Skip to main content

View Diary: polling firm biases (27 comments)

Comment Preferences

  •  Marginal gain per dollar (none)
    You've hit the nail on the head.  My original motivation for all this was the same as yours: to figure out what the marginal gain is per dollar invested.  You're thinking in exactly the right way.

    I don't know if you'd agree, but my summary is this:


    (1) Primary objective: maximize the probability of taking at least 1 of Presidency, Senate, House.
    (2) Where is the marginal gain per dollar invested the highest en route to achieving (1)?
    (3) Where is the marginal gain per man hour (e.g. GOTV) invested the highest en route to achieving (1)?
    Figuring this out requires:
    (a) A probabilistic model of all 3 races.
    (b) Data on return per dollar invested in each media market
    (c) Data on return per man hour invested in each geography
    Each of us seems to have made some progress towards (1).  I saw some posts on dkos about estimations of how much a dollar is worth in terms of voter turnout, in various media markets.  [Darn it, I can't find that data right now.]  I also saw similar data for GOTV mentioned on dkos.

    Still much work to do ... perhaps too much, given that we have to hold down our dayjobs :)

    I agree with your assessment re Senate race.  The Prez race is already saturated.  Meanwhile, we have a sparsely populated area like Alaska, with incredibly cheap media prices.  In turns of ROI, it seems entirely logical to hand Tony Knowles 10 million dollars.  That would surely be enough to almost guarantee the seat.  Similarly with the other key senate seats.

    And if more GOP House candidates retire (as reported on the main page for PA-8 today), that starts opening up possibilities.

    Regarding the senate race, other than the pollingreport subscriber polls (very sparse), the closest thing to a data source I have found is tradesports. This link should take you to the "Quote Board".  (If not, "Quote Board" the fifth tab accross the top.)  Then click "Politics" in the left tab.  There's the senate race data --- though not complete, if I recall.  Here are the closest seats (as of July 12th, bid/ask midpoints):

     SENATE.AK.MURKOWSKI    46.5
     SENATE.LA.VITTER       58.1
     SENATE.SD.DASCHLE      62.95
     SENATE.NC.BOWLES       66.5
     SENATE.WA.MURRAY       82.5
     SENATE.PA.SPECTER      83.75
     SENATE.AR.LINCOLN      85
     SENATE.IL.OBAMA        87
     SENATE.CA.BOXER        87.5
     SENATE.MO.BOND         88

    "There are lies, damned lies, and the GOP."

    by Winger on Mon Jul 19, 2004 at 06:57:41 PM PDT

    [ Parent ]

    •  We are on the same page! (none)
      I think we are pretty much in complete agreement here. Friends in my circle (nearly all scientists) are asking me the same questions that you list.

      A full analysis would indeed require what you outline. To simplify things I have been thinking about the presidency, Senate and House as separate events. My gut feeling about the probability of these going Democratic is:

      • White House >80%
      • Senate 50%
      • House 10%
      Of course those guesses are very speculative, but the rank ordering is hard to dispute. In any event, it's clear that the WH race is saturated. If I give $100 to a $100 million campaign, that's 10^(-6) of the total budget. Therefore I have given little to the Kerry campaign. I have given more to Tony Knowles (Alaska) and Inez Tenenbaum (South Carolina).

      The Tradesports numbers are interesting. SC is missing, which is too bad. The listed candidates who are most marginal are Knowles (AK) and Vitter (LA). Runners-up are Daschle (SD) and Bowles (NC). So your diagnosis to give $10M to Knowles is dead-on.

      In regard to the House, I am not ready to sign on to that goal yet, though a swing to the Democrats might change my mind. If I did give it would be to the DCCC, since they are better positioned than I am to calculate (and have more time!).

    •  Diary entry (none)
      Error on my part: do we know yet who is the favored Democratic candidate in Louisiana?

      To reiterate, your points are extremely cogent. I wonder if a diary entry is in order.

      •  Yes, (none)
        This kind of material should appear in a diary at some point.  I'm happy for you others (e.g. you!) to do it, because I really don't care about being credited for any ideas around here.  The important thing is that the ideas get out there.

        As for LA, here are the last 3 polls.  I don't have time to format the cut and paste; hopefully you'll decipher it without too much trouble.

        Not clear whether Kennedy or John is the favorite.  My money's on us holding the seat though, when it comes to a runoff (which might be Nov 16?).


        Market Research Insight for private clients. May 19-26, 2004. N=600 likely voters statewide. MoE ± 4:

              .

        Election Trial Heat:

              .

          David
        Vitter (R) John
        Kennedy
        (D) Chris
        John (D) Arthur
        Morrell (D) Unsure
          % % % % %
         5/04 35 22 18 5 20
             .
         .

              .

        Multi-Quest International poll. May 21-25, 2004. N=413 registered voters statewide (MoE ± 6), including 305 likely voters. (Multi-Quest International works for Senate candidate David Vitter; this poll was not conducted for the Vitter campaign.):

              .

        "The candidates for U.S. Senate include Chris John, John Kennedy, Arthur Morrell and David Vitter. If the election were held today, for which candidate would you vote?"

              .

          David
        Vitter (R) Chris
        John (D) John
        Kennedy
        (D) Arthur
        Morrell (D) Unsure
          % % % % %
         All registered voters 26 15 9 2 48
         Likely voters 29 16 8 2 45
              .

            .
         .
         .

        Wilson Research Strategies (R) for a corporate client. May 4-5, 2004. N=400 registered voters statewide. MoE ± 4.9:
              .

        Election Trial Heat:

              .

          David
        Vitter
        (R) John
        Kennedy
        (D) Chris
        John
        (D) Arthur
        Morrell
        (D) Unsure
          % % % % %
         5/04 24 22 10 5 39
              .

              .

        "There are lies, damned lies, and the GOP."

        by Winger on Wed Jul 21, 2004 at 02:57:40 PM PDT

        [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site