Skip to main content

View Diary: MN-Sen Recount: Update v. 1.1 (261 comments)

Comment Preferences

  •  Here's my math (8+ / 0-)

    There are 8900 undervotes on the presidential race. I'd argue that any extra undervotes in the senate race are deliberate, so let's start with that number.

    A 2006 audit in the Minnesota senate race showed a .05% rate of machine error. That would explain 1500 of the undervotes, leaving 7400. A 2004 study in FL counties which looked at optical scan machines, like MN's showed, of the undervotes which were not machine error (different machines, so that part is meaningless), 24% were correctible voter error (marking an X in the wrong place, etc.). So that would be about 1800.

    If the 461 disqualified absentee ballots, and others like them across the state, are eventually validated, that could be as much as about 1400 more votes.

    There are two ways Franken could win: through random fluctuations (luck) in those ballots, or if they were systematically more Democratic. Random fluctuations would lead to a standard deviation of not more than 35 votes; that means there's a 98% chance that the deviation is under 70 votes.

    So he needs a systematic advantage. To add 206 votes out of a pool of 1800 (the machine errors can have no bias), those votes must break for Franken by about 11%. With Barkley in the picture, that means that they must be about 47/35 F/C. On exit polls, the best Franken demographics - low income, minority, and first-time voters - only make about 52/33. So these good demographics must be about 80% of the undervotes. Since they are under 1/3 of the voters, that would mean they would be 8 times as likely to make a correctible voter error than other voters. That is a massive difference. On the other hand, in FL-2000, correctible voter error on optical scan ballots tilted about 20% more democratic than the counties in question, so it is not out of the question.

    Still, those absentee ballots are key. I can't really see that Franken can have more than a 50/50 chance, even under the best possible assumptions, without them.

    Opinions are like assholes. I spend way too much time looking at them on the internet.

    by homunq on Thu Nov 13, 2008 at 06:25:58 AM PST

    •  But you are assuming that none of the counted (3+ / 0-)
      Recommended by:
      rincewind, mcartri, Don Enrique

      votes change.  They recount every vote, not just undervotes.

      •  Didn't talk about that (1+ / 0-)
        Recommended by:

        but those votes should change in a pretty unbiased way overall. That would increase the standard deviation (double, because each changed vote is a 2-vote shift). But there's no reason to believe that there would not be fewer changed votes than machine-caused undercounts, so the stdev is still somewhere (significantly) south of 70, which means a 98% chance that he'll get fewer than 140 votes on that basis alone.

        The Absentees are The Keys.

        Opinions are like assholes. I spend way too much time looking at them on the internet.

        by homunq on Thu Nov 13, 2008 at 07:08:11 AM PST

        [ Parent ]

        •  My god are you wrong (0+ / 0-)

          I've lived through several recount'sin Minnesota's downticket races.  And I've seen dozens of votes change in state house races.  If you extrapolate that to the entire state for a US senate race, I think it is highly likely that several hundred up to two thousand votes may be added or subtracted.

          •  I said <1500 (0+ / 0-)

            which is not far off your estimate. The 70 number is the standard deviation based on random fluctuations, which is the square root of the total of vote changes from all sources. It's only what's left over after most of the vote changes cancel out.

            Opinions are like assholes. I spend way too much time looking at them on the internet.

            by homunq on Fri Nov 14, 2008 at 07:37:27 AM PST

            [ Parent ]

    •  Not just undervotes (4+ / 0-)
      Recommended by:
      buddabelly, mcartri, brein, Don Enrique

      Some portion of the ballots may not have been counted at all. At the precinct I was doing Voter Protection (VA), one voter noticed that scan machine's counter didn't change. Her ballot might not have been counted. And, she's the one who actually noticed this. How many others didn't even notice?

