Luck, not skill?
Given the worldwide interest in the referendum, and the number of polling organisations that were involved, I think it would be more surprising if there wasn't at least one organisation that "predicted" the correct result.
The EU referendum was a catastrophe for opinion pollsters. Remain’s official pollster, Populus, predicted a ten point margin of victory. Banks commissioned their own private polls, and also predicted victory, only to lose billions within hours as the strength of the Leave vote became clear. However one pollster stands out – …
Seems to me they also used a fair amount of knowledge to produce unbiased results. Recognising that their polling skewed the results due to education and social status seems to be the key factor here, whereas other polling companies have assumed that sentiment would be the same regardless.
"We polled 2000 degree educated people and came to result X"
"Seems to me they also used a fair amount of knowledge to produce unbiased results. Recognising that their polling skewed the results due to education and social status seems to be the key factor here, whereas other polling companies have assumed that sentiment would be the same regardless."
Unfortunately, what seems to you is false. Education and social status are definitely parameters in every pollster's model. I've written a short article about this last week, so looked into how these things are computed at a slightly greater depth than available in this article. Every polling company uses age, sex, education, social class, political party affiliation/support and newspaper readership (if applicable). There are other parameters used as well, but these are all standard.
True, but I thought the explanations given for why they got this one right while others didn't sounded reasonable. There is method to what pollsters do, they're not just guessing, and it stands to reason that some methods will work better than others.
The fact that most polls got the result wrong, rather than a roughly even split between right and wrong, suggests that there were common errors or biasing factors affecting most of those results, so it's certainly possible that the few who did get it right did so because they avoided those errors, rather than simply getting lucky. And I think the luck aspect was alluded to by Mr Taylor in the last paragraph anyway;
"It therefore needs to be caveated that there is no guarantee that this will also be the case in the next General Election!".
He's basically saying that no matter how good a job pollsters do, voter opinion could change after they sample their data; in this case it didn't, but next time it might well do. You pays your money...
"The fact that most polls got the result wrong, rather than a roughly even split between right and wrong, suggests that there were common errors or biasing factors affecting most of those results, so it's certainly possible that the few who did get it right did so because they avoided those errors, rather than simply getting lucky."
If you look into it in more detail, you find that Isos/ComRes's telephone polls were much less accurate than online polls like YouGov's, which more or less said it was within the margin of error. The telephone polls were included into poll of polls, and so this skewed things towards Remain. Excluding phone polling and it was 50/50 with Leave in the lead on several occasions. If you look at the two graphs on this page:
you'll see that the online polls were pretty accurate.
"If you look into it in more detail, you find that Isos/ComRes's telephone polls were much less accurate than online polls like YouGov's, which more or less said it was within the margin of error."
There was a report on the Beeb saying much the same thing. However we're told that the Leave vote majority was due to all those old folk who wouldn't recognise the internet if it bit them in the leg, etc. etc. That doesn't quite fit.
"There was a report on the Beeb saying much the same thing. However we're told that the Leave vote majority was due to all those old folk who wouldn't recognise the internet if it bit them in the leg, etc. etc. That doesn't quite fit."
It does, because the samples are weighted. Some old people do use the net, and those people are chosen in YouGov's online panel. If they cannot find enough oldies, they just double the significance of their response. (Roughly. In reality, YouGov use what's known as Mr P: Mulitlevel regression and poststratification.)
"voter opinion could change after they sample their data"
Or the way a question is worded.
A person might tell the pollster they feel the UK should remain with the EU, but intend to vote "leave" as a protest.
So the difference between asking, "Should we remain in the EU or leave?", or, "Are you voting leave or remain?", could give totally different results.
It would be interesting to see them do a follow up poll now the referendum is over to see exactly what the swing *after* the referendum has been given the lies that both sides have told have come to light. My guess would be a slight swing back to Remain, but probably not enough to change the result. The alternative is an elective that will believe anything even after proven to be false. Mind you, considering how many people believe in UFOs, the Loch Ness Monster, that politicians keep promises...
Indeed, but I personally would still be interested in a poll asking how people would vote if they could do it all over again. Here in the US, the media is full of stories about how most of the crazy Leave voters are already regretting their votes; I'm really curious to know if that's true or not.
Poll for Daily Mail, reported in Metro, was that 7% of Leave voters had changed their mind, as had 4% of Remain voters, which if translated into a vote at that time would have resulted in a Leave vote with the margin below 1 million votes.
@pdh I've seen a couple of polls looking at this, but I'm not sure where to lay my hand on them at the moment.
The first which I think was closer to the initial result was 7% of leave voters would have switched, and 4% of remain voters (not quite enough to switch the outcome). The second suggested 5% of leave and 2% of remain voters would switch (again not quite enough to swing things).
True, but it could influence HOW we exit. Although judging from comments from the EU any chance we had of remaining in the EEA as part of our exit may be slim to none. It would appear the EU has decided that the UK has thrown one petulant tantrum too many and has been relegated to the role of that friend you never invite to parties because they get drunk and start a fight.
