Waiting for the results

[RandBlade]

Or cite the fact that you don't like the arguments presented by Remain as a reason you're voting leave, despite having been involved in an intense running debate about the EU for over a decade, and being easily informed enough to make up your own mind?

Because I took it as an automatic, unquestioning truth that membership of the union was a good thing. It was only when we had this debate that stopping and taking stock to actually question what underpinned that - and drawing a blank - made it change. Hazir and I never debated EU "in or out" until the final few weeks, I was arguing for my version of the EU, he was arguing for his version. At the end of the day his version is what we have and my version isn't available. So goodbye.

[Flixy]

With respect, you can't let the fact that morons and dumbfucks are going to grab an excuse to act out to decide whether a piece of public policy is the right path for the country or not.

I'm sure that if Trump wins, there will be a rise in racist incidents in the US. That's not a reason to oppose Trump. His horrible policies and politics are. I expect there were pieces of "black power" crime after Obama was elected. Not a good reason to regret supporting him or to justify opposing him. To do otherwise is to. . . well to oppose Manchester United because of football hooligans identifying with the team.

..You are aware that football clubs are actually held responsible for the actions of their supporters with sanctions ranging from fines to matches without any supporters or even expulsion from tournaments? And that by law they are required to ban hooligans from their grounds and are required to protect the rival fans? That racism by fans in particular often leads to sanctions?

[Hazir]

Because I took it as an automatic, unquestioning truth that membership of the union was a good thing. It was only when we had this debate that stopping and taking stock to actually question what underpinned that - and drawing a blank - made it change. Hazir and I never debated EU "in or out" until the final few weeks, I was arguing for my version of the EU, he was arguing for his version. At the end of the day his version is what we have and my version isn't available. So goodbye.

Your way of looking at that debate is totally not the way I see it. It was very much about in and out. You at no point during the past decade have ever offered a positive idea of how to make the EU better, other than with some crazy ass idea that would have been tantamount to abolishing it.

[LittleFuzzy]

With even more respect, you can't make barely concealed dog-whistles about immigrants, or even outright racism, a central part of your campaign to enact a piece of policy, then act like it's nothing to do with you when hate crimes suddenly jump 50% immediately after your victory.

But that has absolutely nothing to do with Brexit as public policy. Condemn racist bigots who participated in the campaign all you want, but be aware that you're engaging in basic ad hominem fallacy if you try and use them to condemn the idea of Brexit. Or to target people who aren't like that and still voted for the policy.

[Loki]

Except the campaign pretty clearly focused on immigrants, particularly brown ones.

[LittleFuzzy]

Except the campaign pretty clearly focused on immigrants, particularly brown ones.

And since when did you start thinking election rhetoric was the same as actual policy, Loki? Or is that just something that happens when you decide you don't like the policy?

[Flixy]

I have to agree with Fuzzy, I'm fine with blaming the campaign leaders etc., but that doesn't mean that people who voted for a Brexit for other reasons should be lumped with racists. Of course I do think it's silly he tried to deny it's even happening when reports strongly suggests otherwise. But especially in a referendum with just two choices it's a bit unfair to lump everyone in each group together. As fun as blaming Randblade for British Nazism may be it's not that far from pulling a Godwin.

[RandBlade]

..You are aware that football clubs are actually held responsible for the actions of their supporters with sanctions ranging from fines to matches without any supporters or even expulsion from tournaments? And that by law they are required to ban hooligans from their grounds and are required to protect the rival fans? That racism by fans in particular often leads to sanctions?

Within the grounds yes, because the club has responsibility to control it's own ground. Outside the grounds though the club is not responsible for.

In the same way Vote Leave excluded Nigel Farage from all official capacities. He was treated as a hooligan excluded from the grounds. Doesn't stop him grabbing a loudhailer and making a noise outside but that can't be prevented.

