Identifying Communities of Interest Using Contact Graphs

A community of interest is generally defined as a territorially contiguous community of people with distinct, shared interests worthy of representation in a legislature. Preserving communities within voting districts gives communities the power to form voting coalitions, self-determine needs, and effectively organize and engage with their political representative. However, map drawers, community advocates and NGOs, and researchers face extensive problems identifying these communities, hindering community representation and academic research. I introduce a systematic method for identifying communities of interest using contact graphs and community detection algorithms and evaluate this method using three datasets: (1) 2020 Census population counts, (2) Census population estimates from the 2016-2020 5-year American Community Survey, and (3) cellular device geolocation data from four separate months across nearly 2 years. I explore and discuss validation, calibration, and use cases for these graphical communities before discussing policy implications, recommendations for use, and avenues for future research.