The Primacy of Race in the Geography of Income-Based Voting

The forthcoming article “The Primacy of Race in the Geography of Income-Based Voting” by Eitan Hersh and  Clayton Nall is summarized by the authors here:

For decades, researchers have studied the relationship between income and voting, investigating where, why, and how much the rich and poor diverge from each other in their political preferences. In American politics, scholars such as Gelman et al (2008) and Alesina and Glaser (2004) have paid particular attention to how geography, and geographic context, shapes the relationship between income and mass politics. They have asked whether rich and poor voters have more divergent preferences in rich states or poor states, racially diverse states or racially homogenous states. Due to the limitations of survey samples, most research on income, voting, and geography has focused on variation across large-scale boundaries like states. However, at the scale of a state, it is difficult to disentangle explanations for the geographic variation in income-based voting. For instance, if the rich and poor have divergent preferences in a state that is both poor and racially diverse, is it more likely the economic context or racial context that contributes to the income-based stratification?

This question brought us into a partnership that connects Nall’s research agenda of political geography with Hersh’s research agenda of using granular public records to study politics.  In our article, we study sources of geographic variation at a much lower-level of geographic aggregation than previous studies. We employ 72 million individual-level party registration records and 185,000 precinct-level election returns. With these data sets, we assess factors that predict stronger or weaker income-based voting.  With an abundance of data, we estimate the relationship between income and partisanship in small geographies, like state house districts, mostly non-parametrically and with few modeling assumptions.

We discover several descriptive facts that have been obscured in prior research based either on county-level data alone, or on individual-level data analyzed at the state level:

• Most Americans live in state house districts with very few African Americans (i.e. less than 5%). In these areas, the relationship between income and Republican support is, on average, quite flat. The modest relationship does not vary much across states or regions.
• In the racially diverse districts outside the rural South, the income-party relationship is also very weak, typically with both affluent and less affluent voters supporting the Democrats. Rich whites who live near poor blacks in urban areas are surprisingly Democratic. This is even true in urban areas of “red states.”
• In the rural South, the pattern looks much different. There, we find that the unusual politics of the Black Belt extends to income-based voting. These areas exhibit, by far, the strongest relationship between income and partisan support in the country. Similar differences appear in other places known for having extensive rural, minority poverty, including the Rio Grande Valley and California’s Central Valley. The sharp differences between rich and poor voters in these areas of rural, minority poverty account for much of the state-by-state differences in income-based voting.
• Understanding the link between income inequality and partisan polarization within small-scale areas is, in short, inseparable from the often historically persistent race-based class conflicts in those areas.