• Our booking engine at tickets.railforums.co.uk (powered by TrainSplit) helps support the running of the forum with every ticket purchase! Find out more and ask any questions/give us feedback in this thread!

How did the polls get it so wrong?

Status
Not open for further replies.

TheNewNo2

Member
Joined
31 Mar 2015
Messages
1,008
Location
Canary Wharf
Leading up to the general election we just suffered, the polls were clear - neither Labour nor the Tories would be a majority, and while Tories would likely be a plurality their vote share was roughly the same as Labour. Judging by the current results, the Tories had a 6% advantage in the end result.

This isn't the first election which the polls have got significantly wrong recently. Nate Silver points to three other recent ones: the US 2012 presidential (Obama underestimated by 3 points), US 2014 Senate (Republicans underestimated by 4 points) and the Israeli general election (Likud support significantly underestimated).

While election polling will always be problematic, the fact that polls seem to uniformly say the same (wrong) thing is somewhat surprising, as each will have its own way of asking things and adjusting for the sample. So what's going on that all the polls are so wrong?
 
Sponsor Post - registered members do not see these adverts; click here to register, or click here to log in
R

RailUK Forums

pemma

Veteran Member
Joined
23 Jan 2009
Messages
31,474
Location
Knutsford
Polls are an estimate based on a representative sample. They'd be more accurate if we had proportional representation.

In 2010 the polls expected a big increase in votes for the Lib Dems - which happened. However, the increase in votes caused Lib Dems to come second in a lot of areas where the Conservatives won which didn't do anything to increase the number of seats they had.
 

crehld

Established Member
Joined
1 Nov 2014
Messages
1,994
Location
Norfolk
Polls are subject to a 95% confidence level and a 3 percentage point confidence interval.

In lay man's terms this means the result of the opinion poll is expected to be correct 95% of the time; there is a 5% chance it won't be. When correct it is still subject to a 3 percentage point margin of error.

The other problem with opinion polls is that our electoral system is not proportionate. So you have to model disproportional results based on a proportional sampling technique. One way to get around this would be to conduct opinion polls in every constituency, which would be much more accurate for a nation-wide picture but this isn't very practical from a pollster's point of view, and it is very resource intensive!
 

Simon11

Established Member
Joined
7 Nov 2010
Messages
1,335
or the fact that the people who voted for Labour in the polls, didn't actually bother to go in and vote.
 

TBY-Paul

Member
Joined
9 Feb 2013
Messages
327
The Majority of the Country don't want a Party from the extremes, Too far Left or Too far Right. The opinion polls were showing the possibility of the SNP having to much influence. So a lot of votes were switched to make sure that did'nt happen.

The SNP being expected to win so big in Scotland made a big differance to how the Middle ground voted in England.

Opinion Polls should be outlawed in the month leading up to an election, because they have such a negative influence on how people vote on the day,and PR introduced. It would remove tactical/negative voting.
 

Yew

Established Member
Joined
12 Mar 2011
Messages
6,539
Location
UK
Because we live in a country where less than 1/6th of the population counts as a 'majority'?
 

telstarbox

Established Member
Joined
23 Jul 2010
Messages
5,928
Location
Wennington Crossovers
Opinion Polls should be outlawed in the month leading up to an election, because they have such a negative influence on how people vote on the day,and PR introduced. It would remove tactical/negative voting.

Why should opinion polls be banned and what's the difference between a poll and asking round your friends or colleagues?
 

TBY-Paul

Member
Joined
9 Feb 2013
Messages
327
Some Countries already ban opinion polls because of the negative effect they have on the final vote. It's one thing having a straw poll amongst your mates. It's a completely different matter having opinion polls published on a daily basis. All these individual polls plus Poll of Polls etc, being constantly published have got to have had some influence, otherwise they wouldn't be so wrong.
 

TheNewNo2

Member
Joined
31 Mar 2015
Messages
1,008
Location
Canary Wharf
Opinion Polls should be outlawed in the month leading up to an election, because they have such a negative influence on how people vote on the day,and PR introduced. It would remove tactical/negative voting.

I based my tactical voting on the 2010 result.
 

bb21

Emeritus Moderator
Joined
4 Feb 2010
Messages
24,151
Is there any statistical evidence of the so-called "negative effect"?
 

crehld

Established Member
Joined
1 Nov 2014
Messages
1,994
Location
Norfolk
Opinion Polls should be outlawed in the month leading up to an election, because they have such a negative influence on how people vote on the day,and PR introduced. It would remove tactical/negative voting.

The removal of opinion polls would not stop tactical voting. It is a psychological effect inherent with the mechanical design of single member plurality electoral systems (ie first-past-the-post). If you want to remove tactical voting you need to remove first past the post (which isn't going to happen anytime soon)

--- old post above --- --- new post below ---
What do you mean by a 'negative effect?'

