Thursday, June 5, 2014

SNP retain commanding Holyrood lead over Labour in Ipsos-Mori poll

I was beginning to wonder if the Holyrood numbers from this week's Ipsos-Mori poll were ever going to appear, but they've finally turned up a day later than expected. They show the SNP retaining exactly the same 9-point lead over Labour that was reported in the poll three months ago, with the SNP vote creeping up by 1%.

SNP 39% (+1)
Labour 30% (+1)
Conservatives 14% (-3)
Greens 5% (+3)
Liberal Democrats 5% (-4)
UKIP 4% (+1)

We're in an odd situation at the moment where I don't find myself instantly celebrating a handsome lead for the SNP, because what I'm most preoccupied with is the implications for the accuracy of referendum polling. In some ways it would be better if Ipsos-Mori - the most No-friendly of all the credible pollsters - were showing the SNP miles behind Labour, because then we could point to the disconnect between the poll and the European election results as an obvious reason for suspecting the firm is overstating the No lead. But, no, Ipsos-Mori clearly aren't using a suspiciously "Nat-light" sample.

But fortunately there's another way of looking at this. One or two concerns were raised about the accuracy of the Yes-friendly pollsters after both ICM and Survation significantly overstated the SNP vote in the European elections. (We can still count ICM as a Yes-friendly pollster for the time being, because their most recent poll will look like a bit of an outlier until and unless it is replicated.) The biggest potential issue is that both firms have recently introduced weighting by 2011 vote, and although the logic for doing so is beyond reproach, it's still an untested method and there could be a flaw in it that we don't know about yet. So, theoretically, the overestimate of the SNP's vote in May could have been the first sign of a problem. However, what makes Ipsos-Mori unusual (apart from the fact that they're the only active referendum pollster that uses telephone fieldwork) is that they don't weight by past vote at all, so their methodology is essentially unchanged from the one that produced a fairly accurate result in 2011. Accordingly, there's no reason to be sceptical when they show an SNP lead for Holyrood that is three times bigger than the lead actually achieved by the SNP in the European elections. Indeed, if we compare this poll (and Ipsos-Mori's last poll in March) to the Holyrood numbers from Survation's most recent poll, the SNP lead over Labour is only three points lower. So that lends greater credence to the theory that the overestimate by both ICM and Survation of the SNP's strength in the European elections was to a large extent not caused by misconceived methodology, but instead by differential turnout - ie. SNP supporters were proportionately less likely to go to the polls last month than Labour or Tory supporters. I'm not entirely sure why that would have been the case, although it has to be said that the SNP were the only major party I didn't receive a leaflet from, and my part of the world (the local council ward is a traditional SNP stronghold) was very noticeably not festooned in the traditional yellow posters. So maybe the SNP were deliberately conserving their resources for the bigger fight to come, and as a result failed to get their vote out as effectively as some of the other parties.

Talking of the European elections, this poll is also the first opportunity to see whether the results in May have themselves had any impact on the balance of opinion. Although the changes are not huge, the answer seems to be that, if anything, the SNP might have received a modest boost. The party may only be up 1% on the headline numbers which are filtered by certainty to vote, but the figures for the entire sample give a slightly stronger indication of the direction of travel -

SNP 38% (+2)
Labour 32% (n/c)
Conservatives 15% (-1)
Liberal Democrats 6% (-4)
Greens 4% (+2)
UKIP 4% (+2)

The Scottish subsamples from YouGov's daily GB-wide polls also support the impression that the SNP are on the up at the moment. No fewer than three out of seven subsamples since the European elections have shown the SNP ahead on Westminster voting intentions, which is highly unusual for YouGov - it's not very long since the SNP had gone for months without enjoying a single lead.

1 comment:

  1. What the European results say about different weighting procedures is very hard to make sense of.

    I notice that YouGov, although they guessed the gap between SNP and Labour most accurately, showed that 77% of the sample were going to vote, which of course is a million miles away from the 34% turnout in the election.

    Online volunteer panels having samples that are unusually politically aware could be a cause of distortion.