[RandBlade]

Your way of looking at that debate is totally not the way I see it. It was very much about in and out. You at no point during the past decade have ever offered a positive idea of how to make the EU better, other than with some crazy ass idea that would have been tantamount to abolishing it.

I have argued for a meaningful implementation of the concept of subsidiarity which is Article 5 of the Treaty of the European Union so should not be a crazy ass idea. But despite being a key principle officially the reality is that subsidiarity hasn't occurred. If this had been taken seriously the EU would be infinitely better.

[Hazir]

I have argued for a meaningful implementation of the concept of subsidiarity which is Article 5 of the Treaty of the European Union so should not be a crazy ass idea. But despite being a key principle officially the reality is that subsidiarity hasn't occurred. If this had been taken seriously the EU would be infinitely better.

The report drafted by your own experts told the opposite of what you state here. The UK government experts concluded that the balance of powers resting in the EU and national government was 'about right'. Nicely reflected in the fact that Cameron merely negotiated a deal allowing the UK to cope better with some incompatibilities of its welfare state and freedom of movement.

[RandBlade]

A report drafted by a government that was laying the foundations for a remain vote found not much needed to be changed in order to justify a remain vote. Quelle surprise.

[Hazir]

A report drafted by a government that was laying the foundations for a remain vote found not much needed to be changed in order to justify a remain vote. Quelle surprise.

That's not the way your civil service works.

[RandBlade]

Well yes you're right, the way the civil service works is to take the terms of reference given by the government and then filter it through Sir Humphrey's received wisdom. Considering the civil service received wisdom was for there to be a Remain vote and the terms of reference were set such that a remain vote could be sought, of course the output came as it did.

Did you seriously think the civil service was going to come out with a report that made the EU out to have major flaws that Cameron couldn't fix in his renegotiation? How would that assist either Sir Humphrey or Cameron?

[Aimless]

400k racists is about 80 times the amount of BNP voters last year and is less than a third of the Leave side's winning margin. Remove the 400k Leave votes and Leave would have still won.

I took a leaf out of your own book and noted that 564k voters voted for the outright racist BNP in the 2010 election even though your electoral system actively discourages people from wasting votes on such parties. Where did those voters go? Even if 164k of those voters died between then and now, some of them would have been replaced by new racists. Moreover, research conducted since then indicates that at least 20% of UKIP voters are racists. In the latest general election, UKIP received over 3.88 million votes, likely an underestimation of their total support. 20% of that is around 776k votes. For the purposes of this discussion we can take that to be the absolute lower bound for the number of voting racists in the UK. What portion of those do you think voted for Remain?

[RandBlade]

As I've said in the past the 2010 BNP along with 2010 Lib Dem voters were 2015 UKIP voters. So would fall under the 776k total that you mentioned.

Even if we round it up to a million that is still less than Leave's winning margin.

[Loki]

https://yougov.co.uk/news/2015/04/20...ory-defenders/

A plurality of UKIP voters are ex-Tories.

Kind of sad to see the Greens cannibalize most of the Lib Dem vote.

[RandBlade]

If you trust YouGov further than you can throw a sumo wrestler which given their absolutely appalling track record in election after election is an interesting theorem.

Makes me really wonder how intelligent people can despite repeated failures people treat opinion pollsters as providing the gospel truth.

If YouGov were believable then not only would the UK be remaining in the EU but the referendum would never have happened as Prime Minister Miliband wouldn't have called the referendum in the first place.

[Loki]

This is pointless. I've had more rational debates with GGT.

[Aimless]

If you trust YouGov further than you can throw a sumo wrestler which given their absolutely appalling track record in election after election is an interesting theorem.

Makes me really wonder how intelligent people can despite repeated failures people treat opinion pollsters as providing the gospel truth.

If YouGov were believable then not only would the UK be remaining in the EU but the referendum would never have happened as Prime Minister Miliband wouldn't have called the referendum in the first place.

Not long ago, you trusted YouGov and I pointed out to you that their forecasts should be taken with a pinch of salt.

