Skip to main content

View Diary: Beyond the Margin of Error (5): This is Why We Can't Have Nice Things (12 comments)

Comment Preferences

  •  More misrepresentation in one direction? (1+ / 0-)
    Recommended by:
    David Nir

    It looks to me from the above that more Romney supporters than Obama supporters are mis-indicating their race as a minority, is that correct? (Although the percentages are higher in terms of "true" support by non-whites of Obama - does that suggest the reverse is true for Obama supporters? I.e., more  Obama supporters mis-identify as white? This is of course assuming, which is a big assumption, that responders are not making a critical choice about these categories and their own backgrounds - among other things.)

    I know one can't know anything about cause from this data, but, to the extent I've understood correctly, it makes me wonder if there is a deliberate effort to represent Romney's base as less white than it actually is - maybe something they know to be an issue.  Cf. the attempt to identify nonwhites at the GOP by JJP earlier this year?

    We might be heading into the McCarthy era. But I hold out hope: after the 50s came the 60s, and that's when they invented sex.

    by marynyc on Thu Sep 20, 2012 at 09:44:40 AM PDT

    •  Er, (0+ / 0-)

      Not that I endorse JJP's attempt, necessarily - for the record, I think it's problematic to try and visually i-d someone's race, on a few grounds.

    •  Oh boy (1+ / 0-)
      Recommended by:
      David Nir

      That is a very funny link.  Thank you!

      On to the meat of your comment. I cannot tell from the data if more white Romney voters than white Obama voters are choosing the wrong race. For the simulation, I assumed all respondents had an equal chance of choosing the incorrect race.

      This assumption may not be correct, but there is no way to test it with the data available.

      Personally I would suspect that white Romney supporters are more likely than white Obama supporters to provide intentionally misleading answers to the race question, cause it's kind of a jerk thing to do and being a jerk is kind of the foundation of the Republican mindset: I've got mine, screw you. But it's not just my personal bias that leads to this supspicion.

      Why? In an informal study, of 6% of drivers who intentionally ran over a rubber animal, 89% of them were driving an SUV. (Ya gotta click through this link.)

      And, large SUVs were favored by 76% of Republicans compared to 5% of Democrats. Make of that what you will.

      •  What's the probability Romney supporters choosing (0+ / 0-)

        Native American as their 'origin' or ethnicity, did so because they firmly believe they indeed are native-born full citizens of the U.S. of A (versus, say, Kenyans or 'redskins' born on a reservation or illegitimate illegal immigrants)?  I'm be willing to bet that sort of view could explain a fair chunk of these.

        Might need to factor in what percent of the population may have some sort of disability like dyslexia which might lead to incorrectly linking words & meanings to spatially located check boxes or bubbles.  

        It might seem shocking but some conservatives may know identifying as a non-white race on a census form might mean more federal funding coming to their districts and they may feel that's just peachy to help that's as easy to take money from the 'liberal' socialist government as it is to take candy from a baby...and they'd justify it thinking we'd just be takin' back some of the real tax payers money and putting it to more righteous use for 'real' Americans.

        Consider as well that the labels 'White' and 'Caucasian' can be confusing.  How much White ancestry makes someone white?  If I am 'mixed' but can 'pass' for white walking down the street, isn't that sufficient?  Ask a less educated White person if they are Caucasian and you may get slapped in the face if not a look of confusion.  Ask an average White person who actually knows they are technically labeled 'Caucasian' why that name applies and you may get total bafflement.  Ask a college educated White where to find the Caucas mountain region on the globe or to name the continent where it is found and you'll probably get more blank looks and embarrassment...or hear random mumblings about Christopher Columbus, Portugal, Spain, England, Puerto Rico, Cuba, wait, the Mayflower...Puritan Pilgrims, Plymouth or Chrysler Rock or GM rock...  the French...Lousiana..., Jamestown, Catholics Protestants... slaves...  turkey... kool aid... oatmeal... maybe it's Quaker Caucasians?

        When life gives you wingnuts, make wingnut butter!

        by antirove on Thu Sep 20, 2012 at 01:44:02 PM PDT

        [ Parent ]

        •  That's a possibility. (2+ / 0-)
          Recommended by:
          antirove, MichaelNY

          I'll be exploring the Native American poll responses in a separate post, but you bring up one possibility for sure.

          As far as a learning disability increasing error rates on a push-button poll response, that too is certainly possible. But in the previous post in this series, I showed a maximum error rates on the gender question of about 0.2%, and this should include the sort of error you mention. So we know it will not have a huge impact on responses.

    •  Smaller subgroups get skewed more (2+ / 0-)
      Recommended by:
      tapu dali, MichaelNY

      from the randomness, as the respondents incorrectly claiming to be in the group are from larger groups, where the same rate of error produces larger raw numbers of wrong answers.

      It's the same small group/large group imbalance that makes lie detectors unreliable for distinguishing liars; since very few taking lie detector tests are liars, the truth-telling tests that err load a group into the "liar" set that is as large as the actual liar set, correctly identified.

      Let's look at an example:

      Say there are 1000 respondents in two groups, one of 900 (A) and one of 100 (B).  Say there is a 10% chance that a respondent is mislabeled into the wrong group.  Then 90 A's are labeled B's, and 10 B's are labeled A's.  So the sample would be identified as 820 A's and 180 B's.  The slide from 900 to 820 (under 10%) is minimal in comparison to the climb from 100 to 180 (nearly double).

      1. Corporations control our democracy and do not have our interests at heart;
      2. The media is not neutral -- and not blameless;
      3. Ordinary people have extraordinary power.

      by MooseHB on Thu Sep 20, 2012 at 11:02:35 AM PDT

      [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site