Dani Rodrik’s piece on the reasons why political economy needs a proper account of ideas has provoked a debate among international relations bloggers. See Phil Arena, Moonhawk Kim, Thomas Oatley, Kindred Winecoff and Jeremy Wallace. Rodrik also has written a follow up blogpost calling for a new research agenda on ideas.
As some of my readers complained, this argument is not unfamiliar to “constructivists” and others within political science who have long made the case for the supremacy of ideas. Economists too have recently begun to focus on the formation of political ideology (through media and other formative experiences). But what I have found missing in all this literature is an overarching framework that parses out the respective contributions of ideas and interests, without necessarily giving supremacy to one or the other. What is also missing is empirical work (“systematic empirical evidence” as an economist would call it) that takes this approach to the data.
This is something I and other people have been thinking about on and off for the last couple of years. Below the fold, some brush clearing (a lot is needed) and thoughts, presented as a numerically ordered list, as to what a different and better account of the role of ideas in political economy might look like. NB that these thoughts owe a lot to this piece with John Quiggin (a cleaner version of which is about 70% written), an as-yet-not-fit-for-human-consumption piece with Cosma Shalizi and possibly one other co-author, and various conversations with people, in particular Mark Blyth and Kathy Thelen, over the last couple of years. None of these other people are responsible for any stupidities below; they are surely responsible for any insights. All of these contentions are open to argument. But here goes …
(1) It’s a bad idea to start this debate by thinking of idea and interest based explanations as competing with one other. They are not. There may be a few sincerely deluded structuralists out there, who genuinely believe that interests may easily and automatically be read from some kind of external structure (whether this be the international system, the class struggle or market position is irrelevant), but they can reasonably be ignored. Interests are the product of individual perceptions, which in turn means that they are inevitably intertwined with ideas, and vice-versa. It may sometimes be analytically convenient to focus on interests and pretend that ideas don’t exist, or vice versa, but one should never mistake assumptions-for-convenience for ontological claims.
One implication of this is that constructivist-v-rationalist disagreements in international relations are not nearly as fundamental as they might appear (for a slightly different take on this, see Fearon and Wendt, passim). When game theorists talk about the influence of outcomes that are off the equilibrium path of play, they are really saying that people’s beliefs or ideas (which term one uses seems to me to be a matter of disciplinary sociology rather than conceptual clarity – the two fade into each other) about things that will never happen have consequences for the things that they actually do. The argument between rationalists and constructivists isn’t one over whether ideas do or do not have consequences. Both agree that they do. It is over when it is appropriate to apply one approach to understanding the consequences of ideas, and when it is appropriate to apply another. Game theory makes strong and often unrealistic assumptions about actors’ beliefs and ideas (i.e. common knowledge with all of its attendant weirdnesses ), but can very nicely model how actors will respond to each other given these constraints. Constructivist approaches are often more realistic but more fuzzy. The choice over which to use is, or should be, a pragmatic one.
(2) Ideas are not, actually, collective phenomena. When Moonhawk Kim says:
Any ideational theories most likely require some methodological wholism, in which individual actors are not the sole ontological unit. Ideas are collective phenomena and those that are not are not interesting. This is clearly at odds with the predominant although possibly waning focus on experimental methods in political economy.
he is expressing the general wisdom of the field of political economy. Constructivists typically emphasize holism, the collective nature of ideas etc etc. Rationalists for their part also assume that the ideas underlying actors’ strategic interactions are collectively shared (albeit indirectly, through assumptions like common knowledge and complete information).
But these, again, are convenient assumptions which have hardened into implausible ontological claims. Political scientists desperately need to read Dan Sperber’s wonderful book, Explaining Culture, which takes on accounts of culture as a collective phenomenon, and points out that they are obviously wrong. Simplifying his argument a bit, when I think about the concept of ‘deterrence,’ I associate it with a specific set of characteristics which likely have a lot in common with the ideas of other people who have read Thomas Schelling’s work. However, my internal mental representation of deterrence is necessarily going to differ, perhaps importantly, from that of these other people, given that processes of ideational transmission are inevitably flawed, I come to the concept of deterrence with a different set of prior experiences and associated beliefs than other people and so on.