      Coming Soon -- to an Internet connection near you:

      by FischFry on Thu Nov 13, 2008 at 08:15:58 AM PST

      [ Parent ]

    •  Many "undervotes" are disputable (2+ / 0-)
      Recommended by:
      mcartri, brein

      as are the validity of disqualified absentee ballots.  Therein lies the importance of Franken lawyers at every count site.  This is by no means a given.  Franken is pleading online for donations.  

      Coleman, it seems, has them already due to the generosity of Mitt Romney.

      Link to Franken recount contribution page:

      Money is more important than math.

    •  This assumes no voter error, which can't be right (0+ / 0-)
    •  9,848 Pres Undervotes (1+ / 0-)
      Recommended by:

      There were 9,848 Presidential undervotes and 34,678 Senate undervotes.

      I agree that most additional Senate undervotes are likely deliberate.

      Machine error can cause missed votes, but it apparently can add spurious votes.
      The hand inspection audits this week resulted in lost votes in some cases.

      I also don't think that .05% is the correct machine error rate, won't random machine errors generally cancel each other out?

      Although it IS a small sample, the recently counted 32 absentee ballots heavily favored Franken (he gained 11 votes), more so than the county at large.

      •  Good catch (1+ / 0-)
        Recommended by:

        9,848 not 8900. That lowers the margin Franken needs without the absentees to 9-10%, still pretty high - it would mean the "good" populations would need to have been over 5 times as likely to make voting errors.

        As you say, the absentees should be strongly Franken, and if he (has the money to) get these counted his chances are good. But without them, it is feasible, but a bit of a stretch.

        0.053% is correct, it was something like 48 votes out of 97,000 in the 2006 audit. And you are right, these mostly cancel each other out; random fluctuations lead to a margin of something less than twice the square root of the number of random errors. That is what I was talking about with the standard deviation above.

        Opinions are like assholes. I spend way too much time looking at them on the internet.

        by homunq on Thu Nov 13, 2008 at 11:13:37 AM PST

        [ Parent ]

        •  jmho (2+ / 0-)
          Recommended by:
          homunq, brein

          I don't think the 2006 audit numbers will necessarily hold for this election. All mid-term elections drop off from Pres elections, and this year MN increased its traditionally-high turn-out rate. Mid-term voters tend to be the hard-core committed voters, and the increased turn-out in Pres-election years are the newbies/intermittent voters (more susceptible to making mistakes?).

          IMPEACH " that no future president may infer that we have implicitly sanctioned what we have not explicitly condemned." John Conyers, 1974

          by rincewind on Thu Nov 13, 2008 at 12:29:55 PM PST

          [ Parent ]

    •  systematic advantage (1+ / 0-)
      Recommended by:

      since new voters are more likely to make errors than experienced voters I would expect that votes counted in the recount will come disproportionately from:
      -young voters
      -people who registered for the first time this year

      both of which are friendly demographics... so the presence of a systemic advantage seems likely to me.

      •  Also absentee (1+ / 0-)
        Recommended by:

        Folded ballots are less likely to be read correctly.

        I believe Franken won with absentees but I don't have numbers.

      •  yes (0+ / 0-)

        friendly demographics are overrepresented in the pool. But as I said, for franken to win without the absentees, they would have to be overrepresented by a factor of 5 to 10, which is a lot.

        With the absentees, it becomes much more plausible.

        Opinions are like assholes. I spend way too much time looking at them on the internet.

        by homunq on Fri Nov 14, 2008 at 07:39:46 AM PST

        [ Parent ]

Subscribe or Donate to support Daily Kos.

  • Recommended (154)
  • Community (61)
  • Environment (36)
  • Culture (34)
  • Memorial Day (31)
  • Republicans (31)
  • 2016 (28)
  • Elections (26)
  • Civil Rights (25)
  • Bernie Sanders (25)
  • Science (23)
  • Media (22)
  • Climate Change (22)
  • Education (22)
  • Labor (21)
  • Spam (21)
  • GOP (20)
  • Law (20)
  • Rescued (19)
  • Trans-Pacific Partnership (16)
  • Click here for the mobile view of the site