An unweighted Yes vote of 60.4% in the Norstat poll became 53.8% in the weighted results

Similar to that sensation of 'feeling your ears burning', I can sometimes sense when Professor John Robertson of Global Ferry News fame is about to write a snarky blogpost about Scot Goes Pop, because it's often presaged by him attempting to leave a comment here.  There was a comment yesterday or the day before with all the usual hallmarks, so I took a look at his blog, and this time there was no post about SGP, but what I did find was a post with the following headline: "Did the Sunday Times hide around 60% support for independence by dramatically reducing the number of 2014 Yes voters in their poll sample from 387 down to only 278?"

That's what John Rentoul would call a QTWTAIN (Question To Which The Answer Is No).  Of course the Sunday Times did no such thing - they have no role in determining the weightings used in any poll, they would never think of making such a request and Norstat would refuse such a request if it was made.  That said, I've looked at the data tables and it's true that the difference between the unweighted and the weighted numbers is pretty extreme.  The Yes vote in the unweighted numbers is 60.4%, whereas in the weighted numbers it's 53.8%.  OK, even normal demographic weightings can often bring down the Yes vote, because there might be too many young people in a sample, or too many SNP voters, or whatever.  But I very much doubt if that sort of thing would have had quite such a dramatic effect - the major explanation in this case is almost certainly direct weighting by each respondent's recollection of how they voted in the 2014 referendum, which of course took place more than a decade ago, thus opening up a risk of significant levels of false recall.  That's one of the reasons Ipsos have cited for not weighting by recalled 2014 vote, incidentally.

In the overall Norstat sample, before the likelihood to vote filter is applied, 43.9% of respondents claim to recall voting Yes in 2014, and only 30.0% claim to recall voting No.  A drastic adjustment has been made to bring those numbers into line with the actual 2014 result.  In fairness, there was initially a very good reason for introducing 2014 weighting, because polling companies had systemically overestimated the Yes vote by a small amount in the 2014 campaign.  But after more than a decade that adjustment is getting harder and harder to justify, and it's impossible to rule out the possibility that it may be artificially skewing poll results towards No and giving us a totally false impression of the state of play.  It must be very unusual to weight poll results by an electoral event that took place more than ten years ago - I'm struggling to think of any other examples of that happening, even internationally.  Somewhere in the deepest recesses of their minds, the heads of polling firms must be gearing up towards a review of this problem, sooner or later.  

*  *  *

We should now think about calling a national holiday, because that rarest of rare things has just happened - Chris McEleny has actually responded to an email.  Apparently the clerk of the committee will be in touch in due course about the arrangements for my appeal against expulsion from the Alba Party.  I'll be interested to see what "in due course" actually means, because from my recollection of the party constitution (which I can no longer read because I'm barred from the party website), the appeal is supposed to be heard within an extremely tight timetable.

*  *  *

Poll commissions, poll analysis, election analysis, podcasts, videos, truly independent political commentary - that's Scot Goes Pop, running since 2008 and currently the fifth most-read political blog in Scotland.  It's only been possible due to your incredibly generous support.  If you find the site useful and would like to help it to continue, donations by card payment are welcome HERE, or alternatively donations can be made direct by PayPal.  My PayPal email address is:  jkellysta@yahoo.co.uk