Political polarization in the U.S. mirrors its spatial divide.
Setting the nuttiness of Donald Trump and his candidacy aside, the 2016 presidential campaign serves as a fitting reminder of America’s deepening partisan divide. Americans have not only grown more ideologically polarized over the past couple of decades, Republicans and Democrats are drawn to very different kind of places. Back in 2004, Bill Bishop dubbed the self-segregation of Americans into like-minded communities, “The Big Sort.” (My next post for CityLab will dig into how these differences are shaping the 2016 election results. So stay tuned!)
A recent study by Ron Johnston, David Manley, and Kelvyn Jones in Annals of American Association of Geographers takes a deep dive into the big sort in American politics, examining the past six presidential elections, from 1992 to 2012, spanning Bill Clinton to Barack Obama. While most research examines America’s partisan divide across states—Andrew Gelman’s book, Red State, Blue State, Rich State, Poor State being perhaps the best example—this study does so across three different geographic levels or scales: nine broad Census divisions or regions, 50 states, and more than 3,000 counties.
Democratic Share of Vote in Regions, States, and Counties
The chart above shows the distribution of votes for these divisions, states, and counties using the minimum, maximum, and average percentage of Democratic votes that each geographic level garnered. The takeaway is that there’s substantial evidence of a growing spatial polarization across all three. Aside from a blip in 1996, there’s been a general decline in the minimum number of Democratic voters and an increase in the maximum number of Democratic voters at each scale.
This pattern is even more amplified as the geographic scale gets smaller. In 1992, the gap between divisions with the least and greatest support for Democrats was ten percentage points. By 2012, the largest gap had grown to more than 20 percentage points. At the state level, it grew from 26 points to more than 46 points. And at the county level it expanded from 75 points to 90 points. In 2012, Democrats carried just 3.5 percent of votes in the minimum county and 94 percent of the votes in their maximum county.
Median Odds Ratio Within Divisions, States, and Counties
The study then borrows from research on geographic segregation to create a broader measure it dubs a “median odds ratio” of political polarization for each of these three levels. As the graphs above show, these odds ratios also consistently increase over the two decades between 1992 and 2012. Indeed, the study finds that polarization increased by 29 percent across Census regions, 12 percent across states, and 14 percent across counties over this period.
Regional Difference from National Trend by Census Region
At the regional level the shift is distinctly bi-coastal, with New England, the Mid-Atlantic and Pacific regions having grown more Democratic, while the West North Central, East South Central, and West South Central have become less so.
The same trend is happening across states. Polarization has increased in the lion’s share of U.S. states. Just three had less political polarization in 2012 than in 1992. And polarization is increasing within states as well. In-state polarization increased significantly in 29 states and declined in only three (it stayed more or less the same in 17 others). Below is a chart that shows how much more predictive residence within a particular state has become for guessing how someone from a certain county will vote.
Median Odds Ratio for States Between Counties
Ultimately, the study provides substantial evidence that the Big Sort has deepened over the past couple of decades. The study notes:
Over the two decades and six elections between 1992 and 2012 there has been greater spatial polarization in the percentage voting for the Democratic Party candidates in presidential elections, at all three spatial scales analyzed.
Two interrelated factors appear to be driving the Big Sort, according to the study. On the one hand, like-minded people cluster together or with other like-minded people, and on the other, such clustering together makes people more like-minded. “There is clear evidence of significant spatial polarization of support for the country’s two main political parties across recent presidential elections,” the authors of the study write, “as like people tend to vote the same way, and like people tend to cluster together, such clustering increases greater polarization in voting patterns is the consequence.” As this study’s evidence and the run-up to the 2016 election show, the Big Sort is only getting bigger.