10 November 2020

Vote Share, Population Density, Religion, and Regional Culture

The good baseline for predicting the Democratic-Republican vote shares of states in Presidential elections involves two main factors: urbanization based upon population density measured at the census tract level, and religiosity. Democrats do better in more urban, more secular places, while Republicans do better in more rural, more religious places.

Throw in education as a factor (with more educated states favoring Democrats and less educated states favoring Republicans), and you'd probable do even better.

Looking at urbanization alone, Vermont, Maine, and to a lesser extent New Mexico, New Hampshire and Wisconsin are vulnerable to GOP pickups, while Florida, Texas, Utah and Ohio are vulnerable to Democratic pickups.

Vermont, Maine and New Hampshire show a New England regional shift to the left.  New Mexico's lean is influenced by significant Hispanic and Native American populations. Utah's lean right is a product of it being predominantly Mormon.

Still Maine, New Hampshire, Wisconsin, Florida, Ohio and Texas which are out of step between their urbanization and their party leaning are among the closest to even in terms of partisan lean in Presidential elections.

While Wyoming and Montana are both safely Republican and the two least urbanized state in the United States, Montana is much less Republican leaning than Wyoming, despite the fact that Wyoming is significantly less religious than Montana. This may be due to the stronger influence of the coal industry in Wyoming. It could also be that urbanization is underestimated in some of the mountain states because it isn't sufficiently fine grained an analysis and fails to capture the extent to which people actually live in densely populated small towns rather than spread out as in most rural areas.

From Politics.StackExchange.

The urbanization index used in this chart and the one below is calculated as follows (which rules out some conjectures I'd had about problems with methodology):

As a somewhat random aside, the best metric I've been able to find for how urban or rural a given state/community is—i.e. what tends to best correlate with political outcomes where urbanization is a factor—is this: How many people live within 5 miles of you?
If you calculate this number for each person (well, each Census Tract) in the state, take the natural logarithm, then average them together (weighed based on the Census Tract's population), you can come up with a nifty "urbanization index" that looks like this.

This has a formal name. It's the . . KLD of the population and land distribution plus the log population density of the state.

The correlation coefficient between Presidential election two-party vote share and urbanization is about r= 0.66.

As the illustration demonstrates, regional differences account for almost all of the reset of the difference. "Northern" states are usually more Democratic than their urbanization index would indicate. "Southern" states are usually more Republican than their urbanization index would indicate.

Without the New England outliers of Maine and Vermont, the correlation would be significantly stronger and more tightly correlated. The Southern and Appalachian states, in contrast, are below the line.

The correlation was r= 0.68 in 2016 as shown below:

The correlation was r= 0.55 in 2012.

Religiosity, is another alway to explain the deviation from urbanization in the outcomes that captures the Southern, Utah and New England deviations from an urbanization based distinction.

The religious connection isn't a coincidence, it is causal (Mormons generally vote like Evangelical Protestants despite being theologically rather distinct):

Adjusting for religiosity, New Mexico, Wisconsin, Florida and Ohio look like the most vulnerable states. Race and ethnicity, again, takes New Mexico largely out of the calculus, leaving Wisconsin, Florida and Ohio which are indeed, perennial swing states.

No comments: