“If Joey jumped off a bridge would you jump too?” An Introduction to the Selection Versus Influence Problem in the Social Sciences.
In my previous article I wrote about the work of Nicholas Christakis. In his book, Connected, Christakis argues that variables such as gaining weight or becoming happy are party due to a social process called contagion. Broadly speaking, this theory states that your position in a social network can affect the sorts of traits that you develop. If a friend gains weight then you become more likely to gain weight too—effectively the obesity has spread through the network.
In this article I’m going to explore a potential problem with Christakis’ work. The problem is that there is another social process—homophily—that might act to confound social contagion. Homophily is the tendency for people exhibiting similar traits to form ties which each other. If I am a big fan of skiing, and you also like skiing, then there is a good chance that we will form a tie with each other since eventually I am likely to bump into you on the slopes.
One recent article to explore the link between homophily and contagion— “Homophily and Contagion Are Generically Confounded in Observational Social Network Studies”—was published by Andrew C. Thomas in 2011. Although the math that Thomas explores in his article is fairly complex his basic premise is fairly easy to understand.
Thomas provides what he takes to be the classical hypothetical situation of social influence: Imagine that there are two friends named Ian and Joey. Ian is asked the question: “If your friend Joey jumped off a bridge, would you jump too?” Why might Ian say yes?
The significant point about this example is that it can be explained through a number of processes. Ian might say yes because he is inspired by Joey’s example (social contagion); he might say yes because Joey has infected him with a virus that inhibits fear of failing (biological contagion); he might say yes because Joey and Ian are friends based on their shared fondness of bridge jumping (manifest homophily); or he might say yes because Joey and Ian became friends because of their shared liking of roller coaster rides, which was caused by a common thrill seeking disposition, which also causes them to want to jump of bridges (latent homophily).
The problem is that it is difficult to distinguish exactly what’s the most significant process in any situation. As Thomas explains, latent homophily and contagion are a particular problem: “when there is latent homophily, contagion effects are unidentifiable, and even the presence of contagion cannot be distinguished observationally from a causal effect of the homophilous trait.”
When conducting research in the social sciences Thomas argues that we must be wary of this possibility of observational confound. Although it is tempting to provide a “Just-So” story for any phenomenon we witness— Joey jumped because of Ian’s influence, Man United won the Premiership because of Alex Ferguson, the stock markets crashed because of corporate greed— it is important to remember that there could be some latent effect that we haven’t take into account.