Trump's recession in 2020 in Trump 2016 country counties

One strange phenomenon in 2020 was that Trump's vote share dipped slightly in many small counties where he had done outstandingly in 2016. (However, this didn't happen across the board; for the most part, Trump did continue to improve in such counties.)

For example, if we look at the Trump 2016 country counties (the counties in which Trump received over 60% of the vote and a higher vote share than Reagan in 1984 or any subsequent nominee) listed here, Trump's vote share dropped in 2020 in Morgan, Limestone, St Clair, and Coffee, AL; both Fremont and Delta, CO; Sumter and Walton, FL; and both Allegany and Garrett, MD. (In the italicised counties, 2020 was the first election in the 21st century in which the GOP vote share fell.)

That doesn't seem like that many, but we didn't systematically look at all the Trump 2016 country counties, just those in select states (and we also looked only at those casting over 10,000 votes). If we look at simply the first 20 counties of Georgia alphabetically (Appling through Camden), there were three (of any size) that were Trump country counties in 2016 and in which Trump's vote share declined in 2020:

Appling, Bartow, Ben Hill

And seven which were Trump country counties in 2016 and in which Trump's vote share rose in 2020:

Atkinson, Bacon, Banks, Berrien, Bleckley, Brantley, Butts

A clear majority were Trump country counties again in 2020, but three (out of just the first twenty) is a non-trivial number in which Trump's percentage went down after being unprecedented in 2016.

(In Appling, incidentally, 2020 was the first election this century in which the GOP vote share declined--that includes the 'generic Democratic year' of 2008.)

Regionally, northern Georgia is a rather more evenly divided patchwork of Trump 2016 country counties where Trump's vote share declined (Rabun, Union, Gilmer, Catoosa, Walker, Bartow, Madison, Oglethorpe) or rose (Towns, Fannin, Murray, Dade, Chattooga, Floyd, Polk, Haralson, Gordon, Banks, Stephens, Franklin, Hart, Elbert, Morgan) in 2020. (Out of the first group, 2020 was the first election this century to see the GOP vote share decline in all save Bartow, Madison, and Oglethorpe.)

Similarly, if we look at the first twenty counties by alphabet in Kentucky (Adair through Carlisle), one were Trump country counties in 2016 in which Trump's vote share declined in 2020:

Boyd

and sixteen were Trump country counties in 2016 in which Trump's vote share rose again in 2020:

Adair, Allen, Anderson, Ballard, Barren, Bath, Bell, Bourbon, Bracken, Breathitt, Breckinridge, Bullitt, Butler, Caldwell, Calloway, Carlisle

(And to Boyd we could add, at a minimum, Pike County, Kentucky, noteworthy for being the biggest Kerry-McCain county outside Pennsylvania, and McCracken County, Kentucky, noteworthy for being the biggest Bush 2000 breakthrough county in Kentucky. In the case of all three, 2020 was the first election this century to see a decline in the GOP vote share.)

Out of Tennessee's first twenty counties (Anderson through Decatur), the corresponding figures are two:

Bradley, Carter

and 16:

Anderson, Bedford, Benton, Bledsoe, Campbell, Cannon, Carroll, Cheatham, Chester, Claiborne, Clay, Cocke, Coffee, Crockett, Cumberland, Decatur

(In addition, Sullivan County, Tennessee--the county that displaced Knox County as the GOP's biggest vote trove in Tennessee in 2020--was a Trump 2016 country county that saw Trump's vote share slip in 2020. Again, for all three of Bradley, Carter, and Sullivan, 2020 was the first election this century that saw the GOP vote share slip.)

And out of Texas' first twenty counties (Anderson through Brazoria), two:

Blanco, Bowie (Bowie also being a 2020-first-slide-this-century county)

and nine:

Anderson, Angelina, Aransas, Archer, Armstrong, Atascosa, Baylor, Borden, Bosque

In each case, Trump did continue upping his vote share in 2020 more often than he did not, but in each case, there are already a couple places, again just out of the first twenty (except in Kentucky), where Trump declined in 2020 after having outdone every nominee for the prior 30+ years in 2016. (It wouldn't be that surprising to see Trump slip in 2020 in a county where he was already doing worse than Romney in 2016.) 

Perhaps this was primarily a Southern phenomenon? It does seem as though Trump continued his upward trend more broadly in other regions of the country. Every Trump 2016 country county casting more than 10,000 votes in Michigan, Indiana, Kansas, Arizona, Idaho, and Nebraska, for example, went on to be a Trump 2020 country county. That said, we already saw above that Trump declined in the two Trump 2016 country counties in Colorado that cast over 10,000 votes (and there was at least one more that cast fewer than 10,000 votes: Moffat County). In a number of Mountain West states, there were relatively few Trump 2016 country counties to begin with; and in the states west of Beatson's Line* in general, there was a relatively high third-party vote in 2016 that meant that it was more likely that both parties would increase their vote shares in 2020.

If we look at Kansas' first twenty counties (Allen through Decatur), we do find one Trump 2016 country county in which Trump's vote share declined in 2020:

Cheyenne

as against twelve where it rose in 2020:

Anderson, Barber, Barton, Bourbon, Chautauqua, Cherokee, Clark, Cloud, Coffey, Comanche, Cowley, Decatur

And there was at least one other county similar to Cheyenne in Kansas, Kearny.

