Methodology. Stronger weighting to older respondents and those who voted in 2015
Political polling
- Thread starter brogdale
- Start date
Vintage Paw
dead stare and computer glare
Different companies use different weighting methods, and also sample demographics in different numbers.
Some weight more in favour of those who have historically been known to vote more consistently, which tends to be old people. Old people vote Tory by a considerable margin.
There are loads of things that go into weighting algorithms, including likelihood to vote, past voting record, age, etc. The best thing you can do is look at lots of polls, look at the poll of polls (which aggregates them all), and then disregard the whole lot to a large extent as you remember that FPTP renders polling percentages an almost impossible gauge of final seat outcome without doing further maths beyond most of us.
bemused
Well-Known Member
So given we use a constituency based do these polls actually reflect the real result?
ItWillNeverWork
Messy Crimbobs, fellow humans.
The FT have a poll of polls running. This one was updated today. If the current trend continues, and if the number of young people expected to vote is being underestimated, and if the number of old people expected to vote is being overestimated, and if there is a sudden surge of people switching from Tory to UKIP, and if there is a sudden surge of LibDem wankers switching to Labour, then maybe Labour will get ahead in the percentage game. Whether that would then translate into seats is another question.
Lots of ifs there.
Lots of ifs there.
bi0boy
Power User
It's bollocks that's what it is. They may as well give more weight to people who like to go for a stroll in the evenings, and less to those who can't successfully describe the pen or pencil typically used to mark ballot papers.
littlebabyjesus
one of Maxwell's demons
they attempt to model it on how likely various groups are to actually vote. If you don't at least try to do that, your poll really will be worthless. It's all based on past performance, which gives the unknown quantity - dynamics change, and different groups become disillusioned and don't turn out at different times.
bi0boy
Power User
It's just guessing. And it's not even based on past performance because it's a secret ballot so no one knows whether people who said they would vote in the last election actually did or who they voted for.
Just ask a properly selected sample who they are going to vote for if any, and leave it at that. I bet the trend from multiple such polls would be just as useful, of not more so, since each poll wouldn't have been differently affected by the application of a turnout model which is highly likely to be inaccurate.
Either Ch4 news or Newsnight has a major debate between the generations tonight, sorry don't know which.
free spirit
more tea vicar?
Latest Guardian ICM poll is out now from polling at the weekend, it's an example of how the figures can be skewed significantly by the weighting factors used.
The actual figures are 33% tory, 30% Labour, but the adjusted figures blow that difference up to 45% to 34%.
This is presumably because they've somehow managed to still get the 10/10 likely to vote under 25s at only 44%.
I suspect the difference in this figure to the survation figures is down to the wording of the question with the ICM poll prefacing the question with a more confusing and suggestive sentence than survation which is clearly about this election.
ICM
"QA. Some people have said they would not vote in a new General Election, while others have said they would vote. How certain is it that you would actually vote in a general election if it were to be held tomorrow? On a ten point scale where 10 means you would be absolutely certain to vote and 1 means you would be certain not to vote, how likely is it that you would cast your vote?"
Survation.
And how likely will you be to vote in an immediate General Election, on a scale of 1 to 10, where 10 means you would be absolutely certain to vote and 1 that you would be absolutely certain NOT to vote?
The actual figures are 33% tory, 30% Labour, but the adjusted figures blow that difference up to 45% to 34%.
This is presumably because they've somehow managed to still get the 10/10 likely to vote under 25s at only 44%.
I suspect the difference in this figure to the survation figures is down to the wording of the question with the ICM poll prefacing the question with a more confusing and suggestive sentence than survation which is clearly about this election.
ICM
"QA. Some people have said they would not vote in a new General Election, while others have said they would vote. How certain is it that you would actually vote in a general election if it were to be held tomorrow? On a ten point scale where 10 means you would be absolutely certain to vote and 1 means you would be certain not to vote, how likely is it that you would cast your vote?"
Survation.
