An Overview of Political Science (Policy and International Relations Primer for EA, Part 3)
This is the third in a series of posts on International Relations and Policy. First was the introduction to this sequence, which discusses why this matters, and who should care. Second was defining Policy and International Relations, and talking about the sub-fields of political science. That post deferred discussion of different approaches, and instead looked at subject areas. That leads to the current post, which is a more detailed overview of political science specifially, rather than international relations and policy, which will be discussed in later posts.
This post is primarily going to discuss approaches to political science, such as Behavioralism, Rational Choice Theory, and Institutionalism, as well as Constructivism, Marxism, and Feminist Political Science. This background is particularly useful because these different approaches will be important in discussing how to use political science and similar tools in specific types of EA decisionmaking.
First however, I (again) want to deal with the issue of political science as a soft science. In addition to being useful for later understanding how political science can and cannot be used in Effective Altruism, I think it gives useful background on what the different approaches are doing, and why the different types of political science differ in more than just methodology.
The Soft Science of Politics
Political Science attempts to deal with “systems of governance, and the analysis of [activities, thoughts, and behavior] that apply to members of a group.” According to most approaches, it is a science because it attempts to provide a systematic framework and theories for predicting and explaining what happens and what will happen.
It is possible that this scientific approach builds on psychology the same way chemistry builds on physics. If so, once we understand the individual parts, we can discuss what groups of those parts do when organized in different ways. This would make it an exercise in deductive thinking.
Is political science a science in the same way chemistry is? Unfortunately, as I discussed in the introduction, psychology is a soft sciences, where we cannot make sharp predictions. As a result, despite Asimov’s hopes, inasmuch as political science deals with the same issues, political science is a soft science as well. We can build psychological models of behavior, and look at how these aggregate into group decisions. Alternatively, we can look at group decisions directly and try to see how they work, building theories more directly. Both approaches are useful—both in chemistry and in political science, but they are very different. This has implications for how different approaches see political science, so it’s worth explaining.
The model where fundamental understanding like physics or psychology informs chemistry or political science is useful because it has a sort of guarantee that if the underlying model is correct, the results are certain. If we understand how individuals will decide, perhaps we can understand how groups will decide. In addition to the fact that psychological predictions are imprecise, the problem with this is much like the one in physics. Fundamental models are complex and scaling them up requires detailed knowledge of all of the parts, and how they interact.
In some places, we can rely on statistical guarantees the way chemistry relies on themodynamics (i.e. statistical mechanics.) That’s how we predict elections. In most places, however, there are lots of factors that need to be known to move from individual to group behavior. These factors include the institutions, the way different groups or classes of people interact, the importance of individuals in power, and the complex way that even perfectly rational agents can behave. Each of these roughly corresponds to the focus of one of the below approaches. (Institutionalism, Marxism, Psychological theory, and Rational Actor theory respectively.)
The second model is where physics (or psychology) is effectively ignored, and insights about the system need to come from a more direct empirical approach. If we see that, for instance, social cohesion predicts degree of democracy, we can quantify and test that theory directly. This often requires using higher-level abstractions, however, and these are often hard to measure. This model allows approaches that are not contradicted by the first one, and the two are often used together.
Lastly, there are the models where all of the above is bullshit, and people can create meanings and have personal agendas that drive politics that can’t be usefully captured by predictive models, and can only be observed. This is sometimes the view of Feminist, Constructivist, and various Post-Modern approaches. When this is assumed, “theory” is not consistent with the Popperian or Bayesian understanding of what a scientific theory or explanation must be. Instead, theory means a convenient or insightful way of grouping ideas or observations parsimoniously, an explanation in the sense Occam probably originally intended.
Unfortunately, which of the three approaches is being used is often not clear to readers. In some cases, it seems they are not even conceptually distinct to writers and political scientists, who use the term theory in ways that skip back and forth between these approaches. Given that, I’ll discuss the primary “theories” and approaches in political science.
Approaches to Political Science
For this overview, I’m going to treat the approaches as if they are completely separate. They aren’t, and most good political science is not dogmatic, and takes insights from different approaches where they are helpful. I’d even argue that once we get past the fairly major obstacles of ego, narrowness of expertise, and confirmation bias, there are relatively few disagreements between political scientists about the places where each approach is limited, and where each is especially helpful.
But it is helpful to consider them separately, just to understand that the approaches do bring different ideas to the table. This is true even though sophisticated analyses of each type begin to converge in many ways. Rational choice theorists may account for institutional and Marxist arguments in the structure of their models, or the motives of the actors may account for norms and domestic concerns of each actor. Similarly, “modern” behavioral approaches are free to consider Nash equilibria and other rational actor tools. And given that caveat, the below table, also linked, has a good overview of some of the viewpoints I’ll discuss.
