All your polling needs right here...

jmcc

Member
Nov 27, 2018
1,022
618
Did you ever think you’d see the day that SF would have 10x the support of Labour in a poll.
When the Stickies took over Labour, I knew it was destined to be a fringe party. On the other place, I kept saying that Labour was in trouble and that the decision to go into coalition with FG in 2011 was a massive strategic error. SF is about five years ahead of the curve in terms of where I thought it would be. Opinion polling is Astrology with a veneer of Numerology. (I work with large datasets of hundreds of millions of elements so I might be a bit biased when it comes to people prognosticating on the basis of samples of about 1K. The media analysis of these opinion polls, without understanding their serious limitations, often looks like children playing at being grownups. The Brexit and Trump threads on the other place were full of such people who were so convinced that Brexit wouldn't happen and Trump would never win because they did not understand how problematic and non-representative the opinion polls were even when it was explained with drool-proof electrons.) It is far more important to analyse why things happen and how they happen.

Apart from the clientelism model that dominated a lot of Irish politics for decades, there is another factor that has a major effect. That's the incompetent school teacher's Dail sabbatical. These people are not really good at being teachers so they throw themselves into a political career because they have the free time and opportunity to meet more people than the average person. The other teachers are often happy to see the back of them and kids may get better teachers as a result.

The ideal situation for a party would be that its representatives have a wide range of expertise and backgrounds that continue to reflect society and its electorate. The general problem at the time that Labour started to decline was that teachers were over-represented as a group in the Oireachtas with approximately 26% of seats. Society changed but the parties did not. Throw in the whole generational thing of family seats and parties are facing a kind of toxic trio of clientelism, non-representative TDs and political inbreeding. Left unchecked, it overwhelms parties.

Society has become (allegedly) better educated in the last forty years or so and this meant that the Dail, and political parties, were, due to a lack of term limits, gradually becoming less representative of their electorates. When that happens, there is a time where the number of new voters a party gains is overtaken by the number voters that it loses. It isn't just a pattern of election ebb and flow. It is also the point where any mediocrity of its representatives becomes absolutely lethal. (Lenihan, Cowen, Noonan, Varadkar and Harris are good examples of people out of their depth having catastrophic impacts.) There's no new thinking and too few new candidates. The party becomes dominated by aging careerists who cannot adapt and are more concerned with keeping their own snouts in the public trough rather than growing their party. The problem for parties in that kind of electoral death spiral is that they lose relevance and their ability to survive as major players. It even affects larger parties. Labour got demolished. FF is also in trouble. FG nearly got overtaken by the PDs (Provisional FF). The PDs had a chance to replace FG but blew it. If SF does not continue to evolve with its electorate, and with society in general, then it will suffer the same fate as Labour.
 
Last edited:

Seosamh

Member
Nov 29, 2018
10,422
7,162
Thought this was an interesting thread-
Everyone is aware of it but to say it laid out like this…

 

Shaadi

Member
Feb 16, 2019
2,299
2,548

I mentioned how Labour were able to get to 20% a long time ago. They managed to lose it every time.
 

ruserious

Member
Dec 4, 2018
5,138
4,881
New one out.

 

Napper

Member
Dec 4, 2018
354
144
👀👀
That's nearly into SF winning DBS territory.
That would be a mental result
the system would have absolutely and completely collapsed
michael Martyn couldn’t survive this in a real world
leo couldn’t survive this in a real world
the constant sf bashing while failing to put forward measures to deal with the pressing problems
health, housing, and education now joined by mica/pyrite are the pressing issues for people as they were the day of the last election

tha The shinners could be considered as serious contenders in that area, in a by-election has to be a serious concern
if they could win there, they will win anywhere
SF has a message, FG has an anti SF message, after 11 years in government and unwilling to argue on the basis of your record , easy for the shinners as they’ve no record to defend

I think FG will win the seat, but their candidate will damage the FG brand
FF transferring to FG in big num will damage FF, damage Martyn and damage o’Callaghan
FFG are two drowning men clinging to each other

here’s hoping ff implode like the dup
 
Last edited:

Shaadi

Member
Feb 16, 2019
2,299
2,548
To overcome FG's lead in that constituency for three by-election, I reckon FG would need to drop to around 20% and SF would need to be at around 35% nationally.
By-elections are different, turnouts are usually dire. If your candidate is deeply unpopular then there's a Gay Mitchell Presidential Election type factor where the electorate turn on a candidate in a way that doesn't remotely reflect the National party support trends.

