The 15 county types identified in the American Communities Project were derived from a standard clustering method of analysis, conducted by political scientist Iris Hui, PhD, where a set of 36 different indicators – everything from population density to military service members – were sorted using an algorithm that identified like places.
This initial analysis created 13 groupings. Dante Chinni, director of the American Communities Project, then created two more types using statistical measures based upon community differences he understood from his 20-plus years experience as a journalist.
1) In the Urban Suburbs group, we broke out counties that 1) held any of the nation’s 50 largest cities and 2) were above the norm in population density. That led to the creation of the 46-county Big Cities group. Those counties were broken out because in general they are much more diverse (economically, racially, ethnically) than the country as a whole. Politically, they are heavily Democratic.
2) The 13 types created one large category (676 counties) that lacked a strong distinguishing characteristic. It was slightly whiter, slightly better educated and slightly wealthier than the norm. To address the differences inherent in that group, we broke the group in two using population density numbers to create two types: Middle Suburbs and Rural Middle America. Counties with population above the norm were placed in the suburban group and places with population density below were placed in the rural group.
In addition, we went through the county types that statistically organized around specific racial, ethnic or population markers and made sure that individually their demographics fit with the broader type demographics. We wanted to remove obvious outliers. This resulted in the actual moving of only a small number of counties – roughly 50 total – to their closest type. We believed these moves were crucial in that the groups have to have a close semblance to the on-the-ground reality.