Will you vote for independence?

[butchersapron]

I think that support for yes is being underestimated in the polls.

OK. How?

 

[DairyQueen]

Mainly due to turnout and certainty to vote. I think the latest TNS poll was interesting. It suggested that Yes jumped six points if they looked at certainty to vote.

 

[butchersapron]

Mainly due to turnout and certainty to vote. I think the latest TNS poll was interesting. It suggested that Yes jumped six points if they looked at certainty to vote.

Explain that please and why it means you don't trust the polls.

 

[DairyQueen]

The first is turnout. Most polling organisations I have seen rely on weighting by how people have voted in the past. If there is large changes to the voting demographic, then it could be difficult to weight correctly. If it is closer to 80% than 70%, then that there would be more uncertainty around the estimates.

The certainty to vote is another issue, but polling companies do record this. The impact (shown in the data they collect), again tend to show a slight bias in the headline yes-no split.

Whether this amounts to large numbers of yes voters is unlikely, but I suspect (maybe it is hope) that these will work in favour of Yes.

 

[geminisnake]

It's only ever the losing side who question the polls

I'm questioning the polls because I know several hundred people, most of whom can vote and most of whom have stated more than once they are voting yes. I accept that I live in the NE of Scotland but afaik I know 5/6 no voters. I've been to various bits of the country in the last few weeks and still I'm not meeting these nos so how come the polls are as they are? Do all the no people live in one town? Or near danny??

 

[butchersapron]

The first is turnout. Most polling organisations I have seen rely on weighting by how people have voted in the past. If there is large changes to the voting demographic, then it will be very difficult to weight correctly. If it is closer to 80% than 70%, then that there will be a lot of uncertainty around the estimates.

The certainty to vote is another issue, but polling companies do record this. The impact of the data they collect, again tend to show a slight bias in the headline yes-no split.

Whether this amounts to large numbers of yes voters is probably unlikely, but I suspect (maybe it is hope) that these will work in favour of Yes.

They don't poll for turnout. They poll for not voting last time though.

They all rely on past voting intention and re-allocate on that basis. How is that a blow against polls full stop?

Certainty to vote is a post-research analysis - it can't effect how people decide to vote or how they tell the pollsters they will vote.

wtf is this?

 

[butchersapron]

I'm questioning the polls because I know several hundred people, most of whom can vote and most of whom have stated more than once they are voting yes. I accept that I live in the NE of Scotland but afaik I know 5/6 no voters. I've been to various bits of the country in the last few weeks and still I'm not meeting these nos so how come the polls are as they are? Do all the no people live in one town? Or near danny??

Or maybe they are accurate. As they so often have been in the past.

This is getting daft now.

 

[DairyQueen]

They don't poll for turnout. They poll for not voting last time though.

They all rely on past voting intention and re-allocate on that basis. How is that a blow against polls full stop?

Certainty to vote is a post-research analysis - it can't effect how people decide to vote or how they tell the pollsters they will vote.

wtf is this?

I never said that they poll for turnout. I never said that certainty to vote has an effect on how people decide to vote. Both have an effect on the difference between the numbers that the pollsters produce and the actual result.

 

[butchersapron]

I never said that they poll for turnout. I never said that certainty to vote has an effect on how people decide to vote. Both have an effect on the difference between the numbers that the pollsters produce and the actual result.

Seriously - i cannot follow what you are saying what your objections are. Each time you highlight a problem i deal with it, then i do the same with the next one. Then you say that you have objections. Can you put together a more coherent post about your problems with polls and YGs polling in particular?

 

[DairyQueen]

YouGov polls - I do not like their panel. That is their sampling population (i.e., they draw random people from their panel members to poll). I don't like that approach. That is, certain people put themselves forward for the panel and, perhaps, choose the information they include.

The broader question, could the polls for the referendum be wrong? I am saying they could be (although a very large difference is probably insurmountable). Certainty to vote comes into it, people might respond to the survey but then not vote. Then there is a problem with voter turnout. If voter turnout is a lot higher than the previous election (i.e., the core part of weighting), then there will be uncertainty (i.e., the polls will not be correctly weighted). Whether these result in gains for the yes vote is academic, and we would know after the event. It might be unfair to suggest pollsters should not weight as they are, but a much higher voter turnout will contribute to uncertainty since the weightings will not be correct.

EDIT: As I said before, I accept that my hope could be guiding me on concluding these factors have a positive impact for the yes vote (maybe no voters will be galvanised), but to deny they have an effect on the accuracy of polls in predicting the result is statistical illiteracy.

 

[butchersapron]

YouGov polls - I do not like their panel. That is their sampling population (i.e., they draw random people from their panel members to poll). I don't like that approach. That is, certain people put themselves forward for the panel and, perhaps, choose the information they include.