Until we get some credible constituency polling or feedback then we are in the dark, well twilight, maybe we're even in the twilight zone...
 

ainm_eile^2

Member
Jul 30, 2020
208
191
Opinion polling is Astrology with a veneer of Numerology. (I work with large datasets of hundreds of millions of elements so I might be a bit biased when it comes to people prognosticating on the basis of samples of about 1K. The media analysis of these opinion polls, without understanding their serious limitations, often looks like children playing at being grownups.
1000 is the number of observations at which it becomes quite unlikely for the sample average to seriously differ from the population average. At that point you can be confident that random deviations from the population mean in one direction will be substantially cancelled out by random deviations from the mean in the other direction, so that adding everything up and taking the sample average is a decent estimate.
 

jmcc

Member
Nov 27, 2018
1,022
618
1000 is the number of observations at which it becomes quite unlikely for the sample average to seriously differ from the population average. At that point you can be confident that random deviations from the population mean in one direction will be substantially cancelled out by random deviations from the mean in the other direction, so that adding everything up and taking the sample average is a decent estimate.
What the media is trying to do is the predict the outcome of a very complex system based on extremely limited data. It may give an indication of support for the parties at a national level but the seats are contested at a constituency level. These small 1K samples, unless they are run at constituency level , do not accurately capture local activity.

With STV, even relatively small changes in support at a national level can have massive effects. The 2002 GE and FG is one of the best examples of this. FG had 54 seats before the GE and 31 afterwards. It lost approximately 42.6% of its seats. The change in support was only 5.4%. The reverse of that would be FF's gain in 2016. It had 21 seats before the GE but 44 after it. That's approximately a 52.3% seat gain but the the increase in support was only 6.9%. SF's gain was more impressive. It had 14 seats before and and 23 after. That's a gain of 9 seats or 64.3% on an increase of 3.9%. in support.

Using percentages can be misleading when a party is small but there are so many variables in a GE that national polls may only be a guideline to the situation that existed before the GE. The danger is in using these national opinion polls to guess the outcome of local elections without any data on the constituencies. A party's support in an GE, or national opinion poll, may not directly translate into seats won or lost. FG's 2002 result of 31 seats only gave it 18.8% of the seats on 22.5% of the vote. FF won 49.1% of the seats on 41.5% of the vote. (I think it is referred to as the "election bonus".)

The TL;DR version is this: opinion poll percentages do not equal seats.
 

Statsman

The nice one, or so it seemed.
Staff member
Moderator
Member
Nov 28, 2018
9,619
10,380
A quiet retirement home
What the media is trying to do is the predict the outcome of a very complex system based on extremely limited data. It may give an indication of support for the parties at a national level but the seats are contested at a constituency level. These small 1K samples, unless they are run at constituency level , do not accurately capture local activity.

With STV, even relatively small changes in support at a national level can have massive effects. The 2002 GE and FG is one of the best examples of this. FG had 54 seats before the GE and 31 afterwards. It lost approximately 42.6% of its seats. The change in support was only 5.4%. The reverse of that would be FF's gain in 2016. It had 21 seats before the GE but 44 after it. That's approximately a 52.3% seat gain but the the increase in support was only 6.9%. SF's gain was more impressive. It had 14 seats before and and 23 after. That's a gain of 9 seats or 64.3% on an increase of 3.9%. in support.

Using percentages can be misleading when a party is small but there are so many variables in a GE that national polls may only be a guideline to the situation that existed before the GE. The danger is in using these national opinion polls to guess the outcome of local elections without any data on the constituencies. A party's support in an GE, or national opinion poll, may not directly translate into seats won or lost. FG's 2002 result of 31 seats only gave it 18.8% of the seats on 22.5% of the vote. FF won 49.1% of the seats on 41.5% of the vote. (I think it is referred to as the "election bonus".)

