How Races and Religions Match in Online Dating

http://blog.okcupid.com/index.php/2009/09/29/how-races-and-religions-match-in-online-dating/

by noodle • 17 years ago

160 90 17 years ago

blog.okcupid.com

jacquesm6• 17 years ago

Non-hacking, race & religion in one article, that's got to be a record for flamebait.

showerst• 17 years ago

I'd argue that it's quite a hack -- Trying to discern an algorithmic approach to attraction via a large data set, and create a recommendation engine to a very fuzzy problem. =)

potatolicious4• 17 years ago

I'd argue that it's not at all a "hack" (I really dislike HN's tendency to equate everything to a hack), but nonetheless of significant interest to those of us interested in statistical trends in large web systems and social networks.

This kind of statistical kung-fu is well worth a look even if your own interests do not lie in dating sites.

jacquesm6• 17 years ago

Yes, but these are the results of the hack, not the hack itself. OkCupid has a nice thing going but this is not about their software, this is about trends they've picked out of their database of dating profiles.

And it's not like we haven't been here before:

http://news.ycombinator.com/item?id=822782

http://news.ycombinator.com/item?id=803603

http://news.ycombinator.com/item?id=66129

natrius2• 17 years ago

It's not flamebait at all. The data is all about people's responses to generic questions about what they're looking for in a date. They don't say "I don't want to date Muslims" and have that reflected in match scores. The match score is based on things like "I hate people who squeeze the toothpaste from the middle."

It's quite interesting.

jacquesm6• 17 years ago

To put data based on 'race & religion' out in the open to me is already a thing that is not excusable (besides the lousy definition of such things as race and religion, both of which are floating, not discrete).

There is a massive pre-selection problem here (people that frequent dating sites, rather than people in general) and the link to 'hacking' is a very tenuous one.

I've seen Michael Jackson referred to as a hacker here, so I probably shouldn't be too surprised.

This is not a hack, this is some statistical analysis with input data of dubious value, pretty pictures and all, drawing conclusions that are completely off the wall.

"Jews and Agnostics get along better with people"

"Muslims of both sexes and Hindu men get along worse"

"Catholics are more universally liked than Protestants"

Please, if that isn't flamebait I really don't know what is.

Where are the control groups, where are the standard deviations and so on.

Effectively this is a dating advert for programmers.

This is infotainment at it's best, statistical noise at its worst.

shalmanese• 17 years ago

One possible explanation for the low Hindu-Hindu match is that there are many websites dedicated to specifically Hindu dating which means that the Hindus using OkCupid are those who are specifically not looking for a Hindu date.

For other religions, such sites aren't nearly as prevalent.

alanthonyc3• 17 years ago

Not really.

The matches are based on answers users give to questions (both site designed and user designed).

When the blogger says they don't match, it means:

  - Person A: I like X

  - Person B: I don't like people who like X

Where X could mean ice cream, god or tooth-brushing. It has nothing to do with people selecting the race or religion of the people they think they want to date.

ecuzzillo• 17 years ago

Yeah, but if Hindu men like ice cream, and I don't want Hindu men, I might not like men who like ice cream either because a) it's what Hindu men do, or b) I don't like people who like ice cream to begin with, and I don't like Hindu men for that reason.

So it's not totally unrelated.

joeyo2• 17 years ago

Unless there are very few Hindus using the site, in which case there could be (in the extreme) a single bad hindu-hindu match that skews everything.

We need error bars.

gambling8nt• 17 years ago

I was going to post a comment directly disagreeing with this, and citing the large number of Jewish dating sites, but these are really only used by people on the more religious end of the spectrum, so it is possible that the effect is more prevalent in the Hindu-Hindu match calculations than in Jew-Jew match calculations. I would posit, however, that this effect, along with the self-selection bias of using a multi-cultural dating website, is almost certainly the cause for the tendency for higher matches towards non-relgious people.

neilk• 17 years ago

JDate and similar sites aren't exactly religious. There are lots of secular Jews who would prefer to date or marry another Jew, for ethnic, cultural, or national reasons.

btilly• 17 years ago

Given that strongly religious people (of whatever religion) had a harder time matching, removing the most religious Jews would improve the compatibility of the remaining Jews.

narag2• 17 years ago

... which means that the Hindus using OkCupid are those who are specifically not looking for a Hindu date.

Or maybe they are specifically looking for a non-Hindu date. I mean that they might be looking for a uncompromising relation, maybe easier with a person with different roots.

sharpn2• 17 years ago

I liked this on a few different levels:

-Hacking data to reveal the answers to important difficult questions.

-Exploring contentious issues in a neutral & intelligent way.

-Great example of how to attract potential customers by providing valuable information that also gently demonstrates competence related to your service & differentiates your offering from competitors.

