Kanazawa Uses Statistics And Evolutionary Psychology To “Prove” Black Women Less Attractive

Academic, self-labeled evolutionary psychologist, and member of the London School of Economics, Satoshi Kanazawa wrote a blog post on Psychology Today that asked, “Why Are Black Women Less Physically Attractive Than Other Women?” Satoshi Kanazawa

Notice that this was put in the form of a question about a subject to which Kanazawa presumed all agreed. He was not asking if black women were less physically attractive, but why they were so.

The magazine pulled the original post after the inevitable firestorm of “outrage”, but it has been made available here.

Now, in his favor, and misunderstood by many who have commented on the fracas, Kanazawa did not claim that all black women were uglier than all non-black women. His statistical analysis implied that if you grab any black woman and any lined her up next to any non-black woman, then more people would rate the non-black woman as more attractive than the black woman.

Specifically, he did not claim that no person would find black women in these comparisons less attractive, just that most people would find these ladies uglier.

“Of course, it might even be true,” you are now saying to yourself, “But if so, it is entirely due to the curse of racism.” And you might be right, but Kanazawa anticipated your criticism and argued that culture was not the driving factor behind the rating differential.

Instead, black women were considered uglier because—wait for it—evolution made black women uglier. He said his evidence was not only “objective”, but it was hard-wired into our genes.

When you look at a black face and at a non-black face, and you find yourself preferring the non-black choice, this is because it is to your evolutionary advantage to do so. Somehow. Says Kanazawa, not I! So please, no hate mail.

Okay, Mr Kanazawa, why?

Africans have more mutations in their genomes than other races. And the mutation loads significantly decrease physical attractiveness (because physical attractiveness is a measure of genetic and developmental health)….

The only thing I can think of that might potentially explain the lower average level of physical attractiveness among black women is testosterone. Africans on average have higher levels of testosterone than other races, and testosterone, being an androgen (male hormone), affects the physical attractiveness…women with higher levels of testosterone also have more masculine features and are therefore less physically attractive. The race differences in the level of testosterone can therefore potentially explain why black women are less physically attractive than women of other races…

Ah, sweet testosterone! Is there no evil in which you are not complicit!

But this merely explains why the conclusion is what it is, it does not explain how he reached the conclusion itself. To do that, Kanazawa had to use a statistical model.

That is, he did not, as you might have expected, run out into the street and hold up photos of various women and ask people to say who is prettier. Instead, he looked at a database “on a nationally representative sample of adolescents in grades 7-12 who have been followed up to adulthood.” The United States only, you understand.

That database had over 8,000 different variables, some of which were subjective responses on attractiveness. Sort of: “At the end of each interview, the interviewer rates the physical attractiveness of the respondent objectively on the following five-point scale…The physical attractiveness of each Add Health respondent is measured three times by three different interviewers over seven years.” Wait!

The attractiveness of the kids, who turned into teens, was assessed by the data takers! The same small set of data takers, about whom we know very little, except that they were residents of these fine United States.

Well, these are still measures of attractiveness, no matter how silly. So did Kanazawa rely on them? Well, no. He used these, and other variables, to compute a “latent ‘physical attractiveness factor’ by a statistical procedure called factor analysis. Factor analysis has the added advantage of eliminating all random measurement errors that are inherent in any scientific measurement.”

The last statement is mathematically false. In no way can factor analysis “eliminate” error of any kind. Further, there is no reason in the world to use factor analysis on data like this. There isn’t the space here to explain, but factor analysis is a dangerous tool in the hands of the statistically naive, because it can be, and has been, used to “prove” anything.

Kanazawa’s analysis is absurd. But why he thought the attractiveness ratings of a handful of American survey takers was of any interest is the real story. There are already calls for firing Kanazawa. Student groups protest “We support free speech and academic freedom,” but not for thee, Kanazawa-san.

Due to the unique lunacy which accompanies nearly all thinking on racial matters, it is impossible to say how the story will end. But we may safely predict that it won’t go well for Mr Kanazawa.

He should have stuck to writing more papers like “Why beautiful people are more intelligent” (Intelligence, 32 (3). pp. 227-243.)—or to proving that the converse is true.

Categories: Statistics

18 replies »

  1. I always wondered how researchers can assign a numerical value to somebodies verbal or written response, analyze the data using statistics and then claim something is proved when nothing was really ever measured. How do you measure attractivnes? What are the units of measurement? Have the Japenese invented the attractometer or repulseometer?

  2. “Objective attractiveness” is an interesting term. Why would it be less subjective than, say, “artistic taste”? With an 8000 variable database you could probably prove anything provided you are into things like p-factors.

