**** Elections thread *****

[OngBonga]

What you have to understand is that when there is a huge public debate raging, many people just quietly keep their thoughts to themselves and only express their opinion in the form of voting. I think such behaviour has a tendancy to skew the polls in favour of the candidate or option that the media favours, because those who oppose the media's idea of the "positive" option are less likely to engage in public discussion or polls.

This is what I think happened during our referendum. Those who wanted out were sick of being accused of racism, so just kept their thoughts to themself until the vote, making it seem that the "remain" camp were ahead. I see parallels when it comes to Trump.

[Poopadoop]

What you have to understand is that when there is a huge public debate raging, many people just quietly keep their thoughts to themselves and only express their opinion in the form of voting. I think such behaviour has a tendancy to skew the polls in favour of the candidate or option that the media favours, because those who oppose the media's idea of the "positive" option are less likely to engage in public discussion or polls.

I don't have any detailed information about the whole polling issue, but I assume people are being polled with anonymity. It doesn't guarantee they're going to give an honest answer, but it seems to remove (or at least reduce) any possible embarrassment about who or what a person says they're voting for.

On an aside, I think an issue was raised with the accuracy of polls in general after one of the polls last year in some election. The poll itself ended up being way off. Well, no statistician is going to argue a poll is 100% accurate, and there's always a chance any given poll will be way off because it's drawing a sample and is not an exhaustive measure of the entire population.

There may be issues with how samples are drawn in some cases and whether they're representative. One example I recall hearing about was a telephone poll conducted in some distant year when home telephones were still only owned by relatively well-off people. Not surprisingly the sample wasn't representative because it only included people who owned telephones, and the poll was miles off the mark. But I think sampling issues are much more widely controlled for these days and so without knowing too much about how these election polls are being done I'd hazard a guess that most of them are as accurate or nearly as accurate as they claim to be (i.e., fairly accurate but not precise).

This of course assumes the pollsters aren't setting them up to give biased results, either deliberately or out of ignorance. This probably does happen a fair bit. That's why the source of the poll is often your best guide as to its reliability; the more disinterested the pollster, the more likely the poll is to be giving an accurate estimate.

[Poopadoop]

If this is the USC polls you're referring to, this is certainly an unorthodox approach to polling.

http://cesrusc.org/election/

The 2016 USC Dornsife / LA Times Presidential Election Poll represents a pioneering approach to tracking changes in Americans' opinions throughout a campaign for the White House. Around 3000 respondents in our representative panel are asked questions on a regular basis on what they care about most in the election, and on their attitudes toward their preferred candidates. The "Daybreak poll" is updated just after midnight every day of the week.

Just from a quick scan of their methods, they seem to be doing something sensible, but in all likelihood they're overcomplicating their analysis and probably skewing their results. For example, they're asking people to give a probability estimate of whether they plan to vote and another probability estimate of who they're planning to vote for. The they combine this with demographic data to equate each person's responses to a certain number of votes for candidate X.

The first probability question is not unreasonable but combined with the latter amounts, mathematically, to counting die-hard supporters of either candidate as more than one voter. My intuition is that what this poll is good at showing is how many people have already voted in their minds.

Finally, they don't seem to have any empirical support to suggest that what they're doing is more valid than traditional polling techniques. It all seems like one big experiment based on someone's reasonably-well-thought-out-but-not-necessarily-true assumptions about how good people are at estimating their own behavior. Certainly people are going to be pretty accurate when they believe 'I'm 100% likely to vote and 100% likely to vote for X', but how accurate they are at estimating probabilities outside 100% is questionable.

Anyways it's an interesting idea.

[Poopadoop]

This method seems a lot more sensible, as it combines data from other polls. It also has a proven track record of being very accurate in previous elections.

http://pollyvote.com/en/

Hillary up by 6 points.

Sorry Wuf.

[surviva316]

What you have to understand is that when there is a huge public debate raging, many people just quietly keep their thoughts to themselves and only express their opinion in the form of voting. I think such behaviour has a tendancy to skew the polls in favour of the candidate or option that the media favours, because those who oppose the media's idea of the "positive" option are less likely to engage in public discussion or polls.

This sounds like a groundbreaking argument for a dissertation/study, and something that wouldn't be all that difficult to verify with research. Maybe your armchair brilliance is unmatched by anyone in the field ... or maybe you're just coming up with slapshod theories for why the polls with your favored results are better than other polls. I don't know, just spitballing here.

Not sure why you're so obsessed with Brexit upset. The polling average was well within the margin of error, especially considering they were consistently reporting double digits for Undecided voters. Seems a little over-reactionary to completely blow up your idea of the value of polls over that one result.