In this post, we'll look at the counties that switched from each major party to the other in every election from 1920 on (as well as a few earlier elections). In some cases, there is a source that gives the number of counties, in which case I link to the source. In some cases, I deduce the number of counties that one of the nominees flipped, from the number that the other nominee flipped and from the number of counties each party carried in the previous election (the figures for which are here). These can be slightly (hopefully only slightly) off for a variety of reasons, so in these cases, the figure is preceded by a tilde. In cases where I count the counties myself, my principal source is the Crystal Ball database linked here (although I correct for any errors I spot, of which there are a non-trivial number; Wikipedia takes precedence but is not practical to use simply to count turnover counties in a given election).
We'll look at the percentages of each party's counties from the previous election each nominee flips, as well as the number of states that each nominee's turnover counties are in (as Tom Schaller does here for the Obama 2012-Trump 2016 counties), and how the number of turnover counties of nominees in different types of elections compares over time. In particular, we'll pay attention to 'anti-trend' counties. Albert J. Menendez characterised the Stevenson '56-Nixon '60 counties as
'anti-trend counties', which I've generalised to counties flipping from
the out-party to the in-party when the out-party succeeds in taking
power away from the in-party (as well as to some elections with
particularly large swings even if the in-party succeeded in staying in
power; this is a list of the largest swings in US presidential elections).
It appears that 1932 is the only election in which there were no counties that flipped from one of the two major parties to the other (as suggested by Calthrina950 and 'Dr RI'). There are some other elections in which it appears highly improbable for there to be such counties; most of these are in the 20th century, but 1860 is one (especially because those two forum posters are, more exactly, saying that 1932 is the only election in which one of the two major nominees flipped no counties). It doesn't appear improbable that Douglas in 1860 should have flipped some, say, Fillmore '56 counties. But it does (at least to me) that he should have flipped some Frémont '56 counties.
I hadn't been counting Breckinridge as a 'major party nominee' in 1860, but he kind of was, or at a minimum, the situation was ambiguous. And it is Frémont-Breckinridge that has been seen, by those who are interested in this, as the truly unusual voting pattern from that period, rather than Frémont-Douglas. More to the point, Frémont-Douglas doesn't appear to be interesting at all.
And indeed, this does appear to have been one of the 'most unusual' voting patterns. According to the Crystal Ball database, there was one Frémont-Breckinridge county: San Luis Obispo, CA. (Information about how the county voted in those elections is unavailable, but, even though it was his home state, Frémont almost certainly would have had to have run significantly better in the county than across the state to carry it, as he got only 18.78% statewide.)
Bryan '08-Taft '12 counties are another seeming improbability, but it seems there were six, in two states (Kentucky and Oklahoma): Trigg and Caldwell, KY, and Dewey, Ellis, Grant, and Pawnee, OK. Grant cast the most votes of the six, at 3,649.
Wikipedia doesn't have information about how Oklahoma's counties voted in 1908, but of the two Bryan '08-Taft '12 counties in Kentucky, Taft did increase his vote share in Trigg County, from 43.65% to 46.42%.
At first thought, Taft '12-Wilson '16 counties might also seem an improbability, given how badly Taft did in 1912. But 1912 was a highly chaotic election, with many counties being carried with very low pluralities, and in 1916, Wilson won one of the two states Taft carried with a plurality in 1912, by over 20% (Utah). So it's not actually surprising, but still worth noting.
Manistee cast 4,570 votes, and Polk, NC, 2,687.
Going by Wikipedia's numbers, Harding carried 973 more counties than
Hughes, which would suggest that Harding flipped 974 counties, or about 47.7% of the 2,039 counties Wilson carried in 1916--although, at this time, there were still counties going into and out of existence, or not reporting results, on a not infrequent basis, as well as the phenomenon of third-party candidates who were not necessarily that significant on the national stage carrying some counties. Harding flipped at least one Wilson county in 44 states, with none available to flip in Rhode Island, as Hughes had swept that state's counties in 1916. They included 16 'mega-counties', with the largest being Manhattan (which cast 464,420 votes).
The Cox-Coolidge counties are so few in number that we can list them: Pinellas, FL; Clark, Johnson, Putnam, and Shelby, IN; Bracken, Fayette,
Henderson, Kenton, and Pendleton, KY; Cecil, MD; Daviess,
MO; Luna, NM; Johnston, NC; Butler, Crawford, and Fairfield, OH; and
Anderson, Bandera, Bee, Comanche, Crockett, Kerr, McMullen, Real,
Terrell, and Val Verde, TX. (I am here counting Manistee, MI as a Harding '20 county.)
The two Cox-La Follette counties were Clark, NV, which cast 1,636 votes, and Mineral, MT, which cast 1,032.
Looking either just at the two major nominees, or at all flips, the county flips in 1924 do not paint a picture of an election that is of a roughly similar type as the preceding election (i.e., a 'landslide'). Significantly (given Robert Wheel's implication that landslides are
those elections that tend to be most responsible for interrupting what
would otherwise have been 'durable streaks of voting for a party on the county level'), all the Cox-Coolidge counties went on to vote for Hoover in 1928.
Not only were the Cox-Coolidge counties far fewer than the Harding-Davis counties, but they were also not particularly disproportionately populous. Here were the turnover counties in 1924 that cast over 20,000 votes (with the colour representing the party that carried it in 1924):
(We don't have a figure for how many Coolidge '24 counties there were, so we can't estimate what proportion of the Coolidge '24 counties Smith flipped.)
Again, if 1924 is a 'landslide', the county flips in 1928 should reflect a basically similar election as 1924. Do they?
Well, on the one hand, Hoover flipped far more counties than Smith. (He also inherited the lion's share of the La Follette counties.) However, in this case, the counties that flipped in one direction were certainly more disproportionately populous than in the other, at least down to a certain level. (The italicised counties voted for La Follette in 1924.)
Certainly, at the very top, the turnover counties are vastly disproportionately--indeed essentially exclusively--Smith turnover counties. From Oklahoma County down, however, they seem to be disproportionately Hoover turnover counties. (And this isn't because we ran out of Smith turnover counties--there were 19 counties Smith flipped that cast more vote than Oklahoma County, less than a quarter of his 82 total intra-major-party turnover counties, and less than a tenth of his overall turnover counties.)
--three
elections in which there was no mistaking that there was a particular
national trend in one party's direction, and in which the anti-trend
counties were, as in 1928, disproportionately large but less numerous than the
'pro-trend' counties. In all of 1900, 1904, and 1908, the pro-trend counties were only between
three and four times as numerous as the anti-trend counties (whereas in
1928, they were over five times as numerous). In 1900 and 1904, the
absolute number of anti-trend counties was larger than the number of
Coolidge-Smith counties, despite that there were probably fewer overall
counties at the time. And none of McKinley in 1900, Roosevelt in 1904, or Bryan in 1908 flipped nearly as many counties as Hoover flipped Davis counties in 1928; while this may partly be a function of there having been fewer counties in the '00s, it doesn't seem from the numbers that this can be the entire explanation.
Perhaps the county that came closest to being an anti-trend county in 1932 was Berkshire, MA, which voted for Smith in 1928 49.98%-49.5% (or by 220 raw votes), and for Franklin Roosevelt in 1932 48.2%-48.1% (or by 66 raw votes).
The 14 Landon-Roosevelt counties were Magoffin, KY, Assumption, LA, Kennebec, ME, Sagadahoc, ME, Calvert, MD, Somerset, MD, Sullivan, NH, Clay, NC, Davie, NC, McMinn, TN, Roane, TN, Essex, VT, Montgomery, VA, and Wythe, VA. The largest, Kennebec, cast 30,765 votes, followed by Sullivan, NH, which cast 11,455.
1944Willkie '40-
Roosevelt '44: 23 (2.01% of Willkie '40 counties) (in
14 states)
Roosevelt '40-
Dewey '44: ~ 219 (11.2% of Roosevelt '40 counties) (9.52 times as many as Willkie '40-Roosevelt '44 counties) (in
35 states)
I wouldn't call the Willkie-Roosevelt counties 'anti-trend', given that 1944 featured only a small Republican swing (of 2.96%), but the turnover counties certainly seem to reflect a strong swing in favour of the Republican Party. The 23 Willkie-Roosevelt counties were: Martin, IN, Dubuque, IA, Howard, IA, Baraga, MI, Houghton, MI,
Keweenaw, MI, Aitkin, MN, Grant, MN, Texas, MO, Greeley, NE, Cavalier,
ND, Ramsey, ND, Slope, ND, Towner, ND, Traill, ND, Cherokee, OK, Elk,
PA, Kent, RI, Bennett, SD, Ziebach, SD, Lee, TX, Benton, WA, and
Florence, WI. The largest was Kent, RI, but, unlike in 1940, four cast more than Sullivan, NH cast in 1940:
Kent, RI (27,826)
Dubuque, IA (25,458)
Houghton, MI (19,238)
Elk, PA (11,809)
Benton, WA (8,172)
They were also distributed more evenly across the US than the Landon-Roosevelt '40 counties.
I actually counted 220 Roosevelt-Dewey counties (which are listed
here), which was pleasantly surprisingly close to the deduced figure. Two were mega-counties: Fairfield, CT (205,297 votes) and Franklin, OH (188,686). However, there is a fairly large dropoff in terms of number of votes cast at a certain point:
Fairfield, CT (205,297)
Franklin, OH (188,686)
Lucas, OH (153,356)
Schuylkill, PA (76,744)
Sedgwick, KS (73,698)
Peoria, IL (67,251)
Lehigh, PA (61,033)
Baltimore County, MD (60,322)
Still, this is a big differential. Perhaps the most striking thing is how few counties overall flipped in either direction: about 242 or 243. Counting just intra-major-party flips, this was fewer counties than flipped in any election since 1924, and it was fewer than would flip in any election again until 2004.
1948Dewey '44-
Truman '48 (anti-trend?): 259 (19.3% of Dewey '44 counties) (2.3125 times as many as FDR '44-Dewey '48 counties) (in
29 states)
FDR '44-
Dewey '48: 112 (6.40% of FDR '44 counties) (in
27 states)
This election, too, was highly unusual. I labelled the Dewey '44-Truman '48 counties as 'anti-trend' but with a question mark, since they clearly don't look like anti-trend counties, even though 1944 saw Dewey go down to a defeat of a similar magnitude as Dukakis in 1988 or McCain in 2008, and 1948 saw him come within
30,000 votes in three states of winning the Electoral College.
In fact, Truman flipped just over twice as many counties as Dewey, which is a greater ratio than that which is normal for elections that are simply of a roughly similar character to the previous one.
Truman flipped counties in slightly more states than Dewey, and he even flipped the single largest turnover county (Middlesex, MA).
Well, what if we look, as with previous elections, at the numbers of votes cast by the top n vote-producing turnover counties? Well, here are the turnover counties that cast over 30,000 votes:
Middlesex, MA (485,908)
San Diego, CA (205,459)
Lucas, OH (142,853)
Luzerne, PA (135,611)
Stark, OH (100,150)
Santa Clara, CA (99,502)
Oneida, NY (97,613)
San Bernardino, CA (95,838)
San Mateo, CA (86,272)
San Joaquin, CA (59,361)
Atlantic, NJ (58,071)
Pinellas, FL (44,524)
Lorain, OH (43,638)
Linn, IA (43,098)
Burlington, NJ (42,432)
Greene, MO (39,664)
Stanislaus, CA (38,371)
Brown, WI (36,558)
Clark, OH (35,917)
Gloucester, NJ (35,765)
Scioto, OH (34,852)
Santa Barbara, CA (34,400)
Scott, IA (34,081)
Monterey, CA (34,063)
Marin, CA (32,855)
Cumberland, NJ (32,313)
Kent, RI (31,785)
Clackamas, OR (30,083)
Jasper, MO (30,078)
Of these, Dewey flipped about 72.4%. That is a very large majority, and mitigates the impression given by the numbers of counties flipped somewhat. Nevertheless, it is still somewhat strange.
What this might suggest is that, much
like Biden in 2020, Dewey's strength was propelled by changes in the margins in large red (and blue) counties, more than by currents that were likely to outright flip a large number of counties.
I'll give some examples of such counties, but examples can be misleading. Let's first do an exercise. Let's colour in the electoral map, colouring each state with the colour of the party that netted the most vote out of any one county in the state. For example, if we do this for 2020, ruby-red Tennessee would be blue, because Biden's 116,290 plurality out of Shelby County was greater than Trump's raw vote margin out of any county in the state.
If we do this for 1940, 1944, and 1948, we get these three maps:
1940,
1944,
1948. The red states in 1940 represent 11 electoral votes; in 1944, 25 electoral votes; and in 1948, 85 electoral votes.
85 is still a minority (but in close elections, the red states, in this exercise, will generally constitute a minority of the Electoral College). But it is a vast improvement on the 25 in 1944. (On the other hand, 1944 was not much of an improvement on 1940.)
Now, the fact that this happened at the same time as more counties flipped from red to blue than vice versa, would indicate that a lot is happening that is not demonstrated by the actual turnover counties. For example, the title of the best Republican raw vote margin county in the state shifted from a smaller to a larger county in
12 states (doing the reverse in five states). Conversely, the title of
the best Democratic raw vote margin county in the state shifted from a larger to a smaller county in
ten states (not counting the Thurmond states, although it actually stayed the same in South Carolina), and from a smaller to a larger county in two states.
In California and Pennsylvania, this actually leads us to two examples of truly enormous counties that almost flipped to Dewey. Dewey closed the gap in Philadelphia County from nearly 18% in 1944 to 0.76% in 1948, and in Los Angeles County from just over 14% in 1944 to 0.49% in 1948. In the city-county of San Francisco, he similarly closed the gap from over 20% to 2.16%, and in Alameda County, from just shy of 16% to 1.23%. Either of the former two would easily have bested Middlesex, MA as the largest county to flip in 1948, and either of the latter two would easily have bested San Diego County as the largest Roosevelt '44-Dewey '48 county.
In terms of expanding his own margins, Dewey expanded the Republican margin in Orange, CA by 10.1%; in Arlington, VA by 7.1%; in Fairfield, CT by 10.6%; and in Montgomery, MD by 8.8%. (Each of these counties played a special role in 1948: Orange and Arlington were Dewey's best raw vote margin counties in two states where the single biggest county raw vote margin was Republican in 1948 but had been Democratic in 1944; and Fairfield and Montgomery handed their state to Dewey in 1948--in both cases, despite Dewey flipping no county in the state.)
In another example of much going on under the surface of unflipped counties, the state in which Dewey increased his vote share relative to 1944--by far--was
Texas. And yet Dewey flipped only three, small counties in Texas (Austin, Kerr, and Washington, whereof Washington, the most vote-casting, cast 3,742 votes). Probably far more significant was that he closed the gap in Harris, Dallas, and Tarrant Counties (the state's largest, second-largest, and fourth-largest counties) by 44.36%, 29.82%, and 32.44%, respectively. Again, he did this without coming remotely close to actually flipping any of these (Truman won all three by at least 12%).
It is tempting to say that the county flips in 1944 look like what 1948 should have looked like, and vice versa. At a stroke, that would explain the idiosyncratic nature of the county flips of both elections relative to the kind of elections they respectively were. It may also not be completely invalid--many of the counties in which Dewey enlarged his margin significantly in 1948 were counties he had flipped in 1944 (including Arlington, Fairfield, and Montgomery [MD]).