I enjoyed reading this, I learnt something, and I thought favourably of the website based on this. No idea how good they are, but that post was a good marketing lesson.

vaksel2• 17 years ago

okcupid is pretty much considered the hip version of plenty of fish.

sharpn2• 17 years ago

I don't know what that means, but I did think this was a brilliant marketing post from a company, and an interesting read.

Goladus• 17 years ago

I don't think it is really as neutral as it seems. Match percentage is a very preliminary survey of how compatible two people are, and the article itself is fairly judgmental.

ilyak3• 17 years ago

Really-really interesting.

I'm curious why asian females scored worst of all when asian males scored descent.

hughprime4• 17 years ago

The differences on the race graph are so small as to be insignificant.

codexon4• 17 years ago

The reason is because at that point they already decided to interact with people of a different race.

If you want a more interesting result, take a look at my other comment for a diagram that shows response rates between different races. It illustrates a greater schism that one would come to expect.

hughprime4• 17 years ago

Come to think of it, the differences on the race graph are amazingly small, given how strongly race is correlated with certain religions -- (e.g. middle-eastern people are much more likely to be Muslim, and almost all Hindus are Indian.)

pyre3• 17 years ago

Keep in mind that the 'Match %' is based on OKCupid.com's algorithms that number crunch against quiz answers that the users answer. This has nothing to do with 'success rate' of users on the site hooking up with other users.

ilyak3• 17 years ago

Nope; the differences on the zodiac graph are.

Differences on the race graph want to tell us something. Not necessarily something relevant, but something, anyway.

codexon4• 17 years ago

They left out the axis on the "message reply rate" diagram as a teaser for the next blog post.

I just copy and pasted the image onto the axis. It shows that white and native American males are more likely to get responses than the other races.

http://www.codexon.com/temp/mrr.png

And I wrote a small analysis here if you can be bothered to take a look.

http://www.codexon.com/posts/okcupid-dating-statistics-shows...

camccann• 17 years ago

Assuming they didn't deliberately shuffle the ordering just to confound people like you.

Your results look depressingly plausible, though, so I'm guessing not.

potatolicious4• 17 years ago

As a single male Asian I am praying that the first row isn't me. But odds are it is anyway :( It certainly conforms with stereotypes about Asian girls...

codexon4• 17 years ago

I addressed those concerns in a comment on my website.

Gormo• 17 years ago

As an aside, I can't help but wonder if OKCupid intentionally left the reply rate chart unlabeled and unexplained, knowing that some people would post an analysis on their own blogs and generate some viral goodness.

Brilliant strategy if so.

roundsquare3• 17 years ago

Nice, thanks for this.

However, I think one critical component is missing (from what OKCupid posted, not from your analysis). Who sends the first message? For example, we are seeing a lot of green in the black female column. Does this mean that:

a) Black females reply back a lot?

b) Lots of men reply to black females?

These two results would mean entirely different things. The first means that black females aren't picky, the second that men like black females.

bdr• 17 years ago

I'm loving this blog. Have we seen match percent vs reply rate anywhere?

jrockway2• 17 years ago

I think this is probably too biased, as messages from users with a close match percentage are highlighted in your "inbox". The data might not be good enough to discern between "replied to message because it was highlighted" and "replied to message becuase match % was high".

(Personally, I have a feeling that "pretty profile picture" is more influential than "high match %". But of course, it's very hard to measure what one considers physically attractive.)

dmoney• 17 years ago

They need to make it a little more clear that they're not trying to say anything bad about a particular group.

alextp• 17 years ago

This is hilariously interesting. Data mining can always bring good insights if you have good data.

I wonder, though, (a) if most people search for similar people or for different people, (b) how well does the matching percentage really reflect a "liking probability" (ie, there is always some wishful thinking when designing date profiles, and some of that might be counter-productive, ruling out good matches).

elai• 17 years ago

It seems like recently immigrated groups who are not established and similar disadvantaged groups of people do worse than the established.

Chukwu5• 17 years ago

While I love the number crunching ethos, and think these guys are giving it one hell of a go - I will go on the record and say that they will ultimately fail because they are operating under many false premises, a few being:

1. People do not change their minds.

By algorithmically matching based on answers to questions, you must assume that those answers are meaningful. And meaningful means CONSISTENT. And people are NOT consistent. Especially on complex issues; in fact there are whole professions (sales (insurance, real estate, car, etc) comes immediately to mind) where success is based on one's ability to get people to change their minds. (That example is offered as evidence, not proof - but proof probably isnt a complicated exercise)

2. Like Thinkers make Great Daters

Assuming you are even GETTING like thinkers, (which is tenuous at best) where is it written that like thinkers make great companions? You don't want to have absolutely nothing in common, (at least a common language is necessary) but it isn't linear and i would be surprised if there were any correlation let alone one that was non-asymptotic. Simply - more like thoughts does not entail more compatibility.

