Polls and the media - real news or fake!?

[Pavilionaire]

What we know is that Tory will get at least 30% of the vote, Labour will get at least 25%, Lib Dem will get at least 7%, Greens and the Celtic parties will get at least 6%, so for two-thirds of the electorate it's a done deal.

The 12% that voted UKIP 2 years ago will vote something like 8% Tory, 4% Labour.

It's the remaining 20% of floating voters that will be key, but to get to parity with the Tories Labour will need two-thirds of them to vote with Corbyn. It is impossible to tell how they might vote, and that's about 9 million voters.

 

[jakarta]

I also wonder how true this 'shy Tory' adage is too. Are there really that many people who feel embarrassed to divulge their voting intention?

Only time I have ever been asked by a pollster I told him to f*ck off, how I vote is my affair and I'm certainly not going to tell a complete stranger about it.

 

[Mr Smggles]

Only time I have ever been asked by a pollster I told him to f*ck off, how I vote is my affair and I'm certainly not going to tell a complete stranger about it.

I'd like to think you didn't tell him to "**** off" and actually gave reasons behind you not wanting to give your decision.

 

[Weststander]

By Friday morning, we will know the truth about which polling companies and their various methods were nearer the mark and which got it so wrong.

The rival pollsters has been consistently wildly divergent. Yougov and another are the media darlings, because they show a collapse of all non-Labour voting intentions, giving the shock of a Corbyn government. Very news worthy. Whilst all the other polling figures barely get a mention.

The polls got it so significantly wrong in recent elections, but the obsession in following them has bizarrely intensified.


Against that, the betting industry on fresh bets even now, still offers long odds on Corbyn being the PM.


Who to believe?

 

[Scotchegg]

Polls vary so much because they all use different methodology and weighting to adjust the figures. For example percentage of expected young voter turnout Vs historical records etc. I think I read that the one poll with the 1 point difference (is it survation?) has the young voter turnout to be about 81% which is kind of... well, most other polls disagree. There's also two different types of modelling. After that it's all just spin I guess.

there's a really good blog all about polls here: https://marriott-stats.com/nigels-b...-5-steps-to-making-sense-of-the-latest-polls/

He explains quite a bit, though it's way heavy on the statistics.

 

[Machiavelli]

Polls vary so much because they all use different methodology and weighting to adjust the figures. For example percentage of expected young voter turnout Vs historical records etc. I think I read that the one poll with the 1 point difference (is it survation?) has the young voter turnout to be about 81% which is kind of... well, most other polls disagree. There's also two different types of modelling. After that it's all just spin I guess.

there's a really good blog all about polls here: https://marriott-stats.com/nigels-b...-5-steps-to-making-sense-of-the-latest-polls/

He explains quite a bit, though it's way heavy on the statistics.

I suspect the young vote/turnout will be quite substantially up on recent elections, but there might be some hat-eating going on if it gets as high as 81%.

 

[Weststander]

I suspect the young vote/turnout will be quite substantially up on recent elections, but there might be some hat-eating going on if it gets as high as 81%.

Deja vu on this very subject from 2010:

https://www.google.co.uk/amp/s/amp.theguardian.com/politics/2010/apr/22/surge-voter-registration-election-volatile

A mass registration of young voters in marginals.

 

[Eeyore]

When does the law allow the last poll to be published ?

 

[Machiavelli]

When does the law allow the last poll to be published ?

I think it's the day before the election, so one more day to go.

 

[seagulls4ever]

Saw an article on the Guardian recently that addressed this quite well. Yes, YouGov was set up by pro-Tory people, but the narrow gaps they are reporting are pretty easy to see are due to fairly transparent methodology reasons: they're predicting a higher than normal turn out from students, and students are leaning towards Labour; hence, their poll results prop Labour up. YouGov are, from what I've seen, one of the few polling companies that's fully trusting what the students are telling them. That's risky, but it's a valid approach.




They're all (except Survation, apparently) adjusting - independently - their methodology after they all ballsed it up in the 2015 GE and the Brexit polls. The key differences are in how they're treating the turn out prediction. Some are more trusting (eg YouGov) of a student body that's claiming they'll definitely vote, while others are ignoring what the students are saying and using the turn out figures from the last two major elections as their guide. The latter are producing a more favourable result for the Tories.

The exit poll got things right in 2015, but it's worth noting that an exit poll is always going to be more accurate because they don't need to make the turn out prediction. They're polling people who have actually voted, so as long as they poll representative samples they should be pretty accurate.




Yeah, I really don't think any of the polling companies are deliberately attempting to skew their results. I think it's far more plausible (and thus far more likely to also be true) that they're all trying to figure out how to adjust their methodology to deal with the failures from the 2015 GE and Brexit votes.

The key to who gets it right this time around will be in who has correctly predicted the turn out figures, and the demographic breakdown of the turn out. The Tory-leaning "grey vote" is pretty stable and easy to predict in terms of turn out: they'll go vote in high numbers, and will favour Tory. At the other end of the spectrum we've got the students, who are clearly favouring Labour but have historically not had the turn out rate.

There's rumblings out there that this election could see that trend for students to not vote get broken. If that happens, we could see the Tories lose their majority and YouGov will be justified. If it doesn't, YouGov will have egg on their faces.


Edit to add: YouGov's also being really, really brave and attempting to model the most likely result on a seat-by-seat basis. Will be really interesting to see how accurate it is, although their error margins are HUGE. Current prediction is a hung parliament with Tories on 304 seats, but the error range is anywhere between 265 and 342 Tory seats. I'd suggest that if they've got it wrong, it's most likely favouring Labour so I'd suggest the most likely variance is that the Tories end up closer to that 342 figure.

2nd Edit to add even more: There's also the complete unknown of how effective the "Tactical Voting" drive will be at this election. There's never previously been an organised effort anywhere near the scale of what we're seeing this time around. It's highly unlikely, but plausible, that Labour and Lib Dems could hoover up a bunch of marginal Tory-held seats on the back of strong tactical voting.

A good summary but YouGov aren't fully taking the youth at their word - they are down weighting which produces a turnout similar to the 2010 GE, whereas other pollsters are weighting according to the 2015 GE when turnout was lower.

https://yougov.co.uk/news/2017/06/01/pollsters-experimental-election/

 

[Audax]

Interesting coming back and re-reading this after seeing what actually happened.

Two things that sunk the Tory majority:

1. The students voted.
2. Tactical voting may have had an impact in some seats.

And one thing that saved them from an even worse result:

3. The SNP stuffed up (IndyRef2) and opened the door for 12 extra Tory seats north of the border.

Survation have to be the winners as far as advance polling is concerned, and would have been closely followed by YouGov if they hadn't bottled it at the last minute (haven't looked in detail, but seems they switched to 2015 turnout figures instead of 2010 for the students and thus weighted in favour of the Tories?)