Data from the OKCupid project.

In light of a recent paper examining who prefers to date within their own religion, I recalled that there was a question about this in the OKCupid dataset, except that it is for race: “Would you strongly prefer to go out with someone of your own skin color / racial background?” It’s a dichotomous outcome, where yes = 1, so I used a logistic model. I added the usual predictors. Results:

$coefs
                                            Beta   SE CI.lower CI.upper
age                                         0.03 0.01     0.00     0.06
gender: Man                                 0.00   NA       NA       NA
gender: Other                              -1.85 0.59    -3.01    -0.70
gender: Woman                               0.12 0.03     0.06     0.19
race: White                                 0.00   NA       NA       NA
race: Mixed                                -1.15 0.06    -1.27    -1.03
race: Asian                                -0.66 0.07    -0.80    -0.52
race: Hispanic / Latin                     -1.40 0.09    -1.57    -1.22
race: Black                                -1.49 0.09    -1.67    -1.31
race: Other                                -2.06 0.15    -2.36    -1.76
race: Indian                               -0.60 0.15    -0.89    -0.31
race: Middle Eastern                       -1.30 0.24    -1.78    -0.82
race: Native American                      -0.28 0.26    -0.78     0.22
race: Pacific Islander                     -2.93 0.72    -4.35    -1.52
CA                                         -0.41 0.02    -0.44    -0.38
ideology: Liberal / Left-wing               0.00   NA       NA       NA
ideology: Centrist                          0.74 0.04     0.66     0.82
ideology: Conservative / Right-wing         1.49 0.05     1.39     1.60
ideology: Other                             0.76 0.04     0.69     0.83
religion_seriousness: Not at all important  0.00   NA       NA       NA
religion_seriousness: Not very important    0.42 0.04     0.35     0.49
religion_seriousness: Somewhat important   -0.18 0.03    -0.25    -0.12
religion_seriousness: Extremely important  -0.09 0.03    -0.15    -0.04

$meta
        N pseudo-R2  deviance       AIC 
 34519.00      0.12  30143.27  30183.27
$meta N pseudo-R2 deviance AIC 34519.00 0.12 30143.27 30183.27

The betas are standardized. CA = cognitive ability. Measured from an ad hoc collection of 14 items.

Some conclusions:

  • Women are slightly more ethnocentric.
  • All non-White races are less ethnocentric.
  • Smarter people are less ethnocentric.
  • Non-liberals and particularly conservatives/right-wingers are more ethnocentric.
  • Religion has a non-linear relationship in that the somewhat religious are more ethnocentric, but the more strongly religious are less ethnocentric.

See also: heritability of ethnocentrism and related traits (sorry egalitarians, racists were born that way!)

R code: gist.github.com/Deleetdk/a64dcfa8ae93e5b918b2f64024f6bdf4


Previously I found that religiousness predicted ethnocentrism fairly linearly, but I can’t reproduce the result — even using the same code! No idea what caused this result.