But that goes for any pollster, really. The question is how much salt you need. The image Loki provided is from an analysis I posted previously together with an analysis of another dataset, both of which indicated that a plurality of UKIP's supporters in 2015 supported Con in the preceding election. Similar results were obtained with a larger YouGov sample of 40k people.

Let's say for the sake of discussion that the findings are inaccurate. If that is the case, how inaccurate are they? How inaccurate would they have to be for that specific finding to be completely wrong?

YouGov got the latest general election wrong wrt Lab and Con, but they nailed (well, almost, for the purposes of this discussion) most of the other parties:



They were also better than most other pollsters:



It is likely that the sample--though it was very large--was unrepresentative (too many young people, too few old people, too many politically interested people) and that the models used to estimate turnout for different voters were flawed. Either way, I see no evidence supporting the belief that the sources of error--in this dataset or ipsos-mori's--are so great that we can completely dismiss the finding that a plurality of UKIP's 2015-supporters previously supported Con. Problems with voter recall is a relevant concern but the impact of that kind of error should be less significant when discussing support instead of actual voting.

I think the problems with the YouGov findings are such that you'd run into problems if you tried to do any real science with them eg. to develop future models for forecasting election outcomes, but they are not so great that they can't be used as the basis for an informal discussion. Same goes for similar findings based on other data.

[Aimless]

As I've said in the past the 2010 BNP along with 2010 Lib Dem voters were 2015 UKIP voters. So would fall under the 776k total that you mentioned.

Even if we round it up to a million that is still less than Leave's winning margin.

As I said before, that is the absolute minimum number of outright racists we should expect. I was responding to the claim that 17 million non-racists voted to Leave. Based on the foregoing discussion, the number should be lower than 17 million. If we exclude xenophobes the number should be even lower. Of course, even racists and xenophobes can have other reasons for wanting to Leave the EU, but I don't imagine those reasons would be strong enough to even make them vote had they not been racists and/or xenophobes.

[RandBlade]

This is pointless. I've had more rational debates with GGT.

Because I keep trumping your claims with facts? YouGov were systemically broken in 2015, the so-called "margin of error" on an opinion poll is 3% but in reality that is meant to be the margin of error for one poll if scientifically conducted with sound procedures. Technically speaking the margin of error is actually below 3% based on the figures they were reporting but lets be generous and say the full 3%

They got the Tory share wrong by 3.8% that is outside the margin of error!

However YouGov were like a broken record consistently providing the same figures. Scientifically what are the odds of a scientifically conducted poll with sound procedures getting an incorrect result outside the margin of error not just in the final poll but on a daily basis for months? It isn't possible, the figures were fundamentally broken, they were garbage. Anything based on subsamples of garbage is in itself garbage too.

[RandBlade]

Not long ago, you trusted YouGov and I pointed out to you that their forecasts should be taken with a pinch of salt.

All pollsters should be taken with a pinch of salt I agree. I don't recall saying I trusted YouGov, I have viewed ICM as the Gold Standard but even it has performed terribly in recent years. No pollster is worth more than chip wrapping paper currently. Even if I trusted YouGov in the past, when the facts change I change my mind. The fact is that YouGov have not performed to an adequate level to be taken seriously.

The image Loki provided is from an analysis I posted previously together with an analysis of another dataset, both of which indicated that a plurality of UKIP's supporters in 2015 supported Con in the preceding election. Similar results were obtained with a larger YouGov sample of 40k people.

Indeed from a sample that if it was scientifically accurate would mean Ed Miliband is Prime Minister today. Garbage In, Garbage Out.

How inaccurate would they have to be for that specific finding to be completely wrong?

I would say outside the margin of error is completely wrong!