Some Countries already ban opinion polls because of the negative effect they have on the final vote. It's one thing having a straw poll amongst your mates. It's a completely different matter having opinion polls published on a daily basis. All these individual polls plus Poll of Polls etc, being constantly published have got to have had some influence, otherwise they wouldn't be so wrong.

Is there any statistical evidence of the so-called "negative effect"?

I believe the phrase we might be looking for is "potential to influence the result of an election", of which some academic research exists. But the question of whether this influence is "negative" is subjective for which no impartial empirical evidence (on either side of the debate) could ever be presented.
--- old post above --- --- new post below ---
The Majority of the Country don't want a Party from the extremes, Too far Left or Too far Right. The opinion polls were showing the possibility of the SNP having to much influence. So a lot of votes were switched to make sure that did'nt happen.

We don't know this, and neither do the opinion polls or the actual electoral result shed any light on the matter. That said expect to see some research conducted over the next few months about the motivations of voters.
 
Last edited:

TBY-Paul

Member
Joined
9 Feb 2013
Messages
327
The removal of opinion polls would not stop tactical voting. It is a psychological effect inherent with the mechanical design of single member plurality electoral systems (ie first-past-the-post). If you want to remove tactical voting you need to remove first past the post (which isn't going to happen anytime soon)

--- old post above --- --- new post below ---






I believe the phrase we might be looking for is "potential to influence the result of an election", of which some academic research exists. But the question of whether this influence is "negative" is subjective for which no impartial empirical evidence (on either side of the debate) could ever be presented.
--- old post above --- --- new post below ---


We don't know this, and neither do the opinion polls or the actual electoral result shed any light on the matter. That said expect to see some research conducted over the next few months about the motivations of voters.

I agree with everything you just said, only you put it better than me :oops:
 

TheKnightWho

Established Member
Joined
17 Oct 2012
Messages
3,184
Location
Oxford
The overall poll numbers have been bang on - 34%, 34%, 12% UKIP and 8% LibDem last I looked (may have changed since then), but the constituency polls were off.
 

St Rollox

Member
Joined
2 Jun 2013
Messages
650
The opinion polls didn't get it wrong in Scotland.
They said labour would be wiped out and guess what, they were wiped out.
 

Metrailway

Member
Joined
1 Jun 2011
Messages
575
Location
Birmingham/Coventry/London
The overall poll numbers have been bang on - 34%, 34%, 12% UKIP and 8% LibDem last I looked (may have changed since then), but the constituency polls were off.

No that is wrong - the overall polls underestimated the Tories and overstated Labour significantly.

The popular vote share is as follows:

CON - 36.9%
LAB - 30.5%
LIB - 7.8%
UKIP - 12.6%

I suspect it is like 1992 again - the massive shy Tory factor and people making up their minds on the polling day itself.
 

crehld

Established Member
Joined
1 Nov 2014
Messages
1,994
Location
Norfolk
The overall poll numbers have been bang on - 34%, 34%, 12% UKIP and 8% LibDem last I looked (may have changed since then), but the constituency polls were off.

Indeed, it's the translation from the overall proportion of votes cast to seats gained which is where opinion polls come apart. It's not the first time it's happened, nor will it be the last!
 

Xenophon PCDGS

Veteran Member
Joined
17 Apr 2011
Messages
32,270
Location
A semi-rural part of north-west England
I suspect it is like 1992 again - the massive shy Tory factor and people making up their minds on the polling day itself.

Call it what you like, that particular 1992 polls fiasco showed the fallibility of those who claim to be able to project matters forward in terms of possible results, whilst always making reference to the "uncertainty percentages" so applying....<(
 

crehld

Established Member
Joined
1 Nov 2014
Messages
1,994
Location
Norfolk
Call it what you like, that particular 1992 polls fiasco showed the fallibility of those who claim to be able to project matters forward in terms of possible results, whilst always making reference to the "uncertainty percentages" so applying....<(

The polling companies are quite willing to admit fallibility. The problem is with the media who demand the polls to be undertaken and have become far to accustomed to using them to predict results and using it as the basis to inform the election debate. Past experiences, and indeed the polling organisations themselves, tell us they can be unreliable, yet those who commission them insist on treating them like gospel truth.
 

Tetchytyke

Veteran Member
Joined
12 Sep 2013
Messages
13,305
Location
Isle of Man
Why should opinion polls be banned and what's the difference between a poll and asking round your friends or colleagues?

Opinion polls are already banned on the day of voting itself, because they are perceived to have a negative effect on voting behaviour. The exit polls cannot be published until voting has closed for the day, which is why you hear nothing until 10pm on election night.

