1. We have added a Gift Upgrades feature that allows you to gift an account upgrade to another member, just in time for the holiday season. You can see the gift option when going to the Account Upgrades screen, or on any user profile screen.
    Dismiss Notice

[RD] US 2016 election: Poll watching thread

Discussion in 'Off-Topic' started by dutchfire, Jul 14, 2016.

  1. uppi

    uppi Deity

    Joined:
    Feb 2, 2007
    Messages:
    4,782
    For the presidential election he is right: There will not be enough of them to prove him wrong anytime soon. But there is more data on the individual states and the senate races . And there I guesstimate that his predictions are too correct: According to his numbers he should be wrong more often. That would mean that his assumed distribution is wider than the actual polling error and he should make predictions more aggressively.

    Yes, this is one of the points of making a numeric model: You can see trends and estimate the impact of events. In theory, a probabilistic model also gives you the chance to evaluate and compare different models. But for the latter, the usual problem of data analysis applies: There is never enough data.
     
  2. AdrienIer

    AdrienIer Deity

    Joined:
    Mar 4, 2010
    Messages:
    3,379
    Location:
    Paris
    But most national polls have between 5 and 10% undecided voters (and between 5 and 10% third parties, whose numbers are usually higher in polls than in actual ballots). That's much higher than usual, and has to be represented as uncertainty in the model.
     
  3. uppi

    uppi Deity

    Joined:
    Feb 2, 2007
    Messages:
    4,782
    Of course, and I would agree that this election is harder to predict than the last presidential election. But if my suspicion is correct, the absolute uncertainty should be lower for all elections.
     
  4. onejayhawk

    onejayhawk Afflicted with reason

    Joined:
    Jul 6, 2002
    Messages:
    13,625
    Location:
    next to George Bush's parents
    I doubt it was that much harder. The pollsters did little better in 2012. The pollster that did best now does the LA Times poll.

    One thing that has become clear is that Republican support is more reluctant to admit it than Democratic support. I have heard of one pollster that found a good trick question to spot the so-called hidden Trump supporter. If anyone has details, please post. In any event, self-screening is a growing issue in the business. See below.
    https://hbr.org/2016/11/why-pollsters-were-completely-and-utterly-wrong

    The other issue is the Electoral College. A popular spread of 2%-3% produces a dead even EV projection. This is because of overconcentration of Democratic support in NY and CA. Of course, that was a big part of Trump's message--Clinton only cares about the big city and the big money.
    http://fivethirtyeight.com/features...percentage-points-makes/?ex_cid=2016-forecast

    I found the pollster. It was Trafalgar Group, a C rated firm in Silver's rankings. They called Pennsylvania and Michigan. This is a small firm but it did up to the last minute polls. http://investmentwatchblog.com/brea...ate-poll-has-trump-winning-pa-fl-mi-ga-ut-nv/
    http://www.politico.com/story/2016/11/how-could-polling-be-so-wrong-2016-231092
     
    Last edited: Nov 9, 2016
  5. Archbob

    Archbob Ancient CFC Guardian

    Joined:
    Oct 25, 2000
    Messages:
    11,774
    Location:
    Corporate USA
    The National Polls weren't that wrong, actually. She just lost the EC. She will probably win the popular vote by 1-2% when its all said and done(CA hasn't been counted in full yet).
     
  6. Algeroth

    Algeroth 8 and 1/2

    Joined:
    Sep 10, 2006
    Messages:
    1,466
    Location:
    Прага
    I've reread this comical gem from the start. Now when I would like to commit suicide, I need just reread it once more and take a shot whenever someone predicted a landslide here.
     
  7. metalhead

    metalhead Angry Bartender

    Joined:
    Apr 15, 2002
    Messages:
    8,031
    Yep, this is an important point. The "great polling miss" wasn't much of a miss on the national level.

    The state polling was siginificantly worse, though. Even Trump's data team saw little more than a blip in the data suggesting there might be an opening in the Midwest. Nobody anywhere was getting data that indicated he could win Wisconsin, or Pennsylvania, or win Ohio by 8 points. The strategy was basically to pound those states and pray the data were wrong.
     

Share This Page