Has the Philosopher-Tory successfully philosophised his way out of his most cynical lie? (Spoiler: no, he hasn't.)

He's Ready for Rishi.  Potty for Penny.  Batty for Badenoch.  Rabid for Raab.  Yup, the Scottish political world is still reeling from the former pro-independence blogger Stuart Campbell's shock announcement that he is backing the Tories at the next general election, although the man himself and one or two of his fans seem to be mostly furious with me for bringing the news to slightly wider public attention.  Which is odd, in a way - if he's concluded that voting Tory is morally justifiable and strategically wise from the point of view of whatever the hell his current objectives are, you'd think he'd be keen for as many people to know as possible.  

As per usual, the main outlet for the anger against me is an attempt at amateur psychoanalysis, with the burning question of the day being *why* I started my much-requested Wings-Watch fact-checking service.  They haven't as of yet considered the most obvious and straightforward possibility, namely that Campbell writes a very prominent blog which has a notoriously strained relationship with the truth, and that a fact-checking service will thus remain necessary until he stops regularly lying to his readers.  (Of course all bloggers and journalists make the occasional inadvertent factual blunder, but that's not the sort of thing we're talking about here.)

Instead, they ascribe Wings-Watch to two main factors.  Firstly, the old favourite that I or anyone else who disagrees with the great man must somehow be "deranged".  (If Campbell still owns a paperback thesaurus, you can be sure that the one page that has long since fallen out due to extreme overuse is the one containing synonyms for "mentally ill".)  But their second explanation is much more interesting, because although it's hopelessly misconceived, it's unwittingly quite revealing.

Since I started Wings-Watch, the lie I've had to correct by far the most often is Campbell's dodgy graph purporting to show that support for independence has remained absolutely static at 47% every year since either 2015 or 2016 (depending on which version of the graph is being used on any given day).  By this stage, Campbell knows the graph is a lie, I know it's a lie, you know it's a lie, even the dogs on the street know it's a lie, and the only people who don't know it's a lie are the unfortunate souls who never step outside the Wings bubble and are naive enough to believe that everything Campbell tells them is honest.  However, I think we may now have stumbled on how he philosophically justifies that lie to himself.  Although he knows the graph itself is fraudulent, it may be that he genuinely thinks it's a lie that contains a 'poetic truth' because he's labouring under the misapprehension that the standard 3% margin of error in polling renders the increase in independence support we've seen over the last few years statistically meaningless.

To be clear, though, he has no excuse for that erroneous belief.  As long ago as 2016, when I was still on good terms with him, I and a number of others (including Dr Morag Kerr, who is normally one of his stoutest defenders) pointed out to him where he was going wrong about the margin of error in polls.  He had been repeatedly insisting that there was no systemic error in the polling for the Trump v Clinton presidential election, which on average showed a 4-point lead for Clinton.  If that had been the actual result on the popular vote, it almost certainly would have translated into a Clinton win in the electoral college, and Trump would never have become president.  Instead, Clinton's real lead was only two points, and we all know what the consequences of that proved to be.  But Campbell was adamant that the polls had not been wrong, because a 4-point Clinton lead was "within the margin of error" of a 2-point Clinton lead.

The problem is that the margin of error only applies to each individual poll.  If individual polls are only slightly inaccurate due to normal sampling variation and not because of methodological failings, you'd expect the errors to be randomly distributed - in other words you'd expect roughly as many polls to underestimate the Clinton lead as to overestimate it, and for the average error in all of the polls to be far less than 3%.  That simply didn't happen - the vast majority of polls in fact overestimated the Clinton lead.

Similarly, if Campbell is correct in his belief that Scottish independence support has remained static at around 48% for several years and only the margin of error has been masking that, we should see in any calendar year with a large number of polls that roughly as many polls have Yes above 48% as have Yes below that figure, and that the annual average should always work out at pretty close to bang on 48%.  Is that what's happened?  Let's remind ourselves yet again of the real numbers.

Yearly support for Scottish independence in the Scottish Social Attitudes Survey:

2014:  33%

2016 (a):  39%

2016 (b):  46%

2018:  45%

2020:  51%

2021:  52%

Average yearly support for independence in conventional opinion polling:

2016:  47.7%

2017:  45.3%

2018:  45.5%

 

2019:  47.6%

2020:  53.0%

2021:  49.6%

2022:  49.8%

Presumably even Campbell would have to acknowledge that if the Scottish Social Attitudes Survey is right, there has been a massive increase in support for independence since 2014 - so presumably we must infer that he thinks those figures are not valid for some unspecified reason.  (Which again is distinctly odd, because another of his dodgy graphs blatantly depends on the use of a cherrypicked figure for independence support from the 2007 Social Attitudes Survey.)

Let's turn instead, then, to the averages from conventional polling.  Even leaving aside what Campbell dismisses as merely a "Covid blip" in 2020 (something that lasted for almost a year is quite some "blip" by any standards), you can see for yourself that the Yes average was as low as 45% (after rounding) in 2017, which is three points lower than what Campbell claims to have been the constant underlying Yes figure.  It was as high as 50% after rounding in both 2021 and 2022, which is two points higher than Campbell's claimed steady figure.  Given the sheer number of polls that were conducted in all of those years, the changes simply can't be explained by random sampling variation.  Unless there is some reason to believe that there was some systemic error in the polls in 2017 that does not apply now, or vice versa, the only conclusion it is possible to draw is that Yes support was substantially higher in 2021 and 2022 than it was in 2017.  And no, a four or five point increase cannot be dismissed as trivial or underwhelming, given that the Yes vote recorded in the 2014 referendum was only five-and-a-bit points shy of victory.

Incidentally, it might amuse you to discover that Campbell did not exactly make an effort to listen and learn when we tried to politely explain where he was going wrong about Clinton v Trump in 2016.  His reaction instead was to angrily insta-block me, thus automatically placing me on the block-list he exported to hundreds of his fans.  Some things never change....