MikeyWellFan Posted July 3, 2014 Share Posted July 3, 2014 This place will be hilarious if Yes wins. 0 Quote Link to comment Share on other sites More sharing options...
renton Posted July 3, 2014 Share Posted July 3, 2014 Wow. Impressive whinge. Not an actual rebuttal to Kellner's argument, indeed he accepts that Kellner may even be correct in his analysis that incorrect voter recall is only statistically significant for SNP voters. The sum total of this blogpost amounts to "BUT IF HE'S WRONG YOUGOV ARE PURE SHITE EH?" I think, the issue here is that YouGov has said that 'about half' of SNP support in 2011 came from what you might call 'passing nats' - in any sample YouGov has of this lot, there may only be a very few respondents who match this 'passing nat' description who will be upweighted dramatically to make up YouGov's 'passing nat' quota. Now, if said subsample is small, then it cannot be taken as statistically relevent or accurate, and that error is then mulitplied when the subsample is upweighted. That could lead to dramatic changes in the headline numbers. Why do it anyway? Why not simply rely on a single SNP group and asking them yes/No - as the other pollsters do. They all show a 15-18% No rate amongst the SNP on 2011 recall. Why suspect that as wrong? It feels like his assumption is that there must be more No voters, so let's go find them, akin to ICM and their search for shy Nos. Yet I can think of a few problems with Kellner's proposition, his assumption is that if he has a small subsample of 'passing nats' he needs to upweigh, then any error he finds is irrelevent since this group must all be closet Unionists lending a vote to the party of competent administration. we all know these people exist, like HB and Lex. However, Kellner doesn't seme to factor in that some folk voting SNP for the first time in 2011 did so, because between '07 and '11 they'd genuinely begun to believe independence was viable and desirable - as I did. no doubt there are other hues of the same colour, Kellner's binary assumption seems to be that the 2007 vote is the SNP's only true pro indy constituency and that any additional 2011 votes gained were merely unionists in favour of competent government. That seems to me to be the only way he could get away with multiplying up fractional subsamples without introducing an error - if he assumed said sub group was very uniform in it's outlook. Now, it may be that there is a big enough sub group in there such that any multiplied error would not be egregious. If so, publish the data tables showing the two SNP groups split - you'll note that they don't. Thus we can't interrogate this subsample for it's size or demographic outlook - if Kellner is uplifiting 45 55-65 year old women from the top two percentile, then no wonder the No lead is amplified. Not to stray from the basic point, fractional subsamples dramatically upweighted must introduce error. It's an issue all pollsters will run into from time to time, particularly in terms of age groups - ICM had previously produced massively varying headline figures base don trying to get more or less youth voters in to their polls. In this instance I see no valid reason why Kellner would introduce this further subsample. 0 Quote Link to comment Share on other sites More sharing options...
ICTChris Posted July 3, 2014 Share Posted July 3, 2014 I really doubt that a company like YouGov would deliberately skew the polling results. They are a business, based on producing accurate public polls, and if they were to get the referendum result significantly wrong it would be damaging to them as a company. Survation have responded here - http://survation.com/response-to-yesterdays-times-yougov-articles-and-yougovs-published-research-about-survations-scottish-independence-methodology/ 0 Quote Link to comment Share on other sites More sharing options...
Ad Lib Posted July 3, 2014 Share Posted July 3, 2014 Renton, the point is Survation, in failing accurately to tell the difference between SNP loyalists and floating voters via the recall, has already proved woefully wrong on actual current voting intention within Scotland in other polls, such as the European elections. They, specifically, over-estimated the levels of support of the SNP, suggesting that the people they have been getting through to identifying as 2011 SNP voters are likely to give an inaccurate reflection of broader voting intention. The reason for the up-weighting is because it restores the balance of voters so that it actually reflects a fair demographic of who voted what in 2011. This does not seem methodologically unreasonable. 0 Quote Link to comment Share on other sites More sharing options...
