What do you learn in a year of a Poli-Sci PhD
As I wrap up my first year of graduate school, I’ve been thinking about what I’ve learned. In prosaic terms, I’ve learned a ton. If you had told me a few years ago that I’d be fluently coding in R and working out linear algebra proofs, I would have laughed. And I’ve gotten much more familiar with the body of knowledge in political science, but have a great deal left to learn. I’ve also learned that chronic wrist pain can improve your work ethic and self-scheduling a great deal, because you can no longer count on pulling frantic all-nighters at the last minute. So those are nice, but I’m thinking more serious lessons.
I’ve gone from a complete skeptic of social science to a guarded appreciation of it. As I’m in political science, I’m mainly talking about political science. As with many people coming from a qualitative background, I had always derided the social sciences, particularly political science. The real world is very complex, and to a historian the social sciences seem like a hubristic effort to quantify the unquantifiable. The phrase, “When all you have is a hammer, everything looks like a nail”, is invoked a lot.
Qualitative types have a clear mental image of a political scientist – let’s call him Bill. Bill is a wild-eyed dilettante, with a copy of Stata and only the vaguest idea about the political and historical context of the kitchen-sink regression he’s running. Bill doesn’t need to know much more about the context, but as his model will show, it’s really quite simple – those dry old historians need to shut their books and go find a real job.
Bill does exist, but he doesn’t run the field if he ever did. In modern political science, the degree of empirical caution is striking and the field has moved far beyond simply running regressions. Current training in the field emphasizes thoughtfulness about quantitative inference much more than pure statistical machinery. Today, it’s nearly impossible to get a paper published without demonstrating causality much more clearly through a variety of sophisticated strategies. Scholars in the field are thinking extremely carefully about how to meet the burden of proof, and usually are much more humble about the strength of their conclusions than historians.
This drives my main problem with the field. The increasing sophistication of strategies to show causality are mostly driving the focus of investigation smaller. The hot thing in political science right now is the survey experiment, where the experiment actually consists of varying parts of the survey. For example, you might run a survey about voters’ preferences on taxes, but the experiment is that for half of the subjects you insert a question about welfare before, the “treatment”. If the treated group reports more anti-tax attitudes, then voila! You’ve demonstrated that talking about welfare makes people more anti-tax, and because it was a controlled experiment there’s no way that anybody can question your results. It’s a good way to get a paper published, and so this is a very popular area of research right now.
The problem here is that this might not be generalizable, or even very important. If a simple question about welfare can measurably alter people’s opinions on taxes, how strong are those opinions in the first place? More to the point, how persistent is the effect? Will, for example, ads about welfare queens help build anti-tax sentiment? Probably not, if the effect evaporates within 30 minutes. In fact, does this tell you anything really important about politics or does it just demonstrate that people’s policy opinions are mostly pretty diffuse and weakly held? Because we have known the latter for a long time. This problem extends beyond just survey experiments. It’s not a formal mathematical tradeoff, but usually the cleaner an identification of a causal effect is, the less clear it is that the effect is generalizable to other situations. Political science today is leaning much more strongly in the direction of clearly identified and not very generalizable.
Which makes for a lot of boring papers.