Slightly surprisingly, the Liberal Democrats have released the datasets for the internal poll of East Dunbartonshire that they leaked yesterday - one of what Ashcroft famously referred to as their "comfort polls". So they've helpfully made it very easy for us to spot the working of the conjuring trick which has somehow got them (just about) into the lead in the poll, in spite of the word on the ground being that they're well behind, and probably not even in second place.
Here's how it was done...
* The results were weighted by 2010 vote recall. On the headline numbers, this resulted in respondents who said they voted SNP being weighted down from 53 to 33 (a drop of almost two-fifths), while respondents who recalled voting Lib Dem were upweighted from 73 to 90. Assuming that a significant proportion of people are probably getting their 2010 and 2011 votes mixed up, this factor alone is sufficient to introduce a huge distortion, and to transform what would have been a comfortable SNP lead into a small Lib Dem lead.
* No other form of political weighting was added to help balance out any distortion from the 2010 weighting (a practice used by some other pollsters, albeit admittedly not Ashcroft). Respondents were asked how they voted in the independence referendum, producing too big a lead for No, and yet the headline voting intention figures were not weighted by recalled referendum vote. Again, that alone would have been sufficient to put the SNP in the lead.
* The poll departs from standard good practice by not asking the headline voting intention question at the start of the question sequence. It's in fact asked fifth. The big problem is with Question 4, which names the local candidates and asks respondents to rate them. Unsurprisingly, this works in Jo Swinson's favour, because only 3.3% of the sample have not heard of her, compared to 49.7% who haven't heard of SNP candidate John Nicolson. (Although Mr Nicolson is a television personality, he's been off our screens for quite a while, and is probably more recognisable for his face and voice than for his name.)
* When the voting intention question is finally asked, it's posed like this : "If the general election was tomorrow, how would you vote in the East Dunbartonshire constituency knowing who is standing?" (my emphasis). People try to keep their responses to polls logically consistent, and if you've just said that you like Jo Swinson and haven't heard of John Nicolson, it's very hard to suddenly say you're planning to vote for John Nicolson, particularly when the question is insisting that you take into account who is standing. So it's likely that some people who plan to vote SNP (or indeed Labour) without being overly fussed about the candidates will have been coaxed into falsely saying that they are going to vote for Swinson. This is the twisting effect of question sequence that was famously illustrated in a Yes Minister scene, in which Sir Humphrey gets Bernard to say he is both in favour and opposed to the reintroduction of National Service.
What's interesting of course is that the Lib Dems have done a large number of these polls, but have only released a small fraction of them. That suggests even a rigged methodology isn't sufficient to produce a favourable result in many seats.
* * *
Have you ever had one of those nightmare journeys over a ridiculously short distance that leaves you wondering if a celestial power is conspiring against you? Yesterday morning I was due to take part in one of Derek Bateman's audio recordings at around 11.30, but the Stagecoach bus I was planning to catch turned up early, and I missed it by literally twenty seconds. So I texted Derek to let him know I would be late. After waiting an eternity, the bus I finally caught managed to shed its drive shaft midway during the journey. I didn't have the heart to text Derek again to tell him the bus had broken down, if only because it would have sounded like a classic "the dog ate my homework" excuse. So instead I hopelessly tried to make up the time by running when I got to the city centre, and I eventually arrived at the studio 50 minutes late. Thankfully everyone was very understanding, although when the recording is put on the website I suspect you'll still be able to hear me gasping for breath!