3. Looks aren't EVERYTHING

Just based on sheer observation - this is not the case. It is unfortunate, and it speaks badly about human nature, but most human beings make judgments based on appearances. And this "appearance crutch" is only exacerbated when it comes to looking for someone to lay with. How many single, male, bleeding heart liberals would take a swing at Sarah Palin if given an iota of an opportunity? Again - offered as evidence, not proof.

philwelch3• 17 years ago

"By algorithmically matching based on answers to questions, you must assume that those answers are meaningful. And meaningful means CONSISTENT. And people are NOT consistent."

In practice, the user base for OkCupid will change. And old users will reanswer questions sometimes too.

"Assuming you are even GETTING like thinkers, (which is tenuous at best) where is it written that like thinkers make great companions?"

That's not how it works. It's not as simple as "answer match questions and we hook you up with people who answered the same way". You can specify which answers your ideal match would give, even if those answers conflict with the answers you yourself give.

"most human beings make judgments based on appearances"

Which is why you upload photos to OkCupid instead of just using match percentages.

Chukwu5• 17 years ago

Uploading photos isn't what supposedly sets okcupid apart, right? They aren't using physical attractiveness in their matching algorithms and that is the point I was trying to make.

Your other point has been addressed above; people don't know what they want. Asking them what they want their "ideal match" to answer is as foolhardy as asking them to describe their "ideal mate" neuron by neuron.

potatolicious4• 17 years ago

"where is it written that like thinkers make great companions?"

OKCupid does not judge match percentage based on similar answers. For every question on the system it will ask for your answer and your ideal mate's answer. The system knows that some things are important to some people - but maybe not in reverse.

Chukwu5• 17 years ago

4. People Know what they want

Clearly they don't - otherwise they wouldn't be on okcupid in the first place.

CatBaby• 17 years ago

I can only imagine what else OkCupid's data would reveal in adept hands. I could probably write a book if they sent me a spreadsheet.

papersmith• 17 years ago

Am I the only one who had a good chuckle at the zodiac grid?

electromagnetic• 17 years ago

I was laughing before that with this:

> ". . . I point this out now so that, below, when we claim that Jewish women are easier to get along with than Christians, you don’t blame us, you blame Jesus."

padmapper2• 17 years ago

Nope, hilarious.

And that's the first really solid evidence that I've seen shared that zodiac signs are utter BS, which makes me doubly grateful to them. Something to point to when someone says it can't be proved otherwise...

ZitchDog• 17 years ago

The best evidence against astrology I've heard came from my college astronomy professor. Basically, due to the precession of the earth on its axis, the constellations are visible at completely different times than they were when astrology was invented. So even if it wasn't BS back then, it sure is now.

http://www.livescience.com/strangenews/your-astronomical-sig...

kingkawn2• 17 years ago

go jews!

jrockway2• 17 years ago

The atheists and agnostics beat the jews, however.

kingkawn2• 17 years ago

actually because of the response of muslim women to jews being more positive than to ath/agn, I think Jewish men overall come out on top.

Combine that with Jewish men who ARE atheists, which is a good number of em, and you've got a super-match. Which all goes to prove what my mother always told me; "I love you."

dkokelley• 17 years ago

Favorite line of the article: "If religion is a minefield, then race is a field that’s just one giant mine"

The article was not quite as statistically sound as it could be, but at least they aren't trying to pass it off as statistical data.

SeanLuke• 17 years ago

So many statistics.

So little statistical significance testing.

This tells me much about this service.

lamby• 17 years ago

I'm not sure it's valid to infer that their entire service is poor from these series of rather niche blog posts.

(Although saying that, I wonder if this series of theirs is to attract press attention? If so, then it would be nice to see some significance data.)

GrandMasterBirt• 17 years ago

Bullshit, us pastaferians are not represented in the religious statistics! This is heracy!

On the other hand I am glad I'm not Aquarius because statistically speaking they are insignificantly less compatible with an Aquarius. His noodelieness has blessed me well!

RAmen!

stakent• 17 years ago

Brilliant.

Free publicity. Great lesson in PR. Thank you.

ALee• 17 years ago

The article fails to mention statistical significance of race versus religion?

Because this data makes sense, I hesitate to protest. Nevertheless, I'd be interested to see if race is statistically significant versus religion.

known• 17 years ago

Endogamous marriages in Ashkenazi Jews have resulted in serious genetic disorders and hereditary diseases.

http://en.wikipedia.org/wiki/Ashkenazi_Jews#Specific_disease...

stevejohnson• 17 years ago

Could you elaborate on how this relates to the article? I'm sure you have an interesting point to make, but I find it difficult to extrapolate.

How Races and Religions Match in Online Dating

Related Stories