Using Dual Graphs to Model Territorial Communities of Interest

Abstract Communities of interest are foundational to democratic representation in territorial constituencies yet are often un- or broadly-defined. The ambiguity of definition and process of community identification leaves practitioners, legal experts, and academics with relatively little clear empirical evidence from which to draw reliable conclusions. This reveals a problem with governance in territorial districting, where a foundational theoretical concept--the community as the centerpiece of representation--collides with practical limitations, including the characterization and identification of a community. To help mitigate and explore this problem, I introduce a graph-based model of territorial communities based on assumptions derived from available statutory definitions and use Census data to explore how communities can be represented, though the method is broadly extendable to a range of empirical characterizations of interest. This novel approach to community identification builds on existing graph-based methods for computational redistricting to facilitate new theoretical research into the interactions between communities and representation while providing a new tool to support practitioners' needs in identifying communities.