Race and attraction

For many years the online dating site OkCupid produced the OkTrends blog, where they analysed their data to explore people's dating habits. In 2009 they looked at racial preferences — the original blog has disappeared, but a copy of the article can be found here.

The article analysed the data through the lens of American racial politics, and, perhaps unsurprisingly, concluded that their users are racist. However, there are a few problems with their analysis.

First, they measured preferences by looking at how frequently people responded to initial messages. However, this measures not just the preferences of the recipient, but the judgement of the sender. For example, a sender who mostly contacts attractive people will almost certainly get a lower response rate than one who contacts unattractive ones. Does judgement vary by race? We have no way of knowing.

Second, much of their analysis is based on men responding to women's messages. But in online dating — as in the real world — the overwhelming majority of initial contacts are made by men. It is hard to imagine that a popular woman, her inbox full of messages from suitors, would initiate contact with yet another man. Which means that the women who do send initial messages may not be a representative sample, and may be skewed to the unpopular end of the spectrum.

A better analysis would have looked at rates of initial man-to-woman contact, broken down by race, and then the response rates for those messages. That would show whether lower response rates are simply a result of unrealistically high rates of initial contact, rather than racial preferences. And the breakdown of the initial man-to-woman messages would provide a less-skewed idea of male racial preferences.

But we can only analyse the data we're given.

All things being equal

For people's preferences to be considered racist, they would need to vary by race only when all other attributes are equal. But in reality, many attributes that affect attractiveness — height, weight, education, income, etc — vary with race.

Let's start by looking at how often men replied to women, broken down by the woman's race.

• Middle Eastern 49.5%
• Pacific Islander 46.0%
• Other 44.4%
• Asian 43.7%
• Indian 42.7%
• Hispanic 42.5%
• Native American 42.3%
• White 42.1%
• Black 34.4%

Many studies have shown that men like slim women, preferably without children. So could the rankings be explained by their weight and parental status?

For weight, ideally we'd like to find the Body Mass Index (BMI) for each of the nine races, but that doesn't seem to be available. The closest proxy is female obesity rates for some of the races in the US.

• Asian (Asian, Indian, Pacific Islander) 17.2%
• White (White, Middle Eastern, Native American) 39.8%
• Hispanic 43.7
• Black 56.9

The CDC's four racial categories don't directly match OkCupid's, and the Asian and White categories include a diverse range of people, but the response rates do seem to correlate with obesity.

When it comes to children, it isn't known what fraction of American women are single mothers, but the United States Census Bureau does publish the percentage of children living only with their mother.

• Asian 7.8%
• White 13.4%
• Hispanic 24%
• Black 46%

The order for single-mother status is the same as for the obesity numbers, and again seems related to response rates.

Now let's look at how often women reply to men of different races.

• White 29.2%
• Native American 27.8%
• Other 26.8%
• Middle Eastern 25.7%
• Pacific Islander 24.6%
• Hispanic 23.1%
• Asian 22.2%
• Black 21.7%
• Indian 20.8%

Might the rankings be explained by the height of the men? Let's look at the height data for American men.

• White (White, Middle Eastern, Native American) 177.4cm
• Black 175.5cm
• Asian (Asian, Indian, Pacific Islander) 169.7cm
• Hispanic 169.5cm

Height explains the high ranking for white men, but not the low ranking for black men, and Hispanic men seem to rank a bit higher than you'd expect. So it doesn't explain everything.

Maybe income? Here's the median income in the US for men of different races.

• Asian $75,671
• White $65,208
• Black $45,644
• Hispanic $41,519

Given their high income, Asian men (i.e. Asian, Indian, Pacific Islander) do pretty badly. And, once again, Hispanic men do a bit better than expected.

Summary

Assuming the male preference numbers are valid — and given the relative rarity of women initiating contact with men, they may not be — the rankings for women can still be explained by rates of obesity and single motherhood. So the effects of racial preferences are likely to be small — which is backed up by speed dating studies.

The rankings for men are more ambiguous. A combination of height and income data explain most of it, but Asian and black men are slightly less popular than expected, and Hispanic men slightly more. This is also reflected in speed dating research, where women were more likely than men to both express a racial preference in advance and to act on it — although the effect was weak.

So, despite the conclusions of the OkCupid article, race is probably not a significant factor in who Americans find attractive. Other, more prosaic, attributes, such as height, weight, and income, play a much larger role.