Sunday, July 6, 2014

A Farewell to Academia, Pt. IV - Why is Political Science Research So Bad??

Check out Pts. I, II, and III

[In this post, Geist, our intrepid explorer, considers why much of the statistical work being done in political science is quite bad. Will he bore himself to death? Read below to find out!]


To the layperson, statistics is oftentimes tedious and quite boring. So are laws. So is anything that massively affects your life outcomes. It’s why I have no idea what Obamacare actually is. And it’s why I don’t care to find out. From what I’m told, it’s why we need Obamacare in the first place.

Pop statistics sites, with their manifestos and self-proclaimed brio, have tried to gussy up something that’s by its very nature rather unsexy. So inevitably, Five Thirty Eight has succeeded at sucking the fun out of everything from eating burritos to masturbation[1],Vox has us memorizing flashcards for a Vox quiz that’s always seemingly just over the horizon, and the Monkey Cage is, well, fine I guess. I doubt many people ever click the link to the article being referenced or the book being cited.

Because if they did, they’d likely see a lot of statistics. And statistics do two things to high-minded, supposedly well-informed liberals: scare the shit out of them and confer authority on whatever it is being said.[2] Now, I’m not talking about basic findings like “Mitt Romney is responsible for 23% of the unemployment found among Hispanics”. People are pretty comfortable citing these things when it helps their point and conveniently skeptical when they don’t. 

Instead, I’m talking about articles using techniques like maximum likelihood estimation. Then fear strikes: most people will pledge blind devotion to whatever the findings are because “Holy Shit! Look at those equations! I don’t want to be seen as stupid in the company of my STEM overlords.” Others (the liberal arts Left), though equally scared of something they don’t understand, performatively thumb their noses at authority by trashing the whole enterprise altogether.

As I’ll discuss in my next post, there’s nothing inherently wrong with statistical research when it’s properly conducted. It’s something that I think the Left needs more of. That said, the statistics produced by political scientists largely suck.

I’m not going to spend much time substantiating this claim. It requires too much space and it’s boring as hell. That said, having worked with one or two tenured political scientists who are actually quite good at their job, the general feeling among those who are highly statistically literate is sheer wonderment over how most of the articles published ever get through peer review. Most political scientists know a lot of advanced, sophisticated techniques just enough to use them, but not well enough to know if they’re being used correctly. The result is that political scientists bury many of their more suspect assumptions or design decisions under layers of high level statistical techniques.

Perhaps to uncover why this is happening we should conduct a sociological examination of the discipline. While I don’t have any hard evidence for my following hunch, it's certainly testable. 

The problem, I think, is how the graduate school is underserved by the undergraduate curriculum. A political science degree essentially confers a bachelor’s in electives. Very few programs require any statistical or mathematical training. All of the articles we have undergraduates read are either purely theoretical or historical in nature. The findings are presented either as fact or they’re debated like issues are on cable news—which is to say, without any sort of rigor. Very few papers are ever assigned that give the undergrad a huge dose of statistics (cuz it’s scary!) and if they are the statistical section is not assigned. American political science students are presented with a very skewed (and inaccurate) view of how the field works. They think the field is what people were doing in the 80s and 90s.[3]

But since the early 00s, the sophistication of our statistical methods has increased by magnitudes. Everyone has to be trained in a whole battery of quantitative methodological approaches in order to be at all employable. Political science departments began hiring academics with PhDs from economic and business school programs. The field has become massively teched-up as a result. Much of the work done by these academics (largely in American political science) requires a Masters in economics to understand what they’re doing.[4] The other subfields—international relations, comparative politics, and even some of political theory—have since borrowed from their techniques and methods.

Political science graduate schools are not pulling from economic undergraduate programs, however. Instead, they’re selecting primarily from a pool of undergraduates who think political scientists either generalize from news stories and history books, theorize endlessly, or conduct surveys all day.[5] They’re pulling from a group of students who didn’t do much math or statistics in college and then expect them to learn graduate econometrics in roughly a year. This is a recipe for terrible, terrible scholarship.

What usually happens is this: students attend “math camp” for a week or two where they learn the parts of calculus that will later help ground their statistical training. They’re then trained in a variety of statistical methodologies over the next year and a half, frustratingly being taught as if they were statisticians or economists in training. The economists or business school PhDs who teach these courses teach in equations and proofs. These students only understand in prose. They then have a political scientist explain the work in English, which gives them a grasp of the material at a basic conceptual level. Unfortunately, statistics is ultimately a mathematical discipline, so there’s very little sense among graduates students what the mathematical reasons are underpinning any one statistical decision. The result is what I talk about above: bad research.

Unless political science departments change their undergraduate programs, this problem is not likely to be ameliorated. Economists and statistician undergrads are not going to apply to political science programs because the job opportunities are terrible, especially in comparison to those in economic and statistics departments or the private sector. And the field can’t go back to the pre-millennium Stone Age.


Pt. V on the problems with Leftist political science will be posted soon.




[1] I know much of what they do is tongue in cheek, but should we go so far as to read Nate Silver’s manifesto in that way too?
[2] Conservatives on the other hand just don’t give a fuck about numbers
[3] Which is the old literature most tenured faculty are fluent in
[4] They’re immune to my criticism’s here. My main problem with these researchers is addressed in Pt. II of this essay.
[5] I do think that our survey research has been fairly successful, though I do have some normative reservations regarding how it’s conducted. See (pt. II). 

No comments:

Post a Comment