Correct, but whether the result matters is a different question, since it was objectively pretty close and it's Parliament that decides what happens, not the popular vote. Parliament can certainly conclude that 37% of registered voters aren't enough to trigger Article 50, especially given the number of people who apparently want to recant now they've found out exactly what "Leave the EU" actually means. If you agree with that conclusion, I suggest emailing your MP right now.
"Parliament can certainly conclude that 37% of registered voters aren't enough to trigger Article 50"
I think they'd be on shaky ground if they did. A lot of MPs received less than 37% of the votes from registered voters in their constituencies.
Problem is, they were not guessing a number from one to ten. This was an either/or question and all but one polling org were on the wrong side of the final result. This suggests to me that most of the pollsters really wanted the Remains to win, and were quite happy to see their polling show this result.
The same thing happens in the U.S. We're repeatedly told that the Democrat candidate is way ahead, then the real poll often proves otherwise. It's not complicated; the idea is to dishearten the Republican voters so the other side has a better chance. Most of the polling is done by news orgs, which are known to be about 90% self-admitted Democrats.
As for a new Brexit vote, I predict that if it happens, it will be an even bigger Leave win, since the voters will be less likely to believe the biased polling they'll again be subjected to.
Obviously the secret sauce needs to stay secret, but you honestly can't tell me that the competition didn't factor for education, class, marital status etc., those are insanely obvious factors to anyone who ever spent two seconds stumbling over the phrase "representative sample". (Also, it should be pointed out that with estimates of 16% undecided and an undisclosed margin of error, nothing short of a landslide would have proved them wrong.)
It would be more interesting to get a post-mortem analysis from some of the other pollsters. For all I can tell, huge amounts of wishful thinking may have been involved: the curves that made their way into German media looked fairly close, no idea where the 10% lead was supposedly coming from.
I think there should be a complete ban on polls during the week where an event is held, whether this is a vote or a referendum. The problem is that polls and preliminary counts can heavily influence the outcome which I don't consider to be quite fair.
Please note than I'm not taking any sides with this opinion, I don't care if the polls go in favor or against, that's also totally not the point here. But I do think it would be a lot better if we don't get any "predictions" which more than often don't even turn out to be accurate at all.
Let people make up their own mind instead.
Let people make up their own mind instead.
That would work if people actually made up their own minds and not play "follow the leader". From where I sit, people are much like lemmings.
From reports, FB was filled with this type of thing where one person (usually followed by many others) makes a statement like "I'm voting to leave", or in the States: "I'm voting for Trump". Suddenly almost everyone is yelling "me too". Reminds of grade school in many ways.
I saw this happen a lot a work. The "leader" (and not necessarily the manager) would say "I'm voting...." and sure enough, all the followers would follow the lead lemming.
So you see people as lemmings? And that explains the popularity of the ideas and people you disdain, like "Leave" and "Trump?" What about those ideas and people you do like? Are all who agree with you calm and reasonable adults, and not at all lemming-like? I thought so.
Sure must be easy going thru life with nice simple answers for all those vexing "problems."
So you see people as lemmings? And that explains the popularity of the ideas and people you disdain, like "Leave" and "Trump?"
The 1958 Disney film White Wilderness won an Academy Award for Documentary Feature, in which staged footage was shown with lemmings jumping into certain death after faked scenes of mass migration.A Canadian Broadcasting Corporation documentary, Cruel Camera, found the lemmings used for White Wilderness were flown from Hudson Bay to Calgary, Alberta, Canada, where they did not jump off the cliff, but were in fact forced off the cliff by the camera crew. [Source: Wiki-bloody-pedia]
So yes, people are lemmings. They get it that if they don't toe the line, they are going to be forced to commit suicide by some jerk-off film crew, or killed even.
I can't say I disdain "Leave" or "Trump"... same for "Remain" or "Hillary".
If they weren't lemmings would they all still line up for "new shiny"? People want "secure" and will put up a lot for that. Then there's FB.. and the stock market.
But look around, social groups, politics, business. If someone's a "leader", there's an abundance of followers.
Maybe "lemmings" is too strong... maybe "yes men" would be more appropriate.
I have a BSc(hons), but am wise enough not to trust crony-capitalist (Fascist) or socialist (Communist) mainstream media, but rather sift news from independent news sites, including Zerohedge and Market Oracle, which I think are less likely to be biased, corrupted or lie by omission, but not known-corrupt FEAR distraction sites like prison planet. This is one reason why I visit this and other alternative media IT site for IT news rather than waste my time on boring months old news in mainstream junk like "Computing" and "Computer Weekly".
Even some people who mainly get their news from mainstream media are waking up to the worsening reality of what is actually around them and getting fed up with the establishment lies, thus Brexit, support for Trump and the rise of support for Nationalists the crony-capitalists (Fascists) and socialists (Communists) label as far right extremists.
A long time ago (1970s?), Mike Royko, a columnist for the Chicago Tribune, noted that polling was beginning to replace reporting on the candidates (like it has now).He suggested a way to counteract that: when a pollster calls, lie. I do from time to time, just to keep 'em guessing. If we can get to invalidating polls as a news item, somebody might actually get back to asking hard questions of the people trying for office. Well, it could happen.
Biting the hand that feeds IT © 1998–2019