As for the Truman turnover counties, a significant number of them had actually, as of 1940 or 1944, been traditionally Democratic counties--at least, as traditionally Democratic as a county outside the South was generally likely to be at the time, as Solid South counties generally stuck steadfastly by FDR. (One exception, which we'll encounter later, was Ste Geneviève, MO, which had never voted for a losing Republican before 1940.)
As for Truman himself, there were--not counting the Thurmond counties--only 47 counties he was the first Democratic winner in the 20th century to lose. This was only slightly more than the 44 counties Roosevelt was the first Democratic winner in the 20th century to lose in his 1936 landslide (of which four--Leavenworth, KS, Miami, KS, Taylor, KY, and Sullivan, MO--went on to vote for Willkie, Dewey, and then Truman). (In a similar vein, in both 1940 and 1944, Willkie and Dewey became the
first Republican ever to carry a particular county--Franklin, IN and Wells, IN, respectively [of which Truman won back Wells], whereas the longest-ago that any FDR '44-Dewey '48 county had last voted Republican was 1888 [Williamsburg City, VA--which went on to vote Republican in every subsequent election through 1988 save 1964, so its defection likely indicated a secular trend rather than Truman alienating traditionally Democratic voters that another Democrat could win back].)
Given how close the 1916 election had been, the counties FDR was the first winning Democrat in the 20th century to lose in 1940 and 1944 would have been considered, at the time, counties that one would expect a nationally competitive Democrat to be carrying, even if they hadn't voted for any Democratic losers in the prior 44/48 years (which many of them had). So, Truman's carrying them would not necessarily indicate that he was accomplishing anything more than simply being nationally competitive, unless one assumes he was also retaining whatever inroads FDR had made in the prior two elections that had enabled him to more than make up for their loss.
1952Dewey '48-
Stevenson '52 (anti-trend): 1 (
source)
Truman '48-
Eisenhower '52: many (see below) (in
44 states)
Given that Eisenhower flipped many counties in 1952, and a non-trivial number of counties voted for a non-major-party candidate in 1948, I don't really know how many
Truman '48-Eisenhower '52 counties, precisely, there were. I didn't find it worth counting them. I can say Eisenhower carried 912 more counties than Dewey in 1948, and that 19 were
mega-counties. Eisenhower carried at least one Thurmond '48 county or parish in each of the four states in which he flipped no Truman counties.
The one Dewey '48-Stevenson '52 county was Dawson, GA, which cast 1,240 votes. In this case--unlike with Manistee, MI in 1920--there doesn't seem to be any doubt about Dawson actually having voted for Dewey in '48 and for Stevenson in '52.
That said, there seems little in the way of an explanation as to why Dawson County made the switch. The more puzzling of its two votes in the two elections in question was in 1948. Calthrina950 writes that North Georgia, at the time, was
'a region with divided loyalties between New Deal populism and ancestral Mountain Republicanism'. But, despite being close to some unionist counties, Dawson itself seems to have been a Solid South Democratic county--from 1876 through 1944, it voted Republican only in 1904, 1908, and 1920.
Dewey did make some inroads with Solid South Democratic counties in 1944 and 1948 (although not so much in Georgia in particular); but in general, these counties continued along an increasingly Republican trajectory in at least the next several elections after 1948. Of 69 Davis '24-Dewey '48 counties, none (other than Dawson) voted for Stevenson either time, and only five (aside from Dawson) even so much as voted for Kennedy in 1960 (Montgomery, MD, Iberia, LA, King George and Stafford, VA, and Kent, DE). (That said, two of these five--King George and Kent--voted for Eisenhower rather more narrowly in 1952 than one would have expected from their vote for Dewey in 1948. This was also true, to a lesser extent, of another Davis '24-Dewey '48 county closer to Dawson's general neck of the woods, Lincoln County, NC.)
Dawson, in contrast, proceeded to vote Democratic in every election from 1952 through 1964, then for Wallace in 1968. When it voted for Nixon in 1972, it was its first Republican vote in 52 years, bar its vote for Dewey in 1948. Dewey wasn't
quite responsible for Dawson County not having a
'long' streak of not voting for a party, if 'long' means at least 15 consecutive elections (as it seems to for Wheel). However, he
was responsible for the county not having a
12-election streak of not voting Republican.
1956Eisenhower '52-
Stevenson '56: 121 (5.75% of Eisenhower '52 counties) (in
23 states)
Stevenson '52-
Eisenhower '56: 180 (18.1% of Stevenson '52 counties) (1.49 times as many as Eisenhower '52-Stevenson '56 counties) (in
34 states)
Four turnover counties, all Stevenson-Eisenhower counties, were mega-counties: Allegheny, PA, Baltimore City, MD, Hudson, NJ, and Providence, RI. The three largest Eisenhower-Stevenson counties were Macomb, MI (121,419), Pierce, WA (115,544), and Stanislaus, CA (54,927)--the last of which had voted for Dewey in 1948.
1960Stevenson '56-
Nixon '60 (anti-trend): 97 (10.5% of Stevenson '56 counties) (in
17 states) (
source)
Eisenhower '56-
Kennedy '60: 408 (19.04% of Eisenhower '56 counties) (4.21 times as many as Stevenson '56-Nixon '60 counties) (in 46 states [
source]) (
source)
Kennedy flipped 21
mega counties; the largest county Nixon flipped was Greenville, SC, which cast 36,633 votes. Nevertheless, 97 is an abnormally large number of anti-trend counties; 408 was, until the 21st century, an abnormally small number of turnover counties for a nominee successfully taking the presidency back for his party; and, again, until the 21st century, 19% was an abnormally small percentage of the other party's counties from the last election to flip for a nominee successfully taking the presidency back for his party. The Stevenson-Nixon counties were somewhat concentrated in the South, but were by no means limited to the South (and expanded fairly broadly across the South). (I didn't bother finding the two states where Kennedy flipped no counties.)
1964Kennedy '60-
Goldwater '64 (anti-trend): 266 (22.2% of Kennedy '60 counties) (in
14 states)
Nixon '60-
Johnson '64: many (see below) (in
45 states)
As was the case in 1952, there was a combination of a very large number of counties flipping to one of the two nominees in particular, and a non-trivial number of counties voting for a non-major-party candidate in the previous election, such that I didn't bother counting the number of counties to flip specifically from Nixon to Johnson. There is also some disagreement about how many counties each nominee carried in both 1960 and 1964 (I used Wikipedia's figures to reckon the percentage of Kennedy counties that Goldwater flipped). It is safe to say that Johnson carried over a thousand more counties than Kennedy.
19 of the counties Johnson flipped were mega counties. Ironically, however, the largest county Goldwater flipped--Duval, FL--cast significantly more vote (160,481) than the largest Stevenson '56-Nixon '60 county.
Obviously, the Kennedy-Goldwater counties were particularly unusually numerous for anti-trend counties. They were also, however, highly concentrated, considering their numbers. They were distributed as follows: 107 in Georgia, 53 in Alabama, 31 in Florida, 25 in Mississippi, 13 in
Virginia, 12
in Louisiana, 12 in South Carolina, six in Arkansas, two in Idaho [Camas
and Custer], and one each in Maryland [Dorchester], Missouri [Osage],
North Dakota [Emmons], Tennessee [Haywood], and Texas [Glasscock]. Over
40% (or two-fifths) were in Georgia alone, and there were two different pairs of states that, together, contained over half of them
(Georgia and Alabama, or Georgia and Florida).
1968Goldwater '64-
Humphrey '68 (anti-trend): 35 (
source) (4.24% of Goldwater '64 counties) (in
six states)
Johnson '64-
Nixon '68: many (in
46 states) (see below)
Johnson '64-Wallace '68:
Goldwater '64-Wallace '68:
Again, given the complicating factor of many counties voting third-party, I didn't bother finding out how many counties switched in any of the possible directions except for Goldwater-Humphrey (which I also didn't find out on my own). Obviously, as with elections in which there is a strong trend towards one party, there must have been an overwhelming number of counties flipping in one direction. Like Johnson in 1964, Kennedy in 1960, Eisenhower in 1952, Roosevelt in 1932, and Harding in 1920, Nixon flipped a county in almost every state. (It's worth noting that he wouldn't have been able to flip any counties in Mississippi, where Goldwater had swept the counties; at the same time, Nixon failed to repeat this sweep.) 16 of the counties Nixon flipped were mega counties. Nixon appears to have carried about 1,023 more counties than Goldwater.
There were both Johnson and Goldwater counties that flipped to Wallace. The largest Johnson-Wallace county was Davidson, TN (136,607 votes). The largest Goldwater-Wallace counties were Jefferson, AL (205,033 votes) and Duval, FL (166,978 votes).
The biggest Goldwater-Humphrey county was Chatham, GA, which cast 53,545 votes.
There weren't many Goldwater-Humphrey counties, and they were concentrated in six Southern states. One wouldn't expect many Goldwater-Humphrey counties. In fact, one might not even expect as many as there actually were. 1968, like 1912, 1920, and 1932, is an election in which it would seem plausible that there would be no counties flipping in one of the two possible directions between the two major parties.
In general, people looking at the Goldwater-Humphrey counties seem to find it dubious to attribute them to actual Goldwater-Humphrey
voters. They are instead generally explained by some combination of the Voting Rights Act and Wallace splitting the vote. For example,
one poster writes,
Some of those can rather easily be explained by passage of the voting
rights act shifting political majorities from white voters in power due
to jim crow to black voters now able to vote due to the voting rights
act of 1965.
And, in response to the above poster asking about Goldwater-Humphrey counties with relatively small Black populations,
one poster responds,
IIRC, even in the Goldwater-Humphrey counties with few blacks, it was
about the vote being split three ways rather than two ways, and Humphrey
still getting a lower % than Johnson.
I wouldn’t say [these counties] have “few” blacks, they’re white majority with
sizable black minorities. In South Carolina especially, the white vote
was divided between Nixon and Wallace, enabling black voters to deliver
plurality wins for Humphrey in certain counties.
For the most part, Humphrey's vote shares in the Goldwater-Humphrey counties, and those counties' subsequent political trajectories, tend to bear this out. Humphrey got a majority in only five Goldwater counties: Greene, AL, and
Jefferson, Claiborne, Wilkinson, and Holmes, MS. All of these went on
to vote Democratic in every election from at least 1976 through 2020. There were a further twelve Goldwater counties in which Humphrey got a
plurality of at least 45%. Of these, eight voted Democratic in every
election from at least 1976 through 2020. The four that didn't were West
Feliciana, LA (where he got 48.8%), Madison, MS (where he got 47.7%),
Clarendon, SC (where he got 45.6%), and Beaufort, SC (where he got
45.5%).
Madison and Beaufort never went on to vote Democratic again except for Carter--Madison twice, and Beaufort, once. This tends to give me doubt about Goldwater-Humphrey voters significantly contributing to those counties' flips in 1968. One would think that a Southern county that voted for the Minnesotan-Mainer ticket of Humphrey-Muskie would vote, not only for Carter, but also for Bill Clinton. (That said, Humphrey's plurality in Madison was fairly high, and it voted for Carter not only in 1976, but in 1980 [and not narrowly in 1980: it gave Carter a majority of 54.3%]. It grew quickly in the decades after 1968, and it seems suburbanisation may have played a role here.)
West Feliciana and Clarendon, on the other hand, did not vote for a Republican loser again until the 21st century. West Feliciana voted Democratic in every election from 1976 through 1996, and then Republican in every election from 2000 through 2020; the Republican vote share cracked 60% in 2020. Clarendon voted Democratic in every election from 1976 through 2016, and then voted for Trump in 2020.
Also worth noting in this connexion is Jasper, SC, which gave Humphrey a plurality of 44.994%, and has gone on to vote Democratic in every election from 1976 through 2020, but did so in 2020 by less than 1% (49.9%-49.2%).
There is evidence that a number of traditionally Democratic voters broke Republican one time for Goldwater in 1964 and then returned to the Democratic Party; but in general, on their way back, they appear to have made a detour via George Wallace's third party candidacy in 1968. In evidence of this, there are a number of counties that voted for Goldwater in 1964, for Wallace in 1968, and then, at a minimum, for Carter twice, for Dukakis, and for Clinton twice, before turning Republican sometime in the 21st century. Just to name some where the Republican vote share has reached 60%, for example: Lawrence, AL; Crawford, Marion, and Pulaski, GA; and DeSoto and Pointe Coupée, LA.
But there are a number of these counties, whereas West Feliciana and Clarendon are somewhat isolated. (Especially since we don't know what will happen to the Republican vote share in Clarendon, which Trump carried in 2020 with a plurality of 49.97%. There are some Republican breakthrough counties in the South where the Republican vote share has hovered around 50% rather than climbing ever higher,
e.g., Peach, GA and Caroline, VA, indicating that, very likely, these were counties
'that are around 50/50 white/black, where enough white voters used to be
yellow-dog Dixiecrats to put even Mondale barely over the edge, but
collapses in white rural southern voting for Dems at the presidential
level has allowed Republicans to take those counties more recently.' If this were to happen to Clarendon, that would cast doubt, again, on the idea that actual Goldwater-Humphrey
voters were responsible for the county's flip--Biden lost the county with a higher vote share than that wherewith Humphrey carried it.)
It seems implausible that there would be one parish in Louisiana where traditionally Democratic voters who opted for Goldwater in 1964, and who would return to a loyally Democratic voting pattern within a couple more elections, would return forthwith to Humphrey, when this doesn't seem to have happened anywhere else. But it's hard to say for sure either way, of course. At a minimum, there appears to be no Goldwater-Humphrey county that really demonstrates the likely existence of Goldwater-Humphrey voters.
1972Nixon '68-
McGovern '72 (anti-trend): 16 (0.87% of Nixon '68 counties) (in
seven states)
Humphrey '68-
Nixon '72: many (in
47 states) (see below)
Wallace '68-Nixon '72:
Wallace '68-
McGovern '72 (anti-trend?): 6 (in
two states)
Again, I didn't bother counting the Humphrey-Nixon '72 counties. Again, there was at least one in practically every state. Using Leip's figures, Nixon carried about 1,131 more counties than he did in 1968 (some of which, obviously, came from Wallace). 13 were
mega counties.
It may seem odd to call the Wallace-McGovern counties 'anti-trend', but there were only six (a very small minority of all the Wallace counties). One forum poster made the observation that
'There were probably more Wallace-McGovern voters than there were Goldwater-Humphrey voters.' Ironically, three of the Wallace-McGovern counties show the likely effect of Wallace-McGovern voters in a way that none of the Goldwater-Humphrey counties does the likely effect of Goldwater-Humphrey voters. The six Wallace-McGovern counties were Bullock, Lowndes, and Wilcox, AL; and Stewart, Perry, and Houston, TN. The former three have voted Democratic in every subsequent election; whereas the latter three voted loyally Democratic for a period of time following 1972 but, at some point in the 21st century, turned powerfully Republican. All of them gave Trump at least 70% of the vote in 2020.
