Saturday, April 5, 2014

Thoughts of Margo, plus a graph

I always feel at moments like this that I should be paying my own personal tribute, but there's nothing I could say about Margo MacDonald that hasn't already been said a hundred times better by others. I think all of us who have supported independence throughout our adult lives know exactly how each other feel tonight. And as long-term readers can probably imagine, I also thought that Margo was a tremendous force for good on the subject of prostitution laws and regulations - or at the very least a force for pragmatic common sense, which amounts to the same thing when a powerful and ideologically-driven lobby is threatening to do a great deal of harm.

*  *  *

RevStu has revealed that he has a new referendum poll up his sleeve for the coming weekend, presumably from Panelbase. As far as I'm aware, there have been no hints yet about whether the results are good, bad or indifferent. I know a lot of people will be hoping for (or even expecting) more progress for Yes after the recent crisis in the anti-independence ranks. However, we don't know when the fieldwork took place, and if memory serves me right, I actually saw someone mention on Twitter that they'd just completed a Panelbase referendum poll the day before the now-legendary leak to the Guardian was revealed. In any case, because the recent poll commissioned by Newsnet Scotland marked the first time in months that the No lead reported by Panelbase had fallen below the firm's normal range, it's worth preparing ourselves for the possibility that the Yes vote in that poll was exaggerated a touch by normal sampling variation, and that the new poll will show some kind of modest reversion to the mean. If that happens, it won't of course mean that the No lead has risen in the real world, although doubtless McDougall and co would try to spin it that way!

I'll also be looking out to see if Panelbase have stuck with their new and much more neutral preamble for a third poll in a row.

While we wait, I'd better get this complex and rather wonderful graphic posted before it's out of date. It was sent to me earlier in the week by Scott Hamilton, and it's based (as I understand it) on an extrapolation of the trend from a rolling average of the last eight polls. That differs significantly from the method I use for this blog's Poll of Polls, which is instead based on a rolling average of the most recent poll from each firm, regardless of when it was carried out. In my view, that method produces much more meaningful trend figures in a campaign that's been distinguished by pollsters disagreeing with each other wildly due to different methodologies. To take an extreme example, suppose eight polls had been published in February, and they had all been conducted by Yes-friendly pollsters such as ICM and Survation. And then suppose eight polls had been published in March, all conducted by No-friendly pollsters such an Ipsos-Mori and YouGov. A comparison between the crude averages of the figures for each of the two months would have suggested an increase in the No lead, even though in all likelihood the complete opposite had happened!

However, there are pros and cons in both methods, and it's worth pointing out that Professor Curtice uses an approach similar to Scott's.


(Click to enlarge.)



14 comments:

  1. It seems to me that, what with your analysis and stuff, that the BBC should be beating a path to your door.

    Yave you been approached?

    ReplyDelete
  2. In the six years I've been writing this blog, I've had precisely three expressions of interest from mainstream TV or radio, and two of them came from Argentina! (That's not a joke - there seems to be a genuine interest in how Scottish nationalists view the Falklands/Malvinas dispute.) The other was from Channel 4's 10 O'Clock Live. I sort of got 'auditioned' down the phone for that, but in the end they decided not to run the item (which is just as well, because I realised afterwards that I'd misheard the date and I would have catastrophically double-booked myself).

    ReplyDelete
  3. So if I'm reading this chart right, (which would be a first) Yes should overtake No at the very beginning of July to indicate a result nearer the 15% lead for Yes, Middle of July indicates an 8% lead for Yes, but should Yes take the lead about the 10th of August, we would be looking at nearer an end result of a 3% lead for Yes,...and a whole lot of squeaky bums!!!

    ReplyDelete
  4. Hi, I'm responsible for the graph- thanks for the interest.

    Extrapolation is generally considered to be most accurate for the same period covered by the real data so it should become more reliable as the weeks go by. I will keep this up to date as new numbers are available on whatscotlandthinks.org. Of course the pace of the trend may not continue at the same pace which could be good or bad for both sides but that will be reflected in updated trend lines and projections in my analysis.

    All the data is on John Curtice's website www.whatscotlandthinks.org and it would be good to see more analyses that cuts through the noise of the press and campaigning. All I have done is applied some basic stats to a basic dataset but it does make for interesting reading.

    Whilst I am on the Yes side of the debate, I would still have made this analysis available had it shown the opposite, or even a stagnant trend as this would still have been useful information for both sides.

    I'd interpret it thus- if No lose a similar amount of traction over the next 3 months then we will have pretty conclusive evidence of a very close race (in the polls).

    I do rather suspect that my analysis is lagging behind the mood of the country by a few weeks but I am keen to use a robust approach that resists shocks from rogue data. If the sides are even tied going into the vote the effect of turnout on the result could be dramatic and the result may not be as close as the polls would suggest- voter engagement will be key in the coming months.

    All my data and analysis are available from my dropbox, I'd welcome different interpretations but i would ask that everyone makes their analyses open and public.

    Here's the data for anyone interested https://www.dropbox.com/s/jbx8amn4wc3u69s/Indyref%20analysis%20of%20No%20lead%20Vs%20Yes%20time%20series_confidence.xlsx

    ReplyDelete
  5. Me again

    I will update the chart as new info comes in and publish it here

    http://indycharts.blogspot.co.uk

    cheers
    Scott

    ReplyDelete
  6. James, saw this on the Rev's twitter:

    New Panelbase poll cuts unionist lead to six points in record high for Yes in Sunday Times.

    ReplyDelete
  7. And Rev Stu dropping hints re interesting stiff in ST tomorrow - so prob Panelbase did ST & WoS Qs in tyhe same poll.

    ReplyDelete
  8. Typo "interesting stuff" - though if lead is down to 6 points, Darling's body may be found floating in the Thames - an "interesting stiff".

    ReplyDelete
  9. The full details will be released on the Wings site at 1 am. I shall read them in the morning.

    ReplyDelete
  10. The trend lines shown are very nice and optimistic but take no account of the fact that in the last few weeks of all VIP votes there are huge variations in the slopes of the curves. See polling results for Devo Referendum polls here. This suggests to me YES vote will be higher than trends shown above. Mind you that could wishful thinking. pic.twitter.com/QN15X9aIBw

    ReplyDelete
  11. I've deleted my earlier comment as I have a feeling it'll be proved inaccurate. I assumed that the 6-point No lead referred to the headline numbers, but it may have referred only to the position with Don't Knows excluded. If so, it's an even better poll than it first appeared.

    ReplyDelete
  12. What goes around comes around. See the link below for more info.

    #goes
    www.mocsbar.com

    ReplyDelete
  13. Thanks for sharing this kind of information to the public. Try to visit my site. Thank you and have a good day!.

    triciajoy.com

    www.triciajoy.com

    ReplyDelete
  14. I enjoy most of your articles,the articles are so nice for readers.

    Nersha
    www.gofastek.com

    ReplyDelete