News for hackers. Stuff that matters.

A comment on How (not) to forecast an election: Analysis via a hierarchical Bayesian model.

Re:How (not) to forecast an election: Analysis via a hierarc... by sgt101 on Sunday November 27, 2016 @11:28AM (#13049934)
"We have found that even a modestly sophisticated statistical model does very little to counter unreliable data"... so.. garbage in... garbage out?

I have an alternative explanation for this : people who buy newspapers and mainstream web sites are liberal. Pollsters and analysts who wish to make a living must get clicks and sell copy. The ones that create models that agree with the general zeitgeist of the liberal milieu get clicks and sell copy, other models are gradually ground to extinction.

Markets are not efficient mechanisms for creating insight.

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by dandersh on Sunday November 27, 2016 @11:57AM (#13050091)
Unfortunately your explanation is not correct. For example, it does not explain the extensive coverage of Donald Trump that was used to get clicks and sell copy, despite his violation of everything liberals hold dear...

Early in the campaign, Trump acknowledged his media advantage. “I’ve gotten so much free advertising, it’s like nothing I’d have expected,” he told the New York Times in September. “When you look at cable television, a lot of the programs are 100% Trump, so why would you need more Trump during the commercial breaks?”[1]

[1]http://www.marketwatch.com/story/trump-has-gotten-nearly-3-b...

http://www.politico.com/blogs/on-media/2016/02/les-moonves-t...

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by nilkn on Sunday November 27, 2016 @11:58AM (#13050100)
The coverage of Trump was entirely negative, though. It makes perfect sense that one could sell a magazine to liberals by attacking the main competitor to liberals in the Presidential race.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by dandersh on Sunday November 27, 2016 @12:10PM (#13050148)
And the coverage of Hillary wasn't?
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by kobeya on Sunday November 27, 2016 @12:52PM (#13050333)
The coverage of Hillary tended to criticize her actions (e.g. regarding emails). The coverage of Trump tended to criticize his character. Objectively speaking, one was more of an ad hominem attack, which reflects poorly on the media.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by icebraining on Sunday November 27, 2016 @01:48PM (#13050638)
Ad hominem is a fallacy when evaluating an argument. When you're evaluating a person (as a candidate), arguments are necessarily ad hominem, whether they're about temperament or conduct.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by kobeya on Sunday November 27, 2016 @02:55PM (#13051074)
As someone who thinks elections should be decoded by positions on policy not moral character or temperment, I disagree.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by icebraining on Sunday November 27, 2016 @04:53PM (#13051772)
You may disagree with the metric, but that doesn't make it a fallacy, just a difference of opinion.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by kobeya on Sunday November 27, 2016 @09:48PM (#13053065)
I didn't use the word fallacy. You did.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by rbritton on Sunday November 27, 2016 @12:42PM (#13050294)
Trying not to venture too far into a political discussion, I think the majority of the coverage of both was negative. Towards the last couple weeks, the quantity of coverage was very disproportionate depending on the source of the news [0]. People of the same political affiliation tend to gravitate towards the same news sources, and those sources appeared to be appealing to the affiliation of their viewers with the selection of information given.

The particular example I have provided is from Fox News because it was quick to search for, but I suspect you could find similar elsewhere about other news organizations.

[0]: https://twitter.com/FoxNews/status/787078152895901696/photo/...

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by oh_sigh on Sunday November 27, 2016 @03:09PM (#13051193)
Much less so than Trump. HRC also had much less coverage overall than Trump.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by eli_gottlieb on Sunday November 27, 2016 @02:36PM (#13050962)
It certainly was, but there was less of it.
[ Parent | on HN ]
[dead]
Re:How (not) to forecast an election: Analysis via a hierarc... by dominotw on Sunday November 27, 2016 @12:56PM (#13050354)
> We need to be elevating the Pied Piper candidates so that they are leaders of the pack and tell the press to them seriously[1]

Most important part of the sentence: 'and tell the press'

http://www.salon.com/2016/11/09/the-hillary-clinton-campaign...

