Skip to content

Controversy over social contagion

Dan Engber points me to an excellent pair of articles by Dave Johns, reporting on the research that’s appeared in the last few years from Nicholas Christakis and James Fowler on social contagion–the finding that being fat is contagions, and so forth.

More precisely, Christakis and Fowler reanalyzed data from the Framingham heart study–a large longitudinal study that included medical records on thousands of people and, crucially, some information on friendships among the participants–and found that, when a person gained weight, his or her friends were likely to gain weight also. Apparently they have found similar patterns for sleep problems, drug use, depression, and divorce. And others have used the same sort of analysis to find contagion in acne, headaches, and height. Huh? No, I’m not kidding, but these last three were used in an attempt to debunk the Christakis and Fowler findings: if their method finds contagion in height, then maybe this isn’t contagion at all, but just some sort of correlation. Maybe fat people just happen to know other fat people. Christakis and Fowler did address this objection in their research articles, but the current controversy is over whether their statistical adjustment did everything they said it did.

So this moves from a gee-whiz science-is-cool study to a more interesting is-it-right-or-is-it-b.s. debate.

This is a controversy that I’m pretty well-qualified to comment on, except that I haven’t taken a careful read of any of the articles under discussion. I could imagine a world in which I was a science journalist and was paid to definitively track these things down. That would be fun, but it’s not the job I actually have.

Actually, I think I blurbed Christakis and Fowler’s “Connected” when it came out–I liked the book a lot–but, as you can see, I was careful when summarizing their claims:

Christakis and Fowler are best known for their work connecting social networks and epidemiology, in particular the fact that obese people are more likely to have obese friends. As one wag put it, they find that obesity is environmental and voting is genetic. I guess that sort of interpretation is the inevitable outcome of man-bites-dog reporting, with the real story being that obesity is more determined by social behavior than we might have thought, while voting behavior is more tied to genes than we might have thought.

If Christakis and Fowler did make a mistake, I’d cut them some slack–it’s not like they proved a false theorem or anything. I expect that tonight, while I sleep, these guys are hard at work on the computer, running all sorts of regressions and making all sorts of graphs to try to figure out what their data are saying.

A historical perspective

One thing that was cool about Dave Johns’s article was that he didn’t just write about the current controversy. He also put it in historical context of how contagion has been viewed in earlier periods.

In one way, though, I think Johns goes too far in his criticism. He writes:

As Fowler told Stephen Colbert in a January interview, the research suggests that people don’t really make individual decisions at all but, instead, function as part of a “human superorganism”–like a herd of buffalo or a flock of birds.

Johns describes this as “an assault on a lot of traditional social science,” but I wouldn’t quite say that. There are some behaviors that are really hard to change individually but do seem to change en masse. Consider smoking or even something simple such as seat-belt use. The superorganism or herd-of-animals analogy doesn’t seem too far off-base to me, especially when modified to allow for social networks, so that different countries, for example, can have similar trends to the extent that there are many links between them. Friends and family aren’t the only things–the news media (and even bloggers?) play a role too–but the general idea of studying network patterns makes sense to me. I guess what I’m saying is that the network-research paradigm is more general and important than any specific claims about contagion. John talks about homophily (“birds of a feather flock together”) as an alternative to social contagion, but homophily is itself a consequence of network behavior, no? You can’t “flock with” someone if you don’t know where he or she is.

Anyway, it’s good stuff. A great piece of scientific reporting. I’m looking forward to the follow-up in a year.

P.S. Apparently Fowler was on the Colbert Report. Perhaps Colbert can have another show pitting him against one of the researchers who claims that his findings are all artifacts. If this ever airs, could someone tape it for me? You can record this on the same tape that already has the Laurie Abraham vs. Malcolm Gladwell debate, the Bartels/Frank WWF bout, and the one where they challenge my namesake to see if he can read two full pages from his oh-so-well-reviewed opus of some years ago without the entire studio audience falling asleep. Oh, and if there’s room, you could throw in that clip of Johnny Carson and Zsa Zsa Gabor’s cat. . . . It’s starting to get pretty full, but if you record in super-long-play mode, I think you can fit 6 hours on a standard VHS.


  1. Steve Sailer says:

    I've never grasped what value the phrase "social contagion" adds to the usual modes of analyses.

    Consider obesity. So many of the rituals of friendship revolve around exercise, sports, eating, drinking, or taking drugs, all of which affect weight. Consider three young women who are friends because they go out frequently to Manhattan dance clubs together and try to lure men into buying them cocaine. If they weren't slender, they wouldn't have taken up this hobby, and, in turn, this hobby keeps them skinny (until one, or both, goes into rehab). If one gets fat, she'll get dropped by the other two.

    Consider two male friends who get together most nights to down a six-pack of beer each and watch ESPN. If one of the pair got into triathaloning, they'd probably drift apart.

    To my mind, perhaps more interesting than friendship ties are the social influences of kinship ties, which bring together people of different ages, sexes, and personalities. (You can choose your friends, but you can't choose your relatives.) In-law ties are particularly complex and interesting, but understudied. One incomplete but illuminating definition of "class" is "the kind of people your relatives tend to marry."

  2. Sascha says:

    I think you are wrong when you are saying that "homophily is itself a consequence of network behavior […]. You can't "flock with" someone if you don't know where he or she is". Homophily-sociologists like J. Miller McPherson emphasize the difference between induced and choice homophily. Induced homophily refers to the way groups restrict opportunity structures to interaction, e. g. if you are living in Sicily there are just no protestants to mary even if want to mary one etc. But even in these structured groups individuals make decisions to associate with even more similar others (same education, age, political opinion etc.) which is referred to as choice homophily…

  3. Will says:

    On the side issue of the relatively sudden virtually en masse banning of smoking, Saul Levmore provides an alternative perspective: that this was an unintended consequence of legal incrementalism, where an interest group (the antismoking lobby) pushes a change in one sector (restaurants) but then once changed that weakens the pro-status quo lobby so that banning in other places (next bars, then hotels, etc) becomes progressively easier. As described from 17:30 on

  4. Andrew Gelman says:

    Steve: Good points.

    Sascha: Yes, I see your point. But I still think the framing of homophily vs. contagion is too simple. Real homophily is complex and involves social networks in many ways.

    Will: Interesting story. But my impression was that smoking levels in the U.S were already low even before any of the smoking bans.