The broader question, could the polls for the referendum be wrong? I am saying they could be. Certainty to vote comes into it, people might respond to the survey but then not vote. Then there is a problem with voter turnout. If voter turnout is a lot higher than the previous election (i.e., the core part of weighting), then there will be uncertainty (i.e., the polls will not be correctly weighted). Whether these result in gains for the yes vote is academic, and we would know after the event. It might be unfair to suggest pollsters should not weight on this basis, but a much higher voter turnout will contribute to uncertainty since the weightings will not be correct.

EDIT: As I said before, I accept that my hope could be guiding me on concluding these factors have a positive impact for the yes vote (maybe no voters will be galvanised), but to deny they have an effect on the accuracy of polls in predicting the result is statistical illiteracy.

The political polling is not done by the commercial panels. Seriously if you're going to take them down you need to actually offer some reasons.

What about the rest of it? I don't care whether they could be wrong (they rarely are) - it's just a mess as regards polling isn't it? Turnout being a core part of weighting - not it isn't. What is is past voting behaviour.

 

[DairyQueen]

The political polling is not done by the commercial panels. Seriously if you're going to take them down you need to actually offer some reasons.

What about the rest of it? I don't care whether they could be wrong (they rarely are) - it's just a mess as regards polling isn't it? Turnout being a core part of weighting - not it isn't. What is is past voting behaviour.

YouGov panel - https://yougov.co.uk/opi/ - they don't use this?

The rest of it. The polling companies are rarely wrong in UK elections, largely because most elections are decided in a few swing seats. They were poor at predicting the proportion of votes which is what we are discussing, by the way, for the Scottish Parliament elections.

 

[butchersapron]

YouGov panel - https://yougov.co.uk/opi/ - they don't use this?

The rest of it. The polling companies are rarely wrong in UK elections, largely because most elections are decided in a few swing seats. They were poor at predicting the proportion of votes which is what we are discussing, by the way, for the Scottish Parliament elections.

Hang on so they're not wrong, but you don't like YG because it's wrong?

Use what?

 

[DairyQueen]

Hang on so they're not wrong, but you don't like YG because it's wrong?

Use what?

I don't understand what point you are making. Are you suggesting that it is easy for them to predict the results of an election if the number of people participating increases from the last one by some undefined amount (possibly as high as 20%) because that is all I am saying?

http://research.yougov.co.uk/services/panel-methodology/

Over the last ten years, YouGov has carefully recruited a panel of over 350,000 British adults to take part in our surveys. Panel members are recruited from a host of different sources, including standard advertising, and strategic partnerships with a broad range of websites.
When a new panel member is recruited, a host of socio-demographic information is recorded.

That sounds like the panel I posted a link to. That is their sampling population, its certainly useful (for a lot of polling purposes) but I don't like it in this case. I am a member of the link I posted.

 

[danny la rouge]

I'm a YouGov panel member. They always prepare the ground by asking how you'd vote if a general election were held tomorrow, and how you voted in the last general election.

I always answer truthfully, so I say I wouldn't and I didn't, respectively.

I assume that this means they suspect I wouldn't vote in the referendum (although I've told them I would, but they never answer the feedback they gather at the end of each poll).

I suppose that if my suspicion is right, and if they are underestimating the number of habitual non voters who will vote in the referendum (and RIC thinks it has found a lot who intend to vote Yes), then we could be looking at a similar effect to the last time polling companies got their methodology wrong, which was when they underestimated the Tory vote in 92. They all changed their methodologies after that (to the "how did you vote last time" weighting). So it is *possible* that this very method could lead them agley this time.

However, I don't actually think so; I think the polls are right. But that isn't to say that the numbers will stay the same for the next 5 weeks.

 

[DairyQueen]

I suppose that if my suspicion is right, and if they are underestimating the number of habitual non voters who will vote in the referendum (and RIC thinks it has found a lot who intend to vote Yes), then we could be looking at a similar effect to the last time polling companies got their methodology wrong, which was when they underestimated the Tory vote in 92. They all changed their methodologies after that (to the "how did you vote last time" weighting). So it is *possible* that this very method could lead them agley this time.

However, I don't actually think so; I think the polls are right. But that isn't to say that the numbers will stay the same for the next 5 weeks.

That's not how it works. You would have been participating in a Westminster poll. You would have been in a different sample. Habitual non-voters will be less likely to be on the panel in the first place, so won't be sampled, this will be true for many strata of society. That is why I don't like the methodology.

I think the press will ramp up the vitriol. I suspect we will see basically the debate (hopefully just in the press) boil down to whether or not you like Alex Salmond. So, a very British referendum.

 

[butchersapron]

I don't understand what point you are making. Are you suggesting that it is easy for them to predict the results of an election if the number of people participating increases from the last one by some undefined amount (possibly as high as 20%) because that is all I am saying?

http://research.yougov.co.uk/services/panel-methodology/



That sounds like the panel I posted a link to. That is their sampling population, its certainly useful (for a lot of polling purposes) but I don't like it in this case. I am a member of the link I posted.

I have no idea what the first paragraph means.

