How the pollsters got it wrong



the secretary

Well-known member
Joined
Jan 29, 2013
Messages
5,953
I suppose a nationwide poll in the US is pretty pointless.
Candidate performance in each state is more important. State by state polls may have been more accurate.
 

GDPR

1
Joined
Jul 5, 2008
Messages
217,846
Fascinating, jmcc. Before we get too carried away, Trump got no landslide and there is no Trump Tendancy in the USA. The same voters who voted for him are just as likely not to vote at all next time.

Brilliant points here though, even if I admit Im ready to accept them because they chime with my gut.

(1) The people you need to reach dont fill in poll forms or chat with pollsters.

(2) That is largely to do with the fact that every form they ever filled in or formal feedback session they engaged in was either a waste of time or used to punish them in some way.
 

Mercurial

Moderator
Joined
Jun 4, 2009
Messages
86,861
Thus, Civis posits, while virtually all of the data-heavy sites like Nate Silver’s FiveThirtyEight and the New York Times’ Upshot forecast a narrow-but-solid Clinton victory by Election Day...
That's not really accurate, though. 538 gaveTrump a much better chance than NYT, and the eventual result wasn't far off their numbers.

iirc, they actually suggested polls were more accurate this time than in 2012, but nobody points to that, because it was a case of Obama winning by more than was expected, rather than one candidate losing instead of winning.
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,355
I suppose a nationwide poll in the US is pretty pointless.
Candidate performance in each state is more important. State by state polls may have been more accurate.
It has to do with the size of the population being polled. The lower the number of people being polled, the lower the potential accuracy. Thus when you have a few thousand (or even a few tens of thousands) in a survey trying to determine the outcome where the population is in the millions of voters, things could go badly wrong. The localised state by state polls are better because the population being surveyed is smaller. The single digit response rate on some polls as mentioned in the article should have immediately flagged poll results as being unreliable.
 
Last edited:

President Bartlet

Well-known member
Joined
Jun 17, 2006
Messages
9,044
They got it wrong because they believed that what people was telling them was the truth when it clearly wasn't
 

gleeful

Well-known member
Joined
Feb 7, 2016
Messages
7,520
Voter suppression won it for Trump. Exit polls show Hillary winning Florida by 4%. Yet when they counted the votes somehow 1 in every 10 Hillary voters accidentally voted Trump.

Move along. Nothing to see here.
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,355
Fascinating, jmcc. Before we get too carried away, Trump got no landslide and there is no Trump Tendancy in the USA. The same voters who voted for him are just as likely not to vote at all next time.
I'm not so sure about this. In January 2016, NPR mentioned the rise of the independent/unaffiliated vote. That's the floating vote emerging. So what may have just happened in the US is an election where the vote dynamic has switched from the old ping-pong dynamic between a Democrat and Republican candidate to one where the floating vote decides it is potentially larger and more strategically important than either the Democrat or Republican vote. This could have been an abnormal election because there were Democrats and Republicans shifting. Some Republicans were against Trump and some Democrats despised HRC. The party faithful may be hoping for a return to the status quo with Democrats supporting Democrats and Republicans supporting Republicans but it may not happen.
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,355
Voter suppression won it for Trump. Exit polls show Hillary winning Florida by 4%. Yet when they counted the votes somehow 1 in every 10 Hillary voters accidentally voted Trump.

Move along. Nothing to see here.
Keep telling yourself that. It is the same reason that the talking monkeys in the Irish media thought that Labour was going to get 15 seats and that FG was going to get over 60. Some of the Irish polls excluded those who had a likelihood to vote of under 80%. That takes a special kind of f*ckwittery to do as it limits the polls to the HBAPs, the hammers and the rest of the party activists and their families.
 

Half Nelson

Well-known member
Joined
Dec 12, 2009
Messages
21,415
A key point, which is being avoided, is that Obama helped lose the presidency for HC.

In fact, the election result is as damning an indictment of Obama and his eight years as you're likely to see.
 

gleeful

Well-known member
Joined
Feb 7, 2016
Messages
7,520
Keep telling yourself that. It is the same reason that the talking monkeys in the Irish media thought that Labour was going to get 15 seats and that FG was going to get over 60. Some of the Irish polls excluded those who had a likelihood to vote of under 80%. That takes a special kind of f*ckwittery to do as it limits the polls to the HBAPs, the hammers and the rest of the party activists and their families.
Theres a massive difference between polls of intention and exit polls. When the UN does election monitoring they watch exit polls. If 47% of people walking out of the booth say they just voted Hillary but the count says only 42% did - theres something fishy about the count.
 

