Tuesday, October 13, 2009

Online dating and race at OKCupid

This recent post on the OKCupid blog has been widely linked, but since I covered this previously, at some length (also related), I feel obligated to post it.

If you have any interest in this subject, you'll no doubt read the whole thing, but if you need any further suasion, here are the key figures. Reply rates by race when men send messages...

...and reply rates by race when women send messages...

Read it and weep. (Or rejoice, I suppose, if you're in the favored classes and you guiltlessly enjoy racial privilege.)

BTW yes, they controlled for algorithmic match rate, which is essentially flat across all race/gender combinations (see the "Match % by Race" figure in the original post).

As usual, I find the result itself sad — I think this goes far beyond a moderate amount of understandable homophily (look at the diagonals!) — but what I find much more sad is the degree of self-deception that people engage in when discussing these results. Unfairness is annoying; deception about unfairness really brings on the facepalm.

See, for example, the 459-comment Metafilter thread, where about half the educated, literate, liberal MeFi crowd doesn't seem to get the following simple proposition: although diversity of aesthetic preferences, including preference for racially marked features, may be a simple personal choice, systematic statistical skew in aesthetic preferences across a large population strongly indicates socialization to racially biased standards of attractiveness.

Note, by the way, that racial preferences don't mean merely visual discrimination. The degree of racial discrimination is considerably stronger and more widespread for women than men, even though (as folk wisdom has it) women are less visually focused than men. (Personally, I think folk wisdom overstates this sex difference, but I do think it's real.) I think this implies that part of the racial discrimination effect — possibly even the dominant part — is due to people making assumptions about personality or character based on race, rather than preference for a certain physical appearance alone. Which is even more damning.

Standard caveats w.r.t. all such social science analyses apply blah blah blah. On the other hand, the fact that this result essentially replicates, at finer granularity, the results of the Hitsch et al. study I blogged previously, as well as anecdotal evidence gathered from friends and acquaintances, does not incline me to skepticism.


  1. I also made a post on my weblog about the OkCupid findings, which was something I knew about for a long time and was working against me. I agree that the big problem isn't the racial bias, but the deception involved.

  2. After I read the article, I wondered why the article was couched as yet another "Black women as the world's worst victim" piece. Clearly, according to OKCupid's own numbers, it is black men who come off as being in the worst dating situation, not black women. But then, they couldn't grab headlines and the attention of black women if they were to say that could they? Or maybe the owners of the OKCupid website (and the editors and managers of American media) simply don't give a sh*t enough about black men to even notice or care.

    Let's go back and do the numbers again, this time comparing OKCupid's figures for the response rates black men and women achieve when sending messages to potential mates of various racial and ethnic groups:

    Native American - 41% / Black - 28%
    Other - 41% / Native American - 27%
    Middle Eastern - 40% / Pacific Islanders - 25%
    Black - 37% / Other - 24%
    Indian - 37% / Indian - 21%
    Hispanic - 36% / Middle Eastern - 21%
    Pacific Islanders - 32% / White - 21%
    White - 32% / Hispanic - 19%
    Asian - 31% / Asian - 17%

    In other words, even the LOWEST response rate received by black women from various racial and ethnic groups (31% from Asian men) was STILL higher than the HIGHEST response rate received by black men (28% from black women). But apparently the dating plight of black men is of no interest to OKCupid. Or, even worse, they assumed that the dating plight of black men would be of no interest to their readers- white, black or whatever. AND THEY WOULD PROBABLY BE RIGHT IN THAT UNFORTUNATE ASSUMPTION. After all, what is the value of black men in American society? Our value is close to ZERO (except, of course, when it comes to sports and entertainment). We can't even claim "worst position" even in situations that clearly demonstrate that we are in the "worst position."

    OKCupid engaged in what the magicians call "misdirection." They suckered the public into looking in one direction when they should have been looking in another direction. American media does this all the time when it comes to the plight of black men as compared to that of black women. For example, based on the media stories to come out in past years about rates of HIV/AIDS you would be right to assume that black women suffer from higher rates of HIV/AIDS than black men-- and you would be wrong. In fact, the rate for black women is 60.6 (cases per 100,000), while that of black men is 136.8 (see the CDC website and this page in particular: http://twurl.nl/b6qxvp).

    Moreover, the folks at OKCupid even decided to point out that black women were not even valued by black men, when it comes to the responses they received from potential mates ("Essentially every race—including other blacks—singles them out for the cold shoulder.") Yet a perusal of OKCupid's own numbers would show that when it comes to responding to potential mates, black men respond LEAST to black women AND black women respond LEAST to black men. The rate of response from black women to black men is higher than vice-versa, but black men and women are BOTH LAST on each other's agendas. See for yourself:

    Pacific Islanders - 39% / Pacific Islander - 57%
    White - 38% / Asian - 55%
    Middle Eastern - 37% / Middle Eastern - 55%
    Native American - 34% / Other - 52%
    Asian - 34% / Indian - 51%
    Indian - 34% / White - 51%
    Other - 32% / Native American - 50%
    Hispanic - 31% / Hispanic - 46%
    Black - 28% / Black - 37%

  3. I won't deny that things are pretty dire for black men, but I think your analysis on this particular study needs to be refined with some context.

    First, comparing response rates for black men vs. black women alone is not going to yield much insight. Women overall simply receive responses at a much higher rate than men. See the labels for totals below the lower-right corner of each square: women get a 42.0% response rate overall, vs. 27.6% for men. So all else being equal, you'd expect that any given group of men would receive a lower response rate than any given group of women. And indeed that's very nearly the case: the highest response rate for any male subgroup (Pacific Islanders contacting black women) is 39%; but *almost every square* in the second figure is greater than 39%.

    This should be no surprise to anyone familiar with how dating works. Men usually initiate contact, and women are usually expected to select from among the offers they receive. You'd expect women to receive many more messages, and therefore to respond to a smaller fraction of messages.

    To put it another way, the lowest response rate for white women contacting any other group (41%) is still higher than the highest response rate for white men contacting any other group (38%); is this evidence that white men are especially disadvantaged?

    Second, although the rate at which black men receive replies is low (21.7% --- see the totals to the right of each row), it is not even the lowest rate overall (20.8%, for Indian males). Nor is it especially an outlier compared to several other ethnic groups --- Indian, Asian, and Latino are all in the same ballpark of low 20's in response rate. In fact, if you sort the totals for men from worst to best, you get a fairly smooth progression from Indian (20.8) -> Black (21.7) -> Asian (22.2) -> Hispanic (23.1) -> Pacific Islander (24.6) -> Middle Eastern (25.7) -> Other (26.8) -> Native American (27.8) -> White (29.2).

    The situation for black women, on the other hand, is essentially unique. At 34.3%, they receive responses at a rate 7.8% below the next-lowest group, with all the others clustered within a 7.4% band (ranging from 42.1-49.5%). In other words, the difference between black women and the next-worst-off group is greater than the variance among all other groups combined. That is really a striking outlier.

    As for high rate at which black women reject men of the same race, that's a phenomenon they share with over half the female groups studied. Every red box on the top-left to lower-right diagonal indicates a group which exhibits an anti-preference for the same race. This includes Asian, black, Latino, Indian, and Middle Eastern women.

    Again, I won't deny that black men occupy a devalued cultural position in American culture at large, and I can't comment on stuff like the media coverage of HIV because I haven't been following those issues. But I think the OKCupid people have done a pretty good job summarizing these particular results.