Behavioralist Political Science is the attempt to look at actual political behavior (hence the name,) and find falsifiable laws that explain and predict it. Before the “behavioral revolution” that happened around the 1940s and 1950s, much political science was legal theory and speculative and untestable theorizing. Behavioralism (and not “behaviorism,” which is a quasi-related approach in psychology,) instead focused on what individuals did, and what the results were. The new approach was value-neutral, and was in part a result of, and response to, the claim that political science wasn’t a science at all. This became the dominant approach in political science in the 1950s, ’60s, and ’70s.
Behavioralism promoted systematically collecting data and testing ideas. Case studies and experimentation were able to be used at small scales, and these became key tools. Unfortunately, they were not able to answer many key questions in political science, which usually dealt with larger groups. Surveys quickly became a key tool for political scientists. Behavioralists were also a key group in developing statistical approaches to analysis appropriate to social science, where confounders and lagged effects were key challenges. But it turns out that there is really very little that can be done with only data. Instead, generating hypotheses to test requires theory, in the second sense above, to at least drive the choice of data to gather.
In the 60s and 70s, there was a pushback against Behavioralism. It was attacked as mindless empiricism, and the insistence on value-neutrality meant that many interesting questions, such as what approaches would be better for the populace, were ignored. It was also effectively useless for looking at international relations, which was heavily reliant on theory, considering state motives, and looking at institutional decisionmaking. As a consequence of these criticisms, modern Behavioralism accepts the importance of theory, and is far less insistent on only considering value-neutral questions.
Rational Choice Theory was an approach adopted by political scientists from economics, and political economics, starting in the late 1950′s. Starting from the assumption that groups are self-interested and act (approximately) rationally, it can make clear predictions about what groups will do. This is close to the deuctive approach mentioned earlier. Rational Choice was a fairly successful alternative to Behavioralism, and it was still justifiably scientific, while addressing questions that Behavioralism typically did not. This was specifically critical in International Relations, where the equivalent approach is called Rational Actor theory. (This will be discussed in the next post.)
Rational Choice Theory uses mathematical tools, and relies heavily on game-theory, since almost all of political science is about the interaction of people, groups, and states. This means that perhaps the most important question for rational choice in political science is the level of analysis—who the rational actors are. This can also include qualitative evaluation , for example, to directly elicit from experts what the consequences of a given decision will be, rather than to model the scenario directly. And even within quantitative modelling political scientists use expert elicitation and surveys, as well as various other tools from the social sciences, to find inputs to their rational actor models.
In the 1990s, some confidently predicted that all of political science would be subsumed by rational choice theory. And before giving the area more thorough examination, this was certainly my view. But they were wrong—and even to the extent that they are correct, it would be a mistake to ignore other theories. I’ll explain more about why this is a bad idea in the next post on international relations, but at the very least it ignores key issues that matter for actual decisions. That’s at least part of why rational choice theory is currently a minor (albeit influential) approach in political science. (Again, it is much more important in international relations.)
Institutionalism focuses on institutions, norms, and the structure of governments. In an earlier form this was the version of political science that talked about how to write constitutions, and continued the philosophical tradition of Aristotle’s “Politics,” Machiavelli’s “The Prince,” Hobbes’ “Leviathan.” (Interestingly, US President Woodrow Wilson had a PhD in political science, and was very much an institutionalist.)
The approach has been reinvigorated more recently as a way to respond to some of the shortcomings of behavioralism and rational choice theory. Early behavioralists wanted to move past legal fictions and org charts, and get to what was happening based on “real” behavior. When taken to an extreme, this ignored the extent to which those real behaviors we influenced by laws, norms, and the structure of government. Rational choice theorists, on the other hand, start with the idea of preferences. This meant that they would sometimes ignore the way that these preferences were turned into decisions by the structure of the system.
Institutionalists argue that institutions are more than aggregates of individual preferences and individual behaviors, and they should be treated as such. In short, reductionist approaches were unhelpful because they were not the correct level of analysis. The problem is defining what the correct level of analysis is, and what institutionalism has to say.
Modern Institutionalism has dozens of viewpoints about what should be analyzed, each taking a different but plausible set of ideas and questions. A few examples of these approaches should give a sense of what types of approaches these are, Historical institutionalism looks at how institutions developed, and how that influences their structure and the way decisions are made. Network Institutionalism looks at how informal interactions between groups and individuals can shape decisions, and how power is wielded by those perhaps without official positions. Normative institutionalism considers how norms and beliefs shape decisions and constrain of guide institutions. And rational choice, Feminist, and Marxist institutionalism takes each of those viewpoints and applies them to the study of institutions.
Psychological political science uses a psychological lens to understand political actors. This can involve understanding why and how individuals make political decisions, how leaders react to stimuli, and how systematic departures from rationality influence decisions in ways that change the types of ideas espoused by rational choice theory. Political approaches are different than approaches like constructivism, but can be related by the fact that they view psychology and subjective beliefs as a key factor.
