Embracing the limitations of modeling choices: dyadic design, theory, and simplifying assumptions

Cranmer and Desmarais (2016) highlight a number of critiques they deem sufficient to abandon the dyad as a unit of analysis. Not wishing to cover the same ground as Poast (2016), as well as Diehl and Wright (2016), I focus on two other issues:  i) Cranmer and Desmarais’s critique of dyadic theory, and ii) their implicit equation of assumptions with assertions.


First, Cranmer and Desmarais focus primarily on empirical issues, but they also posit that “thinking dyadically’’ poses a a “serious challenge to the development of sound, consistent, and complete theoretical explanations…’’  I take their point to be that dyadic theorizing produces narrow and incomplete theories.  But a theoretical model is nothing more than a tool used to simplify the world and isolate relationships of interest, of ``limited accuracy, [only] partially represent reality, and are purpose build [by the user]’’ (Clarke and Primo 2012, 53)


Researchers structure theories around the questions they wish to ask, snd we should judge them as such, not by our own standards of what type of theory we believe the research should have pursued.  Cranmer and Desmarais appear to advocate the latter.  But scholars should be able to theorize about dyadic relations, with a focus on a narrow set of conceptual relationships, and assume away, for analytical simplicity, the influence of other factors. This is not to imply that extra-dyadic factors don’t matter. It’s a use of theory as theory was intended—to abstract away from the real world in ways that are useful if not completely accurate.  We eventually may develop organizational models encompassing national, dyadic and system level factors to explain the same phenomenon.  But that is not a standard by which to judge any individual theoretical model. 


When scholars translate their relational or dyadic theories into empirical models we rightly see that reflected in their choice of design—the dyad. Assuming away other dyadic or systemic influences is easier in a theoretical model, and key assumptions underlying our statistical tools must be taken seriously, but it is reasonable for scholars interested in dyadic outcomes to focus on dyadic analytic models.


Second, Cranmer and Desmarais argue that a “focus on the independent dyad neglects the fact that what happens at other levels of aggregation of the day has implications for which the higher-level outcomes are comprised.”  They appear to be treating what are clearly reasonable assumptions in our models as assertions.  No reasonable scholar asserts that these other factors do not matter at all. They develop theories at a level of analysis they think most appropriate for engaging that question, then choose an empirical strategy that fits the concepts and levels of analysis dictated by the theory.  And that very well may be the dyadic approach.


Take Ethiopia and Somalia rivalry, two countries engaged in an enduring rivalry over control of the Ogaden region.  Scholars might quite naturally wish to theorize about bilateral relations, decisions to seek outside alliances/support, emergent rivalry, arms competition and war. The rivalry was undeniably influenced by extra-dyadic factors. Somalia’s 1969 coup potentially contributed to a renewed pursuit of the Ogaden region, and Ethiopia’s engagement in Eritrea  presented Somalia an opportunity to launch a war in 1977.  Likewise, Cold War superpower competition provided opportunities for each to acquire arms, advisors, and allies.  Scholars can rightly choose to focus theoretically and empirically on any or all of those relationships and levels of analysis.  But a theoretical decision to focus on the bilateral relationship does not represent an assertion that a coup or the Cold War did not “matter” empirically.


This is not to trivialize the concerns highlighted by Cranmer and Desmarais, but merely to note that scholars have good reason to pursue dyadic theory and, under some proper conditions, dyadic empirical analysis. We can hope that their critique pushes us to think more carefully about the other factors we may need to account for when specifying our dyadic analyses.  And if there is knowledge to be gained about relationships from network analysis, then by all means we should pursue them. But we should pursue them not instead of but in addition to the dyadic approach.  Different theoretical and methodological approaches need not be in competition if they can each contribute to our collective body of knowledge. 



Clarke, Kevin A. and David M. Primo. 2012. A Model Discipline: Political Science and the Logic of Representations New York: 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