Saturday, June 13, 2009

Networks and political science

The Berardo project I referenced earlier found some interesting things about social networks research in political science. If I remember correctly, centrality in the network was inversely correlated with age and time in the discipline. Translated into English, this means that the leading scholars using networks research in political science tend to be graduate students and junior faculty members. It’s a tool of the young. I suppose this is to be expected with any innovation, but it’s still kind of interesting. It's also kind of disorienting, as I'm increasingly turning to people more junior than myself for advice.

There’s a bit of a disconnect between networks tools and the expectations of the discipline. What networks research is really good at is description, but the discipline generally frowns on description and encourages hypothesis testing. A lot of the papers I’ve seen here (including one of my own) are hamstrung by this problem.

If you’ve watched “The Wire,” you understand the value of description of a network. Season Three was a particularly good example. The Barksdale/Bell drug cartel had a very complex structure, with a bunch of pretty poorly-educated guys at the bottom selling heroin on the streets, a group of savvy lieutenants coordinating them, and then the two principals at the top. The cartel did a lot to make it appear like there was no coordination between these levels, mainly by communicating over disposable cell phones and then dumping the phones or switching SIM cards frequently. Meanwhile, the police are trying to map out the whole network out, using increasingly sophisticated tools to overcome the cartel’s efforts to hide the links.

In this kind of case, description of the network is enough. It’s enough to arrest the key players, and presumably enough to convince a jury. Description also works nicely for teasing out the structures of terrorist organizations, figuring out who the key players are and who you need to eliminate to cripple a network.

Some of my research has attempted this same sort of approach with parties. And, in all seriousness, there are important similarities between parties and criminal networks. Party leaders are often attempting certain levels of coordination to do things like advantage a candidate in a primary or convince other candidates to drop out of contests. The things they do along these lines are either illegal or would appear unsavory if run in the newspaper, so they try to keep them hidden. But we still know this coordination occurs.

Greg Koger, Hans Noel, and I have an article coming out soon with BJPS that uses the buying and selling of names and addresses for direct mailings to trace out the party networks. Below is one of the pictures we produced which shows the structure of the Republican Party network. It’s kinda cool, and it’s an image you wouldn’t get by following more formal descriptions of what a party is. But again, there’s not much hypothesis testing here.
I’m not sure what the best path is for networks research in the social science. Should the journals become more accepting of networks descriptions? Or should network researchers be going that next step to test important hypotheses?

No comments: