homophily in sexual networks

Today in my social networks class, we covered Bearman, Moody and Stovel’s 2004 AJS paper. Virtually every time i have heard that paper presented (by Peter or Jim, i’ve never actually met Stovel in person), and the couple of times i’ve talked about it in class, some variant of the following has been part of that discussion:

So, we know that there are any number of factors on which homophily operates in the formation of social networks in general. And in the particular case of romantic networks, the story’s much the same. However, there is one particular trait on which homophily is a much more robust predictor of a relationship than any others. Any guesses what that might be?

At which point the audience guesses a long list of traits, and virtually never hits #1 on the head (my class today got it on about guess 14, when i made it multiple choice from about 15 options). So, i pose the question to you, fair reader. What do you think it is?* If you’ve read the footnotes/appendices to the paper closely you can likely guess, but i’m not sure whether its actually explicitly stated anywhere in that particular paper or not.

*If no one gets it, i’ll post the answer in the comments in a day or two.


9 Responses to homophily in sexual networks

  1. eduardo says:

    Please define “robust relationship”.

  2. Michael Bishop says:

    @eduardo, that is “robust predictor of a relationship” and I agree that it would help to have it defined. I assume the outcome of interest is the binary variable relationship/no-relationship. Do you mean, highest standardized regression coefficient when all fifteen predictors are included, or removing it from the list of fifteen results in the greatest reduction in adjusted R^2, or something else.

  3. jimi adams says:

    all i mean is the one most able to withstand the introduction of other controls.

  4. Andrei Boutyline says:

    My guess is music tastes. They are emotive, finely differentiated, take effort to acquire, and are right for the age group.

  5. Michael Bishop says:

    I would guess being of opposite sex, but that might be represented as two variables rather than one, depending on the specification. Someone would have guessed it if it was on your list of 15. What do you have on there? Age, grade, race/ethnicity, religion, being in the same classes, having mutual friends, having a similar number of friends (indegree/outdegree), SES, GPA, alcohol/drug/cigarette use, vocabulary test score. Maybe its GPA. I never read the paper carefully but I am familiar with the data.

  6. jimi adams says:

    Yes, you are correct in which is actually the single strongest predictor of high school romantic relationships (opposite sex pairs), but that would be heterophily, not homophily.

    My question was about the strongest homophily effect. In terms of homophily – your list includes most of the “prime suspects” including the #1. But it hasn’t been guessed yet.

  7. jimi adams says:

    i seem to have gotten more interaction on this offline than on, but it all still confirms that people remain unable to nab this one, even with a “people typically miss it” prod.

    The answer is smoking. While other estimates waver a bit with the introduction of seemingly reasonable controls, this one remains strong (i haven’t done the analysis myself, but it’s what i’ve heard reported several different times. Also, i should note that this is not specific to the one focal school in that article, but is true when all of the saturated schools are examined, if i recall correctly.)

    But, my favorite part of it all is their – oh so scientific – explanation of that finding: “…perhaps because only smokers can tolerate kissing smokers” (p.68).

  8. Michael Bishop says:

    In hindsight, this makes a lot of sense. Smoking cigarettes sends pretty strong social signals, and it is probably less highly correlated with the other variables likely to be in a regression.

    The “stinky smoker kissing” hypothesis could be tested by comparing it to friend-nomination, or taking into consideration frequency of smoking.

  9. andreiboutyline says:

    Makes a lot of sense to me too. I would interpret it in terms of smokers having to frequently mill around together in areas where smoking is permitted. It’s also an inherently social activity–smokers bum cigarettes and matches off each other, etc. A lot of ex-smokers reminisce to it as an easy way of meeting people. Like Mike, I would like to compare it to friend nomination. I bet it’s equally high.

    P.S., I remember someone (Wasserman?) talking about how smokers often have the highest centrality in social networks inside organizations. I think he was using it as a cautionary tale against reading too much into centrality indices.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: