2016 Postmortem
Related: About this forumHere is a link to several poll aggregation and prediction sites
http://www.nytimes.com/interactive/2016/upshot/presidential-polls-forecast.html?action=click&contentCollection=upshot®ion=rank&module=package&version=highlights&contentPlacement=2&pgtype=sectionfront&_r=0http://projects.fivethirtyeight.com/2016-election-forecast/
http://elections.huffingtonpost.com/2016/forecast/president
http://predictwise.com/politics/
http://election.princeton.edu/
http://elections.dailykos.com/app/elections/2016
http://cookpolitical.com/presidential/charts/scorecard
http://rothenberggonzales.com/ratings/president
http://www.centerforpolitics.org/crystalball/
http://www.latestpollresults.com/
http://pollyvote.com/en/
livetohike
(22,157 posts)DemocratSinceBirth
(99,711 posts)Farmgirl1961
(1,493 posts)It seems like 538 has gone rogue. What the f*ck is going on over there?
DemocratSinceBirth
(99,711 posts)- Nate is good but he is not the only person doing it and he is not the only person who has been successful at doing it.
- He still has her at 75%.
Farmgirl1961
(1,493 posts)It seems like all the other polls have her in the mid 90s or low 80s
And those polls have remained relatively stable through this FBI/Comeygate. But Nate polls keep showing Hillary decreasing and Trump gaining. Don't worry, I'm not putting too much stock in his stuff. But just wonder why his site is the only one that shows such volatility and her number is dropping.
Jim Lane
(11,175 posts)One big reason is that he doesn't assume that each state result is an independent event. For example (my example, I don't remember whether it's his): Suppose Trump were to hit Clinton on trade (support for NAFTA, past support of TPP then flip-flopping on it) and to campaign heavily on the issue of job losses. If that attack caught on, he could conceivably win several Rust Belt states that went for Obama.
Suppose there are two blue states and Trump's chance of flipping each is 20%. If they're independent events then his chance of flipping both is 4%. If, however, the question in both states is whether a trade-based attack will swing enough votes, then the results are correlated. The chance of flipping both will be only a little less than 20%. Silver uses the latter type of model.
GopherGal
(2,009 posts)Also mentions that 538's model allows for more uncertainty in the polling. This will lead to it looking less favorable for the frontrunner than other models.
But yes, the idea of state outcomes being correlated is a big one. And it would operate in the same direction - to make the less likely winner more probable.
[link:http://fivethirtyeight.com/features/election-update-why-our-model-is-more-bullish-than-others-on-trump/|
Demsrule86
(68,643 posts)Gothmog
(145,496 posts)Jim Lane
(11,175 posts)No reports on polls of the nationwide popular vote, which doesn't elect anyone; just a state-by-state breakdown based on polls there, updated daily. Click on the "Senate" link at the top for similar state-by-state analysis of all the Senate races.
Along with the poll aggregation, each day's update includes half a dozen or so items about the latest electoral news.
The site proprietor is a Democrat and has displayed a pro-Clinton bias but tries to keep his electoral projections strictly data-based.
Demsrule86
(68,643 posts)Nate Silver's is pretty much the lowest of the ones I am familiar with...he got brexit wrong and Michigan...just don't know if his new model works.