Dyadic Research Designs: Progress or Postmortem?

To what extent are international relations scholars constrained in their ability to answer important questions by what has become the workhorse unit of observation for analyzing relational data: the dyad? Skyler Cranmer and Bruce Desmarais argue that focusing on dyadic data has led scholarship to fail at properly characterizing relationships of interest in international relations data, while Paul Diehl and Thorin Wright (2016)and Paul Poast (2016) offer conditional defenses of using dyadic data. However, all three acknowledge that many of the problems identified in Cranmer and Desmaris (2016) carry with them both inferential and substantive merit. Journal space prevents a fuller exploration of this debate in the pages of International Studies Quarterly. Thus, to facilitate this important discussion, we invited eight scholars to weigh in on the question of dyadic data.

 

Our contributors (listed alphabetically) represent a range of traditions in empirical work—from network analysis to standard dyadic analysis—and in substantive orientation—from international conflict to civil conflict to international political economy. Allison Carnegie and Tara Slough acknowledge that all models are essentially that—models—and that researchers must be aware of what limitations their empirical models impose when they attempt to measure features of or characterize relationships between variables in their data. Sarah Croco wonders about the extent to which, by perhaps overfitting dependencies, analysts can trade one kind of bias for another if the use of network analysis is not sufficiently grounded theoretically. Kathleen Cunningham notes that, while dyads have improved the quantitative study of civil war, the unit faces specific challenges (say, of the endogenous observability of rebel groups, who are defined by their value on a possible outcome variable) for which a move to network analysis might to be a solution. Cassy Dorff addresses the problem of how to ensure that students are made aware of the costs and benefits of different research designs, how they can develop facility with network approaches and thus be equipped with the tools to choose among different research designs responsibly.

 

Next, Brandon Kinne clarifies the role of a particular model, the ERGM, in properly characterizing relationships in one’s data, estimating dependencies (while not forcing them), and showing the value of such work in IPE. Patrick McDonald emphasizes some theoretical reasons to model interdependencies, focusing especially on the problems of pairing dyadic data with systemic theories, which suggest that the statistical relationship known as the democratic peace is an artifact of failing to model hyper-dyadic relationships. Toby Rider argues that Cranmer and Desmaris seem to conflate theoretical and empirical models of dyads. He cautions against associating theories with any single research design, and notes the inherent problems posed by confronting theoretical models with empirical models in any meaningful “testing” relationship. Finally, Kindred Winecoff defends Cranmer and Desmaris (2016) against specific challenges from Diehl and Wright, as well as Poast. But he  also notes an additional potential problem with dyadic data (also discussed by McDonald) that stems from its inflation of the number of observations in international-relations data.

 

This amounts to what will likely prove an long-lasting debate for the field. We hope that these collected contributions can contribute further to an important issue in the advancement of the field.

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