And how likely will you be to vote in an immediate General Election, on a scale of 1 to 10, where 10 means you would be absolutely certain to vote and 1 that you would be absolutely certain NOT to vote?
free spirit
more tea vicar?
One other point in that poll is that it breaks the figures down into what's happening in the marginals.
In the key marginals that ICM poll has Labour winning vs the Tories by 4-5% in both Labour held and Tory held marginals in the unweighted version, but losing in both sets of marginals in the weighted version.
Basically highlighting that it's all to play for and which party can get their vote out the most in the marginals of really going to be the crucial factor.
I can't help thinking that the party with the 600k membership ought to be able to mobilise their vote better on the day than the party with 100k or so membership.
In the key marginals that ICM poll has Labour winning vs the Tories by 4-5% in both Labour held and Tory held marginals in the unweighted version, but losing in both sets of marginals in the weighted version.
Basically highlighting that it's all to play for and which party can get their vote out the most in the marginals of really going to be the crucial factor.
I can't help thinking that the party with the 600k membership ought to be able to mobilise their vote better on the day than the party with 100k or so membership.
Vintage Paw
dead stare and computer glare
From Goodwin
That's interesting.
Labour are going to rely on a decent fraction of that UKIP vote to hold on to quite a few of its seats where in 2015 the Tory & UKIP vote combined was greater than the Labour vote.
Dogsauce
I’ve got Skol but I’m not a Scholar
They really need to target that UKIP vote with a 'On June the 8th, take back control - of your railways, NHS and public services' message. I vaguely recall from previous polling that rail nationalisation was more popular with UKIP voters than Labour ones.
I wonder if the UKIP redistribution is just seeing people going back to old allegiances, so labour gets a boost in safe labour seats and the Tories get back some of the blazer-wearing contingent in their safe southern strongholds.
I wonder if the UKIP redistribution is just seeing people going back to old allegiances, so labour gets a boost in safe labour seats and the Tories get back some of the blazer-wearing contingent in their safe southern strongholds.
I think they may well be doing this to an extent. Certainly graphs like the above suggest that Labour will be safer in seats such as Doncaster Central than has been suggested by some
Smokeandsteam
Working Class First
That does appear to be the case. And if it is the Tories will struggle to win the seats in the midlands and the north they need for a landslide. Their assumption was that the UKIP vote would come over to them as they owned 'Brexit'. It still might of course.
J Ed
Follow Back Pro Expropriation
Would be interesting to see the data on 2010 UKIP voters as well.
butchersapron
Bring back hanging
For those who don't like weighting.
butchersapron
Bring back hanging
Times have a first overall projection from their YouGov figures and have tories 16 short of majority - labour increase seats from 229 to 257. Tories 310 from 330.
It may but as Labour have basically accepted it, a lot of the Labour->UKIP voters could well see no problem with returning whence they came
Silly question: that one just uses the raw data from each, yes?
butchersapron
Bring back hanging
TheHoodedClaw
acknowledging ur soup leg
Here's the headline numbers from the Times/YouGov thing
Any idea what's the score with the NI numbers there?
Any idea what's the score with the NI numbers there?
redsquirrel
This Machine Kills Progressives
In a referendum about a single issue that was highly polarising. In contrast the turnout for the 2015 GE in Scotland was 71.1%, the 2016 Scottish Parliament elections 55.6% and the EU referendum turnout 67.2%.
brogdale
Coming to terms with late onset Anarchism
At this rate I'm assuming that the vermin are putting in considerable efforts to defend the 5k majority in Hastings & Rye.
Corax
Luke 5:16
So - assuming the cons will get in - is a minority govt a realistic hope?
If so, should I be happy about this? I need schooling...
brogdale
Coming to terms with late onset Anarchism
Green / SDLP?
TheHoodedClaw
acknowledging ur soup leg
Yeah, but add in the NI seats. Also does anyone have an inkling what the NI seat numbers mean here? Previously 8, now 18?
lazythursday
Well-Known Member
It was 18 last time. They've made a mistake.
Similar threads