Four Wallace-Nixon '72 counties cast at least 100,000 votes:
Jefferson, AL (198,528)
Duval, FL (169,204)
Davidson, TN (134,797)
Jefferson, LA (101,841)
The sixteen Nixon '68-McGovern '72 counties were: Pitkin, CO, Jackson, IL, Washtenaw, MI, Stevens, MN, Athens, OH,
Rusk, WI, and ten in South Dakota: Bon Homme, Clay, Davison, Deuel,
Edmunds, Hanson, McCook, Moody, Sanborn, and Union. The most vote-producing was Washtenaw, MI, which cast 107,575 votes. In this case, the general explanation for them is not so much regionally-oriented, as it is about their being a certain type of county that exists in all regions:
'Jackson, Athens, and Washtenaw are all college towns. I think that has something to do with it.' Apparently
this even could apply to Stevens, MN, one of only two Nixon '68-McGovern '72 counties outside South Dakota that voted Republican in 2020. I haven't been able to find any institute of higher education, even a small one, in the other one (Rusk, WI), or anything else that would otherwise mark it as a
'quintessential "liberal elite"' 'type of [place]'. That said, they do seem largely concentrated in the Upper Midwest (although the one each in Illinois and Ohio are in the southern parts of their states). This is especially true when one considers that both of the Nixon '68-McGovern '72 counties (outside South Dakota) that voted for Trump in 2020 were in the Upper Midwest (Stevens, MN and Rusk, WI)--in Stevens County's case, apparently, its
being home to UMN-Morris wasn't enough to keep it in the Democratic column in the much more competitive elections of 2000, 2004, 2012, 2016, or 2020 (or to stop Trump, not only carrying the county, but winning a > 20% blowout in it, in 2020).
1976McGovern '72-
Ford '76 (anti-trend): 3 (2.31% of McGovern '72 counties) (in
three states)
Nixon '72-
Carter '76: 1,531 (
source) (51.4% of Nixon '72 counties) (in
48 states)
Carter flipped a county in every state that reports election results by county-equivalent save Vermont, where Ford carried every county. This was the last time a nominee would flip over 1,000 counties. 13 were
mega counties.
Ironically, despite how few counties Ford flipped, McGovern carried so few that they were actually more than 1% of the counties McGovern had carried. The three McGovern-Ford counties were Washtenaw, MI (which cast 111,689 votes), Pitkin, CO, and Clay, SD.
(Incidentally, both Nixon in '72 and Carter in '76--and, if I had to guess, no-one else, although I haven't been able to confirm this--likely flipped a ward in DC, as
Nixon carried DC's Ward 3 in 1972.)
1980Ford '76-
Carter '80 (anti-trend): 13 (0.93% of Ford '76 counties) (in
six states)
Carter '76-
Reagan '80: ~ 823 (48.1% of Carter '76 counties) (63.3 times as many as Ford '76-Carter '80 counties) (in
48 states)
Again, Reagan flipped a county in every state that reported election results by county-equivalent save Vermont, where Ford had carried every county in 1976 (although Reagan failed to repeat this sweep in 1980). Seven were
mega counties--perhaps a non-trivially smaller number than in previous elections with a clear and strong national trend (1952, 1960, 1964, 1968, 1972, 1976). Furthermore, unusually, one of the anti-trend counties was a mega-county (Monroe, NY).
The 13 Ford-Carter counties were: Cumberland County, Maine; Franklin,
Grenada, Marion, and Yazoo Counties, Mississippi; Marquette and
Washtenaw
Counties, Michigan; Monroe and Niagara Counties, New York; Chittenden
and Grand Isle
Counties, Vermont; and the independent cities of Franklin and Lexington
in Virginia. Aside from Monroe being a mega-county, two others cast over 100,000 votes (Washtenaw, MI--casting its third losing vote in a row--and Cumberland, ME); and Carter flipped the largest Ford county in Maine and Vermont (which were also the largest counties in their respective states full stop).
1984Reagan '80-
Mondale '84 (anti-trend): 36 (1.63% of Reagan '80 counties) (in
15 states)
Carter '80-
Reagan '84: ~ 604 (67.1% of Carter '80 counties) (16.8 times as many as Reagan '80-Mondale '84 counties) (in
32 states)
The 36 Reagan-Mondale counties were somewhat numerous for anti-trend counties, and also less geographically concentrated than anti-trend counties tend to be. They included no mega counties, but four of them did cast over 100,000 votes:
Denver, CO (219,741)
Polk, IA (147,848)
Lane, OR (125,852)
Marin, CA (116,718)
Those four have gone on to vote Democratic in every subsequent election, as have the relatively sizeable counties of Santa Cruz, CA (which cast 92,769 votes) and Arlington, VA (which cast 72,242 votes). The city-county of Denver was the seventh-most vote-producing turnover county overall in 1984, after Hartford, CT, Shelby, TN, Monroe, NY, Montgomery, OH, Summit, OH, and Travis, TX.
Three Reagan '80-Mondale '84 counties were mega-counties (Hartford, Shelby, and Monroe).
All that said, Reagan flipped a super-majority--over 2/3--of the blue counties from 1980, and over 16 times as many counties as Mondale flipped, which would clearly point to 1984 being a different kind of election than 1980 (unlike the county flips in 1936, 1956, or 1996).
1988Mondale '84-
Bush Sr '88 (anti-trend): 7 (
source) (2.10% of Mondale '84 counties) (in
three states)
Reagan '84-
Dukakis '88: ~ 493 (17.7% of Reagan '84 counties) (70.4 times as many as Mondale '84-Bush '88 counties) (in
42 states)
The county flips in 1988 were commensurate with a strong national trend towards the Democrats (belying that 1988 and 1984 were similar sorts of elections). Dukakis flipped counties in over 40 states, and the percentage of the red counties from the previous election he flipped was large, although it fell short of Kennedy's percentage in 1960. (That said, he didn't have nearly as many counties flipped against him as Kennedy did.) Eight were
mega counties.
Bush's turnover counties were, by number, breadth, and size, clearly anti-trend counties. They were located in three adjoining Southern states. The largest was Bibb, GA, which cast 44,396 votes.
1992Dukakis '88-
Bush Sr '92 (anti-trend): 23 (
source) (2.80% of Dukakis '88 counties) (in
six states)
Bush Sr '88-
Clinton '92: ~ 722 (31.5% of Bush '88 counties) (31.4 times as many as Dukakis '88-Bush '92 counties) (in
45 states)
Once again, the anti-trend counties were distinguished by their number, regional concentration, and size. (The source I gave gives '24', but it breaks them down by state, and says there were two in North Dakota; I could find only one, Nelson.) This time, the anti-trend counties were concentrated in Texas and the Upper Plains (although the one in Montana, Lincoln, was in the western, more mountainous part of the state)--such that the forum poster listing them says,
'Looks like a few disgruntled farmers changed their minds on Bush and a few Bentsen fans didn't like Clinton.' The largest, Woodbury, IA, cast 42,864 votes.
Clinton, as usual for a nominee successfully taking the presidency back for his party, flipped a county in nearly every state (with Dukakis having already swept the counties of Hawaii and Rhode Island, all of which Clinton kept). 14 were
mega counties.
1996Clinton '92-
Dole '96: 170 (11.2% of Clinton '92 counties) (in
31 states)
Bush Sr '92-
Clinton '96: 174 (11.0% of Bush '92 counties) (in
34 states)
This was very even; not only were the number of counties each candidate flipped almost even, but so were the percentages of the other party's counties they flipped. The number of states in which each candidate flipped counties was also very even (although Clinton also flipped counties in Maine--but these were all Perot counties).
Speaking of Perot, if one counts the Perot counties, Dole flipped slightly more counties, as he flipped 11 of the 14 Perot counties.
Five of the Clinton turnover counties, and two of the Dole turnover counties, were mega counties:
San Diego, CA (884,151)
Oakland, MI (505,642)
Suffolk, NY (505,213)
Franklin, OH (400,515)
Riverside, CA (391,613)
Fairfield, CT (352,227)
Dole flipped five counties that cast over 100,000 votes; four of these were in California (San Luis Obispo and Fresno, along with the two named above) (the fifth was Wake, NC). Clinton flipped no counties of any size in California, all of which is perhaps a reflection of the partial efficacy of Dole's 11th-hour push for California (which, interestingly,
largely focussed on immigration). However, both Dole and Clinton flipped counties throughout the country (with each having a particularly weak area: Clinton, the West; Dole, the Northeast). Dole was the only nominee to flip counties that had given the other nominee a majority in 1992 (both in the South): Owen County, KY and Macon County, TN (counterintuitively, however, Macon County was a historically Republican county that had last voted Democratic in 1880). That said, the highest HW Bush '92 vote share in a Clinton '96 turnover county came close to a majority: 49.0%, in St Bernard Parish, LA.
Dole's biggest 'breakthrough' county (that is, his biggest turnover county that had voted Democratic in the prior five elections save 1984) was Hardin, TX, which cast 17,900 votes (followed closely by neighbouring Polk, TX, which cast 14,244 votes). In 1992, Clinton had become the first Democrat to win the presidency without Texas since Texas became a state; in 1996, apparently, he was still
'convinced he could win' Texas, but instead, Dole--slightly--increased the Republican margin of victory in the state, as it reprised its new role as the best Republican raw vote margin state in the country.
One thing, tentatively, that seems to become apparent from 1956, 1984, and 1996, is that the West seems particularly prone to turn on incumbents seeking another term. (This wasn't so in 1972; it might have been so in 1980 and 1992 but would be difficult to tell simply from maps of states where each nominee flipped counties, given that the out-party nominee won. [It probably wasn't true in 1976.]) We see this again, to some degree, in 2004, and, although it is hard to tell just from the data we've gathered here, it seems possible to have been the case to some degree in 2012 and 2020 as well. (Of course, it also depends a bit on what is meant by the 'West'; sometimes the
Haga stack states behave like the more mountainous states to their west, and sometimes they don't.)
2000Dole '96-
Gore '00 (anti-trend): 2 (0.13% of Dole '96 counties) (in
two states)
Clinton '96-
Bush '00: 854 (
source) (55.96% of Clinton '96 counties) (427 times as many as Dole '96-Gore '00 counties) (in
45 states)
In both 1992 and 1996--
somewhat unusually for a victorious nominee--Bill Clinton carried slightly fewer counties than his opponent. Despite this, George W. Bush flipped a larger number of counties, it appears, than any nominee since Carter in 1976; and although it was not as great a proportion of the blue counties as Reagan's turnover counties in 1984, they were greater than any non-incumbent since at least 1968.
No election since 2000 has seen even half as many counties change hands, as changed hands in 2000. 2000 is really, in some sense, the last election in which nominees can be held to the usual standards for what is normal for non-incumbents winning the presidency back for their party. No nominee since has flipped remotely as large a proportion of the other party's counties as Bush did in 2000, nor has any nominee flipped counties in as many states (although Obama in 2008 came close).
There was even something already somewhat unusual about Bush in 2000; only three of his turnover counties were
mega counties, the fewest in some time for a nominee successfully taking the presidency back for his party.
Gore flipped only two counties--an unusually small number, given that the national swing in 2000 was significantly smaller than in 1920 or 1952--but one of them, Orange, FL, was the sixth-most vote-casting overall:
San Bernardino, CA (454,893)
Bexar, TX (412,726)
Hillsborough, FL (360,354)
Travis, TX (301,263)
Ventura, CA (282,692)
Orange, FL (280,155)
Mecklenburg, NC (263,036)
The other Dole-Gore county was Charles County, Maryland, which cast 44,592 votes.
Bush flipped four counties in which Clinton's '96 margin had been both over 30% and over 1,000 raw votes:
Arkansas, AR (63.00%-28.52%) (Clinton '96 margin: 34.48%)
Prairie, AR (61.85%-28.67%) (Clinton '96 margin: 33.17%)
Avoyelles, LA (59.0%-27.0%) (Clinton '96 margin: 32.0%)
Red River, LA (61.5%-31.3%) (Clinton '96 margin: 30.2%)
(He also flipped two counties where Clinton's '96 margin had been over 30%, but not over 1,000 raw votes [Foard, TX and Fisher, TX].)
No nominee since has flipped a county carried by the opposition party in the previous election by even as much as Prairie County, Arkansas had been carried by Clinton in '96 (whether the margin was over 1,000 raw votes or not).
The largest Clinton '96 margin in a Bush '00 turnover county casting at least 100,000 votes was 17.75%, in Anoka, MN.
2004Bush '00-
Kerry '04: 64 (2.67% of Bush '00 counties) (in
24 states)
Gore '00-
Bush '04: 156 (23.7% of Gore '00 counties) (2.4375 times as many as Bush '00-Kerry '04 counties) (in
29 states)
(A complete list of the 2004 turnover counties is
here.)
Kerry's turnover counties were very few, both on their own and relative to his opponent's, for a nominee successfully taking the presidency back for his party. They were also very few as a proportion of the counties the opposition party had carried in the previous election. And they were found in only 24 states, whereas previously, nominees successfully taking the presidency back for their party had flipped counties in nearly all the states.
All that said, of course, in most cases, nominees taking the presidency back for their party were not doing so in one close election after another close election.
There were two
mega counties that flipped in 2004, one to Kerry and one to Bush. However, the very largest turnover county went to Kerry, and amongst the very largest turnover counties, Kerry has a slight edge (he flipped four of the six most vote-casting turnover counties, five of the eight most vote-casting, and six of the ten most vote-casting). However, his advantage doesn't extend very far down the list. These are all the turnover counties in 2004 that cast over 100,000 votes in that election:
Fairfax, VA (461,379)
Pinellas, FL (455,357)
Macomb, MI (402,410)
Travis, TX (352,113)
Mecklenburg, NC (323,102)
Marion, IN (320,838)
Monmouth, NJ (299,939)
Guilford, NC (199,314)
Pasco, FL (190,916)
Stark, OH (188,459)
Staten Island, NY (160,143)
Rockland, NY (131,231)
St Joseph, IN (108,619)
Caddo, LA (106,595)
Bush flipped one county that had voted for Gore by over 20% (and by over 1,000 votes), and a further six counties that had voted for Gore by over 10% and over 1,000 votes:
DeKalb, TN (60.1%-38.5%) (Gore '00 margin: 21.6%)
Hickman, TN (58.4%-40.1%) (Gore '00 margin: 18.3%)
Rockland, NY (56.7%-39.5%) (Gore '00 margin: 17.2%)
Warren, TN (56.2%-42.3%) (Gore '00 margin: 13.9%)
Giles, TN (54.9%-43.5%) (Gore '00 margin: 11.4%)
Marshall, TN (54.6%-43.9%) (Gore '00 margin: 10.7%)
Okmulgee, OK (54.5%-44.0%) (Gore '00 margin: 10.5%)
Gore's '00 margin in four of these counties was larger than the largest HW Bush '92 margin in any HW Bush '92-Clinton '96 county (or, put another way, the largest prior-election margin in any county flipped by the last successfully re-elected president): 12.18% in Gulf, FL.
