Race and Attraction, 2009 – 2014

September 10th, 2014 by Christian Rudder

We looked at race in one of our very first posts, and today I’d like to revisit the topic with fresh data. This article folds in millions of person-to-person interactions, what one human being thinks of another. As such, it’s different from a look at, say, unemployment numbers or test scores. So much data on race gives you people vs. the system. OkCupid’s gives you people vs. people. Basically, the site’s ten-year history has been one long episode of Judge Judy. Here’s the verdict on race.

Way Back in 2009

Five years ago, the basics of race and attraction on OkCupid looked like this:

– non-black men applied a penalty to black women
– while black men showed little racial preference either way

– all women preferred men of their own race
– but they otherwise penalized both Asian and black men

Here’s how the exact person-to-person numbers shook out:

The values in these tables are “preference vs. the average.” Think of them as how people weigh race in deciding attraction. So, for example, in the bottom-right corner of the lower table, you see that white women think white men are 17% more attractive than the average guy. Move one square to the left, and you see that they think Latinos are 1% above average, and so on.

The color is there to make the big trends easy to see.

Has Anything Changed?

In some ways, no. OkCupid users are certainly no more open-minded than they used to be. If anything, racial bias has intensified a bit. Here are the numbers for 2009 till now—use the slider to move through time. Like a javascript DeLorean.


These numbers reflect different people year-to-year. Together the charts fold in data from some 25 million accounts.

One interesting thing is to compare what you see above with what those same users have told us about their racial attitudes. Answers to match questions have been getting significantly less biased over time:

And yet the underlying behavior has stayed the same.

You can use other online data to see this split personality play out elsewhere. The night Obama was first elected was a moment of catharsis. It really felt like something had changed about the way America perceived and thought about race, and for at least that brief moment, the nation appeared united. No less than Karl Rove captured the moment well: “an African-American candidate who was aspirational and inspirational…is very powerful. It’s a night for our country to celebrate, and for the world to celebrate.”

Meanwhile, that same evening, American Google Searches for the word “nigger” hit an all-time high.

 I learned this from the work of Seth Stephens-Davidowitz, a researcher at Google.

. . .

Anyhow, whenever I talk about this data (or the race data in my book, Dataclysm) a few questions always seem to come up. So I figured I’d just answer them head-on here.

Q: Are people on OkCupid just racist?

No. I mean, not any more than anywhere else. All the dating data I’ve seen fits OkCupid’s pattern: black people and Asian men get short shrift. For example, below are the numbers from DateHookup, a site that we acquired a few years ago (but that still operates independently.) DateHookup has a distinct userbase, a distinct user acquisition model, a distinct interface, yet their data reflects the same basic biases:

While OkCupid is large enough that its demographics reflect the general Internet-using public, DateHookup is a niche site particularly popular with Latinos and blacks (those groups comprise 13% and 20% of the site, respectively.) Other sites in our portfolio, with still different demographics and business models, show the same attraction patterns.

Q: Is it possible that some small number of users is throwing off the averages?

These biases are pervasive in the data I’ve seen. For example, 82% of non-black men on OkCupid show some bias against black women. And, similarly, it’s not outliers among the women driving the results. It’s a wholesale phenomenon: the ratings for an entire population are shifted down.

Q: Are you saying that because I prefer to date [whatever race], I’m a racist?

No. An individual can’t really control who turns them on—and almost everyone has a “type,” one way or another. However, I do think the trend—that fact that race is a sexual factor for so many individuals, and in such a consistent way—raises deeper questions. See below.

Q: All this data is from a dating site. What does that have to do with my life?

There are many situations that might not be explicitly romantic, but are nonetheless a lot like a first date. A job interview. Trying to rent an apartment. When you meet your freshman roommates. Anytime you’re trying to make an impression on a stranger. And science has long known that bonuses accrue to beautiful people: they have better outcomes at work and at school, more success with juries, even live longer lives, and so on. In short, “beautiful people” receive a lot of the same built-in benefits in our society that white people do.

I think that’s no coincidence. Beauty is a cultural idea as much as a physical one, and the standard is of course set by the dominant culture. I believe that’s what you see in the data here. One interesting thing about OkCupid’s interface is that we allow people to select more than one race, so you can actually look at people who’ve combined “white” with another racial description. Adding “whiteness” always helps your rating! In fact it goes a long way towards undoing any bias against you.

So, yes, this is all just dating data, but it shows who and what we define as beautiful. And that’s something that affects everyone.

Leave a Reply


8 Responses to “Race and Attraction, 2009 – 2014”

  1. Ehizzle says:

    What about same-sex ratings?

  2. Russ in TX says:

    I see a log of the bias towards black women skewed b/c of a preponderance of high weight. Black gals who stay in shape, I see dudes falling all OVER them regardless of race.

  3. John says:

    You have failed to control for other variables that are highly correlated with race. For example, have you removed the effects of body type and income level from these interactions?

    No, a women is going to be influenced by income. A man is going to be influenced by body type. You analysis of the data would ve more accurate, if the rating was only based on picture.

    However, that approach would still cause people who dislike certain body types to be described as having a racial preference, when in fact they are driven by other qualities apparent in the photograph.

    Honestly, this is bad work, since it does not adjust for all the social status biases that are communicated through the okcupid profile.

  4. Tanner says:

    Interesting… I wonder how such data changes for different economic classes and intelligence levels.

  5. Regular Guy says:

    This whole post seems racist. How about you normalize for attractiveness? What if a subgroup of users are all fat or have poor photography skills? This data is useless.

  6. Bill says:

    I think its funny that all of the comments so far are pretty much racist against black women and suggesting that they are, by and large, fat, poor, and unattractive.

    Way to be terrible people OKC users.

  7. TJ Leeland says:

    For those who are saying that they didn’t control for things like body-type, i’d say that’s moot. They’re not looking at the attractiveness of different races, they’re looking at real-world examples. So if in the real world black women are more likely overweight (as someone below suggested), then that’s going to be reflected here.

    I’d say controlling for income, weight, and other factors would be valid, but would be a different study.

  8. Kay says:

    Basically white women and the most desirable while black women are the least and white men are winning across the board…bruh!