Microfoundations for Political Science
Over the past two weeks, the economics blogosphere has spent some time discussing microfoundations. The microfoundations approach means, at base, that all macroeconomic models should constructed from the behavior of individuals. This normally means optimizing agents, although it need not be so. So if we think that the business cycle exists, we should be able to describe it as an aggregate consequence of individual choices. If we evaluate policies, like the stimulus, we should start by considering what an optimizing agent would do in response to them. To see what sorts of consequences this has had for macroeconomics, I’d recommend reading (in alphabetical order) Peter Dorman, David Glasner, Paul Krugman, Kevin Quinn (and here too), Richard Serlin, Karl Smith, Noah Smith, Mark Thoma, and Simon Wren-Lewis. I also recommend the FRB Minneapolis on heterogeneous agents in economic models, which is not exactly a defense against the various critiques here, but is worth reading in light of the discussion.
If you’ve spent any time at all in a contemporary political science department, you’ve heard people talking about microfoundations too. Microfoundations are big time these days. Yes, yes, we know that crossnationally, X is associated with Y, but we want to know the microfoundations of this relationship. The reason why we’re collecting very precise data in this godforsaken corner of the tropics is because that will show us the microfoundations. We are subjecting college students to an interesting lab experiment to get at the microfoundations. We should uncover the microfoundations. We should test the microfoundations. Microfoundations.
I am in principle a supporter of building theories from microfoundations. I have to be: my affinity for methodological individualism forces me to believe that microfoundations are important. But when political scientists say microfoundations, they often don’t really mean microfoundations. They mean something like individual level data, or careful research designs, or causal mechanisms. They almost certainly don’t mean microfoundations in the sense that economists use the term.
Here’s what I mean: microfoundations in macroeconomics are something like theoretical primitives. Macroeconomists don’t go out to see if individuals are really optimizing–they know that this is likely to be incorrect for many people. Building a microfounded model is a way of disciplining the theory at the theory-building stage. The challenge, of course, is that the analyst may find herself believing that a real-world phenomenon cannot exist because it is difficult to write a model in which it emerges as a consequence of optimizing behavior. But the strategy is not to uncover an aggregate relationship and then dig down with data to figure out what individual level behavior produces it, it is to stipulate how individuals behave and then see what follows. It’s no accident that most DSGE models are not tested, they are calibrated. Whether or not that’s a good idea, that’s how it works.
Microfoundations for political science are usually not the stuff of theory-building. These days, microfoundations are usually sought later in the research trajectory, to try to ascertain among various possible theories that could explain aggregate patterns. They are empirical findings, not theoretical statements, and because they are “empirical” and “micro” they require very focused data collection efforts in bounded spatial and temporal contexts (one interpretation of the rise of experiments in contemporary political science is that they seem uniquely compatible with this). Microfoundations like this often do not lend themselves to cumulative research: it’s hard to build on some of the empirical findings that a microfoundations oriented research project in political science provides (not impossible, just hard, and too rarely done). Say what you will, but the microfoundations approach in economics produced a lot of cumulative research, at least for a while.
There are exceptions to this description of microfoundations as empirics (the political science way) versus microfoundations as theory. Stathis Kalyvas’s logic of violence in civil war, while inspired by historical evidence, is as close to a “microfounded” political science research paradigm as I can think of, deriving predictions about the distribution of violence in civil wars based on the optimizing behaviors of individuals trying to avoid being victimized by violence. It’s become something of a cumulative research program. David Laitin’s approach to identify formation in the post-Soviet states is perhaps another, although one that I see as having been less cumulative. But such examples are the exceptions rather than the rule. The classic example for most political scientists is Micromotives and Macrobehavior, but that’s really about introducing a perspective on how think about aggregate social phenomena than a theory of anything. And Schelling is an economist.
Why do so many political scientists use the term microfoundations to describe something other than what the economists mean? I’m not sure. One possibility is that we just ape the economists, and poorly. Another is that both “micro” and “foundation” sound awfully scientific and important. “Micro” makes it sound like we’re peering into a microscope like Anton van Leeuwenhoek to uncover the hidden bodies that are driving
sickness politics. “Foundations” alludes to building the edifice upon which all further political science will be done. Still another reason is that “going micro” is just a reaction against the trend in the 1980s and 1990s to “go macro,” and “foundations” are important as a rhetorical gesture because they put micropolitical work on the same level as macropolitical work.
Still, it is worth taking stock of why our microfoundations tend to differ from economists’ microfoundations, and what it means. I have three thoughts. First, if we really mean causal mechanisms when we say microfoundations, then there could be many reasons why X produces Y, because there may be many causal mechanisms at play at once. People can be both greedy and jealous at the same time; if some people are just greedy, then others could still be jealous too. You wouldn’t look for the microfoundations, you’d perhaps describe some microfoundations.
Second, if we don’t mean that we are willing to stipulate optimizing behavior as the foundation for political analysis, then this frees us from devoting effort to finding microfoundations when they are unlikely to be found. That is, you can imagine a microfounded theory of voting or congressional committees, but I don’t see how this would apply to things like constructivist IPE or some aspects of bureaucratic politics. For some topics, the search for theoretical microfoundations may be a waste of scholarly time and resources.
Third, let’s think hard about the differences between the phenomena tackled by macroeconomics and most political science questions. The microfoundations approach makes sense (I know that not everyone agrees, but bear with me) as a way to start thinking about an entire economy. Markets clear, individuals are price takers, etc. This just doesn’t seem like the right starting point for understanding things like parliamentary coalition formation, capital account policy, or revolution. The idea of the political is that we are not talking about a market, we’re talking about politics. There are lots of very, very rigorous theories out there for political analysis, but they need not be microfounded. In that way, they are like most rigorous economic theories, which aren’t microfounded either.
To be clear, I’m not saying that political scientists shouldn’t do careful empirical work, or use individual level data, or look for causal mechanisms as much as possible. I’m also not against microfoundations—I’m actually quite sympathetic to this sort of approach. We should know, however, that our microfoundations are usually not the economists’ microfoundations, and that that’s OK.