Truth.ie

Well-known member
Joined
Oct 5, 2008
Messages
27,354
Motivation was key.
Trump voters were prepared to walk through hot coals to vote whereas Hillary voters were lacklustre.

Likewise in Brexit, I remember the BBC interviewing people at Glastonbury moaning about the result.
They were sitting in shyt, sniffing laughing grass for breakfast and wondering why they didnt win.
 

Half Nelson

Well-known member
Joined
Dec 12, 2009
Messages
21,415
A very illuminating article on what the pollsters missed in the US presidential election and how it influenced the views of Trump's chances:

https://www.wired.com/2016/11/pollsters-missed-bowling-alone-voters-handed-trump-presidency/

It definitely should start the alarm bells ringing with the Irish pollsters.
Opinion polls are supposedly an important tool for influencing elections.

It would be naive to think they aren't being controlled by the various vested interests.
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,355
If 47% of people walking out of the booth say they just voted Hillary but the count says only 42% did - theres something fishy about the count.
The most obvious reason is that they are not questioning enough voters and enough polling stations. Many voters may not provide information on their vote but with more voters questioned, the accuracy should increase. The other aspect is that if they are overwhelmingly questioning voters in strongly pro-HRC or pro-Trump areas, then they may get an inaccurate view of the overall results.
 

gerhard dengler

Well-known member
Joined
Feb 3, 2011
Messages
46,739
Voter suppression won it for Trump. Exit polls show Hillary winning Florida by 4%. Yet when they counted the votes somehow 1 in every 10 Hillary voters accidentally voted Trump.

Move along. Nothing to see here.
B/s.

In February 2016, Dr Helmut Norpoth predicted that if Trump ran against Clinton, he'd win.
Donald Trump has 97% chance of beating Hillary Clinton, claims Helmut Norpoth | Daily Mail Online

Helmut Norpoth in March 2016 predicted a Trump win.
Trump Near-Certain to Defeat Democrat in November According To Primary Model | The Huffington Post

Norpoth reiterated his prediction of Trump win in July 2016
My model shows Donald Trump has an 87 percent chance of beating Hillary Clinton | Newsday

Norpoth reiterated his prediction of Trump win in October 2016 and advised that pollsters predictions me f/a
Political science professor says pollsters have NO IDEA who will vote in US election | Daily Mail Online
 
D

Deleted member 48908

Theres a massive difference between polls of intention and exit polls. When the UN does election monitoring they watch exit polls. If 47% of people walking out of the booth say they just voted Hillary but the count says only 42% did - theres something fishy about the count.
As the article points out, it's down to those who actually take the time to respond to pollsters. Nobody wants to waste their time talking to them anymore. It may be that the Clinton supporters wanted to beat their chest and bang on about how they voted while voters for others weren't bothered/were too busy going to work/ wanted to get home to their families/ disliked the intrusion.
 

jmcc

Well-known member
Joined
Jun 12, 2004
Messages
42,355
Opinion polls are supposedly an important tool for influencing elections.

It would be naive to think they aren't being controlled by the various vested interests.
They also sell newspapers and get advertising for TV and radio. I think that Wikileaks even had some HRC campaign discussions on using polls in that manner.
 

lostexpectation

Well-known member
Joined
Jun 3, 2006
Messages
14,021
Website
dublinstreams.blogspot.com

sadcitizen

Well-known member
Joined
Sep 27, 2011
Messages
4,271
It's been obvious for months that there was something wrong with polls, made obvious for the exact same reason as in Brexit: there was a discrepancy in, not just predictions but total outcome, between anonymous non-live surveys, live surveys, open-source/data surveys, non open surveys, as well as a discrepancy between surveys performed by private companies and public institutions.

The nature of this wrongness was clear when looking at the differences in data between surveys by private companies and universities.

The most galling thing for me was when I was tracking early voting data in a few states (states that describe all early votes openly), and seeing a crystal-clear landslide for Trump in those states, but the media persisted on reporting "landslide for Clinton in early votes" until the end. It was blatant, black-and-white, flip opposite lies by every large media organisation.

It would be so so easy for polling companies to correct these issues - or at least flag them - and I don't know why they don't.
 


New Threads

Most Replies

Top