Skip to main content

View Diary: NC Early Voting Day 6: 135K Minority/Dem '08 Non-Early-Voters Have Voted… Vs 78K White GOPers (96 comments)

Comment Preferences

  •  They're not tabulated or strictly known (3+ / 0-)
    Recommended by:
    highacidity, elwior, distraught

    I am estimating them based off of %s drawn from 2008 exit polls + common sense. Check out the day 1 diary for a full methodology explanation.

    •  OK, so you're estimating based on 2008 exits (0+ / 0-)

      I thought exit polls are notorious for being not that accurate.

      But even beyond that, how are you getting ethnicity percentages for the 2012 early votes, especially if there is no way to account for actual early votes?

      You write this in your methodology, which you pointed me to above:

      What we are comparing is simply voter turnout, but not necessarily who each voter votes for. To get these estimates, we are comparing the demographics and party registration of early voters from 2008 vs. the demographics and party registration of early voters from 2012.   This assumption is not realistic, but it is more realistic for North Carolina than for most other states, because North Carolina has relatively few true "swing voters."
      Bold my emphasis.

      So if you the know the party registration and race of 2012 early voters, you are estimating who they voted for based on exit poll data from 2008 of who those specific demographics voted for.  Which introduces the error of exit polls.

      But if you don't know the ethnicity of 2012 voters and just party registration (i.e - did they sign the Democratic book or Republican book before voting), then you are introducing even more error by assuming that the ethnicity breakdowns will be the same as in 2008.

      And if you don't know any of that (i.e. - how many Dems, Repubs, Indies voted in a day), then how the heck are you computing any of this "data".

      I enjoy the enthusiasm, but this all just doesn't add up to me.  You're estimating vote totals down to individual votes using methods that either introduce one or two steps of error.  And you even admit that doing so is unrealistic, but you justify it saying that the number of undecideds in NC is low.

      Hell, I've even tipped and rec'd some of your diaries in this series, mostly because I didn't have the time to do ask about your methodology.  But now I wish I hadn't to be honest.

      #RomneyRyan2012: Because one white, rich, male Republican asshole is never enough.

      by mconvente on Wed Oct 24, 2012 at 12:27:51 PM PDT

      [ Parent ]

      •  I'm not sure what you are arguing ... (0+ / 0-)

        Here is the break down by race, and others

        The state's racial composition in the 2010 Census:[49]

            White: 68.5% (65.3% non-Hispanic white)
            Black or African American: 21.5%
            Hispanic or Latino (of any race): 8.4%
            Asian: 2.2%
            Native Hawaiian and Other Pacific Islander: 0.1%
            Some other race: 4.3%
            Two or more races: 2.2%

        AA are currently representing about 30% of the people who have voted early. Non-Whites are about 33% of those who have voted. All these numbers are higher than 2008. If we assume that these folks will vote are about the same percentage as they did in 2008 (or even slightly below that), then It is clear that Obama is gaining voters, unless there is really a massive shift White voters.
        Something similar can be said about the Youth vote. Young voters are up almost 40%. They usually favor the President.

        Of course there is no way to know for sure, but what we are saying is, these numbers are reason for great optimism, assuming things remain as is or don't dip too much.

        Don't forget to register to vote here:

        by bepanda on Wed Oct 24, 2012 at 12:54:05 PM PDT

        [ Parent ]

        •  How do you know the racial breakdown #s (0+ / 0-)

          That's really my question.  Is that being official tabulated somehow or is that being estimated based on how many Democratic, Republican, Indie-registered people have voted early.  

          And as I mentioned above, how is party reg even being determined?  By counting directly how many Dems, Repubs, Indies have voted early (i.e. - by counting signatures in the registration books that people sign before they vote), another method, etc.   Or is that estimated as well.

          In order for anyone to even start estimating Obama vs. Romney early votes, you have to have some actual hard data somewhere.  So what is actually the hard data coming from NC for 2012 early voting?  I don't know why that is so hard to understand from my questions above...