The TL;DR version is this: opinion poll percentages do not equal seats.
The blinding obvious stated at length.
 

hollandia

Literally knows shit
Staff member
Moderator
Member
What the media is trying to do is the predict the outcome of a very complex system based on extremely limited data. It may give an indication of support for the parties at a national level but the seats are contested at a constituency level. These small 1K samples, unless they are run at constituency level , do not accurately capture local activity
This is precisely what the media are trying to do. With a properly built model, which takes into account relative party strength in a constituency combined with past performance, one can actually make a relatively accurate inference as to seat forecast.
With STV, even relatively small changes in support at a national level can have massive effects.
Correct, because often the last seat, and occasionally the last two seats, come down to a balance of votes without making the quota. So a relatively small change in percentage can make the difference between three seats and two, or two seats and one.
The change in support was only 5.4%. The reverse of that would be FF's gain in 2016. It had 21 seats before the GE but 44 after it. That's approximately a 52.3% seat gain but the the increase in support was only 6.9%
See above.
SF's gain was more impressive. It had 14 seats before and and 23 after. That's a gain of 9 seats or 64.3% on an increase of 3.9%. in support.
Also see above. Small changes are enough to move one party down the pecking order for the last seat and another one up.
Using percentages can be misleading when a party is small but there are so many variables in a GE that national polls may only be a guideline to the situation that existed before the GE.
Agreed, however reasonably good estimates can be made with sound analysis of previous outings (i.e. who votes and transfers were gained from) even if a party is new to a constituency. The caveat here is nothing other than an educated guess can be made for a brand new party.
The danger is in using these national opinion polls to guess the outcome of local elections without any data on the constituencies. A party's support in an GE, or national opinion poll, may not directly translate into seats won or lost. FG's 2002 result of 31 seats only gave it 18.8% of the seats on 22.5% of the vote. FF won 49.1% of the seats on 41.5% of the vote. (I think it is referred to as the "election bonus".)
Not necessarily true, as indicated above. RTE's exit poll was remarkably accurate as a matter of fact.
The TL;DR version is this: opinion poll percentages do not equal seats.
More accurately, opinion poll percentages can give accurate seat forecasts if the analysis is sound.
 

jmcc

Member
Nov 27, 2018
1,022
618
This is precisely what the media are trying to do. With a properly built model, which takes into account relative party strength in a constituency combined with past performance, one can actually make a relatively accurate inference as to seat forecast.
Marsh generally provides some good commentary on RTE. It would be nice if the media actually had properly built models.

Not necessarily true, as indicated above. RTE's exit poll was remarkably accurate as a matter of fact.
RTE's exit poll was trying to measure something that had already happened. The opinion polls generally don't crystalise until the last week of the campaign but they are still being used to "predict. The time over which some of the opinion poll samples are carried out seems to have extended in recent years. That can make them vulnerable to changing sentiments so that the start of the poll may not be in synch with the end of the poll if some event has happened during the polling period.

More accurately, opinion poll percentages can give accurate seat forecasts if the analysis is sound.
The methodology matters. It is why there is a major difference for FF %s between RedC and B&A.
 

hollandia

Literally knows shit
Staff member
Moderator
Member
Marsh generally provides some good commentary on RTE. It would be nice if the media actually had properly built models.

RTE's exit poll was trying to measure something that had already happened. The opinion polls generally don't crystalise until the last week of the campaign but they are still being used to "predict. The time over which some of the opinion poll samples are carried out seems to have extended in recent years. That can make them vulnerable to changing sentiments so that the start of the poll may not be in synch with the end of the poll if some event has happened during the polling period.

The methodology matters. It is why there is a major difference for FF %s between RedC and B&A.
A rolling average is generally enough to iron out methodology differences.
My point about the exit poll stands, insofar as seat predictions were accurate. That means that if an opinion poll is accurate, seat projections should be accurate, because the model was correct.
 

jmcc

Member
Nov 27, 2018
1,022
618
A rolling average is generally enough to iron out methodology differences.
If RedC is using its 40K panel rather than a random survey, merging it with other polls using random surveys and face to face polling will be problematic. I think that Ipsos/MRBI's last poll reverted to F2F polling. The difference in FF support across online and F2F polling is considerable.

My point about the exit poll stands, insofar as seat predictions were accurate. That means that if an opinion poll is accurate, seat projections should be accurate, because the model was correct.
With an opinion poll taken before the election, there is always the possibility that the person being surveyed could change their mind and vote. With the exit poll, unless the person is not truthful, they cannot change their vote.
 

