Even if it wasn't for my past history at Political Betting, I would have sighed heavily at the latest post from Mike Smithson, which makes a big drama out of the fact that Labour are "only" 3% behind the SNP in a published aggregate of Populus subsamples from the month of November -
Liberal Democrats 9%
This of course only tells us what we already knew, because the individual Populus subsamples were all published and were freely available for anyone to tot up (but only if that person could "be arsed", of course).
Smithson does add a couple of half-hearted caveats to the figures, but still comes to the silly conclusion that Labour "will be delighted". As we've been saying for weeks, Populus stick out like a sore thumb as the one and only pollster that has produced subsamples since the referendum putting the SNP behind Labour - and they haven't just done it once, but several times. By contrast, every other pollster has put the SNP ahead in every single post-referendum subsample to date - and that includes YouGov, who produce at least five subsamples per week. If you aggregated subsamples from every firm other than Populus, or even if you aggregated subsamples from every firm including Populus, you'd see a very handsome SNP lead - as this blog's Poll of Polls demonstrates.
There's no mystery about why Populus are different - they downweight the SNP much more severely than anyone else does, based on target figures for party identification from way back in 2010, when of course the SNP had much less core support.
In other words, the above figures are a red herring. Nothing to see here.
On a vaguely related point of pedantry, Smithson keeps saying that there have only been three full-scale Scottish polls since the week of the referendum. That's quite simply wrong, but naturally I haven't been able to correct him because of my random lifetime banning from his site. There have in fact been six full-scale Scottish polls since September 18th - two from Survation, two from Panelbase, one from Ipsos-Mori and one from YouGov. Even if you discount the first Survation poll because it was conducted immediately after the referendum, that still leaves five.