Of course, these counties are vastly disproportionately in Tennessee, whence one could deduce that Bush's biggest improvements were a function of Tennessean Al Gore no
longer heading the Democratic ticket. However--aside from the presence on this list of a county (Okmulgee) in another 'Scots-Irish zone' state--Bush flipped a further four counties (outside Tennessee) that had voted for Gore by over 1,000 votes and by more than 7.8%, at least three of which were arguably in or adjacent to the 'Scots-Irish zone'. (7.8% was the
margin whereby Val Verde County, TX had voted for Hillary Clinton in
2016; this was the third-largest Hillary '16 margin in any Hillary
'16-Trump '20 county that had given Hillary Clinton a > 1,000-vote
plurality [2020 being the next election in which a narrowly-elected Republican president was seeking re-election].)
Greene, PA (53.0%-43.1%) (Gore '00 margin: 9.9%)
Columbus, NC (54.2%-45.3%) (Gore '00 margin: 8.9%)
Franklin, IL (53.1%-44.2%) (Gore '00 margin: 8.9%)
Cameron, TX (53.5%-44.8%) (Gore '00 margin: 8.7%)
St Landry, LA (52.9%-45.2%) (Gore '00 margin: 7.7%)
Cameron, TX is in the Rio Grande Valley, and has gone on to vote Democratic in every subsequent election (the only one of the above 12 counties that has done so, other than Rockland, NY, which is a geographic outlier and likely reflects
'9/11 "belt voters"').
(It could also be noted that, of the three
'consecutive-election opposite-party landslide counties' in 2004 [all of which flipped from Gore to Bush], only one was in Tennessee--the other two were each in another 'Scots-Irish zone state' [Oklahoma and West Virginia]. Only the one in Tennessee [Marshall] had given Gore a > 1,000-vote plurality, but Bush himself actually carried Wyoming, WV by slightly more raw votes than he carried Marshall, TN by in 2004 [carrying both by over 1,000 raw votes]. Even the smallest of the three, Haskell, OK, has given Republicans > 1,000-vote pluralities in every election from 2008 on [and had given Democrats > 1,000-vote pluralities in 1976, 1988, 1992, and 1996].)
The largest Bush '00 margin in any of the counties Kerry flipped was 14.8%, in Steele, ND. However, Bush won Steele County by just 180 raw votes in 2000. Amongst counties that had given Bush a plurality of at least 1,000 votes in 2000, Kerry flipped two that had voted for Bush by at least 10%. He flipped only one additional one that had voted for Bush in 2000 by a greater margin than Cameron, TX had voted for Gore by in 2000:
Teton, WY (52.3%-38.5%) (Bush '00 margin: 13.8%)
La Plata, CO (48.8%-38.4%) (Bush '00 margin: 10.4%)
Missoula, MT (46.1%-37.0%) (Bush '00 margin: 9.1%)
Danville City, VA (51.5%-44.9%) (Bush '00 margin: 6.6%)
The percent margin in Steele County in 2000 had been 54.1%-39.3%. The county with the highest vote share for Bush in 2000 that flipped to Kerry was Roberts County, SD (which voted for Bush 54.9%-41.7%, or by 13.2%--but by only 537 votes).
Aside from the extreme outlier of Rockland, NY, the major turnover counties that voted for their winners in 2000 by the largest margins were Staten Island (which had voted for Gore by 6.98%) and Travis, TX (which had voted for Bush in 2000 by 5.21%). None of the others voted for their '00 winner by more than 5%.
2008Kerry '04-
McCain '08 counties (anti-trend): 44 (
source) (7.55% of Kerry '04 counties) (in
nine states)
Bush '04-
Obama '08 counties: 331 (
source) (13.1% of Bush '00 counties) (7.52 times as many as Kerry '04-McCain '08 counties) (in
42 states)
This election was relatively close to a 'normal' election for an election in which the presidency changes hands from one party to the other, particularly in contrast to 2020 (the only other election this century in which the Democracy took the presidency back), but in some ways even in contrast to 2016 and 2012 (when Romney lost but enjoyed a strong national trend). Obama flipped counties in fewer states than any nominee taking the presidency back for his party dating back to at least 1920, but he is the last nominee of either party to date to flip counties in even 4/5 of the states. Likewise, the proportion of the opposition party's counties he flipped was smaller than that flipped by any nominee taking back the presidency for his party dating back to at least 1920 (almost certainly--I didn't calculate this figure precisely for Eisenhower in 1952 or Nixon in 1968), as well as smaller than that flipped by Willkie in 1940 or Dukakis in 1988; but it was still recognisably in the same neighbourhood as these (Kennedy, for example, flipping just over 19% of the Eisenhower '56 counties).
Obama also flipped counties that had given Bush over 60% of the vote in 2004; this is the only election this century in which the Democrat has flipped counties that voted over 60% Republican in the previous election. However, it has typically been normal for nominees successfully taking the presidency back for their party to flip counties that had voted for the opposition party at over 60% in the previous election. (For example, the highest HW Bush '88 vote share in any HW Bush '88-Clinton '92 county was 68.00%, in Miller, GA, and the highest in any HW Bush '88-Clinton '92 county that had given HW Bush a > 1,000-vote plurality in '88 was 67.52%, in Pickens, GA. The highest Reagan '84 vote share in any Reagan '84-Dukakis '88 county was 67.65%, in Ellis, KS.)
The largest Bush '04 margin in any of the counties Obama flipped
was 24.5%, in Newton, GA. In all, there were eight counties Obama
flipped that Bush had won in 2004 by at least 20% and by at least 1,000
raw votes, in five of which Bush had gotten over 60% of the vote:
Newton, GA (62.0%-37.5%) (Bush '04 margin: 24.5%)
Douglas, GA (61.4%-38.0%) (Bush '04 margin: 23.4%)
Kendall, IL (60.8%-38.4%) (Bush '04 margin: 22.4%)
Salt Lake, UT (59.57%-37.54%) (Bush '04 margin: 22.03%)
Rockdale, GA (60.4%-38.9%) (Bush '04 margin: 21.5%)
Staunton City, VA (60.3%-39.0%) (Bush '04 margin: 21.3%)
Cass, ND (59.4%-39.0%) (Bush '04 margin: 20.4%)
McHenry, IL (59.7%-39.3%) (Bush '04 margin: 20.4%)
In addition to Newton, Douglas, Kendall, Rockdale, and Staunton City, there was one additional Bush '04-Obama '08 county that had voted over 60% for Bush in 2004 (but had given him a raw vote margin of only 819 votes):
Teton, ID (60.6%-38.4%)
It seems likely that every nominee taking back the presidency for his party in at least decades had flipped a county that had voted for the opposition party in the previous election by more than 24.5%, and that had given the opposition party more than 62.0% in the previous election. But the Bush '04 vote share in Newton was not vastly off from the Clinton '96 vote share in Arkansas, AR, or the HW Bush '88 vote share in either Miller or Pickens, GA. (The margin was a bit more off; Dukakis, Clinton, and Bush all flipped, in their initial elections, at least one county that had voted for the opposition party by over 1/3 in the previous election.)
Obama also flipped a number of large counties that had voted for Bush in '04 by large margins. He flipped 'only' nine
mega counties. This is relatively few compared to previous nominees taking the presidency back for their party, but there simply weren't many available to him to flip, as Bush had carried only 12 mega counties in 2004.
Obama flipped 65 counties casting over 100,000 votes, many of which ended up featuring on lists of counties Romney was said to have to win back in 2012 in order to win the election:
San Diego, CA (1,235,503)
Harris, TX (1,171,472)
Dallas, TX (740,074)
Riverside, CA (647,299)
San Bernardino, CA (606,334)
Bexar, TX (527,492)
Hillsborough, FL (514,501)
Pinellas, FL (465,152)
Wake, NC (442,245)
Hamilton, OH (425,086)
Macomb, MI (420,176)
DuPage, IL (417,973)
Salt Lake, UT (367,444)
Ventura, CA (341,041)
Jefferson, AL (318,524)
Kent, MI (303,799)
Lake, IL (299,900)
Jefferson, CO (295,068)
Will, IL (287,181)
Fresno, CA (273,452)
Arapahoe, CO (266,156)
Chester, PA (255,252)
Douglas, NE (226,701)
Dakota, MN (225,472)
San Joaquin, CA (210,308)
Hillsborough, NH (204,709)
East Baton Rouge, LA (197,349)
Kane, IL (193,363)
Clackamas, OR (191,878)
Clark, WA (183,925)
Berks, PA (180,511)
Washoe, NV (180,414)
Rockingham, NH (167,822)
Larimer, CO (166,375)
Forsyth, NC (166,133)
Prince William, VA (162,446)
Stanislaus, CA (161,512)
Henrico, VA (154,966)
Charleston, SC (154,434)
Orange, NY (151,982)
Somerset, NJ (151,430)
McHenry, IL (139,632)
Loudoun, VA (139,459)
Washington, MN (137,059)
San Luis Obispo, CA (133,155)
Dutchess, NY (132,302)
Rockland, NY (132,193)
Dauphin, PA (129,845)
Fayette, KY (127,648)
Cumberland, NC (127,575)
Lancaster, NE (127,490)
Winnebago, IL (126,704)
Brown, WI (124,754)
Marion, OR (124,563)
Buncombe, NC (123,795)
Lake, OH (121,642)
St Joseph, IN (118,739)
Saratoga, NY (111,387)
Caddo, LA (108,660)
Chesapeake City, VA (107,521)
Jefferson, MO (106,050)
Jackson, OR (101,047)
Osceola, FL (100,925)
Racine, WI (100,642)
Sangamon, IL (100,106)
(These are also all the counties casting over 100,000 votes that flipped in 2008. The largest Kerry-McCain county, Washington, PA, cast 98,047 votes in 2008.)
Two of these were amongst the Bush '04-Obama '08 counties that had voted for Bush in '04 by over 20%. A further six voted for Bush in '04 by over 15%:
Salt Lake, UT (59.57%-37.54%) (Bush '04 margin: 22.03%)
McHenry, IL (59.7%-39.3%) (Bush '04 margin: 20.4%)
Kent, MI (58.9%-40.2%) (Bush '04 margin: 18.7%)
Stanislaus, CA (58.6%-40.4%) (Bush '04 margin: 18.2%)
Sangamon, IL (58.6%-40.5%) (Bush '04 margin: 18.1%)
Douglas, NE (58.3%-40.2%) (Bush '04 margin: 18.1%)
Riverside, CA (57.83%-41.04%) (Bush '04 margin: 16.79%)
Fresno, CA (57.38%-41.68%) (Bush '04 margin: 15.70%)
Conversely, the Kerry-McCain counties were also--somewhat--'normal' for an election with a strong national trend. These were the 'anti-trend' counties in 2008, of course, and for the most part, they looked like anti-trend counties. There were 44--which is towards the higher end of what is 'normal', and, from a perch in 2021, looks like a lot (with a major-party nominee flipping 12, 17, and 15 counties in 2012, 2016, and 2020, and with the nominee flipping more counties, flipping 63 in 2020). However, it is fewer than half as many as Nixon flipped in 1960, and Obama flipped several times more counties than McCain. The Kerry-McCain counties were also substantially more geographically concentrated than the Stevenson '56-Nixon '60 counties, being much more restricted to the South than the latter. They were located in just nine contiguous states, eight of them in the South. Arkansas, Tennessee, and Kentucky (three of Karl Rove's four border states) led the way with 12, 12, and 7, respectively (no other state had more than four).
That said, one somewhat atypical thing about the Kerry-McCain counties is that some of them had voted for Kerry by very large margins. Two of them had given Kerry over 60% of the vote (and > 1,000-vote pluralities)--and greater margins than Newton, GA had given Bush in '04:
Knott, KY (63.4%-35.8%) (Kerry '04 margin: 27.6%)
Floyd, KY (62.2%-37.0%) (Kerry '04 margin: 25.2%)
This was unusual. (By comparison, the largest margin whereby Dukakis had carried any of the Dukakis '88-HW Bush '92 counties was 10.38% [in Adair, IA], and the largest margin whereby Mondale had carried any of the Mondale '84-HW Bush '88 counties was 4.46% [in Bibb, GA].) This made 2008 the only election since 1968 in which both nominees flipped counties that had voted for the opposition party at over 60% (and in 1968, that could well have simply been down to the Voting Rights Act and Wallace splitting the vote, as discussed above).
Nor were Knott and Floyd outliers (aside from the fact that there are two of them). There were three further Kerry-McCain counties that had voted for Kerry by over 10% and by over 1,000 votes; in two of them, Kerry's vote share had been over 55.
5% (which Dukakis had not reached in Adair, IA):
Humphreys, TN (57.6%-41.9%) (Kerry '04 margin: 15.7%)
Mingo, WV (56.2%-43.3%) (Kerry '04 margin: 12.9%)
Mississippi, AR (53.7%-43.3%) (Kerry '04 margin: 10.4%)
(If we take into account all Kerry-McCain counties, regardless of their raw vote margin for Kerry, the next-largest Kerry '04 margin in any of the Kerry-McCain counties, after Floyd, KY, was 16.8%, in Trousdale, TN, which voted for Kerry 58.0%-41.2%--but by only 537 votes.)
Despite the nation trending heavily Democratic, McCain, at a stroke, outdid all but five of the
Clinton '96 margins that Bush rolled back in 2000.
Furthermore--although the comparison might be somewhat unfair because
of the three-way nature of the '96 election--by flipping Knott County,
McCain outdid, narrowly,
the largest Clinton '96
vote share that Bush rolled back in 2000. Kerry's vote share in Knott, KY would stand as the highest prior-election vote share in
any turnover county this century (or since 1992), until 2020--despite, again, the Kerry-McCain counties being anti-trend.
None of the Kerry-McCain counties cast over 100,000 votes in 2008, although, again, Washington, PA came close (and cast over twice as many votes as Woodbury, IA did in 1992, or as Bibb, GA did in 1988). Four cast over 20,000 votes (a threshold we will see referred to in 2020), and a further six cast between 10,000 and 20,000 votes; these six included Al Gore's best raw vote margin county
in West Virginia and
in Kentucky:
Washington, PA: 98,047
Beaver, PA: 84,488
Fayette, PA: 52,560
Pike, KY: 22,653
Floyd, KY: 15,659
Fayette, WV: 15,295
Pointe Coupée, LA: 12,435
Mississippi, AR: 14,010
Logan, WV: 11,685
Assumption, LA: 10,960
Kerry's '04 margin in Washington, PA--the Kerry-McCain county that came closest to casting 100,000 votes (and actually did so in 2016 and 2020)--was 0.5% (
50.1%-
49.6%). His margin in Beaver, PA, which cast 94,122 votes in 2020, was 2.7% (
51.1%-
48.4%).
2012McCain '08-
Obama '12 (anti-trend): 12 (0.54% of McCain '08 counties) (in
10 states)
Obama '08-
Romney '12: 197 (
source) (22.5% of Obama '08 counties) (16.4 times as many as McCain '08-Obama '12 counties) (in
34 states)
Most lists of the McCain-Obama counties I've been able to find are incomplete, like
this one and
this one.
As far as I can tell, the McCain-Obama counties were: Staten Island,
Somerset, MD, Woodbury, IA, Chaffee, CO, Franklin, KY, Benton and
Warren, MS, Conecuh and Barbour, AL, Darlington, SC, Early, GA, and
Nash, NC.
Numerically, 197 is, historically, a small number of turnover counties for a nominee who is taking the presidency back for his party (or coming close). It was the fewness of the turnover counties that
impressed David Jackson:
Consider this factoid from Election 2012: Of more than 3,100 counties in
the United States, only 208 shifted their support between the
Democratic and Republican candidates -- less than 7%.