renton Posted July 3, 2014 Share Posted July 3, 2014 I really doubt that a company like YouGov would deliberately skew the polling results. They are a business, based on producing accurate public polls, and if they were to get the referendum result significantly wrong it would be damaging to them as a company. Survation have responded here - http://survation.com/response-to-yesterdays-times-yougov-articles-and-yougovs-published-research-about-survations-scottish-independence-methodology/ In so much as all pollsters end up upweighting or downweighting their raw data to fit a demographic profile tha tthey've decided is most accurate, then all pollsters are guilty of skewing their results. The results are necessarily based on the asusmptions of those measuring them. In this case, YouGov are intorducing a seperate subsmaple and upweighting them to meet a pre-conceived target. this must introduce an error, if said subsample requires a dramatic upweighting, based on not haivng enough people in that subsample to make it representative of the broader community. 0 Quote Link to comment Share on other sites More sharing options...
Blaven Posted July 3, 2014 Share Posted July 3, 2014 I really doubt that a company like YouGov would deliberately skew the polling results. They are a business, based on producing accurate public polls, and if they were to get the referendum result significantly wrong it would be damaging to them as a company. Survation have responded here - http://survation.com/response-to-yesterdays-times-yougov-articles-and-yougovs-published-research-about-survations-scottish-independence-methodology/ Just keep telling yourself that. You will be back in Kansas in no time. :lol::lol: 0 Quote Link to comment Share on other sites More sharing options...
renton Posted July 3, 2014 Share Posted July 3, 2014 Renton, the point is Survation, in failing accurately to tell the difference between SNP loyalists and floating voters via the recall, has already proved woefully wrong on actual current voting intention within Scotland in other polls, such as the European elections. They, specifically, over-estimated the levels of support of the SNP, suggesting that the people they have been getting through to identifying as 2011 SNP voters are likely to give an inaccurate reflection of broader voting intention. The reason for the up-weighting is because it restores the balance of voters so that it actually reflects a fair demographic of who voted what in 2011. This does not seem methodologically unreasonable. Polling is only accurate on larger turnouts. It's more than possible that the lower turnout contributed to the polling error seen. Just as a small polling set must introduce errors because they are not representative of the broader community, so a low turnout must do the same. As for the upweighting, it's not an insanely stupid adjustment to make, I think - but it's not smart either. Again, if they have a low number of these passing nats in their sample and upweight them, then the result cannot be said to be representative of a broader sample, and you'd be introducing another - different error to the one you say survation is introducing via it's 2011 recalled vote. as I said above, it's fine if you assume said group would be uniform in it's reasoning for switching from Labour to SNP and therefore the small number of folk in that subsmaple would not introduce an error (but why not Liberal to SNP switchers as well?) but that IS an unreasonable assumption to make. 0 Quote Link to comment Share on other sites More sharing options...
Ad Lib Posted July 3, 2014 Share Posted July 3, 2014 Polling is only accurate on larger turnouts. It's more than possible that the lower turnout contributed to the polling error seen. Just as a small polling set must introduce errors because they are not representative of the broader community, so a low turnout must do the same. As for the upweighting, it's not an insanely stupid adjustment to make, I think - but it's not smart either. Again, if they have a low number of these passing nats in their sample and upweight them, then the result cannot be said to be representative of a broader sample, and you'd be introducing another - different error to the one you say survation is introducing via it's 2011 recalled vote. as I said above, it's fine if you assume said group would be uniform in it's reasoning for switching from Labour to SNP and therefore the small number of folk in that subsmaple would not introduce an error (but why not Liberal to SNP switchers as well?) but that IS an unreasonable assumption to make. Most of the polling evidence I've seen suggests that there has been much less switching from Lib to SNP than Lib to Lab. 0 Quote Link to comment Share on other sites More sharing options...
H_B Posted July 3, 2014 Share Posted July 3, 2014 Some tremendous Arslikhan from Scottish Skier in the Comments below 0 Quote Link to comment Share on other sites More sharing options...
ICTChris Posted July 3, 2014 Share Posted July 3, 2014 In so much as all pollsters end up upweighting or downweighting their raw data to fit a demographic profile tha tthey've decided is most accurate, then all pollsters are guilty of skewing their results. The results are necessarily based on the asusmptions of those measuring them. In this case, YouGov are intorducing a seperate subsmaple and upweighting them to meet a pre-conceived target. this must introduce an error, if said subsample requires a dramatic upweighting, based on not haivng enough people in that subsample to make it representative of the broader community. I don't include weighting as 'skewing'. I meant I doubt that YouGov are trying to get certain results. Essentially there's a dispute about how to weight the results. I really don't think that YouGov are trying to reach certain results. We'll see in September who is right. I studied this a bit at University but in nowhere enough depth to have an informed opinion. I do think that there's a risk weighting on the 2011 results, you'd have to imagine that that is the SNPs high water mark. However, you do get elections that see one electoral 'reality' replaced with another - 1983 and 1997 spring to mind. So it could be that between 2007 and 2011 the picture changed in terms of support for independence. 0 Quote Link to comment Share on other sites More sharing options...
Confidemus Posted July 3, 2014 Share Posted July 3, 2014 Most of the polling evidence I've seen suggests that there has been much less switching from Lib to SNP than Lib to Lab. :lol: 0 Quote Link to comment Share on other sites More sharing options...
ICTChris Posted July 3, 2014 Share Posted July 3, 2014 (edited) I asked in the other thread but is there any polling on the voting intentions of English people in Scotland? ETA: I found one. http://www.panelbase.com/media/polls/F4108w11ScottishSundayTimestables.pdf Excluding Don't Knows that poll puts the split for Yes/No among English born voters 29%/71%, with Scottish born voters 51%/49% in favour. Edited July 3, 2014 by ICTChris 0 Quote Link to comment Share on other sites More sharing options...
renton Posted July 3, 2014 Share Posted July 3, 2014 I asked in the other thread but is there any polling on the voting intentions of English people in Scotland? ICM have definitely done a breakdown of this, think it was in the region 20-25% voting Yes. 0 Quote Link to comment Share on other sites More sharing options...
Ad Lib Posted July 3, 2014 Share Posted July 3, 2014 :lol: No really. More Liberal voters have been switching to Labour than to the he SNP, but it has happened at the same time as a lot of Labour voters went SNP. The effect was to protect Labour against an even bigger net swing to the SNP in 2011. 0 Quote Link to comment Share on other sites More sharing options...
Forza Ayr. Posted July 3, 2014 Share Posted July 3, 2014 Rule Britannia. -2 Quote Link to comment Share on other sites More sharing options...
doulikefish Posted July 3, 2014 Share Posted July 3, 2014 I really doubt that a company like YouGov would deliberately skew the polling results. They are a business, based on producing accurate public polls, and if they were to get the referendum result significantly wrong it would be damaging to them as a company. Survation have responded here - http://survation.com/response-to-yesterdays-times-yougov-articles-and-yougovs-published-research-about-survations-scottish-independence-methodology/ So good that they were 9% out in 2011 0 Quote Link to comment Share on other sites More sharing options...
I'm Brian Posted July 3, 2014 Share Posted July 3, 2014 Rule Britannia. Thoroughly insightful input into the debate. Welcome aboard. -1 Quote Link to comment Share on other sites More sharing options...
doulikefish Posted July 3, 2014 Share Posted July 3, 2014 Rule Britannia. Well done -1 Quote Link to comment Share on other sites More sharing options...
doulikefish Posted July 3, 2014 Share Posted July 3, 2014 wow im so upset i got red dotted on an internet forum 0 Quote Link to comment Share on other sites More sharing options...
I'm Brian Posted July 3, 2014 Share Posted July 3, 2014 wow im so upset i got red dotted on an internet forum Yeah me too. Have a green one to cheer you up -1 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.