We examine differences in income within the U.S., and the regions of persistent poverty that have arisen, using a newly assembled dataset of counties that links historical 19th century Census data with contemporaneous data. The data, along with an augmented human capital growth model, permit us to identify the roles of contemporaneous differences in aggregate production technologies and factor endowments, in conjunction with the historical roles of institutions, culture, geography, and human capital. We allow for possible cross-county factor mobility via a correlated random effects GMM estimator that identifies simultaneously the coefficients on time varying and time-invariant determinants of income. We find evidence of significant regional differences in production technologies, but our decompositions of the poor/non-poor income gap suggests that at least three fourths of the gap is explained by differences in productive factors. Persistently poor counties are different (and poorer) primarily because they have lower levels of factors of production, not because they use the factors they have less efficiently. While much of the income difference is explained by contemporary factors, the contribution of historical levels of human capital is surprisingly large. The combined contribution of historical and contemporary human capital is striking: together, they explain nearly 60 percent of the overall income gap between the persistently poor and non-poor counties.

Document Type

Research Paper

Publication Date


Discussion Paper Number

DP 2012-01