Panelbase’s final call poll for the referendum is out and also has a 48/52 split. Full topline figures are YES 45%, NO 50%, Don’t know 5%, which excluding don’t knows becomes YES 48%, NO 52% – the same as in all three of yesterday’s polls. While it doesn’t seem to be their “official” headline figure, Panelbase also included a forced choice question to don’t knows asking them to imagine they were in the polling station, right now, with no don’t know option – with that data it would have been YES 47, NO 53. Full tabs are here.

Still to come later tonight are the final YouGov poll and the penultimate Ipsos MORI poll.

40 Responses to “Panelbase – YES 48, NO 52”

  1. Surely this is evidence of a conspiracy! For what, I don’t know…

  2. It’s interesting that this poll corroborates some of the evidence put forward by some for the claim that the DKs split slightly in favour of Naw.

  3. Who would have believed it…

    h ttp://

  4. Well the polling evidence is certainly consistent! I still think there will be a slight shy No effect to widen the margin a little.

  5. “Panelbase also included a forced choice question to don’t knows asking them to imagine they were in the polling station, right now, with no don’t know option – with that data it would have been YES 47, NO 53. ”

    So if held at gunpoint, the DK go towards NO. That is awful awful news for YES.

  6. Skippy

    So all the no campaign need to do is rustle up a few 100 AK-47s? Seems an extreme way to get the vote out!

  7. A poll of 47 people is still a poll of 47 people. More polling should have been nice done on DKs to gain an understanding months back.

    Too few now to create any real understanding from.

  8. Interestingly, Panelbase shows 35% of labour voting yes, survation shows 23%.

  9. Fraser

    It still would be hard to determine which issues pushed which DK into which camp. I’d assume that the majority of them becoming more certain about which way they had to vote as the campaign wore on came down to more than “Oh heck, I have to vote as date is approaching so… ermmmm. [“yes”|”no”]”

    The undecideds who do find themselves in a polling booth will in effect be faced with a forced decision which seems to split 3:2 (with humongous error bars so not much to read into that). There’s no reason to think than an undecided 6 months out will end up going through the same process so getting a forced answer 6 months previous will be unlikely to produce a prediction of how the campaign will run.

  10. This may be a daft question, but…

    If we have a number of polls showing the same thing, taken over the same or similar period, and as the question is the same.
    Can we pool them and reduce the MOE as if it was a poll of 4000 people?

  11. @ Skippy

    Maybe that proves the case for ignoring crossbreaks but the poll as a whole being accurate?

  12. Lost in the melee, this morning saw the release of the latest Markit Household finance index.

    The headline figure is marginally up, but still denotes contraction. A bright spot was the quickest growth in employment income since last July, but perceptions for the next twelve months have again edged slightly downwards.

    Overall, a pretty quiet set of figures, but certainly no signs of the kind of sharp revival that intuitively one might have linked with a strong swingback in polling terms.

  13. In previous post on the AW question “Will the polls get the Scottish referendum right?” I have looked forward to how the opinion polling community will be able to explain either the uncanny accuracy of a complete success by all pollsters hitting virtually the same numbers with so many unknowns and different methodologies, or why everyone got it horrendously wrong whilst all agreeing the same numbers.

    There seems a very good chance that the problem on Friday will be attempting to explain the degree of unanimity and unbelievable accuracy.

    However there is a finite chance, let’s pull a figure out of the air, say 10%, that all polls get the result on the day completely wrong, even though they are in complete agreement with each other.

    Why is that?

    To stick with the two horse race analogy, that extra risk (to pollsters) is there because this is a race over a selection of low hurdles and some really tough fences with some excitingly unknown unknowns.

    These all depend on some key events today and tomorrow, mostly inside the collective Scottish mind, but also a few outside that hallowed and sovereign territory.

    48/52 again? What a laugh!

  14. RedFishUK

    Once you get over 1000 respondents, the MOE only decreases very slightly with each additional thousand.

    The end result will be all-but exactly the same.

  15. The more precise gap is 4.3% excluding DKs. With the forced choice responses it’s 5.1%.

  16. Phil Haines

    I wish you had posted that before I went to the bother of working that out! :-)

  17. I’m not sure consistent small leads for No means we can call it. It just means that the polls are consistently saying “Too Close To Call”.

    I still have a niggling feeling Yes may take it, based on the degree of motivation many of their supporters have. My gut tells me Yes voters will turn out.

