2016 Postmortem
Related: About this forumsmerconish LA TImes poll guest: predicted the obama 2012 victory within one point
they went into how the poll is constructed, with the same 3200 people being interviewed over and over, basing it on their enthusiasm fo1q2r a candidate on a 1-100 scale.
I know it's an outlier, obviously, but this stuff gave me pause...
I'm not concerned, of course, just scared
I think the same show had another poll analyst who said the margin of error, as compared to actual results, showed a gap of up to 100 percent
brexit, anyone?
is there any way in the world to feel comfortable about this until it's over?
will it be over?
saltpoint
(50,986 posts)campaign at this point, and the sand is slipping through the hourglass faster and faster.
I say 'the Trump campaign,' but at the same time there doesn't really seem to be a Trump campaign. I'm not getting reports of a convincing ground game for the GOP, there've been no newspaper endorsements, his polling is tanking by the hour, and down-ballot Republican candidates are fleeing for the exits.
It's a burning barn, and nobody can wrest the torch from Trump's tiny hands.
Ace Rothstein
(3,144 posts)It was the RAND poll. They used a similar methodology.
Gabi Hayes
(28,795 posts)anyway, I meant to say that the latest poll had them within a tenth of a point, or so, down precipitously
so that's a good thing
Agnosticsherbet
(11,619 posts)Gabi Hayes
(28,795 posts)here's the LA Times electoral map, at this point:
http://www.latimes.com/projects/2016-presidential-election-map/
RBInMaine
(13,570 posts)Adrahil
(13,340 posts)That the poll with a design vulnerable to sample errors and weird weighting can sometimes be very accurate... with the right sample. The problem with the LA Times poll is that it is VERY vulnerable to bad sample errors. So is it possible that the LA Times poll is right and everyone else is wrong? Sure. Is it likely? No. Keep in mind that the average polling error in 2012 was less than 3%. Even with a similar error this year, HRC would still win.
DLCWIdem
(1,580 posts)Individual.
Foggyhill
(1,060 posts)Don't know how the hell they got their initial sample, but it's so very bad
The fact such a large pool of people lead to huge variances from week to week even when nothing of worth occurs tells you there is something very wrong with methodology
Also, they're essentially ressampling the same pool in small batches and adjusting the weight of who they sample according to demo... meaning that if their sub sample is off, the effect is magnified
There is also the whole temporal, small sample and collating issues makes all rolling polls particularly volatile and un trustworthy
This poll is off the state poll, all polls but Rasmussen and upi
Gabi Hayes
(28,795 posts)2012 within a point. that was what I said, and why it's 'concernging'
in answer to another poster, this
http://www.realclearpolitics.com/articles/2016/10/14/why_pay_attention_to_the_la_times_poll.html
LA Times polled USED to be the RAND poll, according to the above
hate to link there, but it's in the body of the site. they were less than a point off in 2012, despite the wild daily fluctuations
read the story, along with its insane wingnut posturing, but they cite the stats from the poll
that's all I was trying to say here
Foggyhill
(1,060 posts)But if they claim prediction with their barely scientific method, they're full of it
If they had 10 elections that close then maybe there would be something there, but they don't have such a track record
Îm betting they're way way off and will have to stfu after
the amount of variance tells you there is an issue; that they are so dishonest to lie about makes them frauds
I got a long experience of stats and even polling and such bad methods make me mad
woolldog
(8,791 posts)Gabi Hayes
(28,795 posts)correct in the last election
this gave me pause, despite laughing at their outlier status during the majority of the race
they are trending even, pretty much, as of this morning, with Hillary within a tenth of a point, IIRC
woolldog
(8,791 posts)was their poll a consistent outlier like it is in this cycle? I suspect not.
I suggest you read the article. Or you can let yourself be consumed with worry over one poll. Makes no difference to me.
ColemanMaskell
(783 posts)If you are concerned about the polls, go poke around the 538 website (link below). They take into account all the reputable polls and come up with a percentage probability of each candidate winning. They update it every time new data is available. They have very scientific methodology and you can find explanations someplace on the web site.
http://fivethirtyeight.com/politics/
Most recent Election forecast always at http://projects.fivethirtyeight.com/2016-election-forecast/?ex_cid=rrpromo
But like they say -- If you are concerned about the election the best thing you can do, besides vote, is volunteer and go do some campaigning, and be sure everyone you know gets to the polls and votes.
geek tragedy
(68,868 posts)for one, they took a sample and weighted it by whom people reported as having voted for in 2012, which is a guarantee to get a bad sample as more people will say they voted for the winner
Charlotte Little
(658 posts)...stop, people, please. Stop!
Yes, the Allan Richtmans & the Nate Silvers and the "this/that/other polls" plus a whole lot of past voodoo predicted presidential winners (I think even a groundhog was credited one year as making the correct prediction )
BFD! This is NOT a normal election. Nothing about it has been decent or civil (and I mean on the Rethug side, not HRC), there is no precedent for it in the last century (especially given the modern invent of the internet & social media), and we have got to drop our clinging to anything or anyone as to how this is going to turn out.
