A network approach to the production of geographic context using exponential random graph models

The notion of context continues to be both an enduring rationale and empirical problem for addressing human agency for geographers. Despite its centrality to geographic scholarship, context has largely been an abstraction in geography with relatively little effort to either clarify what it means or how to formally operationalize it for research purposes. When context has been formally addressed, it has primarily emphasized either impacts on agency at a specific scale through a reliance on a place-based interpretation. This paper takes up the issue of context by developing a multi-scalar theoretical framework that is suited for use with social network-based statistical models called exponential random graph models or ERGMs. The theory of context emphasizes the importance for both geographic and social contexts for agency while also recognizing place specific and larger scale influences. Using network data about World War I, a series of ERGMs are developed to demonstrate the importance of multiple types of contexts to the observed outcomes. The approach used in this paper reinforces the old truism that context matters by demonstrating it empirically. Most importantly, this paper illustrates the value of continued engagement with a wider spectrum of the theories of how and why context matters.