Got something to say or just want fewer pesky ads? Join us... 😊

[Help] Can we trust polls?

Can we trust polls?

  • Yes

    Votes: 10 15.4%
  • No

    Votes: 41 63.1%
  • Don't know

    Votes: 14 21.5%

  • Total voters
    65


Dick Head

⭐⭐⭐⭐⭐
Jan 3, 2010
13,612
Quaxxann
In 2015 polls suggested a hung parliament but we got a Conservative majority. Psephologists wrongly predicted the results of the US presidential election and the EU referendum and in 2017 they forecast a Conservative majority but we got a hung parliament. So can we trust polls any more or not?
 


TheJasperCo

Well-known member
Jan 20, 2012
4,590
Exeter
Polls.jpg

No. No we cannot.
 


Blue3

Well-known member
Jan 27, 2014
5,559
Lancing
In recent years polls have been less reliable as the influence of social media with true and fake news flooding our every waking moment, pollsters will I am sure develop new and better methodology but can they keep up with the social shift from multinationals pressure groups national governments only time will tell
 




Gwylan

Well-known member
Jul 5, 2003
31,324
Uffern
Psephologists wrongly predicted the results of the US presidential election and the EU referendum and in 2017 they forecast a Conservative majority but we got a hung parliament. So can we trust polls any more or not?

The polls for the US presidential election were absolutely spot on - they nearly all predicted a narrow win for Clinton and they were bang on. What foxed them was that the winner of the popular vote didn't win the election but it would have been very difficult to predict that.

The polls for the EU referendum were a bit mixed but all predicted a result of just one or two percent either way, within the margin of error.

The problem for pollsters is that they are working a few days in advance so even polls on the eve of an election are conducted a day or two. In these days of rolling news and social media things can change, what's key to look is the current trend.

We could see this in the last election where Tories were about 20 percent ahead at the start of the campaign and by the time of the last poll, the gap had narrowed considerably and a poll may well have predicted the result accurately.

The 2015 election was a real outlier though but polling companies have refined their methods since then.

I'll say what I did in the other thread: companies spend thousands and thousands of pounds on market research, do you really think they would spend that money if the research was useless?
 






Baker lite

Banned
Mar 16, 2017
6,309
in my house
They got it terribly wrong with The EU referendum, they seemed to be asking millennials and TammyHuff students, on the day of the race the working class silent majority got up and went and voted in their millions, much to the enragement of the Metropolitan elite.
 








Weststander

Well-known member
NSC Patreon
Aug 25, 2011
63,394
Withdean area
Polls in a first past the post system, as in UK parliamentary seats, worked pretty well when there were just two dominant parties and polarised views. Use of the predicted swing of vote, was straight forward.

But it’s a far more complicated affair to predict where there are third parties such as UKIP, taking votes.
 


Dick Head

⭐⭐⭐⭐⭐
Jan 3, 2010
13,612
Quaxxann
The polls for the US presidential election were absolutely spot on - they nearly all predicted a narrow win for Clinton and they were bang on. What foxed them was that the winner of the popular vote didn't win the election but it would have been very difficult to predict that.

The polls for the EU referendum were a bit mixed but all predicted a result of just one or two percent either way, within the margin of error.

The problem for pollsters is that they are working a few days in advance so even polls on the eve of an election are conducted a day or two. In these days of rolling news and social media things can change, what's key to look is the current trend.

We could see this in the last election where Tories were about 20 percent ahead at the start of the campaign and by the time of the last poll, the gap had narrowed considerably and a poll may well have predicted the result accurately.

The 2015 election was a real outlier though but polling companies have refined their methods since then.

I'll say what I did in the other thread: companies spend thousands and thousands of pounds on market research, do you really think they would spend that money if the research was useless?

I suppose they are broadly reliable. I know they have been weighted towards Labour, but what about so called "shy tories"? How many times do people complain that they've never been invited to take part in a poll? They'd probably hang up anyway. I'm prepared to give them another chance (pollsters), especially John Curtice. He's great!
 






Gwylan

Well-known member
Jul 5, 2003
31,324
Uffern
I suppose they are broadly reliable. I know they have been weighted towards Labour, but what about so called "shy tories"?

Except the last election which slightly underestimated the Labour vote. One of the reasons given for that is that polling companies adjusted their methods to take account of this and adjusted too far.

They're constantly refining their methods to improve their sampling otherwise they'd go out of business. Look at the history of the firms: Gallup was founded in the 30s, Harris in the 60s, Ipsos Mori was formed of two companies from the 60s and 70s - these companies wouldn't survive that long if they produced unreliable surveys.
 










Harry Wilson's tackle

Harry Wilson's Tackle
NSC Patreon
Oct 8, 2003
49,337
Faversham
Calling Harry Wilson

Oh goody. Another poll.

Any sample of the population is representative only if it is randomized and blinded. And adequately powered. That is all normally factored in but there can be mistakes. Also, human beings lie about their intentions ('shy tories'; 'labour voters' who don't bother to vote). Predictions are normally pretty good but when the electorate is evenly balanced the outcome, which is decisive in a first past the post election, may not be reflected by the poll (49:51 versus 52:48 - practically identical but the tiny difference may give one party an extra 30 seats and the election).

Yes, this thread is useless.
 


Eeyore

Lord Donkey of Queen's Park
NSC Patreon
Apr 5, 2014
23,380
My immediate work colleague is a Poll. No trust issues at all.
 






A1X

Well-known member
NSC Patreon
Sep 1, 2017
17,522
Deepest, darkest Sussex
It is worth remembering a few things about this. They got the 2015 election wrong, without question, but since then;

2016 referendum - Polls predicted a tight result with more in the final month favouring Leave than Remain
2016 US Presidential Election - Predicted Clinton would win the most votes, she did, it was the Electoral College which screwed that up
2017 General Election - Two of the polling organisations (Survation and Ipsos Mori) called the result near enough correctly, in fact I think Survation were pretty much bang on, and all organisations showed the polls narrowing

Polls are always adapting their methodologies and sample sizes, the 2015 one was a big miss but since then they haven't been as "bad" as they've been portrayed as
 



Paying the bills

Latest Discussions

Paying the bills

Paying the bills

Paying the bills

Albion and Premier League latest from Sky Sports


Top
Link Here