A pro-independence blog by James Kelly - voted one of Scotland's top 10 political websites.
Tuesday, March 24, 2015
A little bit more about the ICM poll
Just a quick note to let you know that I have a short article at The National about yesterday's ICM poll, which put the SNP 16% ahead. You can read it HERE.
YouGov sub-sample: SNP 37, Lab 31, Con 17, Others <6.
Towards the lower end of SNP scores and higher end of Labour scores.
There has been a trend towards slightly lower SNP scores over the last week or so. SNP haven't scored higher than 44 since 13th March. It is probably just variation around the long term average. There has been no sign of movement in the other polls, particularly the more reliable full-scale work.
That's true. There hasn't been an SNP score lower than 37 for at least a month. Labour haven't been higher than 33 in the same period. That indicates that there is still a pretty big gap, because if the gap was as "low" as 10 points, you would expect some sub-samples to put Labour ahead of the SNP.
Westminster, Holyrood, what did I vote in September, independence question, what I've seen / heard about Scotland's position in the UK recently, whether I'm interested in that topic, how well I understand the role of UK vs Holyrood Governments in Scotland, question on powers / whether the matter is settled, national identity (Moreno), country of birth, how interested I am in rugby...
Anyone seen some of the SNPOUT nonsense on Twitter. Haha! They believe the polls are filled with 'Nats' and that come the 7th of May, the SNP will be lucky to hold on to their six seats.
What's even more odd is the demographics of this 'group'. Seem to mostly middle-aged, women, settlers.
I wonder if they think the huge surge in SNP membership is a mirage? Or even that it's isolated, and doesn't reflect a seismic shift in the electorate as a whole?
The "polls are rigged by Nats" meme is not going to go away, because (like most myths) it is based on a small kernel of truth. Panelbase did have too many SNP supporters in their panel, it skewed some of their early results and they then recruited a new panel. Since then their results have been more in line with other companies, apart from the last poll where they used a question ladder.
But for the other companies (and indeed for Panelbase as well since they sorted themselves out), it's nonsense. In fact, Anthony Wells of YouGov specifically debunked this idea on a BBC radio programme ("More or Less") a week before the referendum. He said that they had heard of this allegation that the online panels were stuffed full of nationalists, so they had gone back and checked the composition of their panel and if there had been unusual numbers of people signing up from particular places or backgrounds. Turned out the sign-up rate in Scotland in the year before the referendum was basically the same as the year before (and so on).
The funny thing too is that the online results should be (if anything) *more* accurate as a representation of the unionist / nationalist balance in Scotland now than they were before the referendum. Most of the online firms (Survation, Panelbase, YouGov, ICM) now use the referendum result as part of their political weighting in the polls. Everyone should be able to remember how they voted in the referendum, given its proximity and importance, whereas there was some concern before the referendum that political weighting was misleading due to false recall of how people voted in elections (Scottish and UK) 3-4 years previously.
Then you also have to throw into the mix that other polling methods (TNS face-to-face; Ashcroft and Ipsos Mori over the telephone) have also shown big scores for the SNP. Are these muppets seriously arguing not only that nationalists are more likely to sign up for online panels, but that they are also much more likely to answer the phone?
There is reason to believe that the actual result may be closer than current polls indicate (e.g. if "undecided" voters return to their previous VI, or if turnout is lower than expected), but it won't be because the whole polling industry is systematically over-rating the SNP.
James : When did Panelbase recruit a new panel? Was it a long time ago? I've no recollection of that, although I do recall them blocking new members from taking part in referendum polls (as a precaution against "infiltration"), and also using another company's panel for half of the sample in one of their polls to see if it made any difference.
Not long now, but the polls even could be under-estimating the SNP support. It means zilch in the long-term of things, but I know of 3 people. One, a Tory and the other two former Labour voters who are all voting SNP. The Tory is a tactical voter (he lives in Dundee West) so he'll hold his nose and vote SNP, the other two I think both voted No, but now voting SNP.
