Skip to content

Christakis-Fowler update

After I posted on Russ Lyons’s criticisms of the work of Nicholas Christakis and James Fowler’s work on social networks, several people emailed in with links to related articles. (Nobody wants to comment on the blog anymore; all I get is emails.)

Here they are:

Political scientists Hans Noel and Brendan Nyhan wrote a paper called “The ‘Unfriending’ Problem: The Consequences of Homophily in Friendship Retention for Causal Estimates of Social Influence” in which they argue that the Christakis-Fowler results are subject to bias because of patterns in the time course of friendships.

Statisticians Cosma Shalizi and AT wrote a paper called “Homophily and Contagion Are Generically Confounded in Observational Social Network Studies” arguing that analyses such as those of Christakis and Fowler cannot hope to disentangle different sorts of network effects.

And Christakis and Fowler reply to Noel and Nyhan, Shalizi and Thomas, Lyons, and others in an article that begins:

Here, we review the research we have done on social contagion and the methods we have employed to examine the Framingham Heart Study, the National Longitudinal Study of Adolescent Health, and other observational and experimental datasets. We describe the regularities that led us to propose that human social networks may exhibit a “three degrees of influence” property, and we review statistical approaches we have used to characterize inter- personal influence with respect to behaviors like obesity and affective states like happiness. We do not claim that this work is definitive, but we do think that it provides some novel sorts of evidence regarding social contagion in longitudinally followed networks. Along with other scholars, we are working to develop new methods for identifying causal effects using social network data.

This was about what I was imagining: the work is not definitive but they have found some patterns in this particular idiosyncratic dataset and it will be interesting to see what comes next.

Finally, Lyons wrote in his paper that the Christakis-Fowler data are “are not available to others.” That’s not completely correct; at the time of the writing of his paper Lyons was unaware of this online data source (apparently googlable from “framingham social network data”). Unfortunately, confidentiality restrictions appear to limit the data in such a way that only partial replications can be done using this public dataset.

One Comment

  1. Felix E says:

    Two points are sometimes overlooked in the recent discussion of Christakis and Fowler's work on contagion (including, I believe, on this blog).

    Both relate to the most popular criticism of C&F's work: that the observed associations may be due to homophily (the attraction of likes) rather than contagion.

    (1) C&F have published a set of randomized experiments (iterated public goods games) in PNAS, 2010. C&F find evidence for contagion of cooperative behavior that extends out up to three degrees of separation. Since network structure is exogenous in these experiments, the homophily critique of their observational results doesn't apply.

    (2) The question isn't whether there's homophily (C&F acknowledge homophily in *every* paper I've read) but whether homophily is strong enough to invalidate claims of contagion. Tyler VanderWeele has published a formal sensitivity analysis in Sociological Methods and Research, arguing that C&F's obesity result, alongside some other results, withstand considerable unobserved confounding (i.e. homophily), while other results are less robust. VanderWeele also shows that some of the evidence produced against C&F's results is not robust.

    Lyons, btw, doesn't cite either of these articles. This strikes me as sloppy, especially since he asserts that "C&F have not done an experiment" (p.12).

    (Disclosure: I collaborate with Christakis.)