This post, and particularly the last paragraph, was not predicated on the YouGove poll predicting a hung parliament – I continue to have no faith in the integrity of that company. I started writing this yesterday based on my own feeling that we could be heading into hung parliament territory. I was however motivated to return to and update this draft by the YouGove poll.
I yesterday watched Michael Gove shouting (literally) about Jeremy Corbyn supporting the IRA and Hamas on The Daily Politics, looking like an agitated tomato in spectacles. Because the mainstream media and political class live in the same utterly unrepresentative bubble, they do not realise that the large majority of ordinary people do not share their detestation of the Palestinians.
Subsequently we had Theresa May spouting utter rubbish about Corbyn going “alone and naked into the negotiating chamber”. 99% of the actual negotiating is done by teams of civil servants. Neither May nor Corbyn would be alone, they would have the same civil servants. Plus Corbyn would of course have Keir Starmer QC.
May’s jibe was supposed to echo Aneurin Bevan but it failed entirely, as the possession or otherwise of a nuclear weapon is irrelevant to the EU negotiations. The entirely spurious “alone” was not in Bevan’s quote and I can find no rational explanation of what it was supposed to mean. It is hard to avoid the conclusion that the whole irrelevant jibe was designed just to set up the titter at the image of Jeremy Corbyn naked. This really has been the most appalling Tory campaign imaginable, aimed at nobody but the nastiest kind of UKIP voter.
For the BBC to lead all its news bulletins on Corbyn’s inability instantly to recall the figure on childcare costs was puerile bias. Anyone can forget a figure. Politics is not a memory test. The attempt to reduce it to such is of course made heinous by differential application. When Tories have the same, perfectly natural problem of instant recall – as when the Chancellor was £20 billion out on the cost of HS2 – it gets nothing like the media coverage given to Corbyn and Abbott.
On which point, my last posting was about the SNP’s excellent manifesto. It was perfectly possible to sit here in Edinburgh yesterday, paying a great deal of attention to the BBC, and have no idea whatsoever of the SNP manifesto’s actual contents. Equally mystifying was the Daily Politics’ attack line against the SNP. How dare they have policies for the UK when they cannot form a government at Westminster? Angus Robertson replied politely that these were the policies their MPs would advocate at Westminster, and potentially support the implementation of, depending on the electoral arithmetic. The BBC reporter flared at this and seemed outraged that the SNP have the temerity to stand for election at all. It was truly bizarre television.
We are seeing more truly bizarre television every day as the mainstream media are puzzled and disconcerted that the plebs are simply refusing to ignore their obviously correct preference for the Tory party, instead having this mad desire to think for themselves. The media remind me of the puzzled look on Ceaucescu’s face as the crowd started chanting against him. The utterly talentless Tory hack Anne McElvoy was on BBC Breakfast today oozing contempt for Corbyn and explaining why his forgetting a number on Radio 4 proved he could not govern. She appeared completely divorced from reality.
And finally, it is remarkable that the Mays’ appearance on the One programme last week was featured again and again on BBC Breakfast and even on Sky News the next morning, with BBC vox pops “showing how impressed the public were with her” and Tory commentators speaking about how lovely and ordinary she was. Last night Jeremy Corbyn was on the One Show, and by the starkest of contrasts I have found no coverage of it at all this morning.
As the polls continue to shift, there is one distinct possibility for the result of this General Election looming. The Tories might be the largest party but with no overall majority. In which case they would form either a formal or a de facto alliance with their friends in the Northern Irish unionist parties. This would either force the unionists to take ownership of hard Brexit and the consequent imposition of a hard border between Northern Ireland and the Republic, or force Theresa May to abandon hard Brexit and outrage her supporters. I suspect the former is more likely, and the consequences of unionist enabled hard Brexit for Northern Ireland would be immense.
An interesting addition to democracy could be the complete banning of political polls, for at least 1 month prior to an election. People don’t like the word banning, and start taking about free speech. However, these pollsters aren’t generally neutral egg-heads doing an Oxbridge Research thesis; they are hard-headed businessmen, often with links to the political parties. They can poll for the rest of the time, just have to desist for a month, to ensure there is no push-polling, or manipulation. I suspect pretty much all polls are a form of push-polling, albeit done somewhat subtly.
Naturally, some sort of inquiry into postal votes would be good too. Can’t really ban them, but counting them correctly would be a good start.
At the beginning, the Tories were said to be on 48%. This was so absurd, it was clear the pollsters, YouGove, were on the make. Perhaps PM Jeremy can turn his mighty eye to this subject, at a later date? He will have to first turn his mind to the traitors in his own party, who will have a mighty hard time deciding whether to take a cushy job, or carry on battling Corbyn. Genuinely, I think this will be hard for them, poor diddums. They appear a bit confused, so perhaps we should give them time to reflect on their behaviour.
Yesterday morning, whilst interviewing Caro0line Lucas, Victoria Derbyshire accused the Green Party of “undermining democracy” for not contesting some Tory safe seats, the unsaid ramification of that being that any party which does not contest every seat must be anti-democratic! Ludicrous.
Yes, Victoria has hosted a few dodgy moments this week including the one where the awful Dominic Raab declared that people using food banks weren’t really in poverty but just have a temporary cash flow problem. One can only wonder why Raab’s disgraceful comment was essentially buried by the media.
Easy to confuse the two, really.
Temporary cash-flow problem: can’t afford a Lamborghini this week
Poverty: Can’t afford food.
