Modeling Diffusion through Statistical Network Analysis: A Simulation Study and Empirical Application to Same-Sex Marriage

Abstract The study of international relations by definition deals with interdependencies among countries. One form of interdependence between countries is the diffusion of country-level features, such as policies, political regimes, or conflict. In these studies, the outcome variable tends to be categorical, and the primary concern is the clustering of the outcome variable among connected countries. Statistically, such clustering is studied with spatial econometric models. This article instead proposes the use of a statistical network approach to model diffusion with a binary outcome variable. Using a statistical network instead of spatial econometric models allows for modeling autocorrelation in policy outcomes rather than the corresponding latent variable, and it simplifies the inclusion of temporal dynamics, higher level interdependencies, and interactions between network ties and country-level features. In our simulations, the performance of the Stochastic Actor-Oriented Model (SAOM) estimator is evaluated. Our simulation results show that spatial parameters and coefficients on additional covariates in a static binary spatial autoregressive model are accurately recovered when using SAOM. To demonstrate this model, the paper applies SAOM to original data on the international diffusion of same-sex marriage and gives practical instructions for using such models.