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by the_cat_kittles on Sunday November 27, 2016 @02:23PM (#13050867)
"downvoted by bots already" i live on the same internet as you, how are you sure its bots downvoting you when im not? really destroys your credibility to be certain of something that you cant be.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by rguzman on Sunday November 27, 2016 @01:40PM (#13050588)
there is no such thing as bad publicity. at least early in the race, the battle is not against other candidates, but against going unnoticed.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by dominotw on Sunday November 27, 2016 @12:41PM (#13050290)
Wasn't it done on purpose to push 'pied piper candidates', so dems can have a slam dunk. Trump was identified as one of the 'pied piper candidates' in podesta emails. Even pravda ran story on it [1]. Those 'and tell the press' emails would make even Stalin blush.

1. http://www.salon.com/2016/11/09/the-hillary-clinton-campaign...

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by Clubber on Sunday November 27, 2016 @12:05PM (#13050131)
Fear sells better than anything. You would think the liberal papers would get more clicks talking about how Trump was winning. Also, if Hillary voters think the election is in the bag, more would stay home versus if they thought it was close.

Perhaps polling is just more or less accurate because fewer people take part in polls.

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by icebraining on Sunday November 27, 2016 @01:53PM (#13050673)
Perhaps polling is just more or less accurate because fewer people take part in polls.

Or the opposite - a voting abstentionist might be more likely to take a position when polled.

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by rjvir on Sunday November 27, 2016 @01:37PM (#13050577)
On the other hand, it often seems like news coverage is biased in the direction of a horse race. As in, the media is incentivized to make it seem like a close race with a lot of action, because that is engaging.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by the_cat_kittles on Sunday November 27, 2016 @02:19PM (#13050841)
that is exactly incorrect actually. a literal betting market has historically been the best predictor of outcomes. this cycle was particularly surprising, so everyone is all up in arms. but the markets that i follow (predictit.com) were more bullish on trump than almost all poll models. if you look at past performance they are excellent.
[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by sgt101 on Sunday November 27, 2016 @02:57PM (#13051092)
Do you have the historic data (run up to 8th Nov) for Predictit? I have this http://predictwise.com/politics/2016-president-winner

Predictwise was calling ~90% for Hillary and 320ish in the college.

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by froogle on Sunday November 27, 2016 @07:19PM (#13052513)
PredictWise is not a prediction market. It's an aggregate of them, with David Rothschild's own personal special sauce added on top - which, in this case, is why it was "wrong".

PredictWise's mistake was that it assumed polling errors would not be correlated[1]. (He actually had a fight with Nate Silver over this.)

Nate Silver had, in advance, stated that polling errors would be correlated[2] (rightfully, as it turned out), which is part of why 538's model turned out to be closer to reality than PredictWise's.

The actual betting markets, which PredictWise was pulling data from, were in the range of 20-30% Trump. Much more bullish.

[1] http://predictwise.com/blog/2016/09/poll-aggregation-fight/

[2] http://fivethirtyeight.com/features/election-update-why-our-...

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by 551199 on Sunday November 27, 2016 @12:55PM (#13050349)
Polls are mainly a way to influence people to vote certain way not show who people are actually voting for. Social proof.

This will be obvious if you look at the weighting and questions 'Would you vote for Obama or Trump'. Reuters/Ipsos and other outlets had to change the way they did the polls after Trump was leading.

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by dominotw on Sunday November 27, 2016 @12:38PM (#13050277)
Or you know they are simply polling people who fit their agenda

http://elections.huffingtonpost.com/2016/forecast/president

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by elsherbini on Sunday November 27, 2016 @12:53PM (#13050339)
The polls Huffington Post was using weren't conducted by them [1]. They were using the same polls everyone else had access to. Their model was probably worse than 538's though.

[1]http://elections.huffingtonpost.com/pollster/2016-general-el...

[ Parent | on HN ]
Re:How (not) to forecast an election: Analysis via a hierarc... by dominotw on Sunday November 27, 2016 @01:03PM (#13050396)
>Their model was probably worse than 538's though.

why would that be, though?

[ Parent | on HN ]