If 80% of the 1.6 mil who voted Yes turn out then, well well, fireworks.
Thanks for clarifying James. I do remember them suspending new recruits - I think after the "rogue" independence poll in mid-2013? I must have mistaken the use of an additional panel for them recruiting a new one.
I think the polls probably are over representing the SNP support. Nationalists are quite evangelical and eager to give their opinion. They are also left wing - historically over represented in the polls while the right are underestimated. We saw in the recent Israeli election how polling companies can be miles out due to having big blind spots. I do believe there is an SNP advantage over labour and probably quite a substantial one - but not nearly as large as is being predicted. I've also noticed SNP support dipping a bit in YouGov subsamples the last few days. Tomorrow looks like that pattern may continue. Are the would be SNPers bottling out due to fear of a constitutional crisis or sending the panicked English running straight into the arms of the tory party? Time will tell!
"Nationalists are quite evangelical and eager to give their opinion."
But that's what recalled vote weighting is there to correct for. It shouldn't affect the headline numbers (except maybe in Ipsos-Mori polls).
"They are also left wing - historically over represented in the polls while the right are underestimated."
Up to a point, Lord Copper, and I'm not sure how that would affect the SNP-Labour battle anyway. (Labour may be very right-wing, but that isn't how its supporters see it). What happened in Israel is not unprecedented, but it's pretty unusual.
"I've also noticed SNP support dipping a bit in YouGov subsamples the last few days."
So have I, but then I also noticed it being unusually high the week before last. That may well just be meaningless sample variation, and I certainly can't see any sign of narrowing in any other firm's subsamples.
"Tomorrow looks like that pattern may continue."
That may be a hasty conclusion to jump to. Sometimes subtracting the Tory, Labour, UKIP, Lib Dem and Green percentages from 100 can give you a good idea of the SNP's GB-wide share, but not always. Rounding issues can produce all sorts of quirks.
Seems significant. A 'bad' reading for about 6 months for the SNP was a lead of anything less than 10%. Now its 2%, having travelled to that point over several days. If this lasts another couple of days I think we can call it a trend - although they can both still be expected to bounce around a fair bit.
I tend to give YG a bit more credence as they run a daily tracker poll and have a better record of accuracy than a lot of the other polling companies.
Nope, not significant at all - you're conveniently ignoring the larger than usual downweighting of SNP and Plaid identifiers (from 65 to 38). Without that the SNP lead would be enormous.
There have actually only been two subsamples recently that could really be classed as narrow SNP leads - to start hoping for a "trend" on that basis seems a tad optimistic.
Interesting. That appears to be an all time high in Yougov UK of people saying they 'identify most closely' with the SNP. Probably down to all the anti-Scottish stuff in the UK media of late.
Another Yougov poll taken partly over the same time frame as the last couple. SNP on a high of 5%. Massive poll too: 8271 respondents so much better than the couple of stnadard Yougov's we've just had.
You must realise that Mr Murphy is a legend in his own mind. Sad, really - he has delusions of adequacy, just like his boss in London.
ReplyDeleteYouGov sub-sample: SNP 37, Lab 31, Con 17, Others <6.
ReplyDeleteTowards the lower end of SNP scores and higher end of Labour scores.
There has been a trend towards slightly lower SNP scores over the last week or so. SNP haven't scored higher than 44 since 13th March. It is probably just variation around the long term average. There has been no sign of movement in the other polls, particularly the more reliable full-scale work.
The SNP lows seem to be getting less low, and the highs higher in terms of the range (since the method change).
DeleteThat's true. There hasn't been an SNP score lower than 37 for at least a month. Labour haven't been higher than 33 in the same period. That indicates that there is still a pretty big gap, because if the gap was as "low" as 10 points, you would expect some sub-samples to put Labour ahead of the SNP.
DeleteCompleting a Yougov scottish poll at the moment.