Barack Obama and Mitt Romney spent more than $1 billion each on their presidential campaigns.
The map above tells you what kind of change a couple of billion dollars will buy you in today’s America.
However, this is misleading. Obama carried historically few counties, for a winning nominee, in 2008. (2012 was the only election this century in which it was even mathematically possible for the Republican to flip as many counties as Bush did in 2000--but if he had, Obama would have been carrying only 21 counties, or only 33 if we assume he still flips the 12 anti-trend counties he actually did flip. By contrast, even George McGovern carried 130 counties.)
In fact, Romney flipped a larger proportion of the Obama '08 counties than Obama had flipped of the Bush '04 counties, or than Kennedy had flipped of the Eisenhower '56 counties. Unlike Obama in '08, in fact, Romney flipped a proportion of the opposition party's counties that was historically normal (if towards the low end of what was historically normal) for an election in which one party is taking the presidency back from the other party.
The 34 states in which Romney flipped counties was also historically few for a nominee taking the presidency back for his party, but it was more states than Kerry flipped counties in in 2004, or than Biden flipped counties in in 2020. It was only slightly fewer than states Trump flipped counties in in 2016.
The McCain-Obama counties were very few and (mostly) regionally concentrated--as is often the case with anti-trend counties, in the South. Both
Republican Michigander
I noticed most of these were Southern and mostly likely due to the black vote outside of Franklin County, KY (Government).
Barbour/Conecuh/Early/Warren/Nash/Somerset/Darlington: Black belt turnout was massive, and there may have been a slight D swing with whites in the Deep South.
give a regionalised explanation for the bulk of the Obama-McCain counties. They also give an idiosyncratic explanation (as was given for the New York area in 2004) for Staten Island, the single largest McCain-Obama county (and over thrice as large as the next-largest, Nash, NC):
Speaking of which, the McCain-Obama counties were, as is usually the case of anti-trend counties, markedly smaller than the counties that flipped in the other direction. Staten Island, the single largest McCain-Obama county, was the eighth-most vote-producing county to flip in either direction in 2012; this may sound impressive, but this was the farthest down on the list of largest turnover counties, that the largest Democratic turnover county had been since 1972 (a timeframe including a number of other elections in which the Democratic turnover counties were anti-trend: 1980, 1984, 2000). And Staten Island was an outlier in terms of size amongst the McCain-Obama counties. It was also a geographic outlier (as were all of those outside the South), making the argument that Hurricane Sandy swung it plausible (although Romney appears to have fallen flat in much of the Northeast, including in upstate New York). Had Romney kept Staten Island, Obama's biggest turnover county would have been truly historically small for the largest Democratic turnover county. (Nash, NC cast 48,492 votes in 2012.)
Altogether, 14 counties casting at least 100,000 votes flipped in 2012. The single largest turnover county was a Republican turnover county, which, again, may not sound that impressive, but 2012 remains the only election since 2000 of which this is the case.
Salt Lake, UT (384,174)
Kent, MI (294,206)
Chester, PA (252,576)
Volusia, FL (235,254)
Douglas, NE (223,927)
Berks, PA (170,676)
Rockingham, NH (170,423)
Staten Island, NY (154,180)
McHenry, IL (134,237)
Brown, WI (128,928)
Lancaster, NE (127,355)
Madison, IL (122,885)
Marion, OR (120,376)
Lake, OH (118,665)
Romney also flipped three counties Obama had flipped in 2008, and that had cast over 100,000 votes in 2008 but not in 2012: Jefferson, MO, Jackson, OR, and Sangamon, IL.
Another thing that marked the McCain-Obama counties as anti-trend was the smallness of McCain's margins in them. The largest McCain margin in any of the Obama turnover counties happened to be in Staten Island:
Staten Island (51.7%-47.6%) (McCain '08 margin: 4.1%)
None
of the others voted for McCain in '08 by more than 3%.
The 4.1% McCain '08 margin in Staten Island was smaller than the 4.3% Dole margin in Charles, MD (the Dole-Gore county which had voted for Dole by the larger margin--and which voted for Dole by over 1,000 votes). It was also smaller than the 7.78% Dukakis margin in Marion, IA (the largest-Dukakis-margin HW Bush turnover county in '92 that had given Dukakis a > 1,000-vote plurality), the 4.46% Mondale '84 margin in Bibb, GA, the 19.58% Reagan '80 margin in Apache, AZ, the 10.73% Ford margin in Monroe, NY, the 4.5% McGovern margin in Washtenaw, MI, or the 4.32% Nixon '68 margin in Washtenaw, MI. (In both 1972 and 1976, Washtenaw was the only anti-trend county that had given the opposition party a > 1,000-vote plurality in the previous election, although it may be worth noting that the largest percent margin whereby any anti-trend county had voted for the opposition party in those elections was 20.14% [/407 votes] and [exactly] 10% [/467 votes], both in Pitkin, CO.) It's hard to say how long it had been since a major party nominee had failed to flip a county that had voted for the opposition party in the previous election by a bigger margin than Staten Island had voted for McCain by in '08.
That said, the Romney turnover counties were unusual in their prior-election margins as well. Romney became the first (and thus far only) Republican since 1996 to fail to flip a county that had voted more than 60% Democratic in the prior election. The largest Obama '08 margin in any Romney turnover county was 15.2%, in Summit, UT:
Summit, UT (56.3%-41.1%) (Obama '08 margin: 15.2%)
The highest Obama '08 vote share in any Romney turnover county was 56.4%, in Ste Geneviève, MO (whose '08 margin for Obama was the second-largest amongst Romney turnover counties; both Summit and Ste Geneviève gave Obama > 1,000-vote pluralities in 2008):
Summit, UT (56.3%-41.1%) (Obama '08 margin: 15.2%)
Ste Geneviève, MO (56.4%-42.3%) (Obama '08 margin: 14.1%)
In fact, if we count all turnover counties (not just those that gave the opposition party a > 1,000-vote plurality in the previous election), the largest-prior-election Dem-margin in a Romney turnover county was the smallest of any Republican since 1992. (The biggest Clinton '92 margin in any Clinton '92-Dole '96 county was 20.23%, in Owen, KY--which gave Clinton in '92 a 722-vote plurality. It was also something of an outlier; the second-biggest Clinton '92 margin in any Clinton '92-Dole '96 county--and the biggest in any county that had given Clinton a > 1,000-vote plurality--was 12.12%, in Lincoln, TN.)
Summit also represented a departure from the previous four elections, in the sense that in 1996, 2000, 2004, and 2008, the largest prior-election-Dem-margin GOP turnover county was a then-historically Democratic county in the South that has gone on to vote Republican in every subsequent election (this is true of both Owen, KY and Lincoln, TN). More specifically, they were all in one of Arkansas, Tennessee, or Kentucky, which, together with West Virginia, were the states Karl Rove described as being the
'four historically Democrat[ic] border states' that the Bush campaign in 2000 had considered crucial to their strategy.
Whether it's altogether accurate to say that all these states were 'traditionally Democratic' as of 2000, it is true that these states retained significant enclaves of rural white Democratic support into and beyond the Clinton years (unlike, in particular, Mississippi and South Carolina, and to a significantly greater degree than most other Southern states). In their 2002 analysis of the 2000 election, Judis and Teixeira refer to Kentucky (which Bush had won by over 15% in 2000) as a
'former Democratic [enclave]'. In an
April 2008 article in which he anticipated the impending collapse of support for the Democracy amongst its most
ancestral constituency should Obama win the party's nomination, Michael Barone wrote that '[i]f Al Gore had carried just West Virginia or Kentucky or Tennessee or Georgia or
Arkansas—all states carried by Carter in 1976 and Clinton in 1992, all
heavy with Jacksonians—he would have been elected president in 2000.' Georgia is an odd suggestion,
as it was a Dole state. The other four were all states that had voted for Bill Clinton twice--and overlapped precisely with Karl Rove's four border states.
Not only the very largest-prior-election-Democratic-margin Republican turnover counties in the last four elections had been in these states, but most of the
several largest-prior-election-Dem-margin GOP turnover counties had been. Many had also been in West Virginia and Louisiana. Tom Schaller described Bill Clinton's two wins in Arkansas and Louisiana, specifically, as
'harken[ing] back to the Democrats' now-dead New Deal coalition'. (He thereby, interestingly, managed to name the precise two states in which Bill Clinton still holds
the record for most votes won by a Democrat for president.) Arkansas and Louisiana, collectively, were home to the six
largest-prior-election-Dem-margin Bush '00 turnover counties. Okmulgee, OK--the only county outside Tennessee (other than Rockland, NY) that Gore carried by over 10% which Bush flipped in 2004--is in 'east Oklahoma', which Barone took the trouble to point out was
'the most Jacksonian part of the state' (and was one of a number of almost entirely contiguous counties in which Hillary Clinton got over 60% in the 2008 primary). In two articles in which Barone appeared to attempt to formalise the states most dominated by 'Jacksonians', he included Oklahoma in both
'Greater Appalachia' and the
'Scots-Irish zone'.
Summit County, Utah, on the other hand, was far away from 'Greater Appalachia'; it was a then-historically Republican county (having voted Republican in every election from 1948 through 2004 save 1964 and 1996) in what had consistently become the
Most Republican State (although its reign as such was about to come to an end). Summit has also gone on to vote
Democratic in every subsequent election.
Of course, Ste Geneviève, MO is much closer to the 'Scots-Irish zone'. Michael Barone never included Missouri in either the Scots-Irish zone or Greater Appalachia, and Karl Rove didn't include it amongst the border states Bush had to win in 2000 (I'm guessing because Rove didn't consider it a border state, not because Rove thought it was a safe state for Bush--Bush won it by less than 4% in 2000). But Aaron Barlow calls native Missourian Harry Truman a
'Borderer' ('Borderer' here referring, almost certainly, to a term used by David Hackett Fischer
to describe essentially what Barone is describing variously as 'Greater Appalachian', 'Scots-Irish', or 'Jacksonian'). Missouri, along with Arkansas and West Virginia, was one of the states
Chris Matthews bemoaned Kerry's inability to keep competitive in 2004, the morning after election night:
They never seemed to have a chance in Missouri, they never seemed to have a chance in West Virginia, or Arkansas; where did they hope to get red states out of that huge clump the President won again?
And, much as Thomas Frank often uses Massachusetts as a byword for the contemporary Democracy, he has at times (
here, for example, and
here) used Missouri as a byword for what the Democracy
used to be.
And, in fact, much like Owen, KY, Lincoln, TN, Arkansas, AR, DeKalb, TN, and Knott, KY, Ste Geneviève was--for the most part--a historically Democratic county as of 2012 (with its historic Democraticness extending back before the New Deal)--although it did vote for Willkie in 1940 and Dewey in 1944. By carrying it, Romney became the fifth Republican to carry it since the Second World War, after Eisenhower in 1952 and 1956, Nixon in 1972, and Reagan in 1984. And it has gone on to vote Republican in every subsequent election.
So perhaps Summit was an outlier? Well, in fact, neither Summit nor Ste Geneviève was particularly representative of the several largest-prior-election-Dem-margin Romney turnover counties, at least those that gave Obama a > 1,000-vote plurality in 2008. Beyond Summit and Ste Geneviève, there were another seven counties that Romney flipped, that had voted for Obama by at least 10% and by at least 1,000 votes in 2008:
Summit, UT (56.3%-41.1%) (Obama '08 margin: 15.2%)
Ste Geneviève, MO (56.4%-42.3%) (Obama '08 margin: 14.1%)
Wood, WI (55.6%-42.5%) (Obama '08 margin: 13.1%)
Tippecanoe, IN (55.1%-43.5%) (Obama '08 margin: 11.6%)
Outagamie, WI (54.9%-43.3%) (Obama '08 margin: 11.6%)
LaSalle, IL (54.6%-43.5%) (Obama '08 margin: 11.1%)
Kewaunee, WI (54.7%-43.7%) (Obama '08 margin: 11.0%)
Oneida, WI (54.3%-43.9%) (Obama '08 margin: 10.4%)
Menominee, MI (54.0%-43.9%) (Obama '08 margin: 10.1%)
(There were also three counties that had voted over 55% for Obama that Romney flipped, but that gave Obama a < 1,000-vote plurality: Vermillion, IN [Obama '08 vote share:
56.1%], Iron, WI [Obama '08 vote share:
55.8%], and Gallatin, IL [Obama '08 vote share:
55.3%]. In light of what follows, it's worth noting that all of these were, more or less, traditionally Democratic as of 2012, although all had become patchy enough in their Democratic loyalties to vote for Bush once. Gallatin's loyalty to the Democracy, furthermore, predated the New Deal and appears likely to reflect a heavy 'Jacksonian' presence. [It was one of the 14 counties Hillary Clinton carried in Illinois in the 2008 primary, whereof Michael Barone wrote,
'all of them originally settled by southerners—more Jacksonian country'.])
If we look at the counties Romney flipped that voted for Obama in '08 by over 10% and by over 1,000 votes, and leave out both Summit and Ste Geneviève, the remaining seven do have a family resemblance. They are all in the Upper Midwest (the one each in Illinois and Indiana are in the northern halves of their states, and have a voting history that reflects sympathy with the Upper Midwest). Four are in Wisconsin alone. None of them could really be said to have been 'traditionally Democratic' as of 2012. Wood, Tippecanoe, and Outagamie can be straightforwardly said to have been traditionally Republican. The most that could be said of any of the others is that they had become, perhaps, Democratic-leaning swing counties in the prior few decades, after having been traditionally Republican in the deeper past. Of all seven, only two (LaSalle and Kewaunee) voted for Dukakis, and only one (LaSalle) had not voted for George W. Bush twice (LaSalle voted for Bush in 2004). All voted for Reagan twice.
Despite being a mixture of historically Republican and historically-Republican-turned-swing, six of these seven counties--like Ste Geneviève--have gone on to vote Republican in every subsequent election, giving Trump a higher vote share in 2016 than they gave Romney in 2012, and giving Trump a higher vote share in 2020 than in 2016. (The exception is Tippecanoe, which gave Trump a plurality in 2016 and then flipped to Biden, an anomaly most likely explained by its being home to Purdue University.)
---
Even though Romney flipped the several largest turnover counties (and even though the biggest Democratic turnover county was lower down on the list of all turnover counties by size than in any election since 1972), the size of Romney's largest turnover counties was also a bit underwhelming. In 2000, Bush flipped the 21st-largest county Clinton had carried in 1996--which was already historically low down on the list of the counties carried by the opposition party in the previous election for a nominee successfully taking the presidency back for his party. Salt Lake was
no more than the 39th-largest county Obama carried in 2008, as Obama carried 38 mega counties. (Indeed, this was the first election since at least the 1800s in which no
mega counties flipped.)
