Sunday, September 26, 2010

Caucasian not white

How strongly does race predict voting? Here is a multivariate model of six predictors:

Voted for Bush in 2004 (standardized OLS regression coefficients)

Black -.35
Other race -.10
Male .07
Years of education -.08
Income .09
Church attendance .22

All predictors are significantly related to the outcome variable--voting for Bush over Kerry or Nader. (The dependent variable is evenly split so I can get away with using OLS).  "Black" is a comparison of blacks versus whites, and is the strongest predictor in the model. Frequency of church attendance is next. "Other race" is in comparison with whites, and is about as predictive as sex, education or income. Another way of stating this is that whiteness is a relatively strong predictor of voting Republican. 

For those interested in getting more people to vote Republican, none of these predictors is easy to manipulate. For example, how do you get more people to go to church?  Setting aside the question of practicality, Republicans might benefit if more people saw themselves as white. If America embraced, say, a Coonian understanding of Caucasian, and focused on facial features rather than skin color, fewer Americans might be drawn to the party of people who feel like aliens--the Democrats. You know--feminists, homosexuals, immigrants, Jews, non-Christians, and non-whites.  

The use of the term "white" does not help since it brings some pasty dude to mind. This has always been a misleading term. My ancestry is 100 percent northern European, and I can give John Boehner a run for his money any day. One of my brothers looks a little like an Arab. A rough rule of thumb would go like this: "Forget the little groups around the world. Does he look like an East Asian? Does he look like a sub-Saharan African? Does he look like an American Indian? No? Then he's a Caucasian guy."     


  1. Anonymous9:49 PM

    In 2004, Bush was running against Kerry.

  2. Boy do I feel stupid! Thanks. I looked at 2000 and got the same results.

  3. Anonymous11:06 PM

    I'd be curious what population density would do.

  4. You should really be using logistic regression not OLS for this.

  5. "I'd be curious what population density would do."

    The GSS has population size. I added it to the model. The effect was not significant, and the results remain unchanged.

  6. "You should really be using logistic regression not OLS for this."

    You're right. I cheated since people are more comfortable with standardized coefficients. The story is the same when I use logistic. Direction and significance are the same.


Meta-analysis of clinical trials: Eat walnuts

I am always looking for easy eating choices that are good for you. This new meta-analysis of 26 clinical trials looked to see if walnuts ma...