Skip to content

Christakis response to my comment on his comments on social science (or just skip to the P.P.P.S. at the end)

The other day, Nicholas Christakis wrote an article in the newspaper criticizing academic social science departments:

The social sciences have stagnated. . . . This is not only boring but also counterproductive, constraining engagement with the scientific cutting edge and stifling the creation of new and useful knowledge. . . . I’m not suggesting that social scientists stop teaching and investigating classic topics like monopoly power, racial profiling and health inequality. But everyone knows that monopoly power is bad for markets, that people are racially biased and that illness is unequally distributed by social class. There are diminishing returns from the continuing study of many such topics. And repeatedly observing these phenomena does not help us fix them.

I disagreed, saying that Christakis wasn’t giving social science research enough credit:

I’m no economist so I can let others discuss the bit about “monopoly power is bad for markets.” I assume that the study by economists of monopoly power is a bit more sophisticated than that! I have studied racial profiling, and I can assure you that this work is not about the claim “that people are racially biased.” Regarding the question of illness being distributed by social class: Is it really true that “everybody knows,” for example, that Finland has higher suicide rates than Sweden, or that foreign-born Latinos have lower rates of psychiatric disorders? As Duncan Watts has written so memorably, it’s easy to say that everything is obvious (once you know the answer).

Christakis responded in an email and in a comment:

I admire and respect your work, and so I am surprised by the tone of some of your comments on your blog.

Have you had a chance to read any of our papers recently? We have many experiments and observational studies in major journals that have survived grueling peer review. We have also responded at length to our critics. Here is the recent response to our critics (with comments and rejoinders):

And here are a few papers (from Nature and PNAS) that further explore social networks using diverse methods, both observational and experimental.

One cannot have it all, Andrew, as you know very well. If we use observational data, nihilists (who you often take on!), criticize the difficulty (in the limit, the impossibility) of causal inference; if we use experiments, critics object to the lack of verisimilitude.

Finally, as someone who spends a lot of time trying to advance the public understanding of science (I mean you, not me!), i don’t think you really should be calling the kettle black.

As to your original complaint, let me say this: my main point (within a constrained 800-word format, for which the editors write the title, not the authors) is that we can learn something from the natural sciences about institutional forms and about ways of doing science (just as they surely can learn from us). This does not mean what we social scientists are doing is bad; but it would be arrogance to assume we have nothing to learn. And re-deploying talent to tackle new (and, yes, old!) topics is fine, but we should be judicious, and, yes, I think we should move to the scientific frontier. Is that a claim you don’t agree with?

My response:

I agree with everything above. What I didn’t agree with in the op-ed was the emphasis on diminishing returns from the continuing study of topics such as monopoly power, racial profiling, and health inequality. I think this stuff is on the scientific frontier too! One source of my strong reaction here could be from my time at Berkeley, when my colleagues in the statistics department were mostly split into three groups: (1) those who did pure mathematics and had no interest in my work, (2) those who felt that Bayesian statistics was low-tech and thus beneath their notice (these people liked to “prove theorems” rather than “make assumptions,” not realizing that a theorem, to be applied, needs assumptions), and (3) those who did applied work in biology and felt that social science was not real science.

And I don’t want people to feel I’ve condemned Christakis and Fowler’s work. Over the years, I’ve pointed to criticisms of their research but I’ve also emphasized that these are difficult problems and that, as the very least, that work is an important first step using information that is available. I just don’t think this sort of work should be privileged over work on topics like monopoly power, racial profiling, and health inequality.

P.S. Also I’d just like to say that I really appreciate that Christakis commented at all. He could easily stayed above the fray and not responded at all, based on the quite reasonable theory that many more people would read his NYT op-ed than would ever see this blog. He could’ve been like Gregg Easterbrook and simply refused to ever acknowledge my existence, let alone my specific comments on his writings. I think it’s great that Christakis stands by his work and welcomes criticism.

P.P.S. Here’s the blurb I wrote for Christakis and Fowler’s book a few years ago:

Margaret Thatcher said there is no such thing as society. Hillary Clinton wrote that it takes a village. In their new book, Christakis and Fowler write, “We don’t live in groups, we live in networks,” and they back this up with dozens of interconnected stories of research findings by themselves and others, ranging from bank runs to suicide prevention, from nut allergies among schoolchildren to epidemics in virtual worlds, from the spread of happiness to the spread of voting. The combination of speculation and science is fascinating and leaves me eager to learn about the next wave of research in this area.

P.P.P.S. To move beyond a pointless who-misunderstood-whom back-and-forth, let me now shift the discussion to an interesting issue raised by Christakis:

What are the (relatively) settled matters in the social sciences? Can we social scientists can ever say that “we have pretty much figured this out” (as in the way biologists have figured out certain topics)?

I dunno. Krugman would say it’s settled that it’s a good idea to expand government hiring during a depression, but others disagree! For an example from psychology, stereotype threat is claimed by some to be very well established, while others have difficulty finding it at all. In various ares of social research, there’s debate about the replicability of all sorts of claimed effects.

Ideally, I think, once something is settled, this can be the staging point for more research. For example, Kahneman and Tversky demonstrated “anchoring and adjustment” and other so-called heuristics and biases. Then later researchers followed up with studies trying to crack open anchoring-and-adjustment, to understand experimentally how it happens and how to alter it.

Overall, I’d say that, if anything, social scientists perhaps don’t spend enough time re-confirming the definitive statements. There’s a real push toward novelty, to the extent that maybe we don’t have enough “gold standards” of well-established social patterns.