I'd agree about banning the publishing of opinion polls for at least seven days prior to voting. If they're a negative influence on election day, they're also a negative influence immediately prior to election day.

As for the polls themselves, they were not that far out. The Tories got 2% more than expected, Labour 2% less, that's within the margin of error. The issue is that the these voting patterns were not uniform across the country. Labour's share of the vote rose in northern England, but as they were already secure there it didn't make the blindest bit of difference. Labour's share didn't rise in the marginals.

Trying to apply a national swing to a marginal constituency is a difficult task at the best of times. I think this election was especially difficult.

I think there are issues with "shy Tories"- Tories tend to be older, and tend to be less internet-savvy, so online polling (as YouGov predominantly use these days) doesn't show these people up. I also think Kellner made a good point in that some people in the marginals simply changed their mind on election day.
 

Xenophon PCDGS

Veteran Member
Joined
17 Apr 2011
Messages
32,270
Location
A semi-rural part of north-west England
The polling companies are quite willing to admit fallibility. The problem is with the media who demand the polls to be undertaken and have become far to accustomed to using them to predict results and using it as the basis to inform the election debate. Past experiences, and indeed the polling organisations themselves, tell us they can be unreliable, yet those who commission them insist on treating them like gospel truth.

In the knowledge of such matters stated by you above, why do people with the ability to think for themselves without the so-called need to make reference to these polls, take any heed of how the media wish to utilise them. Have the media brainwashed a seemingly large and ever-growing percentage of the electorate into submission by their never-ending emphasis of electoral matters in the news programmes at the cost of actual news gathering of matters affecting the world, such as the disaster in Nepal, etc. It takes the trivial matter of a baby being born to a member of the British royal family to give a day's respite to the incessant self-effacing reportage of electoral matters.

I mentioned upon another thread a few days ago of hearing a well-known so-called "media politico" say that he had never been so excited by a forthcoming General Election and I ventured the opinion that "he should get out more and get a life"...:roll:
 

Tetchytyke

Veteran Member
Joined
12 Sep 2013
Messages
13,305
Location
Isle of Man
I've seen this General Election described as "Eurovision for people who were picked last for games at school". It's probably fair comment.
 

edwin_m

Veteran Member
Joined
21 Apr 2013
Messages
24,793
Location
Nottingham
Tactical voting is a symptom of people making a negative decision to keep out the candidate/party they like least, rather than a positive one to vote in the one they like most. I think this is inherent in our electoral system.

The banning of opinion polls wouldn't help at all in my view. There would still be "Vote X - get Y" type comments by candidates, with the difference that voters would have less information to decide whether they were truthful.
 

pemma

Veteran Member
Joined
23 Jan 2009
Messages
31,474
Location
Knutsford
There would still be "Vote X - get Y" type comments by candidates, with the difference that voters would have less information to decide whether they were truthful.

Really no-one with a memory should have listened to the "Vote SNP and you'll get Ed Miliband" argument considering last time Cameron said "Vote Lib Dem and you'll get Gordon Brown" yet when some people did switch to Lib Dem they finished up with David Cameron when they would have been happier with a Labour-Lib Dem Coalition.
 
Last edited by a moderator:

Bertie the bus

Established Member
Joined
15 Aug 2014
Messages
2,781
Polls are an estimate based on a representative sample. They'd be more accurate if we had proportional representation.

Which is irrelevant to the original question which concerned vote share, not number of parliamentary seats.

I have no knowledge of the other results referred to but as far as this election is concerned I believe it was the polls themselves which caused the polls to be incorrect. The polls showed that Labour had no chance of forming a majority government and were very unlikely to even be the largest party. Therefore, for Labour to form a government they would have to have come to some arrangement with the SNP. My belief is that this prospect spooked enough English voters to vote Conservative when they possibly had no intention of doing so even a few days ago.
 

pemma

Veteran Member
Joined
23 Jan 2009
Messages
31,474
Location
Knutsford
I have no knowledge of the other results referred to but as far as this election is concerned I believe it was the polls themselves which caused the polls to be incorrect. The polls showed that Labour had no chance of forming a majority government and were very unlikely to even be the largest party. Therefore, for Labour to form a government they would have to have come to some arrangement with the SNP. My belief is that this prospect spooked enough English voters to vote Conservative when they possibly had no intention of doing so even a few days ago.

There was a suggestion on BBC News on Monday that the Lib Dems had in principal agreed to form a minority government in Coalition with Labour, subject to Labour and the Lib Dems getting more seats than the Conservatives, UKIP and DUP combined. It was also suggested the SNP would vote against a Conservative minority government but might back or abstain from voting if it was a Labour minority government.
 
Status
Not open for further replies.

Top