To get a crude sense of the geographic spread of the US presidential vote, I look at the results by winners of counties. The close similarity of the counties won by the D and R candidates in the recent election and that of four years ago, and the further similarity with that of other recent elections, prompted a quick update of a database I've kept of the county winners since the election of 2000, which is pretty much when the current dispensation of party affiliation settled into place.
I then counted them up.* And here's the result: 78% of US counties have voted the same way in every presidential election since 2000: 67% Republican, 11% Democratic. (The larger Republican number is due to smaller and more rural counties tending Republican, while more populous urban ones tend Democratic. Biden won Minnesota while carrying only 13 of the state's 87 counties, for instance.) Of the remainder, 16% voted for DT in both the last two elections and 4% voted against him, leaving only 2% of the whole - 76 counties altogether - that switched allegiance one way or the other between 2016 and 2020.
That sounds like a high degree of consistency, but has it changed over time? I made a comparison with the previous 6 elections, 1976-1996. Like the latest 6, it had 3 D and 3 R wins, but the latter were closer to blow-outs, which we've had none of since.** The results in the partial set of states I have calculations for is 54% of counties voted the same way in all 6 elections (48% R, 6% D), and 46% split their vote over the 1976-96 period, as opposed to only 22% in 2000-20.
Yes, it looks like we are settling into two nations.
*49 states. Alaska has no counties, and does not count votes by the boroughs and census divisions which are the (roughly) continuing divisions of the state for most statistical purposes. To calculate by those would require painstakingly adding up individual precincts, and nobody has done that for 2020 yet.
In Louisiana the counties are called parishes, for historical reasons dating back to the early 19th century.
A few other states have occasional large independent cities, but Virginia is unique in considering every incorporated city to be independent of its counties, and most Virginia geographic statistics are kept that way. But I dislike this because it makes Virginia statistics geographically incompatible with other states'. So I combine independent cities with their geographical counties, leaving the Hampton Roads ones that have swallowed the entire county. I figure it this way: in some New England states, counties no longer exist as units of government, but they're still considered useful as geographic aggregates for statistical purposes. I'm just doing the same thing with Virginia.
**I haven't calculated the 1976-96 figures for all the states, just about half which I found more interesting to work with, mostly western, Great Plains, and New England states. But they're roughly representative of the whole, with 80% of their counties voting the same throughout 2000-20, as opposed to 78% in the whole US.