One might have thought this a consequence of the increasing Republican weakness in large counties, but here, Romney was outdone even by Trump (who flipped two mega counties, and the 24th-largest county Obama carried in 2012, in 2016). (Romney was also outdone by Dewey in 1948; in his nearly successful attempt at taking the presidency back for the GOP, Dewey flipped the 29th-largest county carried by FDR in 1944.) Amongst the
counties Romney was commonly said to have to win back in 2012, three cast more vote than Salt Lake in both 2008 and 2012: Hillsborough, FL, Wake, NC, and Hamilton, OH. Hillsborough was a mega county in both 2008 and 2012, and was the 32nd-biggest county Obama carried in 2008. Wake became a mega county in 2012. (Of course, there were other counties, in non-swing states, that Romney might also have been expected to flip, that cast more vote than Salt Lake, such as Harris, TX, DuPage, IL, and San Diego and Riverside, CA. The largest county I've seen an analyst explicitly expect Romney to be able to flip was
Oakland, MI, Obama's 18th-largest county in 2008.)
The largest Obama '08 margin in a Romney turnover county casting over 100,000 votes was 9.2045% in Chester, PA, followed closely by two other counties in which Obama's '08 margin had been about 9.2%, and an additional three counties in which Obama's '08 margin had been over 5%:
Chester, PA (54.0%-44.8%) (Obama '08 margin: 9.2045%)
Madison, IL (53.6%-44.4%) (Obama '08 margin: 9.1962%)
Berks, PA (53.8%-44.6%) (Obama '08 margin: 9.1596%)
Brown, WI (53.9%-44.8%) (Obama '08 margin: 9.1%)
Volusia, FL (52.1%-46.5%) (Obama '08 margin: 5.6%)
McHenry, IL (51.8%-46.4%) (Obama '08 margin: 5.4%)
Leaving aside Hillsborough, NH (which was something of an outlier in this respect), the
large counties Romney was commonly said to have to win back had voted for Obama in 2008 by margins ranging from 7.0% (in Hamilton, OH) to 15.9% (in Prince William, VA). A majority had given Obama greater margins than Chester (which was itself one of these counties); along with Prince William, these were Wake, NC (Obama '08 margin: 14.45%), Arapahoe, CO (Obama '08 margin:
12.9%), Washoe, NV (Obama '08 margin: 12.7%), Henrico, VA (Obama '08
margin: 12.2%), Forsyth, NC (Obama '08 margin: 10.4%), and Larimer, CO
(Obama '08 margin: 9.7%)--seven in total. Again, leaving aside Hillsborough, NH, four had given Obama smaller margins than Chester: Hillsborough, FL, Hamilton, OH, Jefferson, CO, and Loudoun, VA.
Obviously, the prior-election opposition-party-margins in Romney's turnover counties were abnormally small for a nominee taking back the presidency for his party (or coming close to doing so), both in general and in large counties. Obama's '08 margin in Chester was far smaller than Bush's '04 margin in Salt Lake or Clinton's '96 margin in Anoka (or Obama's '12 margin in Erie, PA). Even Obama's '08 margin in Prince William would have been a bit low to have been the largest Obama '08 margin in a major Romney turnover county, if Romney were successfully taking the presidency back for his party. (There were, either shortly before or after the election, scattered musings that Romney should be able to, or should have been able to, roll back Obama '08 margins in major counties as great as 18.18% [in
Fairfield, CT] or 20.78% and 21.3% [in
Montgomery and Delaware Counties, PA, respectively].)
However, Romney flipped four major counties that had voted for Obama in '08 by greater margins than any but one of Biden's major turnover counties in 2020 had voted for Trump in '16 by. (Romney also flipped six major counties that had voted for Obama in '08 by greater margins than any of Kerry's major turnover counties in 2004 had voted for Bush in '00 by--Kerry, like Romney, falling just short of unseating a president.)
And, amongst counties of any size, Romney flipped one (Summit) that had voted for Obama in '08 by a larger margin than any Kerry turnover county had voted for Bush in '00 by, and two (Summit and Ste Geneviève) that had voted for Obama in '08 by a larger margin than any of the Kerry turnover counties that had voted for Bush by at least 1,000 votes in 2000 had voted for Bush in 2000 by.
In comparison with Biden--who actually did unseat a president--Romney flipped, again, two counties that had voted for Obama in '08 by larger margins than any of Biden's turnover counties had voted for Trump in '16 by; and nine counties that had voted for Obama in '08 by larger margins (and by at least 1,000 votes) than any but one of Biden's turnover counties had voted for Trump in '16 by (with Obama's '08 margin in Menominee being slightly greater than Trump's '16 margin in Talbot, MD).
2016Romney '12-
Hillary Clinton '16: 17 (0.70% of Romney '12 counties) (in
11 states)
Obama '12-
Trump '16: 218 (
source) (31.5% of Obama '12 counties) (12.8 times as many as Romney '12-HRC '16 counties) (in
36 states)
Some sources (like Schaller) seem to give '206' as the number of Trump turnover counties in 2016. However, they are probably assuming that the so-called
'pivot counties' (those that voted Obama-Obama-Trump--I have no idea why these merit a special name) are the only Trump turnover counties. In addition to these, Trump also won back every McCain-Obama county.
The 218 counties Trump flipped were very few for a nominee successfully winning back the presidency for his party; on the other hand, they were about 31.5% of the blue counties from 2012, easily the largest proportion of the opposition party's counties any nominee has flipped since 2000. The 36 states in which he flipped counties, while few in a historical context, were, again, also more than either Romney or Kerry flipped counties in.
The largest Obama '12 margin in any county Trump flipped was 26.1%, in Franklin, NY (which also voted for Obama by over 1,000 votes). In all, there were four counties that voted for Obama by more than 20% (and by over 1,000 votes), that Trump flipped in 2016:
Franklin, NY (62.1%-36.0%) (Obama '12 margin: 26.1%)
Clinton, IA (60.6%-37.7%) (Obama '12 margin: 22.9%)
Trumbull, OH (60.48%-37.54%) (Obama '12 margin: 22.94%)
Mower, MN (60.03%-37.4%) (Obama '12 margin: 22.6%)
For the second election in a row, then, the largest-prior-election-Dem-margin GOP turnover counties were all outside the South (except this time, none of them was even conceivably in a border state or region).
Unlike in 2012, however, these were all nevertheless traditionally Democratic counties (as of 2016). They had last voted Republican in 1988, 1984, 1972, and 1960, respectively. And, unlike Summit, UT (but like Owen, KY, Lincoln, TN, Arkansas, AR, DeKalb, TN, and Knott, KY), they all went on to vote Republican in every subsequent election to date (which, in their case, means they voted for Trump again in 2020).
The pattern of Romney's largest-prior-election-Dem-margin turnover counties might have given the impression that the Southern Strategy had essentially come to completion in the 2008 election. This would--more or less--be correct.
As Robert Wheel noted just before the 2016 election,
Democrats continue to face an eroding base among the white-working class
that was once their rock — and Trump looks set to win these voters by
an even bigger margin than Republicans typically do. In theory, that
means there are plenty of ancestrally Democratic counties that could
turn against the party. The problem with that analysis is that, for the
most part, Democrats have already lost those counties. In perhaps the
most extreme example, Knott County, Kentucky gave Al Gore 67.3% of the
vote and Mitt Romney 72.5% of the vote just 12 years apart.
As it turned out, Democrats
were still carrying a lot of those counties (as one could infer from the number of possible Trump streak-breakers Wheel came up with)--but not in the South, the region where Wheel found his 'most extreme example' of the erosion of the Democrats' white-working class base. Wheel came up with only one possible Trump streak-breaker in the South, Elliott, KY, whereof he noted that 'it wouldn't be on this list if Romney had won 61 more votes in 2012.' Writing in June 2016,
Robert David Sullivan was even more emphatic about the unlikelihood of the Democracy continuing its winning streak in Elliott: 'The blue speck will probably soon disappear: President Obama’s 49.4
percent plurality was the worst showing by a Democrat in more than 100
years.'
But there
were still a number of fairly powerfully Democratic white-working class counties in the Frost Belt. In general, these areas' Democratic loyalties did not predate the New Deal (which was true of all four Obama '12-Trump '16 counties that had voted over 60% for Obama in 2012). Many of these areas were heavily industrialised and hadn't been won over by the Bush-era appeal to rural America. With the flipping of Trumbull (and the narrowing of the margin in neighbouring Mahoning from a blowout of 28.31% to just 3.25%) in particular, the second part of the veiled hypothesis about the coming Republican coalition put forth by Judis and Teixeira
in their 2002 analysis of the 2000 election finally came into sight: '
The Bush
administration can scour the coal pits of West Virginia or the boarded up steel mills of
Youngstown for converts...'
Trumbull
cast 96,676 votes in 2016, just short of 100,000. It had cast over
100,000 votes in 1984, 1992, 2004, 2008, and 2012, and would do so again
in 2020. Had it done so in 2016, Trump would have won a major county
that had voted for the opposition party in the previous election by a
larger margin than any other nominee this century, as Obama's 22.94%
margin in it in 2012 was slightly larger than Bush's 22.03% margin in
Salt Lake County in 2004. Not only that, but he would have become the
only nominee this century to flip a major county that had voted for the
opposition party at over 60% in the previous election. (Bill Clinton flipped three in 1992, two of which voted for him twice and one of which went on to vote Democratic in every subsequent election: Ventura, CA [HW Bush '88 vote share: 61.64%], San Diego, CA [HW Bush '88 vote share: 60.19%], and Montgomery, PA [HW Bush '88 vote share: 60.20%].)
As it was, the largest Obama '12 margin in a county casting at least 100,000 votes that Trump flipped was 16.03%, in Erie, PA (57.36%-41.33%).
While this was slightly smaller than the Clinton '96 margin in Anoka,
it was easily larger than the Obama '08 margin in Chester. It was even
larger than Obama's 15.9% '08 margin in Prince William, VA--generally
speaking, the largest Obama '08 margin in a large county that Romney was
widely expected to roll back. And, although it may have been the smallest prior-election opposition-party-margin in a major turnover county for a nominee taking the presidency back for his party in some time, that record (whatever record it was) would be broken, ironically, in 2020--by a lot.
Like Anoka (but unlike Chester), Erie was a traditionally Democratic county (having last voted Republican in 1984). Anoka went on to vote Republican in every subsequent election, and even though it appeared shaky for Trump in 2020, even the six-election run from 2000 to 2020 would be fairly impressive. Chester, on the other hand--traditionally Republican as of 2012--went on to vote Democratic in every subsequent election.
Erie narrowly flipped back to Biden in 2020, so we already know 2016 won't have been the beginning of an extended Republican winning streak in the county. However, even in losing it, Trump upped his vote share in the county, from 48.0% to 48.8%. (Trump's 48.0% vote share in Erie in 2016 was higher than he got in a number of breakthrough counties he won again in 2020, such as in Kenosha [47.2%] and Dubuque [47.2%]. It was also higher than Bush's 2000 vote share in Anoka [47.6%], the major Bush turnover county that had voted for Clinton in '96 by the largest margin [and which, like Erie, was a breakthrough county--and which Bush carried again in 2004]. Which is to say it would be fallacious to simply chalk Trump's carriage of Erie in 2016 up to third-party vote. It also implies, however, that Trump was not able to increase his vote share in the county as much as he should have been able to, given that his 2016 vote share there would ordinarily have put him in a good position to carry it again.)
Even if Biden's flipping the county back implied that Trump didn't increase his vote share in Erie in 2020 by as much as one would expect if he were successfully winning re-election, his vote share actually declined, both in 2016 vis-à-vis Romney, and then again in 2020, in Chester.
---
You may notice I didn't mark Hillary Clinton's turnover counties as
'anti-trend', even though they seem clearly anti-trend by virtue of
their numbers, and even though 2016 is the kind of election in which one
expects anti-trend counties (namely, an election in which the
presidency changes parties).
Both are true, at
least at first glance. However, unlike most elections in which the
presidency changes parties, 2016 was one close election following
another. The national Republican swing was slight: just 1.77%. Indeed,
it may be more surprising how few Romney-Hillary counties there were.
The
Romney-Hillary counties were, first of all, disproportionately large.
In fact, for the first time in at least 100 years, a nominee succeeded
in taking the presidency back for his party despite the largest turnover
county overall switching against him--and that county was not an outlier:
Orange, CA (1,197,521)
Suffolk, NY (681,254)
Pinellas, FL (497,485)
Salt Lake, UT (423,743)
Macomb, MI (419,312)
Cobb, GA (334,058)
Gwinnett, GA (330,950)
Chester, PA (272,998)
Anne Arundel, MD (270,081)
Fort Bend, TX (262,066)
Montgomery, OH (259,876)
Douglas, NE (240,433)
Hillsborough, NH (214,157)
Staten Island, NY (180,960)
Stark, OH (176,165)
Orange, NY (152,021)
Northampton, PA (144,566)
St Lucie, FL (141,993)
Gloucester, NJ (141,254)
Lancaster, NE (136,223)
Luzerne PA (135,901)
Erie, PA (125,129)
Saratoga, NY (114,094)
Chesapeake City, VA (112,662)
Trump actually flipped two mega counties, to one for Hillary (as Salt Lake was not yet quite a mega county). Of the top twelve vote-producing
counties to switch sides, however, eight flipped from Romney to
Hillary.
The Romney-Hillary counties are (as is often the case of anti-trend counties) nameably finite: Orange and Nevada, CA; Cobb, Gwinnett, and Henry, GA; Anne Arundel, MD;
Gallatin, MT; Douglas and Lancaster, NE; Watauga, NC; Chester, PA; Fort
Bend and Kenedy, TX; Salt Lake and Summit, UT; Montgomery, VA; and
Whitman, WA.
However, there are some unusual things about the Romney-Hillary counties. First of all, the largest Romney '12 margin in any of them was 20.3%, in Salt Lake County (which was also the largest Romney '12 margin in any of them that cast over 100,000 votes). This was greater than Obama's 2012 16.03% margin in Erie, PA. It was also greater, by far, than Bush's '00 margin in Travis, TX, or than Obama's '08 margin in Chester, PA.
Now, one might say Salt Lake doesn't count, because Hillary Clinton won it with a very low plurality in a genuine three-way race at the county level. The non-fluke nature of Hillary's win there was confirmed by Biden's majority win in the county in 2020 (whereas, if it were just to be 'blamed' on third-party vote, one would have expected Trump to reconsolidate the Republican vote and carry the county in 2020--
which some people did expect). But even if we don't count Salt Lake, then the largest Romney '12 margin in any Romney-Hillary county (which was, again, in a major county) was 12.4%, in Cobb, GA. This was, again, greater than Bush's '00 margin in Travis or than Obama's '08 margin in Chester. Not only that, but it was greater than the biggest Trump '16 margin in any major county Biden would flip in 2020.
The next-largest Romney margin in any of the Romney-Hillary counties was, again, in a major county (9.1964%, in Gwinnett, GA). Even this was larger than all but the single largest Obama '08 margin in any major Romney turnover county, and larger than all but the single largest Trump '16 margin in any major Biden turnover county. (It was also still larger than the largest Bush '00 margin in any major Kerry turnover county.) Indeed, Romney's 6.8% margin in Fort Bend, TX and his 6.22% margin in Orange, CA (both major counties) were both larger than the largest Bush '00 margin in any major Kerry turnover county, and larger than all but two of Trump's '16 margins in any major Biden turnover county.
In flipping Salt Lake County, Hillary Clinton was flipping the single biggest county Romney (or anyone) had flipped four years earlier. In flipping Chester, she was flipping the major county in which Romney had rolled back the largest Obama '08 margin.
