The rumored methodological wars in political science are not the wars actually being fought

Statistical-Analysis-X-Ray-sm

From the position of a political scientist, I commonly hear say, historians or anthropologists summarize what they understand political scientists to believe. Having done a fair bit of participant observation within the tribe of the tsitneics-lacitilop, those descriptions are often frustrating, describing something akin to what I understand were debates within the discipline during the 1990s. It is now 2016.

Personal frustrations aside, such outdated or erroneous views of what “political scientists believe or argue about” are problematic for a couple of more general reasons. For one, they may stand in the way of interdisciplinary collaboration by proposing that political scientists do not study certain things or work in certain ways. They also encourage fence-building between disciplines, by portraying disciplines as having settled debates, doing work that is essentially uninteresting to those elsewhere.

…The most common misconception that I encounter is that political science is divided along a cleavage of quantitative scholars and rational choice theorists versus qualitative or historical scholars. The errors here are two. First, this view lumps together “rational choice theory” with quantitative methodology, which both mistakenly equates theory and methodology and misses that some of the strongest critiques of rationalism in political science come from a quantitative behavioral origin (and vice versa). Second, it misses the extent to which quantitative methods are used in service of historical arguments, and the extent to which rationalist arguments are frequently grounded in qualitative insights. There is probably much more to write on this, but the idea of a discipline characterized by this singular cleavage on this particular axis always makes me cringe.

That is Cornell professor/blogger Tom Pepinsky. Hat tip to Ken Opalo. He goes on to argue that some of the most intense debates are within quantitative political science right now, including:

  1. Is it better to do statistics right, or not at all? It seems easy to conclude that of course we should only do statistics the right way, but if the standards for correct are formidably high, are we prepared to abandon whole areas of inquiry as unstudy-able? There exist quantitative political scientists who believe that we should basically never run cross-national regressions, for example.

  2. Experiments versus observational data. Experiments offer control, but almost always sacrificed realism in service of that control. What is the optimal balance between the two? On what terms should we make tradeoffs between the two?

  3. Microfoundations and macrostructures. Regardless of whether data is observational or experimental, research designs tend to be more straightforward with micro-level data than with aggregate or macro-level data. The problem of reconciling micro-level evidence (what individuals say or do) with macro-level phenomena (how institutions, countries, policies, and/or international systems work) will be, I suspect, one of the core issues that political scientists confront over the next decade.

I would add that I think a lot of disciplinary struggles have to do with the way that innovations in research lead, initially, to a lot of crowding into the new area, combined with the innovators getting a little exuberant with their claims.

For instance, the huge burst of formal theory came on the heels of some innovations in how to write mathematical models (and people with the training to do so).

The waning of formal theory and the waxing of empirical work started with the advent of computers and statistical software, meaning large data could be analyzed cheaply for the first time. This meant high returns to collecting new data.

The fetish for causal identification grew were bolstered by innovations in how to do so, such as instruments and regression discontinuity. Once people showed how you could run field experiments in social science, or labs, and built the institutions and the donor base to implement them, you saw a lot of growth there.

In each case you get a rush of followers, especially among junior professors and graduate students. So it can feel like the profession is marginalizing what others do. Especially when the leaders of the movement, or their followers, make very grandiose claims that seem plausible, for a short while.

I’m not sure what the next technological innovations will be.

  • Maybe in text analysis. This could increase the returns to collecting qualitative data.
  • Psychology seems underutilized. Behavioral economics, for example, focuses on a small number of biases and rules of thumb and hasn’t plumbed the depths of emotion, social identity, and so forth. You’re seeing more work on the formation and effects of personality. Political science is a little further along here but not much, and on in US politics.
  • Maybe there will be neuroscience insights but I don’t see it.
  • Other ideas?

I wonder if productivity improvements in qualitative research are disadvantaged by current  trends. Do we have quant-biased technological change? What would the switch look like when eventually comes? Because the smart social scientist always bets on regression to the mean.

43 Responses

  1. The waning of formal theory and the waxing of empirical work started with the advent of computers and statistical software, meaning large data could be analyzed cheaply for the first time. This meant high returns to collecting new data.

    golu dolls
    golu dolls

  2. I can sum up what I think the big controversy will be in two words: emergent properties. Take cosmology, for instance. We know a lot about how stars work and about their effects on gravity wells around them. However, we can’t reason from that knowledge to any understanding of how galaxies work or, even more to the point, what the structure of galaxies in the known universe is. That’s because galactic scale phenomena are a result of interactions that create emergent properties – i.e. properties at a higher level that can’t be explained by the properties of the objects that constitute the structures. By analogy, we have a lot of social phenomena that are almost certainly like that; social classes are a possible example. Have we got tools that allow us to extrapolate from one level to the other and account for the emergent properties? We wish.

    Problem = a lot of people think we do and go around boasting about how simple methodological decisions will take us to the promised land. Again, we wish.