Sunday, 24 August 2003

Perestroikans and the obsession with "rational choice"

Matthew at A Fearful Symmetry notes trouble brewing at Harvard’s Kennedy School of Government between “rational-choice theorists” and “more traditional political scientists,” according to NYT Magazine writer James Traub.

This description belies a lack of understanding about political science. We can generally break the history of political science into two eras: the pre-behavioral period and the post-behavioral period. During the pre-behavioral period, political scientists basically did two things: descriptive analysis of political institutions and what I’d term “normative political theory,” a nebuolous field that generally involves the study of political phiolosophy, with a nice dash of prescriptivism thrown in just for entertainment value. Woodrow Wilson did a bit of the former, when he wasn’t being a prescriptivist or a racist (the three occupations kept him fairly busy).

Then came World War II, James Gallup, and the sample survey. A bunch of folks at Columbia and Michigan (among other places) decided that it might be a good idea to test whether all these theories people had come up with about the behavior of voters during the “pre-behavioralist” period were valid (and it turns out that they basically were wrong). Thus was born the behavioral revolution in political science—and, arguably, the entire idea of the study of politics using the scientific method (or empiricism).

Note that this account has not used the words “rational choice.” That’s because rat choice really has almost nothing to do with behavioralism. The roots of rat choice come from economics (notably the work of Anthony Downs and Mancur Olson), and in particular the idea of “utility maximization”; rational choice theory generally argues that people behave in a way that has the maximum possible benefit to themselves (“utility”). Utility has proved rather annoying to quantify in political science (in economics, utility maps rather nicely to monetary units; in political science, about the best we’ve done is OxPoints).

However, there are plenty of other ways to explain behavior in political science other than rational choice. Many behaviorists today—including myself—incorporate rationalist explanations with sociological and psychological explanations to formulate their theories of political behavior, which they then test empirically using either experimental or survey-based data with statistical techniques (usually, although not always, borrowed from other fields, including mathematical statistics, economics, psychology, and biostatistics).

More importantly, this ignores other techniques used in other subfields of political science. In international relations (and some other parts of the discipline), many theories are formulated using game theory, which has some links to rational choice (mostly in terms of the institutions the procedures were developed at, most notably the University of Chicago and University of Rochester), or more advanced mathematical modelling techniques; sometimes these techniques are linked with rational choice under the rubric of “formal theory.” Some of these techniques (ironically, like normative political theory) have not historically been subjected to any real-world testing; however, now there is some interest in doing this through the NSF’s promotion of a series of EITM (Empirical Implications of Theoretical Models) workshops and works like that of Rebecca Morton.

So describing what I do as “rational choice” is something of a misnomer. It would be like describing all normative political theorists as “Plato scholars,” or all economists as Keynesians. The truth is that there’s room for many different perspectives on the study of politics in our discipline.

The problem is that I’m not sure most Perestroikans are aware that there are multiple empirically-based perspectives. For example, the book the Perestroikans hold up as validating their point of view comes from two political scientists grounded in the empiricist, Michigan school tradition. Now, I’ll agree that some leading journals often mistake sexy methodology with substantive importance; like in specifying any theory, simpler methods are preferable to complex ones—given similar explanatory power. (This also goes for theoretical articles; if the APSR is to be a journal that everyone with a graduate education can read, surely we should expect prose that is penetrable to non-specialists.) But to reject empiricism outright in a discipline that has science as part of its name is, in my view, a bridge too far.

Daniel Drezner has more on the Summers piece this morning, which will no doubt spark an interesting discussion, while this post (inexplicably) makes John Jenkins is glad he’s a theorist. He writes:

While I'm certain that good work can, has, and will continue to be done this way in political science, more often than not we end up with analysis so blatant in its biases that it's entirely useless. I am reminded of a study that we looked at in one of my methods classes that determined that, on the whole, those who were of [M]exican descent were paid far less than those who weren’t in some city in Texas. It failed utterly to account for how long the people had been in America (i.e. first-generation immigrants tend to work in low-paying jobs because of lack of training and lack of facility with English) and resolutely concluded that the disparity was due exclusively to racism. Anyone think there was a foreordained conclusion there?

Can you say omitted variable bias? Bad research is bad research, whether it’s empirical or not…

More to the point, I don’t think empirically-oriented political scientists claim to be able to make predictions on a par with those made by sciences solely governed by physical laws. In any event, that’s fundamentally not the point of science, which seeks explanations rather than predictions. Just because, as John puts it, “people will do stupid shit sometimes” doesn’t necessarily mean that their behavior isn’t at least somewhat explainable.