I also want to mention Constructivist, Feminist, and Marxist Political Science since I think they often get dismissed in discussions by outsiders, despite being useful. These are all based on the fact that people construct narratives, and their decisions are shaped by their inner viewpoints. These inner viewpoints may be subjective, but they still influence the world—during the Russian revolution, it was far less important whether Marx or Stalin was correct in their view of the world than it was that many people chose to embrace that viewpoint. Given that, specific worldviews can be very influential, and provide useful ways to understand the world.
There are particularly extreme viewpoints that come under these umbrellas, however, that do things like dispute the existence of objective reality, or claim that scientific knowledge is impossible. These viewpoints, while being parts of these movements, have made it easy for people to dismiss the insights more generally, often in ways that make them ignore important issues.
In fact, I would guess that many people who dismiss Marxist and Feminist political science still embrace many of the theoretical viewpoints that they espouse. If you think that we need to consider the impacts of unemployment or automation on politics, you’re using a Marxist lens which looks at different social groups as important political actors. Marxism sees much of politics as a struggle between different groups of elites and segments of the public. (Ironically, the elitism versus the public is often a talking point of the modern American right-wing.) Similarly, if you think that minorities, women, and other groups are distinct from other groups on the basis of their identities, you’re using a feminist lens.
There are other approaches as well, but occupy smaller niches. To name a few, comparative politics looks at differences in political systems, and there are political scientists who specialise in specific types of political behavior, like nationalism or security studies, or specific issues like election systems (psephology) or international development. Each of these is potentially worth further discussion, but my current plan is to focus on other issues, so they won’t be discussed.
(1) This section is much less an overview of political science as understood by practitioners, and much more my personal way of understanding and justifying the (mis)use of terms like “theory” in these areas of social science. When I was first exposed to psychology, political science, and related areas at a graduate level (after doing my BS in mathematics,) I was frustrated that, it seemed, these areas had never heard of Karl Popper or falsifiability. I was wrong, but it took me quite a while to figure out what I was missing—I’m hoping this is helpful to others bridging a similar inferential gap.
(2) Definition adapted from Wikipedia. Their definition of political science cites the Oxford Dictionary, and says it “deals with systems of governance, and the analysis of political activities, political thoughts, and political behavior.” This overuses the word politics, which, according to the link, “involves making decisions that apply to groups of members.”
(3) Chemistry certainly didn’t start with a correct model of physics and build on that—early advances like Boyle’s law (discovered a half dozen blocks from FHI) were based on in-retrospect very insightful but completely wrong models. Boyle thought that the particles of air were static, and held in place via some sort of spring that kept them in place. Newton’s work would later justify the incorrect model of static particles, but it took a few more centuries to discover a proper justification based on kinetic theory.
(4) Bottom-up and top-down models are sometimes discussed, but the terms seem really ambiguous in this context, so I won’t use them.
(5) In part, this was instead of trying to discuss politics in terms of what different institutions did. I’ll talk about that in the below section on Institutionalism.
(6) In the early 1900s, sampling for surveys was first understood and used, and the methods were first used for sociological research in the 1930s. Note all the amazing datasets from surveys performed on an ongoing basis in the US over decades were largely due to the new ideas of behavioralism. This includes the Census Bureau, which in the 1940s starting asking about things other than just population, such as housing, mortgages, utility bills, radio and TV ownership, and indoor plumbing .
(7) It is possible to address many such questions without accepting a set of values as correct, but many of these questions were not in fact addressed.
(8) When predicting individual behavior, rational choice is an economic and psychological model. Applied to groups and institutions, however, it addresses questions in political science.
(9) Returning to the earlier analogy between political science and chemistry, we could in theory build a game-theoretic model of every individual, each choice they make, and how this interacts with all other people to make predictions at the group, state, or international level. In practice, this is as useful as building a model of the human body at the level of quantum chromodynamics. Instead, we assume that some higher-level group is a rational actor.
(10) For example, first-past-the-post voting schemes, gerrymandering, and informal coalitions are an institutional reason that individual choices are not reflected in political decisions. When building naive game theoretic models, it’s easy to miss these considerations, and accounting for them can make the game theoretic models very complex, make the results fragile to exact assumptions and very small changes in relative utilities, or make the models too flexible to be predictive.
(11) I am being unfair by grouping these and giving them so little attention, and will note that they are plausibly worth more attention, as I will explain, but I am far less familiar with them than the other approaches.
(12) It’s worth noting, however, that Feminism was a political movement that later turned into a viewpoint for studying and understanding politics—unlike the other approaches, which were not political movements, or even Marxism, which originated as scholarly ideas, and even once it became a political force, it still had a specific set of academic viewpoints.