American politics has grown increasingly polarized over the past 20 years, degenerating after the 2016 election and in the Trump Presidency into all-out partisan warfare. Observing from the sidelines, a significant portion of the voting public has begun to sense a real danger in this fracturing of partisan politics, and are searching for solutions.
With many competing media narratives, it’s difficult to make sense of it all. Our first instinct is to ask why we are so polarized, while some investigators cite data to question even if we are so divided on our differences. But putting our recent experience into historical context can illuminate how we are actually divided and have been for most of our nation’s history. Since divisions are not necessarily polarizing, we can then examine what specific factors are driving us apart in a follow-up essay.
A number of periods in our history were marked by sharp partisan divisions. We can cite Jefferson’s yeoman farmers against Hamilton’s strong federalism, and Jacksonian democracy vs. New England financial interests supporting the Second Bank of the U.S. Obviously we had the Civil War split between the industrial North and slave-holding South. At the end of the 19th century we again had a split between mid-Western agricultural interests and the financial centers in the East over what William Jennings Bryan coined as the Cross of Gold.
The first thing to notice is that most of these conflicts were regional and associated with certain economic industries, typically agricultural interests pitted against industrial and financial interests. These divisions were driven quite naturally by competing material interests defined by their rural vs. urban geography and, with the exception of the Civil War, were managed and resolved through the democratic electoral process.
Fast-forward to the end of the 20th century and we find that our partisan politics are still largely divided by rural and suburban vs. urban splits. Divisions based on economic sectoral interests persist in our regional economies, but with a more homogenized national economy, they are not as salient. The political calculus today seems driven by something more, and this is the puzzle we are trying to understand. In the contemporary case, once we look at the voting data, state county maps may offer a more accurate representation of the true geographic pattern than state or regional comparisons.
Without getting into Electoral College math, we can see that the big shift in recent years was in the Great Lakes region, where Barack Obama won every state in 2008 and Hillary Clinton lost every state in 2016 except for NY, IL, and MN. We can attribute this to the fate of the Rust Belt economy over the intervening years. (We’ll also leave the contentious debate between the primacy of geography or of population for another day’s discussion, when we discuss the logic and effect of the Electoral College.) Suffice to say that electoral geography offers valuable insights into our voting results.
It would be impossible to capture this pattern in our analysis without employing some kind of geographic explanatory variable. This is where exit polling may fail us. Exit polling is sampling data from survey questions asked of voters after they exit the voting booths. The surveys inquire into numerous classifications of voters, but rely mostly on demographic characteristics such as race, gender, ethnicity, age, income, etc. We can see that exit poll question give us exit poll answers — there is no effective way to capture the urban, rural, suburban divide except by comparing different exit polls. Furthermore, exit polling introduces sampling bias — for instance, pollsters have discovered that liberals are more willing to reveal their votes than more traditional conservatives. Pollsters try to account for this bias.
But county-level data offer a different approach. All votes are reported by county, and counties offer numerous demographic data based on the prior census. More importantly, counties offer a rich population set of more than 3,140 data points without any need for sampling. This allows us to regress the demographic data against the vote shares to determine with a high degree of confidence how county demographics affect voting preferences. In other words, what demographic characteristics: age, income, race, ethnicity, household formation, gender, etc. explain how the county residents vote? Counties also offer such data as median family income, and population density.
The results are revealing. Using regression analysis to run votes shares against the prior census shows that the most significant predictors of how a county’s residents vote are population density and household formation that varies between married and female heads of household. Race, ethnicity, age, gender, children, income, are all insignificant by comparison. In statistical terms these two variables explain roughly two-thirds of the variation in voting patterns. This may seem counter-intuitive but two anomalies help affirm the results.
First, black vs. white, or white vs. non-white, as a race variable is significant until female heads of household are introduced into the equation, then the black variable gets subsumed by female heads of household. According to data from the 2000 census, the correlation between the black population and the female heads of household population is an astounding .8, where 1 is perfect correlation. This suggests that black female heads of household are voting more in concert with female heads of households of other races than they are with married black households.
The second anomaly is the robustness of the population density variable, which is perfectly monotonic across metro county classifications and voting preferences. What this means is that across the entire nation, as population densities increase from rural counties to exurbs to mature suburbs to inner suburbs to core metros, the vote goes from solid Republican to purple swing counties to solid Democrat, without deviation.
There is likely no singular explanation, but one of the more persuasive is the divergence in material and policy interests among communities that vary in accordance with their population density. Urban singles, with or without children, have different policy priorities than suburban and rural married couples. These differences are somewhat reflected in the policy platforms of the two parties. For instance, cheap public transportation and rent control vs. low property taxes, good roads and low gas taxes; more social welfare programs vs. lower income taxes and more autonomy; better public schools vs. more school choice.
Of course, this cannot be the whole story. As Jonathan Haidt outlines in his studies on political ideology, there is likely a strong component ascribed to different orderings of moral values and sensibilities. Blue voters prioritize liberty, care, and fairness, while red voters have a more comprehensive set of value priorities that includes loyalty, authority and sanctity. This naturally leads blue voters to support subnational, multicultural identity groups that are historically disadvantaged in our society, while red voters tend to privilege an “American melting pot” with a distinct national identity and culture.
These value orderings tend to correlate with urban, suburban, and rural lifestyle choices. With the constant influx of new immigrant groups and young people, social change occurs more rapidly in urban communities. Exposure tends to inure us to change and make us more adaptive, whereas culturally traditional rural and outer suburb communities are less exposed to rapid change and thus can be more resistant.
Another way to approach this is by looking at the voting data on religion. The American populace can be described as religious, but the nature of belief that implies varies widely. Orthodox denominations and social conservatives attend church regularly and tend to vote red, while heterodox denominations and progressive secularists lean blue. This explains why churches became Republican messaging centers under campaign strategists Lee Atwater and Karl Rove.
However, we can put too hard an edge on this when judging the “other” and I hesitate to use the labels of liberal and conservative because of their ideological baggage in our political discourse. What journalists have revealed in their narratives of America by crisscrossing the land and interviewing residents, is that we have more in common than not. I would classify most of America as tolerant and traditional, rather than partisan, ideological, or political. This means average Americans have one foot set in the blue camp and the other in the red camp.
The important thing to note is that policy preferences that are divisible, like most of those above, allow for the compromise and convergence necessary to a functioning democratic process. If we are rooted in political camps that have solidified into our immutable personal identities, no such compromise is possible and our democratic governance becomes unworkable. We should probably focus more on what people do and where they live rather than who they are.
We have not yet answered why our plain and obvious differences have caused us to become so polarized into opposing camps when it comes to party preferences. We’ll save that discussion for another day. But our American political experiment has enabled us to self-govern a large, diverse, pluralistic population across a very large landmass for most of our history. There’s no inherent reason why we cannot do so today.