Yes, that's it (what link?) If they're using it for political polling now, then fine.

I'm saying that i don't even begin to get what your critricisms of polling full stop are. I've asked but i can't make head nor tail of them. Just lay out the problems you have with their methodology and why it gives skewed results - alongside true results please.

 

[danny la rouge]

That's not how it works. You would have been participating in a Westminster poll. You would have been in a different sample. Habitual non-voters will be less likely to be on the panel in the first place, so won't be sampled, this will be true for many strata of society. That is why I don't like the methodology.

I think the press will ramp up the vitriol. I suspect we will see basically the debate (hopefully just in the press) boil down to whether or not you like Alex Salmond. So, a very British referendum.

I have been polled on the referendum, and more than once.

 

[danny la rouge]

Oh, and Holyrood elections are also called general elections.

 

[butchersapron]

That's not how it works. You would have been participating in a Westminster poll. You would have been in a different sample. Habitual non-voters will be less likely to be on the panel in the first place, so won't be sampled, this will be true for many strata of society. That is why I don't like the methodology.

I think the press will ramp up the vitriol. I suspect we will see basically the debate (hopefully just in the press) boil down to whether or not you like Alex Salmond. So, a very British referendum.

So the problem is that non-voters, the most apathetic part of the electorate aren't counted (this applies to all pollsters). And they might suddenly get enthused.

This is a misuse of the word methodology. It's also plain stupid. Polls register what they register. They don't register other potential outcomes. That's for you to do.

 

[butchersapron]

Dark days of the end of the YES to AV vote here i'm afraid.

 

[DairyQueen]

So the problem is that non-voters, the most apathetic part of the electorate aren't counted (this applies to all pollsters). And they might suddenly get enthused.

This is a misuse of the word methodology. It's also plain stupid. Polls register what they register. They don't register other potential outcomes. That's for you to do.

butchersapron, you simply do not understand sampling. The panel is not the same as their sample. The point about apathetic voters (including but not exclusive to habitual non-voters) are they will not be in the panel in the first place. What I am saying is that the panel will be unrepresentative, YouGov know this. That is why they will take designed samples from it, and then weight those results.

EDIT: There have been some polls done predicting the turnout. The predictions were high (between 70%-80%).

 

[DairyQueen]

Dark days of the end of the YES to AV vote here i'm afraid.

I still think the referendum has some positives for Scotland. I think it can be used to highlight bias. The need to get the Yes vote out and still convince people is there, regardless of the polls.

 

[butchersapron]

butchersapron, you simply do not understand sampling. The panel is not the same as their sample. The point about apathetic voters (including but not exclusive to habitual non-voters) are they will not be in the panel in the first place. What I am saying is that the panel will be unrepresentative, YouGov know this. That is why they will take designed samples from it, and then weight those results.

I didn't mention sampling. I didn't mention the panel. You did. I dealt with the non-voter thing - politically it's a deux ex machina. Your posts as to why polling and YG in particular are crazy and lacking real critical backing. Say why and how you object to the methodology they use. This is just i don't like the results waffle until you do so.

 

[butchersapron]

I still think the referendum has some positives for Scotland. I think it can be used to highlight bias. The need to get the Yes vote out and still convince people is there, regardless of the polls.

Fuck the bias of pollsters, the potential contact with people who felt they were abandoned - that's a result.

 

[DairyQueen]

I didn't mention sampling. I didn't mention the panel. You did. I dealt with the non-voter thing - politically it's a deux ex machina. Your posts as to why polling and YG in particular are crazy and lacking real critical backing. Say why and how you object to the methodology they use. This is just i don't like the results waffle until you do so.

You don't understand YouGov's methodology. The point I made about was that certain segments of the population will not be in YouGov's sampling frame, i.e. its panel. You suggest that's fine.

This is linked to voter turnout. If you cannot understand why an extraordinarily high turnout will make polls uncertain, I suggest you first read a book on sample design and then a book on political polling.

 

[butchersapron]

You don't understand YouGov's methodology. The point I made about was that certain segments of the population will not be in YouGov's sampling frame, i.e. its panel. You suggest they should not be.

This is linked to voter turnout. If you cannot understand why a very high turnout will make polls uncertain, I suggest you first read a book on sample design and then a book on political polling.

Oh no, non-voters don't appear. So what? Has this impacted their past predictions? Why would it now? Why in this referendum?

 

[DairyQueen]

Oh no, non-voters don't appear. So what? Has this impacted their past predictions? Why would it now? Why in this referendum?

This is linked to voter turnout.

Then there is a problem with voter turnout.

Fuck the bias of pollsters, the potential contact with people who felt they were abandoned - that's a result.

Seriously butchers, you are off-form here.

 

[butchersapron]

?

 

[DairyQueen]

I have tried explaining to you why the a higher voter turnout would make things harder to predict. Also, the point about the YouGov panel missing segments of the population is not confined to non-voters.