The new Scottish Labour and anti-independence propaganda site The Red Robin has until now been best known for the inexplicable decision of CommonSpace (an ostensibly pro-independence news outlet) to actively promote it as if it was practically a sister site. But tonight it's gaining new attention by breathlessly claiming an exclusive about "Labour internal polling" that supposedly shows the party has "dramatically closed the gap on the SNP".
Yeah, up to a point, Lord Copper. The polling was conducted by Moonlight Research, and if the name doesn't sound familiar to you, there's a good reason for that. I cannot recall them ever having conducted a political poll before - or at least not one that's seen the light of day. The Red Robin are correct in stating that Moonlight are members of the British Polling Council [see below for an update on that point], so it would be wrong to dismiss them out of hand as some sort of Mickey Mouse outfit. But with no track-record whatsoever, it's impossible to know how likely their figures are to be even vaguely close to the truth. And because they have never conducted a published Scottish political poll before, there are no baseline figures to measure percentage changes from, which means The Red Robin's repeated claims that Labour have gained ground in the poll are quite simply wrong. Because of house effects and methodological differences, you simply can't look at two polls conducted by different firms and measure percentage changes from one to the other. It just doesn't work that way. You can only measure from previous polls conducted by the same firm - and in this case there aren't any. For all we know, Moonlight may have a Labour-friendly methodology that would have produced very similar results at any time during the last few months.
As a general rule, "internal polling" is always less satisfactory than public polls commissioned by media organisations, because we only get to see internal figures if they're favourable for the party that commissioned them. So, for example, in a series of eight internal polls you might only see the one that happened to be most favourable for Labour, which could well give you a completely distorted impression of the significance of an outlier result. But at least with internal polling that is properly published, it's brought into the open, and we eventually get to see the datasets, allowing us to judge how much credibility to give the figures. This particular case falls well short of that standard, because it looks as if Labour have no intention of publishing the poll, but have just conveniently decided to leak partial figures to a friendly website. We don't even have a clue whether the figures are being accurately reported, let alone whether the sample size was sufficient, or a hundred other questions we'd need to know the answer to.
For what it's worth, these are the purported figures for Westminster voting intention: SNP 34%, Labour 30%, Conservatives 24%. Take those with a bucketload of salt.
* * *
UPDATE: As you may have seen, Nick Moon of Moonlight Research has told Wings Over Scotland that his firm is not in fact a member of the British Polling Council, and is not even eligible for membership as it only has one client (presumably the Labour party). He implies that the confusion has an innocent cause, because as an individual he has a senior role in the BPC. I don't think that's good enough, because the organisation's website specifically lists Moon as the "company representative" of Moonlight Research on the BPC. That gives a grossly misleading impression, to put it generously - any reasonable person would feel entitled to conclude that Moonlight are full BPC members and are bound by BPC rules.