So I’m finally getting to the point where I’ve got time to dive head-first into my first big post-dissertation research agenda, which is all about balancing the ledger sheet, as it were, on the use, effectiveness, and desirability of building foreign policy coalitions. You can find component papers here and here, as well as some recent slides for the latter paper that involve some early empirical tests of the conditions under which coalitions provoke counterbalancing. (On a side note, I put the slides together after the paper, but I still maintain that Hal Varian’s advice that the format of a good paper should follow the format of a good talk is the way to operate. So do as I say, not as I do…) Today, though, I want to talk about why IR scholarship hasn’t examined coalitions in the way I think we should, what my solution is, and what it might contribute to the larger project of understanding war in international relations.
From Iraq in 1991 and 2003, Kosovo in 1999, and post-invasion Afghanistan to the Iranian and North Korean nuclear programs, making joint coercive threats as part of a coalition has become a pretty consistent modus operandi for the United States and other great powers, yet both theoretically and empirically, IR scholarship remains pretty tied to the dyad or dyad-year as a unit of analysis. So not only is the international system a networked, N-adic world with all kinds of relations between units not observed in the dyad, it’s also populated by important actors other than states. Specifically, groups of states often work together to make coercive threats of their enemies, such that observing a state-v-state dyad in a coalitional context makes very little sense, because, at the very least, we’d have inaccurate measures of things like military power, distance, alliance ties, etc., because we’d simply be working at the wrong level of aggregation. States often act together, such that weak ones will act with the expected leverage of a stronger partner’s military, and surely we wouldn’t want to measure that small state’s potential military power in a dispute as simply its own capabilities. Clearly, this isn’t the best way to go, yet the best we typically do as researchers is cluster our standard errors by conflict or side, but if we’re interested in coalitional dynamics, we surely don’t need to treat them as statistical nuisances.
Why does this happen? Why do coalitional disputes get treated empirically as lots of disaggregated dyadic relations? Well, it’s not that we, as a field, are ignorant of the problems posed by particular units of analysis. However, we do like to economize on effort, and since most conflict data is used dyadically, I think we tend to go with what we’ve got and just ask dyadic questions when, in many cases, we probably shouldn’t be. Further, lots of theories of war and peace are cast at least implicitly at the level of the dyad, because complex games with more than three actors are hard to solve and, just as importantly, hard to present. So IR scholars have just economized on both theory and data for quite some time, but at the expense of studying one of the enduring and, arguably, very consequential features of international relations: coalitional diplomacy, threats, and war-fighting.
What’s the solution? Mine is twofold: undeterred by practicality and good sense, I’ve tried to build a general framework with which to model the processes of diplomacy, warfighting, and alignment in a coalitional context, and the first two such attempts at writing down and solving these games are linked above. But lots of other scholars have looked at multiactor models as well, so this isn’t all that big of an innovation. Empirically, though, I think we can gain a lot by just changing the level of aggregation at which we examine conflicts. Rather than look at pairs of opposing states, my initial approach has been to look at pairs of opposing sides in a crises/disputes/conflicts, such that states making joint threats or fighting together are one element in the dyad, and the state(s) they face in the crisis are the other. So we keep the dyad, but we simply aggregate at a level above the state to capture coalitional dynamics in a way that we couldn’t before.
To give a sense of the utility of this approach, you can see the slides above, which include an early attempt to use side-dyads in ICB crises to predict the occurrence of counterbalancing as a function of the characteristics of a coalition. The results are reasonably supportive of the theory’s predictions, though at this point there’s clearly a long way to go. Still, I think that, rather than extend our analysis to brute-force style to triads or some larger number of components, we can use the logic of our theories to make better judgments about units of analysis. In the coalitional case, I tend to think that side-dyads make a lot of sense, but it took theory to get us there and inform exactly where the predicted dynamics should play out.
So what’s the point of this discussion? Mostly that there’s a lot of untapped potential in the study of war and peace if scholars start taking their theories seriously in terms of identifying the proper empirical domain and unit of analysis. If we take the data as given, as I suspect many of us do, then we’re perhaps less likely to think outside the box it gives us, less likely to push inquiry into new areas that would require thinking about data differently. But to the extent that we do push outside that, we’ve got the chance to get at some enduring questions about IR—what causes war? how do states make alignment decisions? when can coalitions cooperate effectively?—in new venues, which, I think, is a good thing.