  18. RedFish

    I suspect differences in how people are selected, polled and weighted might prove to be problematic and if all Polling Companies used the same methodology then they would have no “expertise” to sell to clients. The idea being polling companies want to show that their methods lead to better results than “Brand X”.

    Even if polling companies used similar selection methods, they differ in prompting/ not prompting for minor parties (not an issue for a yes/no referendum but there might be variations in wording that does create a bias still) so even working on a pooled amount of raw data might be problematic.

    Averaging data weighted with different weightings becomes even more problematic and would be pretty unscientific to say the MOE as a result of the 4 polls was 1.5%, yes having polls agreeing with one another gives “some confidence” that that are closer to the truth that a single poll, but over simplifying the process can lead to a pretty unscientific outcome.

    Most polls are around 1k people as that is where the sweet spot between cost of making the poll compared to information gained from the poll lies. Also systematic errors would become apparent if you started doing polls with a MOE of 0.5% so no value is gained by doing them

    In this case it seems that the information is “within MOE of either victory” and so a larger sample size might prove useful as far as predicting the result with confidence goes, as it’d be hard for these polling companies to gain much credit if the result ended up 51% Yes 49% No by claiming they were within MOE.

  19. You also don’t know if each individual was only sampled once. Obviously, they probably were, but then you have to take that uncertainty into account when considering the poll.

  20. @ Neil A

    The old and new systems will run in parallel until, I believe, after the next General Election so if you were already registered under the household system you’ll continue to be so. That system will eventually be replaced by IVR, though.

  21. This is actually more meaningful than last night’s polls, as when the undecideds are counted, the No camp still commands 50%. The other polls have shown less than 50% when counting the undecideds.

  22. “penultimate Ipsos MORI poll”

    So there’s going to be still another Ipsos Mori poll after that? When will that be published (I assume no polls allowed on the polling day itself or am I mistaken?)

  23. Tomorrow will be an historic night as it appears that the BBC has retired Dimbleby.

  24. @Steve2 @Alan


  25. YouGov sample size apparently 3,000 tonight.

  26. Wam, imagine if that showed a Yes lead….

  27. Neil A
    You can get 4/1 on yes so perhaps you want to back your hunch….?

  28. I have a little niggling “what if” thought. Vast numbers of people who have never bothered to vote before are seemingly galvanised to vote this time. What if, when push comes to shove, they revert to type. Those of us who have done a lot of canvassing have all met loads of people who were going to vote, without fail, never fear, definitely voting this time—– and then don’t. If they don’t turn out after all then I would guess NO will win by some margin. We shall know soon.

  29. RMJ1,

    I have dark images in my head of a close Naw/Aye win of 50.01%, with months of outrage and challenges and charges.

    I should try some more of Old Nat’s Nytol.

  30. Here’s another way of speculating about the motivation to vote or not among a section of the electorate…suppose you support yes, would like them to win etc …but at the same time have enough doubt that you can’t face the prospect, should they win, of having yourself to blame for the outcome if things do indeed go tits up afterwards.

    Might you not in those circumstances, leave it for “everyone else” to see the yes vote through and not actually vote yourself…then you can deny all responsibility to yourself if it turns out to be a bad move.

    You can speculate endlessly about imaginary people’s secret thoughts. It probably doesn’t help.

  31. WAM

    That’d correspond to a MOE of ~1.8%, might be enough to make a 95% prediction, although will need digging into the data for a significant figure. The difference between 47.5%+/- 1.8% and 48.5% +/- 1.8% is enormous.

  32. Apparently there is another poll for stv to come out shortly? (according to Aasmah Mir on twitter)

  33. @Neil A and RMH – As far as I know polls don’t measure the efficiency of those getting their supporters out to vote. Enthusiasm may affect this as well as personal propensity to vote oneself. I personally want no to win but I won’t be easy in my mind until they have actually done so.

  34. Gattino

    Or you can vote yes assuming that a proportion of the yes’s are already voting in that manner.

    Or you can abstain assuming a proportion of the yes’s are voting in THAT manner…


    and by the time you worked out whether you should vote or not, the polling stations are closed absolving you of all responsibility.

  35. It’s the ipsos-mori one: 49/51

  36. Kellner on R4 now discussing polling and referendum.

  37. MARTIN.
    I have just seen the IPSOS MORI poll on my twitter.

    Wow, if this is the best poll in terms of accuracy then we are in for a very nervous night.

  38. How can one poll be so wildly out with the others?


  39. @R&D – indeed! lol