Typology of American Poverty

Abstract

This analysis seeks to better understand the geography of American poverty over time. Cluster analysis is used to group 34,908 minor civil divisions according to their similarity in mean-centered poverty rates from 1980 to 2000. Logistic regression is used to asses the groupings' statistical validity and accuracy. Results identify twelve statistically distinct groupings and that over three thousand subcounty places had poverty rates of nearly 20 percent above the national average going back to 1980. However, less than 50 percent of these fall within the U.S. Department of Agriculture's Persistent Poverty Counties. The new typology shows the diversity of poverty in terms and identifies places based on poverty's relative severity. The typology also uniquely identifies places that moved into and out of high poverty and identifies many poor places that are ``statistically invisible'' using existing typologies. Results show that correlates of poverty identified in the literature generally hold true across smaller geographic scales.