among the commentariat on our side to be offering "helpful" advice suggesting that the major Parties and the Better Together campaign should, at this point, be engaged in a collective outbreak of "Corporal Jonesism" seems to me to be unjustified. Not least, but not only, by the polls.

P.S. There is, nonetheless, no room for complacency.