Reconsidering Dyads. Again.

It was with great interest that I read the exchange between Skylar Cranmer and Bruce Desmarias (2016), Paul Diehl and Thorin Wright (2016), and Paul Poast (2016) on the use of dyads in international relations research.  Having written about dyads (and dyad years) in the past, I was sympathetic to Cranmer’s and Desmarias’ concerns. The lack of independence between observations, especially in the context of multilateral events, and other “hyperdyadic” influences merits attention, but I am not fully convinced that the call to abandon the dyad in favor of a more networked approach is the best course of action. My reasoning is twofold.


First, as the other pieces in the exchange argue, the bias problem Cranmer and Desmarias raise is likely not as pernicious as the authors make it out to be. The actions of other states outside of the dyad may be highly relevant in some contexts, but matter less so in others. It is telling that one of the primary examples Cranmer and Desmarias cite to illustrate the potential for bias is World War I. I am guilty of doing the same in my earlier work on dyads—it’s a juicy case that is nearly impossible to resist when discussing interdependence. I wonder, though, if we are putting too much weight on this case and the other (thankfully few) large, multilateral wars throughout history. We know that states behave strategically, considering the moves of potential third parties, but this doesn’t mean the actions (or possible actions) of other states always accelerate the spread of war like they did in June of 1914. As Diehl and Wright point out, this influence can wax and wane. While they were speaking in the context of 3rd parties affecting the behavior of enduring rivals, there is no reason to believe the degree of interdependence between states will remain constant.


Second, I am not confident that the method Cranmer and Desmarias propose to fix the problem—an exponential random graph model (ERGM)—represents a silver bullet.  One could argue that any attempt to account for missing factors is a step in the right direction. Yet applying this fix also requires the researcher to make choices about what characteristics best capture the linkages between states.  Some of the possibilities Cranmer and Desmarias suggest are intuitive (e.g., being sanctioned by the same state), but others (e.g., UNGA voting similarities) strike me as a catch-all that may be imposing ties across states that don’t exist in reality. In this instance, we may be trading one source of bias for another.


In sum, I found myself softening my views on the downsides of dyadic research.  I agree with Poast’s compromise that we should “use dyads, but proceed with caution”. When dealing with phenomena that are obvious multilateral events (e.g., treaties involving more than two states), alternate approaches, such as networks or k-adic designs, may be more appropriate.



Croco, Sarah E. and Tze Kwang Teo. 2005. "Assessing the Dyadic Approach to Interstate Conflict Processes: A.k.a."Dangerous" Dyad Years." Conflict Management and Peace Science. 22.1: 5-18.

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