Skip to main content

View Diary: Daily Kos Elections Morning Digest: Rep. Dan Lungren defends fraudulent voter registration drive (134 comments)

Comment Preferences

  •  That's interesting, but (1+ / 0-)
    Recommended by:
    jncca

    I'm not aware of any relationship between state-level unemployment or change in unemployment and Obama's approval or the change in his approval (see: Nate's "The North Dakota Paradox") nor am I aware of any historical relationship (John Sides has a post or two on this).  You're also comparing one poll to a polling average.

    26, Dem, Dude seeing a dude, CT-04(originally), PA-02/NY-14 (formerly PA-02/NY-12).

    by Xenocrypt on Thu Jun 21, 2012 at 07:35:58 AM PDT

    [ Parent ]

    •  Unemployment was only a possibility. (1+ / 0-)
      Recommended by:
      David Nir

      State level numbers were really twitchy.  Only when I agglomerated into sub regions, and in this case regions, was I able to see any relationship.  But I wouldn't put too much emphasis on it.

      As far as comparing just one poll, well, it's actually 8 polls with N=8000.  So it's not too bad a comparison.  Also, you can see a similar pattern, actually, in the data from all polls recently, but there's many states that are not regularly polled, so I chose the dkos polls instead.

      •  What I meant is (0+ / 0-)

        that it's a comparison between polls from one source and a polling average.  I suppose the former can be seen as an average.  I'm really not a stats guy, so maybe it's fine.

        26, Dem, Dude seeing a dude, CT-04(originally), PA-02/NY-14 (formerly PA-02/NY-12).

        by Xenocrypt on Thu Jun 21, 2012 at 11:34:40 AM PDT

        [ Parent ]

        •  You are right (1+ / 0-)
          Recommended by:
          Xenocrypt

          That that is a potential pitfall.  For instance, if in 2008, all the polls from the South were done by Rasmussen, while all the polls in the Northeast were done by SurveyUSA, you can immediately see the problem.  However, I did quickly look for this sort of effect and did not find one, although a more thorough analysis might find a small effect.

          It would be much better to compare only to ppp polls in 2008 - but there aren't enough of them.  Or, to compare to polling averages in 2012 - but there aren't enough polls in non-swing states.

          •  I suppose all we can do (0+ / 0-)

            is be cautious, which is a good idea anyway.  Thanks for looking.

            26, Dem, Dude seeing a dude, CT-04(originally), PA-02/NY-14 (formerly PA-02/NY-12).

            by Xenocrypt on Thu Jun 21, 2012 at 11:59:36 AM PDT

            [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site