Skip to main content

View Diary: Nate Silver Calls B.S. on U. of Colorado Election Prediction Model UPDATED x1 (139 comments)

Comment Preferences

  •  Paper is here (7+ / 0-)

    Link to paper

    Key to understanding is the graph at back that shows the projected state percentage vote in 2012 against the 2008 outcome.

    Almost perfect line indicating that their model just shifts the 2008 results down by a fixed percentage - only the previous term's results are significant. The amount of the shift is the difference between Carter 1976 and Carter 2008- really that is what they did, they decided that Obama was Carter and reduced his vote share everywhere by the same percentage that Carter experienced.

    •  Thanks! And you can (2+ / 0-)
      Recommended by:
      ericlewis0, angry marmot

      disregard my query below about where you found the paper. I don't teach forecasting, but I use a lot of time-series analysis in my own work; so I look forward to reading it in light of your critiques.

      Grew a mustache and a mullet / Got a job at Chick-Fil-A

      by cardinal on Thu Aug 23, 2012 at 05:11:30 PM PDT

      [ Parent ]

    •  Oops should be Carter 1976 and Carter 1980 (1+ / 0-)
      Recommended by:
      ericlewis0
    •  not quite "just," but yeah (1+ / 0-)
      Recommended by:
      ericlewis0

      As I read the graph, only the win predictions for NV and WI deviate from the simple shift you mention -- which is driven, apparently, by the national unemployment rate.

      So, yeah, using all the state-level data gives them a lot less value than they imagine. It is bizarre that they report OLS results. I don't understand how they could have obtained substantively equivalent results for proper alternative models: I think they must have been assuming independence of variables that obviously aren't independent.

      really that is what they did, they decided that Obama was Carter
      That's a pretty freehand metaphor, no?

      Election protection: there's an app for that!
      Better Know Your Voting System with the Verifier!

      by HudsonValleyMark on Thu Aug 23, 2012 at 06:35:38 PM PDT

      [ Parent ]

      •  Not sure they know what they did (2+ / 0-)
        Recommended by:
        HudsonValleyMark, ericlewis0

        The general level of incompetence is so high, they made so many really basic mistakes, that I can only assume they really thought that the state level data had enough identification power that the estimation was meaningful. If you think you have 102 data points for incumbent Democrats it makes a lot more sense to estimate the impact of unemployment level than if you have 2 (which are highly correlated with about a zillion omitted variables)

        •  (three, I think) (1+ / 0-)
          Recommended by:
          ericlewis0

          I think "incumbent" refers to party, so Dems were incumbent in 1980, 1996, and 2000, yeah?

          With that caveat, I think you're exactly right. This is a hierarchical model, but somehow they didn't notice, and nobody mentioned it.

          Election protection: there's an app for that!
          Better Know Your Voting System with the Verifier!

          by HudsonValleyMark on Thu Aug 23, 2012 at 07:34:58 PM PDT

          [ Parent ]

          •  No, Gore is not counted as incumbent (2+ / 0-)
            Recommended by:
            ericlewis0, HudsonValleyMark

            They distinguish between a second consecutive term for the party (Carter, Clinton) and a third or more (HW Bush, Gore, McCain). The big effects are just for Carter/Clinton. Yes I know this is completely insane.

            We also control for the term that the incumbent
            party is seeking. We consider two possibilities: incumbent parties seeking a second presidential
            term and incumbent parties seeking a third or higher term in the White House. This occurred
            most recently in 2008 when senator John McCain ran for a third consecutive presidential term
            for the Republican party. In our modeling approach, only one binary variable is required to
            capture these alternatives. If a candidate is seeking a second consecutive term for the candidate’s
            party, we code as 1 a second-term contest, 0 otherwise. Notice that if a candidate is of the same
            party as the incumbent in the past two (or more) presidential terms, the second-term contest
            variable is coded as 0. This was the case for president George H.W. Bush, vice president Al
            5
            Gore, and senator John McCain. This variable continues to be coded as 0 if candidates seek a
            fourth or even fifth consecutive term for their party.
            •  umm, I don't think that's right (1+ / 0-)
              Recommended by:
              ericlewis0

              Clearly "First Term Incumbent" in the model is defined in the way you just quoted. It's possible to construe the text as implying that "Democratic Incumbent" means "First Term Democratic Incumbent," but I don't think that is the most likely reading.

              It's a little scary that we aren't quite sure what the independent variables are, but it isn't the biggest problem here.

              Election protection: there's an app for that!
              Better Know Your Voting System with the Verifier!

              by HudsonValleyMark on Thu Aug 23, 2012 at 09:27:15 PM PDT

              [ Parent ]

Subscribe or Donate to support Daily Kos.

  • Recommended (146)
  • Community (56)
  • Baltimore (38)
  • Civil Rights (36)
  • Bernie Sanders (33)
  • Culture (28)
  • Elections (28)
  • Economy (27)
  • Law (24)
  • Texas (23)
  • 2016 (21)
  • Rescued (20)
  • Environment (19)
  • Labor (19)
  • Hillary Clinton (18)
  • Education (18)
  • Freddie Gray (16)
  • Media (16)
  • Barack Obama (16)
  • Racism (16)
  • Click here for the mobile view of the site