In the U.S., COVID-19 messaging and policy implementation (i.e., school closures and stay-at-home orders) are largely administered at the state level. This can be problematic, as functional metropolitan areas can straddle multiple states, and a single state may have subregions that are not well-connected. Much of our messaging for emergencies (such as hurricane warnings) is not at the state-level but at the county-level for these reasons. Such state-level policies have already resulted in friction in local communities—especially in Georgia. To define units for which it is reasonable to apply homogeneous rules, we construct regions that capture core geographies of social and movement behavior.
 
To create effective geographic regions for policy implementation, we apply community-detection algorithms to five large networks of mobility and social-media connections to construct geographic regions that reflect natural human movement and relationships at the county level for the continental United States. We measure COVID-19 cases, case rates, and case rate variation across adjacent counties and examine these dynamics along the boundaries of functional regions and state boundaries.
 
We find that regions constructed using GPS-trace ("trip") networks and commuter networks are the most effective natural partitions for capturing COVID-19 'hot spots.' Conversely, regions constructed from geolocated Facebook friend connections resulted in the least effective partitions. Regions derived from migration flows, Twitter connections, and state boundaries showed mixed results. This analysis reveals that functional regions derived from mobility data are more appropriate geographic units than states for making policy decisions about opening areas for activity, assessing vulnerability of populations, and allocating resources.