Where to meet?

Clarion Congress & Hotel, Luna


David Brown2, Jaclyn L.W. Butler1, Brian C. Thiede1, and Leif Jensen1

1The Pennsylvania State University, 2Cornell University

Topic: Income Inequality Across the Rural-Urban Continuum

Keywords: Income inequality, Pareto Gini Coefficient, Spatial Inequality, Demography, Rural Urban Continuum


Background and Motivation

Income disparities have increased in recent decades, producing an “age of extremes” in which resources are concentrated among a selective segment of the population. This issue is well-studied at the national level, but there has been less attention to spatial variations in the degree iof inequality at the subnational level, including among rural places. We address this gap, with the goal of understanding recent income inequality dynamics in non-metropolitan U.S. counties. Using data from the Census Bureau, we describe and map levels of within-county income inequality in 2016 and 1970. These techniques allow us to (1) examine how relationships between the degree of income inequality and other sociodemographic variables vary between nonmetropolitan and nonmetropolitan counties, (2) how these relationships vary within the nonmetropolitan spatial sector, and (3) examine changes in these patterns and the spatial diffusion of income inequality between 1970 and 2016. We also conduct a multivariate analysis that accounts for the spatial structure of inequality within this time period. Accounting for spatial structure, relationships, and interactions across counties not only more accurately reflects the demographic relationships between these counties, but also improves statistical models.


We contend that our particular focus on the rural United States is merited for several reasons. First, the rural population is sizable. Defined as those living outside of metropolitan areas, nonmetropolitan Americans comprise 15 percent of the US population, and these 46 million people are spread across 72 percent of America’s land area (Economic Research Service 2017). Second, rural communities have distinct and heterogeneous demographic legacies and trajectories which interact with inequality in compelling ways that are inherently worthy of study. Finally, as the 2016 Presidential election reminded the nation, rural areas hold disproportionate political power relative to their population size (Monnat & Brown 2017; Scala and Johnson 2017). Many rural places are both acutely susceptible to recent domestic and global forces affecting increased income inequality and reliable supporters of political movements with platforms that are often antithetical to inequality-reduction.


Data, Measures, and Research Strategy

We draw on county-level summary files from the 2016 and the 2010 American Community Survey (ACS) 5-year samples and the 1970, 1980, 1990, and 2000 decennial Censuses.[1] Our outcome of interest is within-county household income inequality, which we measure using the Gini coefficient. To calculate the Gini coefficient for U.S. counties, the variable of household income, as represented by the number of households falling within income bins, was extracted from the decennial censuses and five-year ACS estimates. We then calculate the Gini coefficient at the county-level using the Pareto method, which assigns a midpoint to all income bins (von Hippel et al., 2017). We use two sets of stratifying variables to disaggregate the non-metropolitan sector into meaningfully different spatial categories. First, we use the binary metropolitan versus non-metropolitan delineations produced by the Office of Management and Budget (OMB). Second, we use the USDA-ERA Rural-Urban Continuum Codes (RUCC) to examine systematic variation within the non-metropolitan sector itself.


The research strategy consists of three major components. First, we examine how levels of income inequality vary between metropolitan and non-metropolitan areas. Regression methods that incorporate spatially explicit variables, such as spatial regimes analysis, are used to help address this question. Second, we examine how levels of income inequality vary across non-metropolitan county types. We will produce demographic profiles of high- and low-inequality non-metropolitan counties to understand whether and how the rural populations residing in such contexts differ. High- and low-inequality counties are defined here as the top and bottom 10 percent of non-metropolitan counties in terms of their Gini coefficients. There were 1,976 non-metropolitan counties in 2013. The demographic profiles include measures of population change, density, race and ethnicity, nativity, age structure, educational attainment, income sources, industrial-occupational structures, and employment status. Finally, we replicate these analyses for 1970, examining levels and patterns of inequality from 1970 to 2016.


Preliminary findings

Overall, we observe a modest uptick in local inequality during 1970-2016, with the average Gini coefficient across all U.S. counties increasing from 0.412 to 0.432 over this period. Local income inequality is, on average, higher in nonmetropolitan than metropolitan counties across both study periods. However, we find evidence of convergence over time, as the metropolitan vs. non-metropolitan gap in the mean Gini coefficient decreased from 1970 to 2016. This diminished gap results from an increase in metropolitan inequality, not a decline in non-metropolitan inequality.


Our examination of variation in local income inequality among non-metropolitan counties in 1970 shows that local income inequality increases as one moves from more urbanized, metropolitan-adjacent non-metropolitan counties to the least urbanized and non-adjacent counties. This gradient is largely eliminated by 2016. Levels of income inequality are uniformly high across the U.S. geography in 2016, suggesting little correlation between the degree of rurality, so defined, and local inequality.


Our county-level maps clearly illustrate the overall increase in income inequality between 1970 and 2016. They also highlight important regional variations, with some places (e.g., the non-metropolitan South) characterized by persistently high levels of income inequality, others (e.g., parts of the central Great Plains and inter-mountain West) by persistent equality, and yet others (e.g., the northwest and northern Rockies) by rapid increases in inequality over the study period.


Future research

To complete this paper, we will build on these preliminary analyses in three primary ways. First, we will test the sensitivity of our findings by using alternative metropolitan (non-metropolitan) delineations. Second, we will also test the sensitivity of our findings to alternative approaches for estimating income inequality using binned income data (von Hippel et al. 2017). As the project progresses, we will expand our repertoire of spatial analyses through Geographically Weighted Regression (GWR) and spatial panel modeling. GWR will be used to examine how the relationship between income inequality and the sociodemographic variables of interest varies across non-metropolitan areas, and spatial panel modeling will be used to examine the spatial diffusion of income inequality over time.



This research was supported by a grant from the USDA National Institute for Food and Agriculture (2018-67023-27646). The authors also acknowledge the assistance provided by the Population Research Institute at Penn State University, which is supported by NIH infrastructure grant P2CHD041025, as well as support from USDA Hatch Multistate Research Project W4001, “Social, Economic and Environmental Causes and Consequences of Demographic Change in Rural America.”



Monnat, S. M., & Brown, D. L. (2017). More than a rural revolt: Landscapes of despair and the

2016 Presidential election. Journal of Rural Studies, 55, 227-236.


Scala, D. and K. Johnson. 2017. “Political Polarization Along the Rural-Urban Continuum? The   Geography of the Presidential Vote, 2000-2016.” Annals of the American Academy of Political and Social Sciences. 672: 162-184.


von Hippel, P. T., Hunter, D. J., & Drown, M. (2017). Better estimates from binned income data:

Interpolated CDFs and mean-matching. Sociological Science, 4, 641-655


[1] We use the summary files because microdata with county identifiers are not publicly available.

Go back to the workgroup WG 19