    Not that I agree with his assessments but I’m willing to bet there would have been no outcry had he compared Japanese women instead of pointing to members of a Those-Who-Must-Never-Be-Dissed group or, even better, had said the reverse. Part of the problem is the fault of those who promote the substitution of statistical models for entrails for the purpose divination.

  3. You left out the most interesting part! What is ‘factor analysis’ and why it is wrong to use it in this case?

  4. Yet another conspiracy to feminize the male, by getting rid of the sweet hormone. Or is it the nasty Pharma companies peddling progesterone.

  5. Well, this is par for the course for pretty much everything coming under the heading of “evolutionary psychology.” Fantasy storytelling, masquerading as science . . .

  6. Kanazawa has also recently proved that the more status-conscious and arrogant you are, the greater your intelligence. — oh wait, he called the traits ‘liberal’ and ‘atheist’. My bad.

  7. Ray:

    Surprisingly enough, you can measure all sorts of “subjective” things with enough care. However, it requires careful development and validation of an ordinal measurement scale, some reliability studies, and some sort of calibration for the observers. Useful statistics can be done with well-chosen random samples and appropriate non-parametric statistics.

    Unfortunately, most folks just bang down a 5-point Likert scale, let the first available doofus tick off values willy-nilly, and then analyze the data using the z-test poorly remembered from Stats 101. (In large organizations, this is called “assessment.”)

    Me? I’d measure attractiveness using some sort of observable physical response. But I’m s-o-o-o Old School.


    Factor analysis is the evil twin of principal components analysis, which is merely a rotation of multivariate coordinate axes that moves from the data space to a similar space where the (orthogonal) axes are aligned in the directions of maximum variability (first axis most variation, second axis secondmost variation, etc.) By itself, this is a useful tool for controlling for collinearity in some regression models.

    Factor analysis takes this one (or two) steps further by allowing the axes to be further rotated in interesting ways, including ways that give sets of non-orthogonal axes. Still not too weird. BUT, at this point, researchers pounce upon these new axes (linear combinations of the measurement variables) and reify them with names like “charisma” or “attractiveness” or “intelligence.” You should see how a social science researcher’s eyes light up the first time a statistician shows him how this gag works (Hey, even statisticians enjoy the occasional Walk on the Wild Side).

    Social scientists eat this stuff with a fork and spoon, and thesis committees love candidates who sprinkle their defenses with multivariate statistical porn. A great, if overwrought, polemic on factor analysis is Stephen Jay Gould’s The Mismeasure of Man, although he eventually goes off the deep end ranting about Hernstein and Murray’s The Bell Curve, which only serves to detract from an otherwise interesting exposition.

    Finally, if you really want to go completely over the edge, check out the economists who dabble in structural equation modeling–Blinding ’em with Brilliance and Baffling ’em with Bullshit in a single explosion of quantitative wonkery.

  8. A proposition: if the population of black women in his sample had originated not in West Africa but instead in the Horn of Africa, people would have howled with laughter at the proposition that other people find the women, on average, to be ugly.

  9. I’ve always wondered if, for instance, Chinese men think that, as it might be, Jennifer Lopez is beautiful? Or do they think she would be but for that enormous nose? Sadly, I’ve learned nothing from Satoshi Kanazawa.

  10. His assumption, that people find black women less attractive than white women, doesn’t match what I’ve always been told by men: that black women are by and large gorgeous. As a matter of fact, a boss of mine (a Jewish guy who was brought up in the Bronx), considered Jamaican-Chinese women to be the most beautiful combination of all. I’ve also read articles — and my personal observation agrees — that black women age better than white or Asian women — they wrinkle less, and keep their muscle tone and other youthful characteristics longer. I worked with a black woman who I could have sworn was in her late twenties at the most, and it turned out she was nearly fifty, and had grown grandkids.

  11. Flawed premise, inappropriate statistical analyses, biased misinterpretation. But Kanazawa, regardless of his level of attractiveness and intelligence, gets attention.

  12. I do believe that black women do exert to darn much testosterone i am a witness

    Edited by the blog owner for style; the content was left to stand for itself; it is on wobbly legs.

  13. Gosh, within 15 secconds I found:

    Q Can You Measure Beuty?
    A Glad you asked…

    Dr Stephen Marquardt has developed a beauty mask and a grid of lines in the form of a face where the lines geometry is determined using the Golden Ratio. By overlaying this grid on a face and seeing how well individual’s features match up to the perfect Golden Ratio face, you can calculate how beautiful a person is.

    But I was looking for:
    Helen of Troy (from the Iliad) is widely known as “the face that launched a thousand ships”. Thus, 1 milliHelen is the amount of beauty needed to launch a single ship.

Leave a Reply

Your email address will not be published. Required fields are marked *