ReplyDeleteWestminster, Holyrood, what did I vote in September, independence question, what I've seen / heard about Scotland's position in the UK recently, whether I'm interested in that topic, how well I understand the role of UK vs Holyrood Governments in Scotland, question on powers / whether the matter is settled, national identity (Moreno), country of birth, how interested I am in rugby...
Anyone seen some of the SNPOUT nonsense on Twitter. Haha! They believe the polls are filled with 'Nats' and that come the 7th of May, the SNP will be lucky to hold on to their six seats.
ReplyDeleteWhat's even more odd is the demographics of this 'group'. Seem to mostly middle-aged, women, settlers.
I wonder if they think the huge surge in SNP membership is a mirage? Or even that it's isolated, and doesn't reflect a seismic shift in the electorate as a whole?
DeleteThe "polls are rigged by Nats" meme is not going to go away, because (like most myths) it is based on a small kernel of truth. Panelbase did have too many SNP supporters in their panel, it skewed some of their early results and they then recruited a new panel. Since then their results have been more in line with other companies, apart from the last poll where they used a question ladder.
DeleteBut for the other companies (and indeed for Panelbase as well since they sorted themselves out), it's nonsense. In fact, Anthony Wells of YouGov specifically debunked this idea on a BBC radio programme ("More or Less") a week before the referendum. He said that they had heard of this allegation that the online panels were stuffed full of nationalists, so they had gone back and checked the composition of their panel and if there had been unusual numbers of people signing up from particular places or backgrounds. Turned out the sign-up rate in Scotland in the year before the referendum was basically the same as the year before (and so on).
The funny thing too is that the online results should be (if anything) *more* accurate as a representation of the unionist / nationalist balance in Scotland now than they were before the referendum. Most of the online firms (Survation, Panelbase, YouGov, ICM) now use the referendum result as part of their political weighting in the polls. Everyone should be able to remember how they voted in the referendum, given its proximity and importance, whereas there was some concern before the referendum that political weighting was misleading due to false recall of how people voted in elections (Scottish and UK) 3-4 years previously.
Then you also have to throw into the mix that other polling methods (TNS face-to-face; Ashcroft and Ipsos Mori over the telephone) have also shown big scores for the SNP. Are these muppets seriously arguing not only that nationalists are more likely to sign up for online panels, but that they are also much more likely to answer the phone?
There is reason to believe that the actual result may be closer than current polls indicate (e.g. if "undecided" voters return to their previous VI, or if turnout is lower than expected), but it won't be because the whole polling industry is systematically over-rating the SNP.
James : When did Panelbase recruit a new panel? Was it a long time ago? I've no recollection of that, although I do recall them blocking new members from taking part in referendum polls (as a precaution against "infiltration"), and also using another company's panel for half of the sample in one of their polls to see if it made any difference.
DeleteNot long now, but the polls even could be under-estimating the SNP support. It means zilch in the long-term of things, but I know of 3 people. One, a Tory and the other two former Labour voters who are all voting SNP. The Tory is a tactical voter (he lives in Dundee West) so he'll hold his nose and vote SNP, the other two I think both voted No, but now voting SNP.
DeleteIf 80% of the 1.6 mil who voted Yes turn out then, well well, fireworks.
Thanks for clarifying James. I do remember them suspending new recruits - I think after the "rogue" independence poll in mid-2013? I must have mistaken the use of an additional panel for them recruiting a new one.
DeleteYou must realise that Mr Murphy is a legend in his own mind. Sad, really - he has delusions of adequacy, just like his boss in London.
ReplyDeleteComRes / ITV poll
ReplyDeletehttp://comres.co.uk/wp-content/uploads/2015/03/GMB_Undecided-Voters-Poll_23rd-March-2015.pdf
It's supposed to be of "undecided" voters at GB level, but it's a very large sample (12K+ across GB) which means there is a decent number in Scotland.
Respondents likely to vote in Scotland: 37% other (inc Greens, etc as well as SNP), 20% Lab, 15% Con, 5% LD, 5% UKIP. 18% refused / DK. Pro-rated gives 45% "other", 24% Lab, 18% Con, 6% LD, 6% UKIP.
