[This is a commentary on racial dating trends from dating site okcupid.]
As far as I’ve been able to tell, America has three Big Ideas about race relations:
and #3, which kind of follows from the other two: American race relations suck.
Obviously, this post isn’t going to change that, but we are going to do something a little different, a thought experiment. Since most thinking about race hinges on the fact that one particular race predominates, what if, using statistical models, you could make that predominance disappear?
Right now, even on a per capita basis, whites gets more messages than non-whites on OkCupid:
But how much of this reflects whites’ “popularity” and how much reflects their population? What would this graph look like in a racially-balanced world? How would people’s choices change, if the people to choose from weren’t mostly of one race?
This post is based on 82 million messages sent on OkCupid over the last few months. Here’s what we found.
Our data set
OkCupid is actually less white than the rest of the U.S. Internet, though it’s hard to make direct group-to-group comparisons because Quantcast (our national demographic source) doesn’t have multiracial data.
For this post, to keep the discussion streamlined and accurate, I’ll only include OkCupid’s largest single-race groups: whites, asians, blacks, and latinos.
Where To Start
When you look at the messages sent by each of our four biggest racial groups, it’s actually kind of amazing how much white people are flooding the market. In raw terms, everybody, whites and non-whites alike, sends white people most of their messages. Broken down by the race of the sender and the recipient, here’s what we have:
The above chart actually undersells how much attention whites get, because I assigned equal widths to the four vertical bars representing the message senders. To properly present reality, we should adjust those widths according the size of each population, like so:
You can think of that whole square up there as the dating “pie”, and, as it stands, whites get 89% of it. But let’s look deeper into the numbers. Thus, our experiment begins.
Here’s the data for Asians.
- As the chart above shows, Asians send about 3 times as many messages to whites as they send to other Asians:
- However, as we saw before, whites outnumber Asians 19:1 on the site:
- Putting this messaging and population information together, we get these ratios:
- Well, what if we rescale these ratios? What if there were 10 whites and 10 Asians instead? How many message would go to each?
- Our experiment tells us that, given equal numbers, Asians would actually overwhelming prefer to message other Asians.
It turns out that if you bust out your mechanical pencil and use it to punch the “on” button on your laptop and then, with that computer, extend the same kind of analysis to the other groups, you get a fresh sense of American race preferences.
In a world without so many white people, where everyone has an equal choice, that first graph we saw—the average monthly messages people get—would look like this:
And, going for a little more detail, this is how it would break out by age.
I built an interactive visualisation of how the Who People Are Messaging chart we were looking at before changes when there aren’t so many white people. You can toggle between states and get a sense for yourself.
supposing racial bias: is like today—the chart will reflect people’s current racial messaging preferences didn’t exist—the chart will suppose people are race-blind when deciding who to message supposing racial composition: is like today—the chart will reflect the population’s current racial composition is equal—the chart will suppose there are equal numbers of each race Total area for each group—a measure of their relative popularity:
Here’s the by-age analysis for each of the four groups:
Back to reality
“What ifs” are an important part of any researcher’s toolkit:
but “what ifs” are only worthwhile if they tell you something relevant in the real world. And here’s the value in our thought experiment.
The kind of messaging imbalance that currently exists has observable effects on how people think about race and dating. Search for “interracial couple” on a stock photo site: you’ll find a rainbow of Asians, Latinos, Black, and Indians, all hanging out with their white significant others. Or just start typing into Google and see how it completes your sentence:
But as we’ve seen, in a scenario where people have an equal choice of messaging options, things are very different.
In other words, person for person:
Asians are the most desired racial group in the country.Then Latinos. Then whites, sort of. As we’re seeing it now, the data is being distorted: a huge part of the country is white, and white people mostly like to talk amongst themselves. Intentionally or not, minorities are left out in the cold.
Nonetheless, people prefer their own race
Given equal choice, every race strongly prefers itself:
And white people actually prefer themselves the least, but right now there’s just so many of them. It’s interesting to think what things would be like if the shoe was on the other foot; if another race outnumbered whites 19:1. We ran a simulation with Asians in whites’ role as the dominant group and everyone else was really on the outside looking in then.
You can actually see groups’ insularity in real life when you look at cities with larger non-white user bases. Minorities tend to become more and more inward-looking as their numbers increase. The dots on the graph below represent the 150 towns with our largest black populations:
For every 1% increase in the black population, blacks send 2.5% more messages to each other. In Baltimore, for example, 1 in 5 users is black, but half the messages from black people go to other black people. The rate is over twice what you’d expect. Other groups show a similar pattern.
For white people, there are also some places where a relatively smaller population still sends a disproportionately high number of messages to themselves—aka the most racist towns in America. The white people in Greenville, MS, for instance, comprise only 59% of our user base there. However, 95% of their messages go to whites; that’s 14× the expected rate.
Cities with the most white-loving white people
#1 Greenville, MS
#2 Jackson, MS
#3 Montgomery, AL
#4 Memphis, TN
#5 Atlanta, GA
The most segregated city north of the Mason-Dixon line is a surprising one: New York. White New Yorkers message other white New Yorkers at about 2× the national rate, probably about Animal Collective.
Soon, there won’t be so many white people
Finally, I just want to point out that the hypothetical we chose for our title isn’t going to be a hypothetical for much longer. The Census Bureau has concluded that around 2050, whites will no longer be the majority in the U.S.
This is from their website:
Along with this news, I think there’s an assumption that at some point all the races will just kind of come together as one, like during Michael Jackson’s “Black Or White” video or like during a lawsuit against the estate of Michael Jackson.
The data we’re seeing, however, just doesn’t support a post-racial future, because even as the races mingle more, people still like to date someone who looks like they do. Asians strongly prefer Asians; Latinos, Latinos, and so on.
Even when white people aren’t the majority, society will be as divided as ever.Endnote: reading level analysis
We analysed the grade-equivalent reading level of each message in our dataset, and here’s a table of how it changes, race to race, message to message.
It seems that generally, people of all races write down to blacks and Latinos and up to Asians and whites. This is a pretty crazy result: proof that race not only affects the quantity of the contacts between people, but the quality as well. For example, the average black person writes at a level almost one full grade-level higher when writing to a white person than when writing to another black person.
Business Insider Emails & Alerts
Site highlights each day to your inbox.