          #RomneyRyan2012: Because one white, rich, male Republican asshole is never enough.

          by mconvente on Wed Oct 24, 2012 at 01:04:37 PM PDT

          [ Parent ]

          •  Did you click on the link in the previous message? (1+ / 0-)
            Recommended by:

            When you register in NC, you have to specify your Name, Age, Address, Race, Party affiliation, ... It is all part of your voter record.

            Race is Mandatory. And of course, you specify "Others".
            When you vote, your vote is tracked. At this point, nobody knows who you have voted for, but people know if you have voted or not.

            Click onthis link.

            Don't forget to register to vote here:

            by bepanda on Wed Oct 24, 2012 at 02:15:58 PM PDT

            [ Parent ]

      •  I don't at all claim (2+ / 0-)
        Recommended by:
        susanWAstate, mconvente

        That my methodology is perfect. You rightly point out some ways that it could be improved (if I had more time to spend on it and if I had more data).

        I do think, however, that it is better than anything else you can get from anywhere else (that I know of), unless you are David Axelrod or someone like that.

        To address what seem to be your main points:

        1) Exit polls are not accurate - I totally agree.

        2) Race/Ethnicity - We do actually know the race/ethnicity of voters, because you have to report your race when registering to vote in NC. So this data is accurate.

        3) This doesn't take into account any possible swing in voter preferences - Totally true, and I said that up front. The point of this is to compare whether voter turnout is more favorable or unfavorable than in 2008 in a more rigorous way than simply looking at party registration or race in isolation. Note I say "more rigorous" - not "ideal." For example, this methodology could be better if we took into account age, gender, and the %s that each precinct voted for Obama in 2008. But I didn't do that because it would have been too complicated, would have taken me more time than I had to spend on this, and I don't really have the data to do that. Basically what we're doing is combining race and party registration in one measure and weighting them based on their historical voting patterns. That certainly doesn't tell you everything, but it is more informative than knowing that e.g. ~30% of voters are African American.

        •  OK, at least you have some hard data (1+ / 0-)
          Recommended by:

          That was my main point with my comments above.  You need to have hard data somewhere.  I guess collecting ethnicity data is a North Carolina thing.  At minimum, it certainly isn't a Pennsylvania thing (which is where I'm currently registered to vote).

          This is probably a lot of work, so maybe just do it for the final early voting total, but perhaps you can run some confidence interval-like scenarios accounting for the actual ethnicity percentages, and then a best case and worst case scenarios (i.e. - say Hispanic voters were 12%, so compute 9% and 15% for worst and best, and see what that does to the estimated total for Obama vs. Romney).  That way we can see what a good guess of the worst-case scenario is and how much of a cushion Obama will have.

          #RomneyRyan2012: Because one white, rich, male Republican asshole is never enough.

          by mconvente on Wed Oct 24, 2012 at 08:11:04 PM PDT

          [ Parent ]

          •  It is different for PA (0+ / 0-)

            Yeah, you're right. It is an NC thing. It is like this for NC and a few other states. Because of the Voting Rights Act, they collect race data from everyone who registers to vote, whereas in PA you just register with a Party (or independent).

            There's no confidence interval needed, however. The data was provided individually on a voter by voter basis, by the voters themselves. So generally it should be about as accurate as census data.

            You're right though more broadly, and especially towards the end I'll run through scenarios. The biggest issue concerns with uncertainty about how White Dems and White Unaffiliateds are voting. Are they are swinging substantially towards Romney or not?

    •  also, "common sense" isn't up to snuff for stats (1+ / 0-)
      Recommended by:

      It just doesn't work that way...

      #RomneyRyan2012: Because one white, rich, male Republican asshole is never enough.

      by mconvente on Wed Oct 24, 2012 at 12:32:43 PM PDT

      [ Parent ]

      •  What I mean by that (0+ / 0-)

        Is, for example, exit polls in 2008 did not have a large enough sample of Hispanics in order to estimate the % of non-African American minorities that voted for Obama in 2008. So I used "common sense" and looked to the national exit polls, and so I assigned them 2 to 1 to Obama. Again, if I had better data and more time, I definitely would do that differently, but all in all, I think it is pretty likely to be accurate enough for our purposes.

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site