Abstract
We aim to find agglomerations of U.S. counties that are partitioned by commuting patterns by representing inter-county commuting patterns as a weighted network. We develop and use a community detection method based on the configuration model to identify significant clusters of nodes in a weighted network that prominently feature self-loops which represent same-county commuting. After we apply this method to county level commuting data from 2010, we find regions that are significantly different from existing delineations such as Metropolitian Statistical Areas and Megaregions. Our method identifies regions with varying sizes as well as highly overlapping regions. Some counties belong to as many as six different statistically significant clusters but some do not belong to any. Our results offer an alternative way of categorizing economic regions from existing methods and suggest that geographical delineations should be rethought.