Hillary Clinton flipped the largest county Romney had carried in
six states (including three large states), the second-largest county (in addition to the largest) that Romney had carried in
two states, and
the largest county in
two states. She flipped Romney's best raw vote margin county in one state--the biggest state in the country (Orange, CA). Orange had also been the second-largest county Romney had carried nationally in 2012.
Much as she flipped the single largest county Romney flipped nationally in 2012, Hillary Clinton also flipped the single largest county Romney had flipped in
seven states (although this wasn't necessarily saying that much in every case--Watauga was the biggest Romney turnover county in North Carolina, but it was smaller than Nash, the biggest turnover county in the state that year; and Romney flipped only two counties in Virginia [although Montgomery, VA is decently-sized]. Romney flipped four counties in Washington, but none of them, including Whitman, was particularly sizeable.)
Nor were the
Romney-Hillary counties regionally concentrated, as one would normally
expect from anti-trend counties. It is true that five of the eleven
states containing Romney-Hillary counties were in the South, and that a
group of five of these states were contiguous. (Texas was quite removed
from the other Southern states.) But there was at least one in every
time zone, and, by most reasonable definitions, there was one in every
region of the country save the Midwest (although some would count the
two in Nebraska as Midwestern). Not only was there one in every region
of the country save (perhaps) the Midwest, but there was a large one
in every region of the country save (perhaps) the Midwest--for example,
Chester in the Northeast, Salt Lake in the Mountain West (or West),
Orange if the Pacific coast is counted as a separate region, and several
in the South (and Douglas in the Midwest, if Nebraska is counted as
Midwestern).
The nine major Hillary turnover
counties had all voted for Bush twice, and their last two losing votes
had been for Republicans. (In fact, this was true of every
Romney-Hillary county save tiny Kenedy, TX.) This means that these
counties could be considered 'Hillary Clinton beachheads', and three of
them in particular--
Cobb and Gwinnett, GA and
Chester, PA--received
a fair amount of credit for Biden's win four years later; perhaps more
than any actual Biden turnover county save Maricopa.
---
Aside from the four counties mentioned above, there was one additional county that Trump flipped that voted for Obama in 2012 by over 20% (although Obama's vote share in it fell short of 60%). It didn't give Obama a > 1,000-vote plurality, but it is
mentioned by David Wasserman, in one of the few contexts I've seen a political analyst give importance to a turnover county's prior-election margin. In this case, the significance is that this was the only county in the country that had voted for Obama in 2012 by over 20%
and for Trump in 2016 by over 20%. This was:
Howard, IA (59.6%-38.6%) (Obama '12 margin: 21.0%/973 votes)
As you can see, Obama's 2012 raw vote plurality from the county fell just short of 1,000 votes (as did Trump's in 2016--he won it that year by 934 votes). I'm not sure where Wasserman would draw the line, but I would imagine he wouldn't have seen it as particularly significant if, say, Kenedy County, TX had voted for two opposite-party nominees by over 20% in consecutive elections. (1,000 votes is the raw vote margin that William Saletan noted in June 2004 that no-one had won Summers County, WV by in the prior six elections.)
In describing the significance of Howard, Wasserman defines a 20% margin as a 'landslide'. He doesn't really give a basis for this definition. While it is a welcome corrective to the 'margin inflation' that has been taking place recently, it is also perhaps a bit restrictive. By Wasserman's definition, there are no other consecutive-election opposite-party landslide counties in the past six elections, apart from Howard, IA in 2016. Does that mean 2016 was a qualitatively unique election? Howard is just one county; it could just mean Howard was odd.
In June 2016, Robert David Sullivan characterised those counties that had voted for either Obama or Romney in 2012 by less than 10% as being able to 'be considered "purple" by the most generous definition'. He doesn't use the word 'landslide', but we can see which counties, for a given election, would not have been able to be considered 'purple' by the most generous definition in the prior election, and then went for the other party in such a way that it would once again not have been able to be considered 'purple' by the most generous definition (and, for the sake of convenience, we can call these 'landslide counties'). A list of such counties for the 2000 and 2016 elections is here (and is still fairly finite, in both cases representing a small proportion of all turnover counties). There were many more for Bush in 2000 than for Trump in 2016, but Trump's, on average, were probably larger--although, at the same time, there were somewhat more of these counties that gave both Clinton in '96 and Bush in '00 raw vote pluralities of at least 1,000 votes (21) than there were that gave both Obama in '12 and Trump in '16 one (13).
It is probably reasonable to say that the Clinton '96-Bush '00 consecutive-election landslide counties illustrate, about the 2000 election, illustrate, about the 2000 election, what Wasserman argues Howard, IA illustrates about the 2016 election (to wit, 'why so many people who
cast a ballot for former president Barack Obama in 2008 and 2012 — and who saw Trump as
unqualified to be president — nonetheless voted for him'). And that 'Democrats can’t credibly blame Howard County’s enormous
41-point swing in just four years on a last-minute letter to Congress, voter ID laws or Russia-sponsored Facebook ads.' (In a county that goes from a
ten-point margin for one party to a ten-point margin for the other, the swing would be
20%--or one out of every five voters changing his mind between two elections. In 2020, after Chuck Todd said that
'Cobb [County] has made a remarkable transformation', he displayed the county results for the previous three elections and continued, 'it is literally swinging almost twenty points.' [But of course, this was over
multiple elections.])
Indeed, arguably, the list of counties that gave Obama an at least 10% margin in 2012 and Trump an at least 10% margin in 2016 illustrates Wasserman's point better than simply the singular county of Howard, IA. That said, even by Wasserman's criteria, Scott County, Arkansas came very close to being a consecutive-election opposite-party landslide county in 2000 (its margin for Clinton in '96 fell just short of 20%, at 19.39%)--and, out of a smaller total electorate, it gave fairly close to 1,000-vote pluralities to both Clinton in '96 and Bush in '00. (If one were to choose a singular county from 2000, it might be Vermilion Parish, Louisiana--the only county to give both Clinton in '96 and Bush in '00 a > 55.5% vote share, making it particularly unmistakable in this case that the swing could not be explained by assuming [and it would be an assumption] that Bush had simply absorbed the Perot vote [although even in the other cases, this would--aside from, again, being an assumption--not be sufficient to explain the swing].)
The two sets of counties illustrate the different foci of where Bush and Trump, respectively, changed the political landscape (with Bush having done so with the greatest intensity in the South, where there were still a significant number of staunchly Democratic 'white counties in the rural South' as of 1996, when, with their help, Democrats could still hope to carry states such as Arkansas, Louisiana, Missouri, Kentucky, and Tennessee [and could almost count on West Virginia]; and with Trump having done so most deeply in the Frost Belt). In each case, one mid-size Mississippi River state in particular is the epicentre of the two Republicans' transformation of the map: Louisiana in 2000, Iowa in 2016.
2020Hillary Clinton '16-
Trump '20: 15 (3.07% of Hillary '16 counties) (in
six states)
Trump '16-
Biden '20: 63 (2.40% of Trump '16 counties) (4.2 times as many as Hillary '16-Trump '20 counties) (in
30 states)
The 63 counties Biden flipped, and the 2.40% of the red counties Biden flipped (a slightly smaller proportion of the red counties than Kerry flipped in 2004), are both deeply historically unusual for a nominee successfully taking the presidency back for his party. The 30 states in which Biden flipped counties is also historically few, although not that many fewer than the 36 states Trump flipped counties in in 2016. It seems almost certain that the > 13% of the red counties that Obama flipped in 2008 was the smallest proportion of the opposition party's counties that any nominee successfully taking the presidency back for his party had flipped in at least a century. But this figure was nevertheless kind of in the same neighbourhood as, say, the > 19% of the red counties that Kennedy flipped in 1960. The proportion of the red counties that Biden flipped, on the other hand, is nowhere near the range of what is historically normal.
Furthermore, the largest Trump '16 margin in any county Biden flipped was 13.37%--smaller than the Bush '00 margin in Teton, WY, or the Obama '08 margin in Summit, UT.
Biden flipped 11 counties that Trump had carried by over 5% in 2016 (all of which gave Trump a > 1,000-vote plurality that year):
Inyo, CA (51.94%-38.57%) (Trump '16 margin: 13.37%) (cast 9,484 votes in 2020)
Talbot, MD (52.2%-42.1%) (Trump '16 margin: 10.1%) (cast 22,555 votes in 2020)
Williamson, TX (50.9%-41.3%) (Trump '16 margin: 9.6%) (cast 290,168 votes in 2020)
Stafford, VA (51.4%-42.3%) (Trump '16 margin: 9.1%) (cast 79,625 votes in 2020)
Lynchburg City, VA (50.4%-41.5%) (Trump '16 margin: 8.9%) (cast 36,363 votes in 2020)
Tarrant, TX (51.74%-43.14%) (Trump '16 margin: 8.60%) (cast 834,697 votes in 2020)
Garfield, CO (49.6%-42.6%) (Trump '16 margin: 7.03%) (cast 30,094 votes in 2020)
Rockingham, NH (49.9%-44.1%) (Trump '16 margin: 5.8%) (cast 199,342 votes in 2020)
Carroll, NH (49.4%-43.9%) (Trump '16 margin: 5.5%) (cast 33,297 votes in 2020)
Tippecanoe, IN (48.6%-43.1%) (Trump '16 margin: 5.5%) (cast 62,515 votes in 2020)
James City County, VA (49.4%-44.3%) (Trump '16 margin: 5.1%) (cast 49,622 votes in 2020)
Trump's highest vote share in any county Biden flipped was 52.2%, in Talbot, MD. This, too, is historically low. It is lower than Bush's '00 vote share in Teton, WY, and easily lower than Obama's '08 vote share in Ste Geneviève, MO, or Romney's vote share in Cobb, GA.
Perhaps the most apt comparison is with Kerry, because, much like Kerry, Biden was seeking to unseat a Republican president whose initial election had been extremely narrow. Again, Trump's 2016 margin in Inyo was smaller than Bush's 2000 margin in Teton, and Trump's 2016 vote share in Talbot was lower than Bush's 2000 vote share in Teton.
But Teton was merely the largest-Bush '00-margin Kerry turnover county that had given Bush a > 1,000-vote plurality in 2000. When considering all of Kerry's turnover counties, it was neither the one with the biggest Bush '00 margin, nor the one with the highest Bush '00 vote share. There was at least one other Bush-Kerry county that had given Bush a greater percent margin than Inyo had given Trump, and at least three that had given Bush a higher vote share than Talbot had given Trump (along with a fourth that had given Bush nearly as high a vote share as Talbot had given Trump, and a greater vote share than Inyo had given Trump):
Steele, ND (54.1%-39.3%) (Bush '00 margin: 14.8%/180 votes)
Roberts, SD (54.9%-41.7%) (Bush '00 margin: 13.2%/537 votes)
Mono, CA (52.53%-40.91%) (Bush '00 margin: 11.62%/508 votes)
Price, WI (52.2%-43.0%) (Bush '00 margin: 9.2%/723 votes)
(Bush's 52.2% in Price was slightly lower than Trump's 52.2% in Talbot.)
Price, Steele, and Roberts have gone on to vote for Trump twice (Price has twice given Trump a > 1,000-vote plurality). Kerry's ability to flip them was evidence that, even as he suffered staggering losses in the rural and small-town South,
Lower Midwest, and parts of the Mountain West
with respect to Gore, there were a few regions of the country (in this case, the Upper Midwest) where he actually had greater appeal to
'Cracker Barrel voters' than Gore (although to such a small degree that it was drowned out by his declines throughout the South and South-adjacent parts of the country).
Biden does not seem to have had a similarly improved appeal to 'Cracker Barrel' voters anywhere (which is saying quite a bit,
given that the nominee he would have been improving on was Hillary Clinton). Nowhere was Biden able to roll back a Trump margin as big as in Roberts,
Price, or Steele in a county that had used to be friendly to Democrats
(as all those three were as of 2004--all three had voted for Carter in
1976, Dukakis in 1988, and Bill Clinton twice). And there is evidence he was expected to be able to--that he was expected to be able to do more than simply swing Erie, PA by 2.6% and Saginaw, MI by 1.5%. On election night,
Juju Chang reported that the Democratic Attorney General of Michigan was 'confident' that Macomb County--perhaps the biggest 'Cracker Barrel county' of them all (and one without which it would be unimaginable for Trump to even compete in Michigan)--would break for Biden. And Chuck Todd said that
'we expect [Macomb] to be very close, whether it's Biden or Trump.' (Macomb County voted for Trump in 2016 by 11.53%, a larger margin than any Biden turnover county of any size save Inyo ended up having voted for Trump in '16 by.)
And when discussing Biden's chances of winning Ohio,
Chuck Todd mentioned that Biden would have to win what he called 'Lake Erie counties', where 'there used to be some union Democrats', and which Sherrod Brown had won 'back' in 2018. Later on, he pulled up a map that highlighted the eight specific counties in Ohio that had voted for Trump in 2016 and for Sherrod Brown in 2018, saying
'Biden needed to carry these.' He then clicked specifically on Erie and and even Ashtabula Counties, which had last voted Democratic in 1988 and 1984, respectively, and which had voted for Trump in 2016 by 9.48% and
18.79%, respectively. Had Biden flipped Erie, OH, that would have made it his fourth largest-Trump '16-margin turnover county; Ashtabula, obviously, would have easily been his largest-Trump '16-margin turnover county.
Even those counties that Biden flipped that might, at first glance, seem like 'Cracker Barrel counties', were often actually
high-tech areas or
'lifestyle communities' (as one might infer from just how few counties he did flip--Kerry flipped almost equally few counties, but they were more clustered [as evidenced by the smaller number of states they were in], and one of the regions they were particularly clustered in was the Upper Midwest, with Minnesota being a rare state whose acreage got noticeably bluer in 2004). Leelanau, MI and Door, WI are
'lifestyle communities'. And, although this begs the question of why Trump flipped the longstandingly Democratic county in 2016 to begin with,
'Madison's tech-driven economy has spilled over county lines, contributing to Biden victories in...Sauk [County].' That, at a stroke, explains both counties Biden flipped in Wisconsin and the only county casting fewer than 100,000 votes that Biden flipped in Michigan.
Mono, on the other hand, showed an impressive ability for Kerry to break
through in a then-reliably red county that, even though it had been
trending slowly blue, was still pretty solidly red as of the previous
election. (Kerry was the first Democrat to carry Mono since 1940; it has gone on to vote Democratic in every subsequent election, giving Biden a > 1,000-vote plurality in 2020.) After
years of supposed growing disenchantment with traditional Republican
voters with Trump, Biden failed to win over any traditionally Republican
but blue-trending county that had stuck with Trump in 2016 as staunchly
as Mono had stuck with Bush in 2000 (at least, going by vote share). (An example of a historically red
but blue-trending county that is a relatively small
'lifestyle community'
[even if substantially larger than Mono], and which voted for Trump at about the same vote share whereat Mono had voted for Bush at in 2000 [52.7%], was Omeena, MI. Yes, it got
'significantly more "blue" in 2020' [or perhaps 'less red' would be more correct], but Trump hung onto it and with an absolute majority at that.)
