Advancing the applicability of exponential random graph models (ERGMs) for the analysis of social and other networks: Algorithms, implementation, and applications

Theoretical ideas, analytical concepts and computational tools made available by recent progress in the analysis of social network are diffusing rapidly across the social sciences. During the last two decades, Exponential Random Graph Models (ERGMs) have emerged as the most promising – and perhaps the most widely adopted – framework for the analysis of networks observed at one point in time. Recent research on exchange and dependence relations among individuals in formal and social organizations, organizations in markets and other institutional arenas, and countries embedded in webs of international relations, have all benefitted considerably from the progressive refinement of ERGMs. Today, ERGMs may be considered as part of a coherent, general and flexible framework for the analysis of data with complex network dependencies across multiple levels. ERGMs are now available not only for the analysis of social networks, but also for the analysis of bipartite networks, multilevel networks, and for the study of social influence. Against this general background, the main motivation for this project is the observation that the growth of interest in these models within the social sciences has outpaced the computational progress needed to scale up ERGMs for the analysis of larger network data sets that digitalization technologies now make routinely available. Despite considerable current efforts, their computational complexity continues to constrain ERGMs to the analysis of relatively small networks with, at most, a few thousand nodes. Building on recent algorithmic advances, the objective of this project is to alleviate or remove this constraint and extend the applicability of ERGMs to the analysis of larger network datasets that social scientists collect with increasing frequency, and at decreasing costs. The objectives of the project are mostly methodological. Expected deliverables include a suite of freely accessible software resources for the analysis of social networks, bipartite networks and multilevel networks, and social contagion.