Slightly higher % in Scotland have definitely decided their vote (76) than GB as a whole (71).
When pressed (i.e. if voting was a "legal requirement"), undecided break as follows: SNP 24%, Lab 17%, Con 14%. 27% still say DK.
I think the polls probably are over representing the SNP support. Nationalists are quite evangelical and eager to give their opinion. They are also left wing - historically over represented in the polls while the right are underestimated. We saw in the recent Israeli election how polling companies can be miles out due to having big blind spots. I do believe there is an SNP advantage over labour and probably quite a substantial one - but not nearly as large as is being predicted. I've also noticed SNP support dipping a bit in YouGov subsamples the last few days. Tomorrow looks like that pattern may continue. Are the would be SNPers bottling out due to fear of a constitutional crisis or sending the panicked English running straight into the arms of the tory party? Time will tell!
ReplyDelete"Nationalists are quite evangelical and eager to give their opinion."
DeleteBut that's what recalled vote weighting is there to correct for. It shouldn't affect the headline numbers (except maybe in Ipsos-Mori polls).
"They are also left wing - historically over represented in the polls while the right are underestimated."
Up to a point, Lord Copper, and I'm not sure how that would affect the SNP-Labour battle anyway. (Labour may be very right-wing, but that isn't how its supporters see it). What happened in Israel is not unprecedented, but it's pretty unusual.
"I've also noticed SNP support dipping a bit in YouGov subsamples the last few days."
So have I, but then I also noticed it being unusually high the week before last. That may well just be meaningless sample variation, and I certainly can't see any sign of narrowing in any other firm's subsamples.
"Tomorrow looks like that pattern may continue."
That may be a hasty conclusion to jump to. Sometimes subtracting the Tory, Labour, UKIP, Lib Dem and Green percentages from 100 can give you a good idea of the SNP's GB-wide share, but not always. Rounding issues can produce all sorts of quirks.
SNP 34, Labour 32.
DeleteSeems significant. A 'bad' reading for about 6 months for the SNP was a lead of anything less than 10%. Now its 2%, having travelled to that point over several days. If this lasts another couple of days I think we can call it a trend - although they can both still be expected to bounce around a fair bit.
I tend to give YG a bit more credence as they run a daily tracker poll and have a better record of accuracy than a lot of the other polling companies.
Nope, not significant at all - you're conveniently ignoring the larger than usual downweighting of SNP and Plaid identifiers (from 65 to 38). Without that the SNP lead would be enormous.
DeleteThere have actually only been two subsamples recently that could really be classed as narrow SNP leads - to start hoping for a "trend" on that basis seems a tad optimistic.
And there is a possibility that there was a genuine last-minute swing in Israel thanks to Netanyahu's loathsome race-baiting scaremongering.
ReplyDeleteComRes UK poll came out at midnight for the Mail.
ReplyDeleteTories 35 Labour 35 Libs 8 UKIP 10 Greens 7 Others 6.
Interesting. That appears to be an all time high in Yougov UK of people saying they 'identify most closely' with the SNP. Probably down to all the anti-Scottish stuff in the UK media of late.
ReplyDeleteAnother Yougov poll taken partly over the same time frame as the last couple. SNP on a high of 5%. Massive poll too: 8271 respondents so much better than the couple of stnadard Yougov's we've just had.
ReplyDeletehttps://d25d2506sfb94s.cloudfront.net/cumulus_uploads/document/wwqytvw1lq/YG-Archive-150325-TheTimes.pdf
Don't give Scottish breakdown but the 2010 transfers to SNP would give ~52% SNP in Scotland.
Will add to my running average. Kinda takes the edge off the straw clutching this morning on here.
Interestingly, the Times is reporting stuff from the Scottish sample of this poll even though it's not in the released tables.
DeleteDo you think the Labour party in Scotland?
56% Does not care about people like me
29% Does care about people like me