Stay sane by only looking at Sam Wang's predictions on the Princeton Election Consortium. He's been the most rational, calm and intelligent member of the political polling world this entire election season.
Can Trump win. Yes.
So, stop fretting, GOYA & GOTV!
Oh, and one last thing. Believe it or not, this year's election is not the nastiest one in history. For you lovers of all things in the past, you might find this an interesting read: http://history1800s.about.com/od/leaders/a/electionof1828.htm
Gabi Hayes
(28,795 posts)OK.....great idea!
Charlotte Little
(658 posts)That's your reply?
Fine, sister. Shiver in your britches if that seems useful to you, but quite frankly, your concern trolling isn't helpful. You say you're working as a volunteer yet you have time to come on here to post (and reply over and over) with fear about the LA Times poll.
In the same vein, I guess we all should just curl up into a fetal position with our thumbs in our mouths since Allan Lichtman hasn't yet declared all 13 keys to White House are in HRC's hands? Trump wins in a landslide!!!
SMDH
Gabi Hayes
(28,795 posts)monster and his depraved family
all I did was report on what I heard on a fricking tv show.
thanks for your supportive response to information provided
enjoy your superior, yet apparently very defensive and threatened, worldview
Charlotte Little
(658 posts)...your stance.
Stop projecting. I'm not the one on DU in the last 3.5 weeks of one of the most important elections of our lifetime putting up a concern thread over what you heard on TV about one poll and its prediction history that clearly has you crapping in your panties.
And to be clear, I'm not on here to support YOU and your "information." I'm not here to provide you with fresh pair of under-drawers.
I'm on here to support Hillary Clinton, Democrats and GOTV.
Now please, head on over here: http://election.princeton.edu/ - to calm your nerves. I think you'll find it most helpful.
Cheerio!
woolldog
(8,791 posts)He or she isn't interested in any analysis or information. I posted a NY Times article that details the flaws with the current iteration of the poll and he or she will not read it. You're better off trashing the thread, which is what I'm doing.
Fast Walker 52
(7,723 posts)I really doubt Trump will win, but it's not a 100% lock for Hillary... sadly given how monstrous Trump is
stopbush
(24,393 posts)Especially in polling, where companies are usually behind the curve, using their old models to predict new events. Look how long it took for them to figure in cell phone users.
Gabi Hayes
(28,795 posts)do with the information, including their track record on the provided link
do with it what you will
seems like all the panty twisting is coming from responders who don't like the message, despite the numbers behind it.
I don't either, but I thought it might be of interest
Gabi Hayes
(28,795 posts)Drunken Irishman
(34,857 posts)1) There's not enough history of this methodology to base any reasonable trend on since it's really only the second election someone has tried to poll this way. If I bowl a 200 my very first time bowling, there is hardly any guarantee I'll bowl a 200 my second time and certainly no evidence to suggest I'm going to be pretty good at bowling heading forward. It just proved, at that point, I did very well. I could turn around and only bowl a 90 something the next game.
2) The RAND poll last go around (it's not the same poll, just similar methodology) wasn't a huge outlier like this poll. It was correct, yes, but there were other major polls that pointed to an Obama win at the end. Here's a list of polling outlets that gave Obama the leading their final polling:
IBD/TPP: +1
ABC News/Wash Post: +3
NBC/WSJ: +1
Pew: +3
National Journal: +5
Polls that had Romney leading:
Rasmussen: +1
Gallup: +1
CNN, GWU and Monmouth had it tied.
RAND was not the lone poll showing Obama in the lead. In fact, most final polls either had him up or tied. Only two had Romney winning. In this go around, only one poll currently has Trump ahead - and that's Rasmussen, one of only two November polls four years ago that had Romney winning.
I'd feel a bit more worried if there was less consensus like four years ago but generally the consensus is the way to go. That doesn't even get into the state polls.
Kashkakat v.2.0
(1,752 posts)question - no, its not possible to feel comfortable til its over.
Mass
(27,315 posts)This test is a new methodology, as the LATimes says on their website.
What they are talking about is the Rand Test, which looked better for Obama than other polls (or rather than the RCP average which included Gallup and Rasmussen) and ended up being fine.
I recommend that you read the NYTimes article about the LATimes poll to see what the problem is.
bluestateguy
(44,173 posts)A supposedly AA supporter of Trump. And the pollster mega weights his responses. They poll the same people each time too.
DemocratSinceBirth
(99,708 posts)The pollsters for the LA TIMES/USC DORNSLIFE POLL claim the accurate 2012 RAND POLL is their model.
However RAND has polled this year and they found Hillary with a ten to twenty five point lead in the same time period the LA TIMES POLL had it tied:
http://www.rand.org/pubs/research_reports/RR1726.html
http://www.rand.org/pubs/research_reports/RR1758.html
http://cesrusc.org/election/