Sunday, May 18, 2014

Support for independence at 42% in new ICM referendum poll

From the moment a few hours ago when the Scottish Tory leader gloated that she was "looking forward" to this month's ICM poll, it was obvious that it was going to show at least some kind of increase on the wafer-thin No lead seen last time around.  But the big question was just how high would the No lead have to be to justify Tory celebrations?  We now know the answer, and it appears that "bad is the new good" for the No campaign, because they are getting excited about exactly the sort of poll results that used to have them screaming "bias" when Panelbase were producing them last year -

Should Scotland be an independent country?

Yes 34% (-5)
No 46% (+4)

With Don't Knows excluded, it works out as...

Yes 42% (-6)
No 58% (+6)

Contrary to the predictable but rather silly spin in the Scotland on Sunday, these figures do not necessarily mean that support for independence has decreased or that opposition to independence has "recovered".  The No lead reported here is within the range that ICM has shown thus far this year, of between 3 and 12 points.  The likelihood is that the last poll from the firm was on the high side for Yes due to normal sampling variation, and that we are now seeing a reversion to the mean - or more likely to something quite a bit higher than the mean.  Unless other pollsters show a significant increase in the No lead in the days and weeks to come (and both Survation and TNS-BMRB singularly failed to do so earlier this week), the changes in this poll will be safely tucked away in the file marked "margin of error illusion".

It's interesting, though, that ICM have proved more volatile than other pollsters.  They've certainly mucked around with their methodology more than others, changing the broad age categories that they use for their weighting procedure at least twice so far this year, and in their last poll they effectively started assuming a 100% turnout by dramatically upweighting non-voters from 2011, thus increasing the reported number of undecided voters.  Ironically, the latter change harmed Yes slightly, and without it the No lead would have fallen even further - so there would actually have been greater volatility.  However, the bottom line is that the more methodological changes there are, the less meaningful it is to make direct comparisons with previous polls.  Time will tell if further tweaks have been made in this poll, although it's certainly troubling that Martin Boon in his commentary last month almost seemed to be looking for an excuse to introduce a mechanism that would artificially boost the No vote.  He had apparently fallen for propaganda from anti-independence campaigners suggesting that No voters are more likely to lie to pollsters about their voting intentions - a claim for which zero evidence exists, and if anything the opposite is much more likely to be true. However, Boon merely suggested that ICM were going to do some research into the matter, so hopefully there's been no jiggery-pokery in this particular poll.

ALERT!  ALERT!  "Better Together" are playing silly buggers yet again : In the wake of the ICM poll, McDougall Central have tweeted a statistical graphic purporting to show that six different pollsters are either suggesting an unchanged No lead, or an increased No lead.  At first glance I thought there was something very fishy about it (well, none of us are going to faint in amazement at that, are we?) and sure enough, the explanation is contained - quite literally - in some hilarious small print at the bottom of the image...

"Source - latest poll by regular pollsters of voting intentions, comparison with previous poll.  YouGov compared with YouGov for Progressive Partnership."

In plain English, what they mean is that they're brazenly pretending that YouGov, who showed a 1 point decrease in the No lead in their most recent poll, are actually showing a 6 point increase in the No lead.  Their flimsy excuse is that the fieldwork for Progressive Scottish Opinion was conducted by YouGov, and that they can therefore count it as a YouGov poll.  This in spite of the fact that YouGov have nothing whatever to do with the weightings that Progressive apply to the raw data, which even the dogs on the street know is much more favourable for No than YouGov's own procedures.

Talk about trying to have your cake and eat it.  It's doubtful whether a non-BPC pollster like Progressive should even be taken into account, but if they are to be included it means that two out of seven pollsters are currently showing a decreased No lead in their most recent poll - and in the case of Progressive the decrease is a whopping 9%.  To get around that little problem, the No campaign are pretending that YouGov and Progressive are exactly the same pollster.  How can anyone take these chancers seriously anymore?

UPDATE : I'm holding off on updating the Polls of Polls (which I'll do in a fresh post) until we've heard the result of the new Panelbase poll.  The Sunday Times seem to have already revealed that the headline Yes vote is unchanged on 40%.  We're still waiting on the figure for No, but that will presumably be in the same ball-park as their 45% share in the last poll (unless of course there has been a dramatic increase or decrease in the number of undecideds).  When taken in conjunction with the Survation poll a few days ago, that would significantly increase the chances that the change shown in the ICM poll is an artifact of the margin of error rather than something real.


  1. Sunshine on CrieffMay 18, 2014 at 12:49 AM

    After a fantastic day at the Scottish Cup Final, this poll has certainly put a dampener on things.

    I appreciate your explanations, but to go from a No lead of 3% to 12% (or 4% to 16% after don't knows are excluded) doesn't look good.

  2. Sure, It's not a great poll but it is still one poll and there is no way on earth the polling (bad or good) is going to stop the grass roots campaigning on the ground for Yes. Campaigning which is only going to get more crucial and decisive the closer we get to September. Just like it did in 2007 and 2011.

    If you want to find comical imbeciles freaking out over a poll or two you go to politicalbetting where the 'analysis' of out of touch twits, a mockney sex tourist and an old man's farts are actually taken seriously along with the cowardly tory moderators. They might as well rename it DanHodgesBetting going by some of the hilariously dimwitted 'output' recently.

    Fact of the matter is if this keeps 'better together' doing exactly the same thing they have been then I for one will be delighted. The same witless fools in SLAB who thought negative campaigning would see them win easily in 2011 are now being joined by out of touch tory twats who also don't know any better. They still haven't work out why there are more pandas than scottish tory MPs.

  3. ICM has shown a lot of volatility in the past few months, and I would suggest that these latest figures have something to do with that. I thought last months' poll was too good to be true and I appear to have been proved right.

    Quite incredibly, ICM have went from being the best pollster for YES with don't knows removed to the 2nd worst - on their whole sample: they show Yes: 40% No: 60% (although the change in weighting procedure means that it can't be compared to polls before last month).