Collectivist accounts of ideas tend to stress the similarities between different people’s understanding of deterrence or whatever, pretending not only that they are more or less identical, but that they have an existence which is somehow independent of the multiple partially overlapping, partly differing representation of ‘deterrence’ to be found in individual heads and representations (such as books, that are also prone to being interpreted in idiosyncratic ways when people pick them up). But this can be a big mistake. In Sperber’s relentlessly materialist framework, ideas and cultures are not collective, because:
there are only mental representations, which are born, live and die within individual skulls, and public representations, which are plain material phenomena – sound waves, light patterns and so forth – in the environment of individuals (81)
This isn’t just a debating point. It has clear research implications. If we keep this at the forefront of our minds, contra Kim, we can build theories of the ways in which these representations differ from individual to individual, and are disseminated from one individual to another, rather than treating them as a homogenous entity that exists independently of the specific representations that individuals hold in their heads. If we look at `ideas’ not as something concrete and uniformly shared, but instead as an imperfect shorthand for a set of mental representations which are approximately similar, which are commonly based on communication between one individual and another and so on, we can begin to think systematically about how (a) differences between these representations, and (b) the processes through which representations are communicated, can have important social consequences.
(3) Given the above, we can gather systematic empirical evidence about ideas. Sperber argues that we can and should think about ideas in epidemiological terms. In other words, we should think about how ideas spread and mutate from individual to individual, and what the forces are that determine their spread and mutation. One possible way to think about this is in evolutionary terms – we can think about the mechanisms that respectively (a) give rise to variation in ideas, and (b) lead to the selective retention of some ideas, while others wither away [Sperber thinks that cultural ideas mutate in too unconstrained a way for evolutionary arguments to have much purchase; Cosma and I are writing a paper which argues that changes in institutional beliefs are plausibly better behaved, so that one can bring in all sorts of fun modeling techniques from mathematical approaches to evolutionary biology]. Another is to turn to empirics. Phil Arena’s post starts from the assumption that idea-mongering is a waste of time, because ideas will only rarely have any influence. The more interesting underlying question, though, is whether there are empirical factors which shape the likelihood that ideas are, or are not taken up, leading to significant variation in outcomes.
One way to think about this is to turn, a la Sperber, to the epidemiological literature. There is a burgeoning literature that political scientists don’t pay any attention to (although sociologists do) about how network structure shapes the spread of epidemics (see here for a useful, if somewhat out-of-date overview). Many of these models pari passu could be applied to understanding the spread of ideas too. Different network structures are likely to be more or less susceptible to the spread of ideas. The crossover with the literature on policy diffusion and papers like this piece by David Lazer is too obvious to need much spelling out.
Another is to look at recent work in computer science. Jure Leskovec, Lars Backstrom and Jon Kleinberg’s work on meme-tracking suggests that one can treat the Internet as something like a cloud chamber, from which one can infer the progress and mutation of ideas through the tracks that they leave behind. Based on this kind of data, one can further reconstruct the networks of influence. To be clear – these are relatively crude proxies for the (unobservable) actual ideas. There is furthermore good reason to believe that we will not ever arrive at the kinds of simple predictive models that many political scientists still aspire to. What we know about the dynamics of cultural influence suggests that there’s a lot of stochasticity. Standard data is never going to allow us to distinguish in any satisfactory way between homophily (in which people are more likely to have social relationships because they’re like each other) and influence (in which people are more likely to become like each other because they have relationships). Finally, without other kinds of evidence, one cannot say much about the consequences of ideas for anything more significant than other ideas.
Even so, I think this points towards a very different way of thinking about ideas than the ways that dominate political economy today. By treating ideas as individual representations, which spread through some process of social contagion, we can get away from an arid dichotomy between constructivist and rationalist accounts. By studying how macro-level factors (such as network topology) shape the ways in which they spread and change, we can start using real data to arrive at something that approximates Rodrik’s demand for ‘systematic empirical evidence.’ To be clear – this research agenda would have its own blind spots and weaknesses. It will be better suited for answering some questions than others. But it surely would help clarify a set of debates that have become increasing murky, confused and self-referential.