Make it so: Interdependence and the Next Generation

I commend all of the authors in this collection (Cranmer & Desmarais 2016, Diehl & Wright 2016, and Poast 2016) on a job well done: each highlights important dimensions of the debate for, or against, critiquing dyadic designs. All of the articles meaningfully contribute to the growing need for scholars in our field to better pair theoretical assumptions with appropriate statistical analyses. 


My first reaction centers around conceptualizing the outcome of interest—be that war, trade, rivalries or nearly any other international relations favorite. Diehl and Wright’s (2016) claim that connections outside of a dyadic relationship might not influence the dyadic relationship itself, such that the effect of state A’s external rivalries on State A’s rivalry with State B might not drive the relationship between A and B--seems to slightly miss the main argument of Cranmer and Desmarais, as well as other work done by myself and scholars of latent networks. That is, those of us who often conceptualize the dependent variable in matrix (network) form.


I accept that, in many cases, the role of multiple relationships might not be the main driver of an outcome of interest. But to argue that this then diminishes the utility of the network approach sorely misses the point, particularly when we imagine how to construct a dependent variable as a network. In these cases (among others), we need to consider A and B’s dyadic relationship in context of all other relationships (or dyads) that actor A or B are involved. Likewise, while some MIDs might only occur between a unique dyad pair, it does not follow that each actor in this unique pair is absent from all other pairs. At least one of the actors likely appears in multiple cases (imagine rows) of the data, violating the critical assumption of independence. Thus, while I agree with Diehl and Wright that we often err in assuming independence, I have a hard time understanding how the assumption of  ``complete interdependence” is similarly misguided.


My second reaction centers on a broader implication of this debate: how can we  learn new research designs? As Diehl and Wright correctly point out, the dyadic research design has a long history in International Relations (IR). But as others note, this dominant framework is merely the the simplest way to analyze a dyad. Other fields—computer science, sociology, social psychology, biostatistics—are far ahead of us when it comes to advancing different ways to assess relational data. Since there is often limited room for pedagogical questioning within research articles, I’ll raise the point here: how can we better lessen the ``start up” costs of learning network analysis? We owe it to our students to answer this question.  We can begin by granting network-oriented IR literature a commonplace role in our IR syllabi (or any conflict-oriented course, as these debates certainly apply to intrastate conflict). Along with this, we can, at the very minimum, teach students the fundamental differences between the accompanying data designs at the structural level. In an ideal world, a network analysis course would be as common as Maximum Likelihood or game theory and would include texts like Jennifer Victor’s Handbook of Political Networks but with a stronger pedagogical orientation. If political scientists fail to seriously embrace the task of learning and utilizing network analysis, they will not only forfeit this opportunity to the many government and commercial professionals already adopting these tools, but they will also systematically limit the field’s ability to engage in this complex, intertwined world of politics.



Victor, Jennifer Nicoll. 2016. Oxford Handbook of Political Networks. Oxford University Press.

Discuss this Article
There are currently no comments, be the first to post one.
Start the Discussion...
Only registered users may post comments.
ISQ On Twitter