Napper

Member
Dec 4, 2018
354
144
A rolling average is generally enough to iron out methodology differences.
My point about the exit poll stands, insofar as seat predictions were accurate. That means that if an opinion poll is accurate, seat projections should be accurate, because the model was correct.
What would ff expect to get on the pole ?
I seat in a four or five seater
certainly nothing in a three seater
 

hollandia

Literally knows shit
Staff member
Moderator
Member
What would ff expect to get on the pole ?
I seat in a four or five seater
certainly nothing in a three seater
Doesn't quite work like that. In areas where FF are strong, it might have a neglible effect. Where they're weaker it might be devastating. What FF's big problem is going to be is deciding how many candidates to run, and the gender quota may force them to run more candidates then they need. This can have the effect of unnecessarily splitting their vote.
 

hollandia

Literally knows shit
Staff member
Moderator
Member
With an opinion poll taken before the election, there is always the possibility that the person being surveyed could change their mind and vote.
Correct me if I'm wrong, but that is the entire point of opinion polling. To give a snapshot of opinion at a particular time.
 

jmcc

Member
Nov 27, 2018
1,022
618
Correct me if I'm wrong, but that is the entire point of opinion polling. To give a snapshot of opinion at a particular time.
It won't stop the media trying to use polls as predictions. The general phrasing is something along the lines of "if a GE was held tomorrow". If a survey is carried out over too long a period, then the "snapshot" element declines.
 

hollandia

Literally knows shit
Staff member
Moderator
Member
It won't stop the media trying to use polls as predictions. The general phrasing is something along the lines of "if a GE was held tomorrow". If a survey is carried out over too long a period, then the "snapshot" element declines.
Of course it won't. Predictions are the entire point of the exercise.
 
Nov 27, 2018
4,639
6,190
Correct me if I'm wrong, but that is the entire point of opinion polling. To give a snapshot of opinion at a particular time.
Opinion Polls would be somewhat useless if people didn’t change their opinions.

If that were the case, all you’d need to do would be a single poll per lifetime, and just add the views of people who subsequently came of voting age, and subtract the views of those who subseqently died.

It’d pretty much destroy the polling industry. Think of their children.
 

hollandia

Literally knows shit
Staff member
Moderator
Member
Opinion Polls would be somewhat useless if people didn’t change their opinions.

If that were the case, all you’d need to do would be a single poll per lifetime, and just add the views of people who subsequently came of voting age, and subtract the views of those who subseqently died.

It’d pretty much destroy the polling industry. Think of their children.
Yes. Or have the results of the next election remain in perpetuity.
 

hollandia

Literally knows shit
Staff member
Moderator
Member
An Phoblacht? When parties do it, there are other motives.
So the purpose of opinion polling is not necessarily the selling of papers and/or advertising. Especially given the recent controversy regarding fake polling, I'd venture that that's actually about number five or six in the list of reasons why opinion polling takes place.
 
Nov 27, 2018
4,639
6,190
Yes. Or have the results of the next election remain in perpetuity.
Well, they'd still have to remove the votes of those voters who passed away, but, yes. One last election, and then for every election after, only those newly eligible to vote could vote.

Essentially everyone only votes once in their lifetime. Which means, after the next (and final "full"). election, parties and politicians would only ever campaign to leaving cert students.

I could see a slight problem with this. Boris Johnson would love it, though.
 
Nov 27, 2018
4,639
6,190
Surely the main purpose of hypothetical voting preference polls are to see how party performance changes (or doesn't change) the electorate's opinion of those parties.

The UK polls are very curious in this regard, as it seems there is no disaster big enough to lessen Tory support. Which is bizarre.
 

publicrealm

Member
Nov 27, 2018
6,895
9,368
Surely the main purpose of hypothetical voting preference polls are to see how party performance changes (or doesn't change) the electorate's opinion of those parties.

The UK polls are very curious in this regard, as it seems there is no disaster big enough to lessen Tory support. Which is bizarre.
It's going to take evidence of widespread Tory bestiality to change public opinion - so it could take months to emerge..
 

hollandia

Literally knows shit
Staff member
Moderator
Member
Surely the main purpose of hypothetical voting preference polls are to see how party performance changes (or doesn't change) the electorate's opinion of those parties.

The UK polls are very curious in this regard, as it seems there is no disaster big enough to lessen Tory support. Which is bizarre.
I'd also add - in the case of private polling - candidate selection in terms of numbers, and sounding out policy changes.
 

Cruimh

Rhubarb fetishist and proud of it!
Staff member
Moderator
Member
Nov 28, 2018
16,391
10,573
Under the blue skies
www.xxx-rhubarb.com

hollandia

Literally knows shit
Staff member
Moderator
Member
Top Bottom