A Critique of Dyadic Design

As social scientists, social systems comprise the objects of our inquiries. International Relations (IR) scholarship, as the subfield’s name suggests, focuses on relational systems. We observe relational systems as relational data, in which data exhibit two levels of observation: the actors (units, or nodes and vertices in network theoretic terminology) and the interactions among actors (relationships, ties, or edges in network terminology). Relational data[1] includes both data on single dyads (e.g., a conflict between two states) and data consisting of several actors and simultaneously established ties (e.g., a multi-state defensive alliance). The relationships of interest (e.g., conflict, alliances, trade) are often reduced to observations of whether pairs (i.e. dyads) of nations are incident to domain-specific connections (e.g., at war or mutually defensively allied). A research design is dyadic when such dyads are considered as isolated components of the larger relational system.

In this article, we discuss the limitations of dyadic designs for developing explanatory, causal, or predictive models of international relational systems. We show that there exist two conditions under which dyadic designs provide an appropriate approach. Both conditions relate to hyperdyadic dependence; a condition under which the state of dyadic relationships depends upon the states of other dyadic relationships. The first condition is simply the assumption that there are no hyperdyadic dependencies (i.e., that dyads do not depend upon each other). The second condition is an absence of interest in hyperdyadic dynamics paired with an assumption of no confounding of systemic patterns and covariates of interest. If neither of these conditions holds then strictly dyadic designs are inappropriate and will often lead to faulty inferences.



[1] We should note that relational data are distinct from but potentially related to relational theory. For example, relational social theory – popularized in voluminous works by scholars such as Tilly (2003) and White (2008) – takes a network oriented perspective on social relations. Relational data are necessary to test relational theories, but one need not have a relational (network) theory to analyze relational data

Full Text PDF   Find Replication Data
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