Quite.
http://ukpollingreport.co.uk/
Noon thursday. As far as I can see “YOUGOVS ANTHONY WELLS” linked above seems makes no mention either of Yougov’s latest Voting Intention Poll or the projection of parliamentary seats which, despite its wide margin of error, has the boldness of independent enquiry. Am I missing something? Craig Murray may be able to tell us what Yougov is up to
Originally opinion polls did not purport to predict how people would act at some future date. They predicted how a whole population would on a given day answer a given question put to a small sample. If a sample was purely random – every possible sample has an equal chance of being drawn – then probility theory could measure with amazing accuracy the probability that any given finding was the result of chance, given some null hypothesis. Even a small sample of truck drivers, if (say) it is ninety per cent male, allows us to reject with measurable confidence the hypothesis that the truck drivers divde fifty-fifty
But Pollsters to not use purely random samples, which are impossibly expensive to track down. They try to exclude unrepresentative samples in advance by an elaborate system of quotas and weighings. And since 2015 they increasingly try to predict what the population will do rather than what it will say. They modify the raw data with a prediction of Turnout. All polling samples since 2015 have been reweighted to fit a model in which many young people who say they will vote Labour wont actually turn out to vote. See that on 30 May Wells speak with obvious approval of a rival company, a day before he ignoring the truly noteworthy findings of his own:.
” However, right now the one really huge difference is turnout. Weighted with all its normal demographic and political weights, ICM would have shown a Tory lead of only 3 points – that was transformed into a lead of 11 points by the turnout model, which predicts how likely respondents are to vote based on the estimates of turnout by age and class at the last election (the change from 11 to 12 points was the reallocation of don’t knows). That’s a big change, but given the errors in the polls in 2015 that may be necessary. On the other hand, if Jeremy Corbyn has managed to enthuse young people and there is a higher rate of turnout among younger voters than in 2015 then it risks understating Labour support. We shall find out next week…”
He is about to say “the only poll that counts….” What is going on in Yougov?
I am disgusted by the way the Tory boys and girls behave at the BBC . I pay .y TV licence and have to continually listen to them slating Jeremy Corbyn but they never slate Theresa May and her appallingly insensitive colleagues and as for Laura Kuenessberg words fail me on her totally biased opinions , another BBC hack who cannot hide her support for her beloved Tory party , hate is a strong word but thats what i feel at present for the biased BBC and the totally out of touch Tory party . God help us if we have to endure another 5 years of these clowns running our country .
As one who dabbles in data science, the new YouGov model appears to be something that makes a lot of sense. Unfortunately, YouGov haven’t released quite enough information to be sure if they’ve avoided some of the bigger pitfalls, especially with regards to pooling regional effects across constituency boundaries.
Anyway, let’s start with the good stuff:
1. They are using Stan, which can do full Bayesian inference and allow you to track uncertainties through the model. This is what allows them to extract a 95% confidence interval from their model in a meaningful way, rather than the usual method used by pollsters (proprietary method, usually pulled from rear end of the marketing department — just look at US elections when some had 95% chance of Hilary winning)
2. They actually display confidence intervals on a seat-by-seat basis. This is another sign they’re not just making up confidence intervals.
3. They seem to be giving results consistent with the picture on the ground. We’ve all heard of shy Tories, but there but I have witnessed first-hand the determined Labour supporters in high UKIP and Tory areas who really don’t want their neighbours to suspect their voting intention. They’re not ashamed of voting Labour, they’re just scared of the consequences of it being known that they are.
4. YouGov have a significant existential interest in this model being right as polling is really bad in the UK. They’ve clearly invested quite a bit of money in its development, far far beyond what is necessary to use as a psychological weapon to trick Labour into taking the rest of the week off. And yet the beauty of this is that, if correct, it will be like the LA Times poll was for the US election.
5. This model is being attacked from the left and the right. In fact, it’s largely being dismissed now. Even if it is not accurate, the fact that it operates with such low latency between samples and results being presented is unprecedented. I think a useful analogy is body fat scales. They physically cannot be accurate and you can’t take the percentage and compare it between scales. But you can see if the percentage on your scales is going up or down over time. Even if the scale is only broadly proportional, you can get an idea for whether you are gaining or losing body fat. Similar with these polls. We can’t know how accurate they are, but the direction of travel tells us something.
Now the bad stuff:
1. No source code. We have no way of checking their model against what they claim they’re doing. YouGov won’t release it because competitors would “steal” YouGov’s methodology. This is probably true. But it really won’t be that hard for competitors to come up with similar models if this does turn out to be correct. In staffing terms, you probably only need one or two decent computational statisticians or data scientists. Stan really does make it that easy.
2. The data source is YouGov’s online panel. This is a self-selected group of people who participate, in exchange for YouGov points to redeem for cash. According to several internet forums, when you get to about 95% of the amount for a payout, YouGov stop asking you as many questions. This is evidently in their interest but does introduce another area of bias. So this is a self-selected panel of people who either haven’t worked out they are unlikely to get paid or are just doing this as an act of philanthropy towards a multinational polling company.
3. In order to correct for YouGov’s self-selected people problem, they profile them and then either sample from the profile or weight the responses according to the likelihood of them having been drawn from the population. Depending on how YouGov have done this will give slightly different results on all but the largest sample sizes. Also, you are reliant on people having been honest with you at the outset.
4. If this polling does turn out to be accurate, then we are completely and utterly doomed. Just think of the mass manipulation possible with accurate low-latency opinion polling.
Yougov have dared to do something different and, unusually for a polling company, have chosen to do something that is quite close to the state of the art in data science and it actually looks like they found somebody with a clue. I’m still undecided on whether it would be better if this methodology was proved right or wrong.
Personally, I’d love to see the DUP lose seats to Sinn Fein at a time when the Tories need them most. For the tories to be locked out of office despite being the largest party due to the bridges they have burned and their toxic rebrand would be pure poetic justice. It looks so unlikely but 5 days is a long time in politics and Theresa May has yet to appear on the not-quite-100%-safe tailor-made red-white-and-blue question time.