happiness, like obesity and everything else, diffuses through networks

Happiness, like obesity and everything else (ok, so, I’m still doubtful), spreads through networks.

Some findings:

A happy friend who lives within a half-mile makes you 42% more likely to be happy yourself. If that same friend lives two miles away, his impact drops to 22%. Happy friends who are more distant have no discernible impact, according to the study.

Similarly, happy siblings make you 14% more likely to be happy yourself, but only if they live within one mile. Happy spouses provide an 8% boost – if they live under the same roof. Next-door neighbors who are happy make you 34% more likely to be happy too, but no other neighbors have an effect, even if they live on the same block. 

Physical proximity seems important.  There must be some virtual, “exposure”-effect too — who, after all, is not instantly in a better mood after getting an email loaded with emoticons ;) or after reading TSS or after seeing this gang cheerfully following Fabio’s fanny pack?

Here’s more, at Edge, summarizing the associated British Medical Journal article by Christakis and Fowler, and including an extension and graph looking at happiness and online effects.  

Written by teppo

January 7, 2009 at 6:03 am

Posted in networks

8 Responses

Subscribe to comments with RSS.

  1. Hmm, the happiness-network effect at least seems to have more face validity than the obesity-network effect. It seems reasonable at least that people who begin associating with happy people would experience a little more happiness in their life. But I can’t tell from reading the LA Times article whether the study is longitudinal, which seems like a necessary study design if you’re going to make this strong causal argument.



    January 7, 2009 at 3:01 pm

  2. I actually think it might be the same data (for obesity and happiness) — so I believe its longitudinal, will look more closely once I have a sec.



    January 7, 2009 at 3:10 pm

  3. I would like to read both papers carefully. I think they are both based on the Framingham Heart Study, which is a decades long study. I was somewhat convinced by the accounts I heard of the obesity study, as it seemed the authors did consider lots of alternative explanations (and the mechanism seemed plausible, as eating food and drinking beer are definitely contagious in my world). However, the happiness one seems trivial to me … almost certainly true but is it really revealed in their data in a way that is convincing? I doubt it. Happiness has got to be an important determinant of the network. We definitely do not have reliable models of network formation yet that really allow us to pick apart such back-door associations. In the end, I feel like the happiness paper was “a paper too far” in the sense that it lessens my belief in the obesity findings because I now doubt the interpretive standards of the author(s). Probably unfair, I know.


    Steve Morgan

    January 7, 2009 at 3:50 pm

  4. Teppo, Justin Wolfers at Freakonomics has some comments on this paper that might interest you. He thinks that the paper misinterprets the findings.


    Mike McBride

    January 7, 2009 at 7:26 pm

  5. Thanks Mike. The Wolfers critique of happiness spreading through networks coincides with the problems I have with obesity spreading through networks: endogeneity and self-selection effects, etc.



    January 7, 2009 at 9:02 pm

  6. People are right to be suspicious. The method used here totally fails to control for homophily. I know the authors say it controls for it, but they’re wrong, and this is easy to see as soon as you draw the graphical model. (Never mind whether there is any reason to hope that logistic regression is even roughly well-specified here.) As Kieran has posted about, there is in fact a paper in the British Medical Journal showing that by this method you can conclude that height is contagious in social networks. This latter paper rather annoys me, because it came out as I was working on one making the same point, but that doesn’t mean they’re not right, and that the happiness and obesity papers aren’t case studies in confounding.



    January 11, 2009 at 1:59 pm

  7. […] the effects of network ties on the diffusion of everything from obesity to happiness (see posts here and […]


  8. […] in the growing literature on social networks, interpersonal influence, and health outcomes (see Teppo’s take on the topic), but the conclusions are somewhat different than its predecessor studies. Obviously, it would be […]


Comments are closed.

%d bloggers like this: