Friday, May 20, 2016

Crikey! ComRes confounds us with another cracking EU contradiction

This week has been the perfect example of what drives me crazy about people's irrational responses to individual opinion polls. We've had seven EU referendum polls since Monday, and this is what they've shown -

* Five reported a Remain lead.

* Two reported a Leave lead.

* Four reported a swing to Remain.

* Three reported a swing to Leave.

As you can see, the overall picture is as clear as mud, and yet there have been big movements on the currency and betting markets, all apparently attributable to just one of those seven polls - the Ipsos-Mori telephone poll showing the Remain lead surging from ten points to eighteen. Now, to be fair, it's arguable that the trend shown by Ipsos-Mori is more important than the modest progress made by Leave in some of the other polls. If there isn't a dramatic indyref-style convergence between telephone and online polling as we approach referendum day, I would imagine the Leave campaign will be hoping to get to mid-June with a clear lead in online polls, and a deficit of no more than a few points in telephone polls. An eighteen-point lead for Remain in a new telephone poll obviously makes the latter aim seem less likely to be realised, so from that point of view Remain are probably happiest with the new batch of numbers. But there's still no excuse for completely discounting the more Leave-friendly part of the story.

One thing is for sure - online polling this week has entirely failed to corroborate the Ipsos-Mori trend. The ICM and TNS online polls both showed Leave in the lead and a small pro-Leave swing - which perhaps can be explained by normal sampling variation, but certainly can't be easily reconciled with a supposedly massive surge for Remain. The only other online poll was from YouGov, and showed the Remain lead increasing from 2% to 4% - but half of that small increase was an illusion caused by a methodological tweak. Essentially, the state of play was unchanged, which again contradicts Ipsos-Mori.

Even the other telephone polls have provided only limited backing for the Ipsos-Mori narrative. In line with YouGov, the ICM telephone poll on Monday showed a statistically insignificant increase in the Remain lead. The ORB phone poll did show a bigger increase, but as I've noted many times, the trend reported by ORB phone polling has from the start looked suspect, because it's been completely out of line with other data. The latest piece of the jigsaw arrived in the form of a new ComRes phone poll last night, which showed a slight improvement for Leave. That, I would suggest, tips the balance and makes it more likely than not that the Remain surge in the Ipsos-Mori poll was at least somewhat exaggerated.

ComRes have started using their turnout-adjusted model for their headline numbers, which makes a comparison with their previous poll more complicated. Here are two different sets of like-for-like comparisons -

Old methodology :

Remain 51% (n/c)
Leave 41% (+1)

New methodology :

Remain 52% (-1)
Leave 41% (+3)


The swing to Leave on both counts is small and isn't necessarily genuine, but it's unlikely the reported movement would be in that direction if Ipsos-Mori were correct about the Remain lead almost doubling.

So is there any particular reason why the Ipsos-Mori poll might have led us astray? Anthony Wells noticed a couple of striking things about the poll - that the percentage of Tory voters backing Remain was much higher than in other polls, and that a disproportionate amount of the boost for Remain came from mere 'leaners', ie. people who initially said they didn't know how they'd vote, but who opted for Remain when pressed by a follow-up question to say which way they'd be more inclined to vote. Ipsos-Mori include those people in the headline numbers for a good reason - past history shows it makes the results more accurate. But all the same, this referendum is an (almost) unprecedented event, and there's a good deal of uncertainty about the turnout. It could be that many of the Remain leaners found by Ipsos-Mori don't really care that much about the issue, and will be less likely to vote at all. A whopping 83% of the full sample gave a firm opinion on the main voting intention question, which is obviously much, much higher than the eventual turnout is likely to be - and further calls into question whether the others will ever make it to the polling station, regardless of their apparent inclination to favour Remain. But even if the leaners do vote in decent numbers, their opinions clearly look more susceptible to change.

* * *

SCOT GOES POP POLL OF POLLS

50/50 ONLINE/TELEPHONE AVERAGE :

Remain 47.8% (+2.0)
Leave 41.7% (+0.4)

ONLINE AVERAGE :

Remain 43.3% (+0.7)
Leave 44.0% (+0.5)

TELEPHONE AVERAGE :

Remain 52.3% (+3.3)
Leave 39.3% (+0.3)

(The Poll of Polls takes account of all polls that were conducted at least partly within the last three weeks. The online average is based on nine polls - four from ICM, two from YouGov, one from TNS, one from ORB and one from Opinium. The telephone average is based on four polls - one from ICM, one from ORB, one from Ipsos-Mori and one from ComRes.)

6 comments:

  1. Groan! No again!

    Do any of them ask any of the circa 2 million so called Brits that live elsewhere in the EU what their voting intentions are? I believe 2 million was the figure used by the EBC today. How many if any of them have the vote? It could certainly mess up any opinion poll and the actual vote.

    ReplyDelete
  2. It's Ipsos MORI, not Ipsos-Mori. MORI is an acronym.

    ReplyDelete
    Replies
    1. Yes, I know, but UKIP is an acronym as well, and it's amazing how much of the media refers to them as Ukip. I'm in the habit of writing Ipsos-Mori, and I don't think it's hugely important one way or the other.

      Delete
  3. Going back through the polls since March (using the data available from whatukthinks.org), I count 20 polls where Remain was ahead of Leave by more than the margin of error (which I set at 3%, which is admittedly a bit crude) and only 6 where Leave was ahead. In fact, I was overgenerous to Leave, since I required Remain to have a lead of 4% or greater, while Remain needed only 3% or greater to count. Setting the thresholds equal for both makes it 23-6 or 20-3.

    If I were in favour of leaving (I'm not), I'd probably be clinging to the hope that the online polls are more accurate than the telephone ones. However, even that would be cold comfort, since those don't show Leave beating Remain by the margins we're seeing in the phone polls. While I agree that the "surge" for Remain is probably an illusion, I think the failure of Leave to have a consistent and comfortable lead in any type of poll is damning enough.

    Besides that, I think that the majority of undecided people will ultimately choose the status quo over leaving. Despite the Remain campaign trotting out the same tired arguments that were used in the indyref (to say nothing of how the Cameorn and Osborne don't exactly inspire public trust or loyalty), the Leave campaign haven't been any better and I doubt they're persuading people in the proportions needed to swing this.

    ReplyDelete
    Replies
    1. Glasgow Working Class 2May 24, 2016 at 12:21 AM

      Holly. The meal ticket is possibly about tae be removed, it is over tae you!



      Delete
    2. 23 stays up late at night, desperate to get the final word.

      Delete