The notion of affinity among countries is central in studies of international relations: it plays an important role in research as scholars use measures of affinity to study conflict and cooperation in a variety of contexts. To more effectively measure affinity, I argue that it is necessary to utilize multidimensional data and take into account the network context of international relations. In this paper, I develop the deep affinity concept and introduce a new algorithm, the three-step graphical LASSO (GLASSO), to infer and recover latent affinity networks. This technique leverages the abundance of monadic and dyadic state-level data to identify the presence or absence of affinity links between pairs of countries. Directly incorporating network effects and using a variety of multidimensional data inputs, I used the three-step GLASSO to estimate latent affinity links among countries. With these data, I examined the implications of affinity for international conflict and foreign direct investment, and found that the measure of affinity generated with the three-step GLASSO outperformed alternative affinity measures and was associated with decreased conflict and increased economic interaction.
Read full abstract