Looking through the Bush breakthrough counties can also give us a sampling of Trump 2016 country counties where Trump's vote share declined in 2020 (italicised ones are ones in which 2020 saw the first decline in the GOP vote share this century):

McCracken, KY
Boyd, KY
Henderson, KY
Letcher, KY
Union, KY
Webster, KY
Navarro, TX
Waller, TX
Limestone, TX

(Perry, KY is an odd case of a Romney country county where the GOP vote share rose in every election from 2000 through 2012, but fell in both 2016 and 2020.)

There aren't many (not that the above are all of them), but that is to be expected--the Trump country counties where Trump continued improving in 2020 seem disproportionately to have been ones that used to be heavily Democratic in the recent past. Still, it's odd that there are even as many as there are

We have been looking just at Trump (2016) country counties, to make the task easier, but perhaps equally worrisome, if not more so, is the fact that Trump failed to match Bush's '04 vote share in a number of Bush country counties (in which Trump nevertheless did improve on Romney), in either 2016 or 2020. Some examples from these two posts would be Autauga, Elmore, Escambia, and Pike, AL; Amador, Shasta, Tuolumne, and Yuba, CA; Camden and Effingham, GA; Grant and Warrick, IN; Shelby, KY; Beaufort, NC; Linn, OR; Isle of Wight, VA; and Okanogan, WA. (In Nez Perce County, Trump improved on Romney in 2016 but failed to match Bush's 2000 vote share [which was better than his 2004 vote share] in either 2016 or 2020.) As we can see, this is a more obviously widespread phenomenon.

In the Northeast, there were relatively few Trump 2016 country counties as well, but we can look at whether Trump increased his vote share in his breakthrough counties from 2016 (that is, those counties he was the first Republican since at least 1988 to carry). In New York, there were five such counties: Broome, Franklin, Niagara, Rensselaer, and St Lawrence. In 2020, his vote share fell in Broome, Niagara, and Rensselaer (whereof Biden won back Broome and Rensselaer), whereas it rose in Franklin and St Lawrence. His vote share also fell in Kent County, Rhode Island, Androscoggin County, Maine, and Northampton and Luzerne Counties, Pennsylvania, although it rose in all his other breakthrough counties in Maine as well as in Windham County, Connecticut, Gloucester County, New Jersey, and Erie County, Pennsylvania. (Elsewhere in the country, his vote share appears to have generally improved in his breakthrough counties, although in some cases not enough to prevent Biden winning them back [as in Montgomery, OH, Saginaw, MI, and Pueblo, CO], although there was at least one similar case far afield from the Northeast: Las Animas, CO.)

In contrast, of the counties casting over 10,000 votes in 2000 in which George W. Bush got over 60% and a higher vote share than Reagan in 1984 or any subsequent nominee, he improved his vote share in 2004 in all but six: Siskiyou, CA, Bonner and Nez Perce, ID, Rankin, MS, Stokes, NC, and Okanogan, WA. Perhaps more importantly than that there were relatively few exceptions, they were concentrated in Idaho: there were only two counties in Idaho that had cast over 10,000 votes in 2000 and had given Bush over 60% and a higher vote share than Reagan in '84 or any nominee since, and both failed to repeat this feat in 2004. They accounted for fully 1/3 of the national exceptions. A third, Okanogan, is in the interior of Washington not far from Idaho (and a fourth, Siskiyou, is in the same general region). In the case of Trump in 2020, on the other hand, the exceptions may be mostly in the South, but they are interspersed with Trump 2016-Trump 2020 country counties relatively haphazardly. 

Bush also retained nearly all of his 2000 breakthrough counties, and again, the exceptions tended to be concentrated in particular states or even regions of states: all but three were in either Kentucky or Minnesota, and of the other three, one each was in Wisconsin, Michigan, or Texas, two of which are Upper Midwest neighbours of Minnesota. The true geographic outlier, in Texas, was Travis County, home to Austin, where Bush's favourite-son status (as yet unattenuated by a White House record) and the Nader vote both probably played a role. The ones in Kentucky were even concentrated in eastern Kentucky, as noted by some politically observant tweeters such as Kaden Collingridge and Jeff Ditzler. I'm not sure what was going on there (and apparently neither does Collingridge or Ditzler), but something seems to have been going on there in 2004. 

And, as a general rule, Bush increased his vote share in nearly all of his 2000 breakthrough counties in 2004. 

2020 wasn't like 2008, when there was a general nationwide trend towards the Democracy (outside the South). There were some McCain breakthrough counties and some McCain country counties, but they were almost all in the South. Elsewhere, the GOP vote share almost uniformly receded. But neither was it like 2004. It was in between, but in a particularly inexplicable way (i.e., not in a region-by-region way, where Trump held onto his inroads in one region better than in another).

* In 1961, James Allen Beatson wrote,

Of the seventeen states west of a line running north-south from Canada to the Gulf of Mexico, formed by the eastern boundaries of the Dakotas, Nebraska, Kansas, Oklahoma, and Texas, Wilson carried all except two: South Dakota and Oregon. The West had spoken with a unanimity... (p. 55)

This line is a potentially useful delimitor of 'the West', as opposed to the 100th meridian. In 2016, Gary Johnson's eight best states were all west of Beatson's Line. (Kansas was 13th.) His 16 worst states were all east of Beatson's Line; and his 17th-worst state was Texas, a Southern state and the oldest state west of Beatson's Line. The only other states west of Beatson's Line to make it in Gary Johnson's bottom 25 were Nevada (his 22nd-worst state) and California (his 24th-worst).