Spatial Effects in Dyadic Data

Political units often spatially depend in their political choices on other units. This is no less so in dyadic settings where, as in much of international relations research, the focus of the analysis is the pair or dyad of two political units. Yet, with few exceptions social scientists have analyzed contagion only in monadic datasets, consisting of individual political units. This article explicates and categorizes all possible forms of modeling spatial lags in both undirected and directed dyadic data. This enables scholars to formulate and test novel mechanisms of contagion, thus paving the way for a whole new generation of analyzing spatial dependence between dyads of political units. We illustrate the modeling flexibility gained from an understanding of the full set of specification options for spatial effects in dyadic data by an application to the diffusion of bilateral investment treaties between developed and developing countries, building and extending on a study published in this journal (Elkins et al. 2006). We find that in addition to the target country contagion modeled in the original article, there is also evidence for positive target-to-dyad and source-to-dyad contagion, whereas source country contagion seems to negatively affect the likelihood of signing a bilateral investment treaty.