Wednesday, September 17, 2014

A fourth successive poll within less than 24 hours puts Yes just 2% away from victory

Panelbase's final poll of the campaign :

Should Scotland be an independent country?

Yes 48% (-1)
No 52% (+1)

(UPDATE : The 5% No lead that you might have seen reported from the figures that include Don't Knows has been exaggerated by the effect of rounding.  On the unrounded figures, it's...

Yes 45.4%
No 49.5%)

The apparent swing to No is not statistically significant, and because this takes us back to the position in the last-but-one Panelbase poll (which was conducted at roughly the same time as the famous YouGov poll that had Yes in the lead), the likelihood - but not certainty - is that we're looking at margin of error noise.

Again, when it's so close we're entitled to ask the question - is there any systemic bias in Panelbase's approach that might be understating the Yes vote?  Unlike the online pollsters that reported last night, they do correct for the over-representation of English-born people in their raw sample, so we can't use that as an alibi.  But at exactly the same time as they introduced country of birth weighting, they also introduced a controversial procedure which reduces the reported Yes vote, and which no other firm uses - namely weighting by recalled European Parliament vote.  252 people who recall voting SNP in May (94% of whom are Yes voters) have been downweighted to count as just 214 people.  The logic for doing this is that the Panelbase sample always has a disproportionately high number of people who voted in May, meaning that it's therefore prudent to make sure those people are weighted in line with the result of that election.  But the problem is that if you have far too few non-voters from May in your sample, the logical conclusion to draw is either that the sample is hopelessly unrepresentative of the population you are trying to poll, or that a lot of people are not telling you the truth (or some combination of those two factors).  That being the case, you can't be sure whether weighting by vote recall is eliminating a bias, or introducing a whole new one.

The fieldwork for this poll is a bit more up-to-date than the three we saw last night - it took place between Monday and today.  So it's tempting to see this as a strong clue that we shouldn't expect any significant shift to No in tonight's YouGov poll, not least because on the unrounded numbers the Yes vote is actually slightly higher than it was in the last-but-one Panelbase poll.  The problem is, though, that Panelbase's trend has become completely decoupled from the YouGov trend - Yes first reached 48% with Panelbase months ago, at a time when YouGov were still showing them on 39-42%.  So it's anyone's guess what the "Kellner Correction" has in store for us this evening.

More details and a Poll of Polls update to follow...

47 comments:

  1. So 4 pollsters sitting on 48Y, 52N, does this mean a high level of accuracy within MOE

    Maybe just playing it safe to avoid egg on face?

    ReplyDelete
  2. Again, the result is within the margin of error. A statistical tie with, perhaps, a tiny advantage for No.

    Note that we have not had a poll under 46% Yes for a few weeks now, but that we have had three that show Yes in the lead or a 50/50.

    Besides, if it turns out that there has been some structural flaw in all polls (as it seems likely, given the fact that many pollsters have too few Scottish-born respondents and that there are many new voters), we are in a place where everything is possible.

    ReplyDelete
  3. 4 with 48% something very suspicious going on!

    ReplyDelete
  4. Panelbase:

    People that think that most of their friends and family will vote Yes - 37%

    People that think that most of their friends and family will vote No - 35%

    That's interesting!

    ReplyDelete

  5. Different methodologies, same result.

    Only one methodology can be correct, but they can all be wrong.

    There is something not right here. Are we seeing a human factor(s) being introduced for such artifical convergence?

    ReplyDelete
  6. MoE works both ways...No could be better...be realistic it's not 50/50...it's close but this is not that good for us (yes I know we've really closed the gap) but we just don't cross the finish line enough :/

    Chris D

    ReplyDelete
  7. @Chris; the race has now been brought to a draw, a tie. For all purposes its a tie race.

    3.8 million will be voting and 1 out of 4 is a new/refreshed former non-voter.

    As James rightly states its a two point gap not four point.

    IF it is true that YES has 35,000 on the ground activists (local) for Thursday who will be getting the votes to the polls, then YES has won.

    Who is more likely to blow a race?

    David Cameron or Alex Salmond

    New Labour or Radical Independence

    I have seen too many political races blown by posh establishment types to not believe that two years of on the ground work will pay off.

    ReplyDelete
  8. Have to say that I'm getting a tad concerned with the consistency of the polls. We are now in a position where we are hoping that all the pollsters have got something fundamentally wrong. It's possible but I'd rather not be in that position.

    I've spoken to a few undecideds today who seem now to be leaning No. Like many others, I thought Yes were keeping something back for the final couple of days which would sway these voters but that doesn't seem to be the case. A tactical error by Yes or do they know exactly what they are doing?

    Anyway, roll on YouGov later this evening and lets hope for something other than 48 / 52.



    ReplyDelete
  9. @Chris, neither YES or NO has crossed the 50% mark.