  1. […] Christakis asks (scroll to the P.P.P.S. at the end of the […]

  2. […] More from Gelman and Nicholas Christakis […]

  3. In re your PPPS: Isn’t there now an empirically solid and highly confirmed connection between income and political opinions (voting)? I thought that was now very well established, as you sometimes remind us when pundits speak of it but get the sign of it wrong! Your work is in the US, but I think there are probably well established results in Europe and elsewhere, no?

  4. Jack PQ says:

    What is settled (or not) in economics can be seen in reliable surveys such as those conducted by the AEA or the Booth School IGM forum:

    But I would caution against trying to find definitive claims in the social sciences that would parallel the laws of nature in the physical and medical sciences.

    Bluntly, in economics, micro and applied micro have been successful (empirically verified and replicated theoretical claims), while macro has not. In the news, one mostly hears about macro claims, and on both political sides economists are too confident. Often, the truth is, We just don’t know. (Stimulus, austerity, tax cuts at a macro scale…) But it doesn’t make for a good soundbite.

  5. Steve Sailer says:

    I’m not a social scientist, but I’ve been a fan of the social sciences for 40 years. I don’t really understand the complaints about the social sciences. I’ve found them entertaining and illuminating, and still do.

  6. Michael says:

    David: elsewhere, these things have been well established, but there is some evidence (not convincing to me, but enough for controversy) that the basic relationship is changing. This points to a bigger issue in Christakis’s argument. The social world and most of the facts we have been able to uncover are changing (indeed, in some cases, change precisely because regularities have been uncovered). Leaving a small “palace guard” behind to study even things we think are clearly established risks having them change under our noses.

    If you are interested enough to track down this book, it is well worth your time:
    Clark, T. N., & Lipset, S. M. (Eds.). (2001). The Breakdown of Class Politics: A Debate on Post-Industrial Stratification. Baltimore: Johns Hopkins University Press.

    Also worth your time:
    De La O, A. L., & Rodden, J. A. (2008). Does Religion Distract the Poor?: Income and Issue Voting Around the World. Comparative Political Studies, 41(4-5), 437–476.

  7. Ashok Rao says:

    You’re right, it’s great that he replied – and was quite engaged in the conversation too.

    But your post raises an interesting question: whether it’s possible for a question in the social sciences to be “settled” at all. Here’s the framework I like to think of it in. There’s a certain “phenomenon” – say, fiscal policy in depression, or racial profiling by authorities. But social scientists never directly address that phenomenon.

    Rather, they “probe” that phenomenon – which is more or less like a black box – with a set of questions. “Do countries where governments engaged in expansionary fiscal policy, controlling for factors to the extent possible, fare better in recovery”. “Are blacks incarcerated at a right not commensurate with their underlying criminal propensity”.

    The questions asked are broadly reflective of ideology – at a greater level of the discipline, but also of the researcher himself. With profiling one might ask whether it prevents crime, liberty and fairness be damned. Another whether there is a disproportionate indictment rate, etc.

    How can there be “settled” beliefs if the questions are always changing? Social scientists can never, ever – by definition – actually understand the underlying phenomenon, but can only ask it various questions, right?

    This isn’t even contrary to what Christakis says. Social scientists are probing the same phenomenon today, perhaps with different questions informed by a superior knowledge of computational techniques. But this intimate link between the two processes reaffirms – in my opinion – the importance of “traditional” research which then can never really be separate from the newer stuff.

  8. Jfly says:

    What are the (relatively) settled matters in the social sciences? Can we social scientists can ever say that “we have pretty much figured this out” (as in the way biologists have figured out certain topics)?


    I think most of the stuff in economics, psychology textbooks are well settled. You won’t expect them to be trashed away in twenty or even fifty years.

    “Krugman would say it’s settled that it’s a good idea to expand government hiring during a depression, but others disagree!” — This is not a good example, even economists who have total agreement on fundamental theories can still have strong disagreement on answering specific questions.

    The illusion of no-agreement-in-social-science mostly comes from the bad way of looking at social sciences. When people talk about social sciences, they like to focus on specific questions (you can see from the post and comments), instead of basic theories.

    • Anonymous says:

      a useful theory has to be able to answer _some_ specific questions. If it can’t answer any specific questions, then it hasn’t really been tested as a theory (and shouldn’t be characterized as ‘settled’).

  9. K? O'Rourke says:

    This might be mostly a matter of time scale and dimensionality.

    Peirce argued that everything evolved including the laws of physics which now evolve extremely slowly (or maybe not at all after an initial period of very short duration?) and that because of this science has to continually redo randomised experiments to keep up to date (his solution to the GRUE riddle?)

    Additionally the purposeful things one might learn about phenomena may have few or many dimensions or aspects.

    So we have physics which changes almost not at all with few purposeful aspects to now about and economics which changes quickly (De Sassure argued that as economics was the science of haw people valued things whenever you discuss it with people it changed) with a myriad of purposeful things that can be worth knowing.

  10. Entsophy says:

    […] better way of doing science (2) find novel data, or (3) stagnate. Evidence suggests they mostly do (3), occasionally do (2), and rarely do […]

  11. numeric says:

    those who did applied work in biology and felt that social science was not real science.

    I’ve heard from people in the Berkeley statistics department that one of the reasons you did not receive tenure was because of your interactions with social scientists. I’m sure this isn’t news to you.