YouGov got the latest general election wrong wrt Lab and Con, but they nailed (well, almost, for the purposes of this discussion) most of the other parties:



They were also better than most other pollsters:



It is likely that the sample--though it was very large--was unrepresentative (too many young people, too few old people, too many politically interested people) and that the models used to estimate turnout for different voters were flawed. Either way, I see no evidence supporting the belief that the sources of error--in this dataset or ipsos-mori's--are so great that we can completely dismiss the finding that a plurality of UKIP's 2015-supporters previously supported Con. Problems with voter recall is a relevant concern but the impact of that kind of error should be less significant when discussing support instead of actual voting.

I think the problems with the YouGov findings are such that you'd run into problems if you tried to do any real science with them eg. to develop future models for forecasting election outcomes, but they are not so great that they can't be used as the basis for an informal discussion. Same goes for similar findings based on other data.

Indeed YouGov did get the other parties right but Lab/Con wrong. Not just marginally wrong, but dramatically and outside the margin of error wrong.

That leaves three possible explanations that I can see:

1: The figures are complete hokum and should be ignored.
2: The figures have an element of truth but the transfer from the Tories to other parties was significantly smaller (or indeed negative) and the transfer from Labour was significantly greater.
3: There was a significant and direct Labour -> Tory swing not picked up by the polling.

Case 3 is extremely unlikely. Case 1 and 2 make your conclusions incorrect. Either way the conclusions are clearly unsound. Most likely is #2 and if the transfers from Tory to UKIP were considerably smaller than estimated (enough to reverse the 3.8% error YouGov polled) then clearly fixing that error would completely adjust the transfer figures.

[RandBlade]

To put another way let's apply Occam's Razor and apply the simplest solution to reconcile YouGov's catastrophic failure to measure 2015 anywhere approaching right. If we assume that they overestimated Tory->UKIP swingers and underestimated Labour->UKIP ones (the idea of a direct Labour-Tory swing goes completely against their own data).

In order to explain an outside margin of error 3.8% error in the Tory polling that doesn't mean reducing the subsample by 3.8% as it is 3.8% of the total not the subsample. 3.8 out of UKIP's 12% that YouGov were recording is 32%

If we correspondingly adjust the transfers taking out 32% of the Tory->UKIP swing (thus bringing the Tory total to what was actually voted for in the election) the share of the vote UKIP's vote that came from 2010 Tories falls down to just 7%! Which is much more in line with real world data.

[Loki]

That's not Occam's Razor. That's you making increasingly untenable assumptions in order to justify your hypothesis. And all the assumptions are intended to support your conclusion. That's what we call dishonesty.

[Aimless]

I would say outside the margin of error is completely wrong!
Indeed YouGov did get the other parties right but Lab/Con wrong. Not just marginally wrong, but dramatically and outside the margin of error wrong.

That leaves three possible explanations that I can see:

1: The figures are complete hokum and should be ignored.
2: The figures have an element of truth but the transfer from the Tories to other parties was significantly smaller (or indeed negative) and the transfer from Labour was significantly greater.
3: There was a significant and direct Labour -> Tory swing not picked up by the polling.

Case 3 is extremely unlikely. Case 1 and 2 make your conclusions incorrect. Either way the conclusions are clearly unsound. Most likely is #2 and if the transfers from Tory to UKIP were considerably smaller than estimated (enough to reverse the 3.8% error YouGov polled) then clearly fixing that error would completely adjust the transfer figures.

To put another way let's apply Occam's Razor and apply the simplest solution to reconcile YouGov's catastrophic failure to measure 2015 anywhere approaching right. If we assume that they overestimated Tory->UKIP swingers and underestimated Labour->UKIP ones (the idea of a direct Labour-Tory swing goes completely against their own data).

In order to explain an outside margin of error 3.8% error in the Tory polling that doesn't mean reducing the subsample by 3.8% as it is 3.8% of the total not the subsample. 3.8 out of UKIP's 12% that YouGov were recording is 32%

If we correspondingly adjust the transfers taking out 32% of the Tory->UKIP swing (thus bringing the Tory total to what was actually voted for in the election) the share of the vote UKIP's vote that came from 2010 Tories falls down to just 7%! Which is much more in line with real world data.

I'd be very interested to see the "real world data" you believe shows that only 7% of UKIP's supporters in 2015 previously supported Con.

I now know of three large samples conducted by different pollsters that have looked at voter-flows and/or supporter-flows between 2010 and 2015, and all of them point towards similar flows between Con and UKIP in addition to the flow from LD to other parties, with flows between Lab and Con cancelling each other out more or less. The "real world data" you provided in a previous discussion does not actually say anything about voter-flows between parties.

The demographics of YouGov's sample led to an overestimation of the Lab vote and underestimation of the Con vote. Their models also overestimated Lab turnout. Those two factors--demographics and turnout-models--explain almost all of the error in predicting the popular vote. That however is much less relevant for the question of support in popular opinion expressed as unadjusted numbers for voting intention.

[Aimless]

In addition to the above, reported estimates of the "margin of error" is less meaningful (verging on meaningless) when dealing with what we now know to be an unrepresentative non-random sample. If we insist on reporting that statistic, in light of the above, the margin of error (in retrospect) should have been higher for YouGov's polls. The margin of error also does not say anything about the error introduced by inaccurate models. It's merely a statistic about sampling error.

[Loki]

And of course the margin of error must be in Rand's direction. Because that's how reality works.

[RandBlade]

And of course the margin of error must be in Rand's direction. Because that's how reality works.

No because that's the facts. The fact is that 37.8% of Great Britain voted Tory not the 34% YouGov polled.

You are ignoring all real world facts and instead arguing with a religious fervour holding the mistaken belief that incorrect polls are evidence. It is like I'm arguing here with a Young Earth Creationist. Just as fossils and other scientific evidence shows the world to be older than 6000 years old, the constituency and aggregate voting evidence shows your claims to be fallacious. But just like a Creationist saying "it's true because the Bible says so" you just keep parroting "it's true because YouGov say so".

I don't believe in Creationism and I don't believe the findings of polls that have been disproven by facts. Try to be less religious and more scientific looking at the real world.

[RandBlade]

In addition to the above, reported estimates of the "margin of error" is less meaningful (verging on meaningless) when dealing with what we now know to be an unrepresentative non-random sample. If we insist on reporting that statistic, in light of the above, the margin of error (in retrospect) should have been higher for YouGov's polls. The margin of error also does not say anything about the error introduced by inaccurate models. It's merely a statistic about sampling error.

That's my point. Margin of error provides a confidence interval if the sample is representative, the sample was not representative.

Given that the sample was not representative, conclusions drawn from an unrepresentative sample are by definition also unrepresentative. Which is why they ate so much humble pie after 2015. Garbage in, garbage out.

How anyone can take something so unrepresentative that so fails to meet actual real life votes and then present it with religious fervour as the Gospel truth is beyond me. It fails at the first hurdle.

[RandBlade]

I'd be very interested to see the "real world data" you believe shows that only 7% of UKIP's supporters in 2015 previously supported Con.

I now know of three large samples conducted by different pollsters that have looked at voter-flows and/or supporter-flows between 2010 and 2015, and all of them point towards similar flows between Con and UKIP in addition to the flow from LD to other parties, with flows between Lab and Con cancelling each other out more or less. The "real world data" you provided in a previous discussion does not actually say anything about voter-flows between parties.

The demographics of YouGov's sample led to an overestimation of the Lab vote and underestimation of the Con vote. Their models also overestimated Lab turnout. Those two factors--demographics and turnout-models--explain almost all of the error in predicting the popular vote. That however is much less relevant for the question of support in popular opinion expressed as unadjusted numbers for voting intention.

I don't believe that 7% were (although I think that's much closer than 39%). Any model that produces Ed Miliband as Prime Minister is a broken model using broken assumptions or unrepresentative data. If you can demonstrate a model that actually lines up with how people DID vote in real life then it would be plausible (but not proof). Any outcome that produces a model significantly out of line with real life does not represent real life.