    Anecdotal political wisdom that I learned in America, if neither party has 50% the entire week before the election the race is fully up for grabs.

    Thats including do not knows, because the deal has not been sealed.

    IF NO was at 51% this last week, with YES at 45% and undecided at 4% then the message would be all over.

    ReplyDelete
  10. Could it be that all the polling firms have the same panel of voters to pick from? If that is the case, then they could simply all be talking to themselves.

    ReplyDelete
  11. In regards to GOTV, let me provide some history. That being Florida in 2012.

    Neither Romney or Obama could break 50%, so the race was wide open.

    Romney had a lead in the poll of polls, guess who took Florida in 2012?

    http://www.realclearpolitics.com/epolls/2012/president/fl/florida_romney_vs_obama-1883.html

    ReplyDelete
  12. @George

    To be fair, Panelbase has too few undecideds.

    ReplyDelete
  13. If the vote is that close then this is not good news for the BBC if No scrapes through. There will be a lot of angry finger pointing at the BBC.

    ReplyDelete
  14. All the polls showing the same in fear we are in for a disappointing Friday sadly. So near yet so far.

    ReplyDelete
  15. Chris : They are indeed showing the same thing so far - a virtual dead heat. As the pollsters themselves are openly admitting to the media, that means that we don't actually know who is going to win.

    ReplyDelete
  16. If the voting intention out there among likely voters is really 48 Y, 52 N, and assuming that 3.800.000 people vote (roughly, 90% of turnout), the distance would be of 150.000 votes, more or less.

    If turnout is lower, or if these polls affect sense of urgency in voting in the No side, then we will be within touching distance or winning.

    And there's still another possibility: the polls maybe structurally wrong, even if it's just for a couple of points.

    ReplyDelete
  17. @xabi, I did an analysis along your lines but a little more specifity. Its on another thread.

    I agree the vote is 48/52 based on the voting pool of 2010 which was 2,542,636.

    The question is how the "new" 885,621 who have never voted before or stayed away from politics will now vote?

    In a nut shell if the base 2.6 million goes like the pollsters predict but YES gets 60% of the 900k, I think I deserve a medium rare steak. ;)

    ReplyDelete
  18. No is going to win, end of.

    ReplyDelete
  19. Oh, what an impressive piece of trolling, Anon. Getting tired?

    ReplyDelete
  20. Also, I think many of us are feeling a bit pessimistic because we have been following the polls for months and a week ago it seemed like we were going to be ahead in the last few days of the campaign. Now we see that we may go into the voting day an inch behind, and we feel unsure.

    We should be realistic: if three weeks ago we had been told that a Panelbase poll showing a 48/52 result would be a bad one, we would probably have felt encouraged.

    Most voters are actually not having this extreme mood changes. They probably cannot distinguish a 48Y/52N result by Opinium (a nice figure) from a 48Y/52N by ICM (a setback, given the previous spectacular 54Y/46N poll).

    So just chill out and carry on. We can win this.

    ReplyDelete
  21. Much as I like what the canvassing data says it is inherently unscientific so a nice to have set of data. No claim their canvassing says they will win. If it's 50/50 why isn't 50/50? Worrying I gotta admit. Hopefully tonight's polls will give us the lead.

    Chris D

    ReplyDelete
  22. Nobody have any thoughts on my earlier post?
    If all the polling firms are using the same few thousand that have registerd with them, then it isn't surprising that they are getting the same results.

    Does anyone know how many folk in Scotland have signed up with polling firms?

    ReplyDelete
  23. Its maybe a bit cynical but do you think that none of the polling companies want to look seriously embarrassed if the final poll is way beyond their predictions?

    As a safety blanket they will ensure that this 48/52 split is consistent across all of them so none can have the finger pointed at them.....hence no dent in their reputations

    ReplyDelete
  24. Juteman : Yes, but the phone polls we got on Friday and Saturday were broadly in line with the online polls. There may well still be systemic issues that are understating Yes, but that makes it much less likely that the culprit is the online method.

    ReplyDelete
  25. Does the fact that the figure with Undecideds included is very close to 50% have any bearing on George's analysis posted above?

    ReplyDelete
  26. Have to say (and hope this isn't concern trolling), worried by them getting 50 including DKs. I was heartened by other polls still showing them failing to get half to back the Union. They also say most undecideds claim they're leaning No. Interesting thing is how many undecideds they're getting. ICM had 14, Opinium 6 I think and now 5. Let's say the real number is 10, I don't think that half of them will have made up their mind in one day.

    ReplyDelete
  27. Well, if it's any consolation it's actually 49.5% including DKs.

    ReplyDelete
  28. In every poll the weighting must use a turnout model. This makes assumptions that the turnout model for the referendum is the same as the model for an election in most cases Holyrood Panelbase uses the EU election. So weigh according to % in the HR election. But they seem to have a much higher % of folk that voted in HR than actually did so that must mean the HR non-voters are under represented in the weighings. So I am not sure how it can work? James do you understand it?

    ReplyDelete
  29. @Niali, YES remains in the race because NO has been unable to break 50% in its own right. It can only break 50% when undecideds are effectively tossed aside.

    Having said that, the margin of error says YES has been able to make up ground in one month and bring the race on paper to a statistical draw.

    I have been personally involved in races at 48/52 and won and lost with 52/48!!

    Its all ground game now. The strategy of Project Fear was rather clever and they succeeded in stopping the momentum of YES. BUT.......the undecideds remain.

    Plus.......we have the 800 pound gorilla in the room and that is the 900,000 new/fresh voters.

    @YES, I have no inside information but in the backrooms of NO, they have the same doubt I am seeing on this board. They think they are going to lose regardless of the lead in the poll. They are likely asking how those hundreds of thousands C2DE newly registered voters are going to vote after being driven to the poll by RIC.

    ReplyDelete
  30. @Callum, If you look at the movement in the polls over the last two weeks, you will see as the actual undecided number drops it is largely going to YES.

    It is actually too odd that the undecideds are going almost one to one to YES once they break.

    ReplyDelete
  31. Just something to think about:

    If there had been no polls in the past 24 hours would you have felt more or less confident of a Yes victory?

    And if the polls all showing Yes behind make us feel less confident, does that perhaps make us less effective at swaying the voters who refuse to think but will go with the flow?

    So in whose interests is it to keep Yes behind in the published polls? Do we really think that polling companies are truly neutral in this campaign?

    It is incredibly easy for them to manipulate their results and these figures are precisely what I was expecting to see published at this stage if Yes was actually ahead. Not ridiculously low as YouGov were before because now they have to retain credibility but just low enough to dishearten.

    Out on the streets the picture is different. I know it's unrealistic to expect this site to ignore the polls but I think we will all be more effective campaigners if we do just that and get out amongst the people who will decide this.

    ReplyDelete
  32. 0.5% doesn't feel like a consolation. Panelbase are vipers that have turned on us

    YouGov should be better.

    Chris D

    ReplyDelete
  33. "Panelbase are vipers that have turned on us"

    WHAT?

    ReplyDelete
  34. As you guys are saying, Panelbase numbers have reached this level for a few months, but I remember I think it was June 2013, Panelbase stopped inviting new panelists after an influx of Yes voters signed up. And they stuck with the same lot they had been using before.

    Now, it's my belief that the people being polled for Panelbase have made up their minds for some time. Hardcore Politico's. That would explain the no changes in numbers. Again, you throw in the 2011 figures instead of 2014 and I think we would be in the lead with PB and have been for some time. Or at the very least equal.

    ReplyDelete
  35. Chalks : They weight by both 2011 and 2014. If it was only 2014 it would have an even more extreme No-friendly effect.

    ReplyDelete
  36. Ipsos Mori for STV... Yes 49% (+7) / 51% No (-7)

    Sample of 1405 polled on 15th & 16th

    ReplyDelete
  37. Panelbase were on our side showing us consistently good Yes results and they bottled it and refused to go over 50% so yes...vipers

    Chris D

    ReplyDelete
  38. @James, @All YES, From the Torygraph-

    This is a must read by someone who likely would not be in favor of YES but is also a watcher of the US Political scene.

    "Don't be surprised if Scotland votes for independence. I think "yes" might have an edge"

    by Tim Stanley

    http://blogs.telegraph.co.uk/news/timstanley/100286651/dont-be-surprised-if-scotland-votes-for-independence-i-think-yes-might-have-an-edge/

    ReplyDelete
  39. Ah, thanks James.

    The VIPERS!

    lol

    ReplyDelete
  40. Ipsos has us on 47 inc. DKs which is fantastic! God, these polls really are helping me develop a manic depressive side (apologies to any actual manic depressives).

    I hope that the YouGov poll with its big impressive sample shows this or something like this. I'd be more than happy going into the day with us polling 2 behind.

    ReplyDelete
  41. Kellner (YouGov) has just said on Radio 4's PM that No will win.

    ReplyDelete
  42. Yeah but what does Kellner know about anything?

    Chris D

    ReplyDelete
  43. Are we beginning to see a very, very late swing to Yes?

    ReplyDelete
  44. So IPSOS now joins the rest of pollsters in converging. 49% for Yes, still too close to call.

    Kellner's words suggest YouGov poll tonight may show a bigger lead for No, but considering every other pollster is stuck on that 48/52 trend, YouGov would be the outlier.

    ReplyDelete
  45. Chris, Kellner knows what the YouGov poll is going to say tonight

    ReplyDelete
  46. Kellner actually says that he can't be sure the No are actually in the lead. YouGov poll biggest sample size of 3237 tonight

    ReplyDelete