Of course, there is a large caveat to this, which is that Biden flipped three major counties that had voted for Trump in 2016 by larger margins than Travis, TX had voted for Bush by in 2000. All three (unlike Travis) are historically Republican (by which I mean, in the context of 2020, that they voted for Romney, and that their last two losing votes were both for Republicans).
Here, the comparison with Kerry favours Biden, although comparisons with other recent nominees does not. Romney flipped four major counties that had voted for Obama in '08 by greater margins than any major county Biden flipped had voted for Trump in '16 by, except Williamson (and Trump's '16 margin in Williamson was not much greater than Obama's '08 margin in Chester). And of course, both Romney's margin in Cobb, and Obama's 2012 margin in Erie, PA, were greater than Trump's '16 margin in Williamson. (And, as noted above, Romney's margin in Gwinnett was greater than Trump's '16 margin in any but one of the major counties Biden flipped, and his margins in Fort Bend and Orange, greater than Trump's '16 margin in any but two.)
The largest turnover counties also favoured Biden significantly more disproportionately than they did Kerry in 2004. Biden flipped 26 major counties (vs. Trump's two), whereas Kerry had flipped six (vs. Bush's eight, although, as we saw, the very largest turnover counties disproportionately favoured Kerry). Biden flipped three of the four mega counties Trump had carried in 2016 (whereas Trump flipped none of those Hillary Clinton had carried, even if one counts Salt Lake, which attained mega county status in 2020). In 2004, Bush and Kerry each flipped one mega county. Biden flipped the 17 largest turnover counties overall; Kerry flipped the one largest turnover county in '04.
Biden also flipped the single biggest county Trump had carried in 2016. For most of the 20th century, this might have seemed, at least superficially, both necessary and sufficient for taking the presidency back for one's party, as, from 1916 through 1984, whenever the out-party flipped the largest county the in-party had carried in the prior election, the out-party took power; and whenever the out-party did not, it did not. This was probably understood as a function of the significant number of times that a change in party happened either in or following a landslide, as well as
LA County's former status as a swing county. Still, before 2020, the last time that any nominee had flipped the largest county carried by the opposition party in the previous election was 1992. Kerry had flipped 'only' the ninth-largest county carried by Bush in 2000. Meanwhile, there's no telling how far down the list of Hillary Clinton's largest counties, Trump's largest turnover county had been. Unlike Salt Lake in 2008, Lorain was not close to being a mega county in 2016; it cast less than 1/3 as much vote as the smallest mega county that year. (And, granted, Trump was not trying to take the presidency
back for his party in 2020, but Bush--in a narrow re-election following a narrow initial election--flipped the 26th-largest county Gore had carried in 2000.)
However, Biden's own winning margins in the major counties he flipped was curiously underwhelming, and here, he actually undershone Kerry (kind of)--he certainly dramatically undershone Obama in '08 (the last time a Democrat successfully took the presidency back for his party).
Here, it's worth setting side-by-side Trump's '16 and Biden's '20 margins in these counties. It's also worth separating the 'traditionally Republican' counties from the rest. (Again, in this context, I am considering a county 'traditionally Republican' if it voted for Romney, and if its last two losing votes were for Republicans. This might seem like an unexacting set of criteria, but any county--at least, any county that Biden might be flipping--that fulfils these criteria, will have voted for George W. Bush twice; for him to carry them in
'such' 'closely contested election[s]', Bush 'had' [or nearly had, in '04]
'to run better in [them] than across the country'.) The large, traditionally Republican counties that Biden flipped would be the candidates to be 'Biden beachheads'.
Major, traditionally Republican Trump '16-Biden '20 counties
Maricopa, AZ
2016: 47.67%-44.83% (Trump '16 margin: 2.84%)
2020: 50.13%-47.96% (Biden '20 margin: 2.17%)
Tarrant, TX
2016: 51.74%-43.14% (Trump '16 margin: 8.60%)
2020: 49.31%-49.09% (Biden '20 margin: 0.22%)
Duval, FL
2016: 48.48%-47.12% (Trump '16 margin: 1.36%)
2020: 51.11%-47.30% (Biden '20 margin: 3.81%)
Kent, MI
2016: 47.7%-44.6% (Trump '16 margin: 3.1%)
2020: 52.0%-45.9% (Biden '20 margin: 6.13%)
Johnson, KS
2016: 46.7%-44.1% (Trump '16 margin: 2.6%)
2020: 52.7%-44.5% (Biden '20 margin: 8.2%)
Morris, NJ
2016: 49.7%-45.5% (Trump '16 margin: 4.2%)
2020: 51.1%-46.9% (Biden '20 margin: 4.2%)
Williamson, TX
2016: 50.9%-41.3% (Trump '16 margin: 9.6%)
2020: 49.6%-48.2% (Biden '20 margin: 1.4%)
Seminole, FL
2016: 48.1%-46.6% (Trump '16 margin: 1.5%)
2020: 50.7%-47.9% (Biden '20 margin: 2.8%)
Virginia Beach, VA
2016: 48.4%-44.8% (Trump '16 margin: 3.6%)
2020: 51.6%-46.2% (Biden '20 margin: 5.4%)
Chesterfield, VA
2016: 48.2%-46.0% (Trump '16 margin: 2.2%)
2020: 52.4%-45.8% (Biden '20 margin: 6.6%)
Rockingham, NH
2016: 49.9%-44.1% (Trump '16 margin: 5.8%)
2020: 50.2%-48.1% (Biden '20 margin: 2.1%)
Marion, OR
2016: 46.3%-42.2% (Trump '16 margin: 4.1%)
2020: 48.9%-47.7% (Biden '20 margin: 1.2%)
Frederick, MD
2016: 47.4%-45.0% (Trump '16 margin: 2.4%)
2020: 53.3%-43.7% (Biden '20 margin: 9.6%)
New Hanover, NC
2016: 49.5%-45.6% (Trump '16 margin: 3.9%)
2020: 50.2%-48.0% (Biden '20 margin: 2.2%)
Deschutes, OR
2016: 46.4%-43.1% (Trump '16 margin: 3.3%)
2020: 52.7%-44.4% (Biden '20 margin: 8.3%)
Hays, TX
2016: 46.9%-46.0% (Trump '16 margin: 0.9%)
2020: 54.4%-43.6% (Biden '20 margin: 10.8%)
Butte, CA
2016: 46.5%-42.9% (Trump '16 margin: 3.6%)
2020: 49.4%-47.7% (Biden '20 margin: 1.7%)
Other major Trump '16-Biden '20 counties
Pinellas, FL
2016: 48.08%-46.98% (Trump '16 margin: 1.10%)
2020: 49.44%-49.22% (Biden '20 margin: 0.22%)
Montgomery, OH
2016: 47.68%-46.95% (Trump '16 margin: 0.73%)
2020: 50.18%-47.94% (Biden '20 margin: 2.24%)
Hillsborough, NH
2016: 46.7%-46.5% (Trump '16 margin: 0.2%)
2020: 52.8%-45.1% (Biden '20 margin: 7.7%)
Gloucester, NJ
2016: 47.8%-47.3% (Trump '16 margin: 0.5%)
2020: 49.99%-48.04% (Biden '20 margin: 1.95%)
Northampton, PA
2016: 49.6%-45.8% (Trump '16 margin: 3.8%)
2020: 49.6%-48.8% (Biden '20 margin: 0.8%)
Erie, PA
2016: 48.0%-46.4% (Trump '16 margin: 1.6%)
2020: 49.8%-48.8% (Biden '20 margin: 1.0%)
Saratoga, NY
2016: 47.8%-44.6% (Trump '16 margin: 3.2%)
2020: 51.6%-46.2% (Biden '20 margin: 5.4%)
Chesapeake City, VA
2016: 48.0%-46.7% (Trump '16 margin: 1.3%)
2020: 52.2%-45.8% (Biden '20 margin: 6.4%)
Saginaw, MI
2016: 48.0%-46.8% (Trump '16 margin: 1.2%)
2020: 49.4%-49.1% (Biden '20 margin: 0.3%)
As noted
here, Obama won a series of electorally critical, then-traditionally Republican counties in 2008, that came to be seen as the counties Romney had to win back in 2012. They were distinguished, however, not only by their size and their previous voting history, but by the margins whereby they had voted for Obama in 2008. Leaving aside Hillsborough, NH (an outlier in this respect), they voted for Obama by margins of between 7.0% (in Hamilton, OH) and 15.9% (in Prince William, VA).
It's also worth noting that Bush himself had won one of these counties--Loudoun, VA--by 12.1% in 2004, a greater margin than Trump had won any of the major counties Biden flipped in 2020 by in 2016. Bush won a further four of these counties in 2004 by greater margins than Trump won any but two of Biden's major turnover counties by in 2016: Forsyth, NC (8.59%), Henrico, VA (8.2%), Prince William, VA (6.4%), and Hillsborough, FL (6.23%).
Biden won only five of his major turnover counties by more than 7.0% (these are the bolded margins)--four traditionally Republican counties, plus Hillsborough, NH. This, despite that he had a less steep hill to climb in any of his major turnover counties than Obama had in Loudoun in '08, and a less steep hill to climb in any but two of his major turnover counties than Obama had in Forsyth (NC), Henrico, Prince William, or Hillsborough (FL) as well.
In fact, not only Obama in 2008, but arguably Hillary Clinton in 2016, did better in her 'beachhead counties', than Biden in 2020. She carried Chester by 9.4%; admittedly, Chester was a narrow Romney win, but this was a greater margin than Biden won any but two of his major turnover counties by (Hays and Frederick). She carried Orange, CA and Fort Bend, TX (which Romney had carried by between 6% and 7%) by 8.59% and 6.6%, respectively. And she carried Gwinnett, which had voted for Romney by more than any but one of Biden's major turnover counties had voted for Trump in 2016 by, by 5.79%. (She only narrowly carried Cobb, but Cobb had voted for Romney by a bit more than Loudoun had voted for Bush by in 2004.)
I mentioned that Biden undershone Kerry in this regard; Kerry's largest victory margin in any of his major turnover counties was 14.01%, in Travis, TX--easily a larger margin than Biden won any of his major turnover counties by. (To achieve this margin, incidentally, Kerry had to do more than simply appear to absorb the third-party vote, as Bush got 46.88% in the county in 2000.)
Travis was not a 'Kerry beachhead', however, as it had been traditionally Democratic as of 2000 (indeed, it was a 'Bush breakthrough county'). Kerry's second-largest victory margin in any of his major turnover counties (and his largest in any then-traditionally Republican turnover county) was 7.31%, in Fairfax, VA--still greater than Biden's winning margin in any but five of his major turnover counties (or in any but four of his major, traditionally Republican turnover counties). That does seem like a lot of counties in which Biden did better than Kerry did in Fairfax, but Biden also flipped many more major counties than Kerry did. (Kerry's third-largest victory margin in a major turnover county was 3.63%, in Mecklenburg, NC, which had voted to re-elect George H. W. Bush in 1992.)
It seems likely that, had Kerry narrowly won the 2004 election without flipping any counties other than those he actually did flip, there wouldn't have been a set of large, formerly Republican 'beachhead counties' that would have been seen as 'stealing the show' in the Kerry era, the way that there were in both the
Clinton and
Obama eras. Across the board, perhaps, Biden did do better than Kerry, both in terms of flipping large counties and in terms of the margins whereby he won them (when taken in the aggregate), but the same seems true of the 'Biden era'. As mentioned above (and in other posts), the most celebrated counties in Biden's win generally seem to have been
Hillary Clinton beachheads,
especially Cobb. Insofar as there were large Biden turnover counties in the Frost Belt states he flipped that shared a family resemblance, they were more likely to be
traditionally Democratic counties that Trump had broken through in and that Biden--narrowly--won back.
This wasn't for
lack of possible Biden beachhead counties, including in most of the swing states. It was clear from commentary on election night (which, naturally, focussed essentially exclusively on swing states) that Biden underperformed
expectations of how much he could move large, traditionally red
counties. On election night, Chuck Todd attributed Biden's failure to flip Ohio partly to the fact that Delaware County hadn't gone
'all the way blue'.
Delaware County voted for Trump in 2016 by 15.8%, a margin larger than
any Biden turnover county (let alone any casting over 100,000 votes)
ended up having voted for Trump in 2016 by (but a smaller margin than
the 'Romney must-win' county of Prince William, VA had voted for Obama
in '08 by--and also a smaller margin than Anoka, Salt Lake, or Erie [PA]
had voted for Clinton in '96, Bush in '04, or Obama in '12 by,
respectively). Todd was speaking of Delaware County as though it could
have become, in the Biden era, what the various 'Romney must-win
counties' had come to be seen as in the Obama years--a large county that
the Republican could not win the state without.
It's not clear what kind of margin Todd thought Biden might have been able to
win Delaware County by, but if he thought Biden could flip a county that had voted for Trump by 15.8% at all, he must have imagined that Biden could not only
flip counties that had been much narrower than Delaware in 2016 (such as New Hanover, NC or Seminole and Duval, FL), but that he could also win them
by substantial margins. (These three counties all did get some attention on election night [
e.g., New Hanover
here, Duval and Seminole
here]--but they appear to have faded from consciousness, as, even though Biden did carry all of them, he lost the states they were in [and, perhaps not coincidentally, he carried them all by small margins].)
There is evidence to this effect when
Chuck Todd talks about Hays County, TX, a traditionally Republican county that had similarly very narrowly voted for Trump in 2016. With Biden leading in the real-time results by 12% (larger than his final 10.8% margin, which itself was his greatest victory margin in any major turnover county), Todd is alarmist rather than sanguine in his interpretation, noting that O'Rourke had won the county by 16% in the 2018 Senate race (and had lost). In addition, when
Chuck Todd checks on Maricopa County at a point when Biden's real-time lead is 9.6% (which Todd rounds up to '10'), he says, 'This is a big lead; look, they needed something in the Sun Belt...' (leaving the thought unfinished, as he often does, but with the implication that, with North Carolina, Texas, and Florida looking doubtful for Biden, and Georgia unclear, the Biden camp could at least be hopeful about Arizona). Biden ended up winning Maricopa--which Hillary Clinton barely lost four years prior--by the much narrower margin of 2.2%. (It was arguably the only 'Biden beachhead' county that was critical to the election, but probably only because of the immense proportion of the state's vote it casts.)
(Bush did, incidentally, win some of the counties that would go on to be 'Romney must-win counties' relatively narrowly in 2004, much as Trump did Maricopa, Seminole, Duval, and New Hanover in 2016--most markedly, he won Wake, NC by just 2.12% in 2004, and won Arapahoe by just 3.9% and Washoe by just 4.2%. But again, Obama won all of them in 2008 by at least 7%.)
It is worth noting that, in 2016, after Romney had just carried Delaware, OH by 23.15%,
Robert Wheel was already seeing this county as flippable. He also hinted (in his entry for Walworth, WI) that Waukesha was just at the edge of what
might be
flippable. In 2016, the Republican vote share in Waukesha fell from
over 2/3 to just under 3/5 [59.995%], although the margin remained
relatively large [26.66%]--no nominee has flipped a major county that
had voted for the opposition party in the previous election by over 25%
in over a quarter century (Ventura, CA's margin for HW Bush in '88 was
24.45%).*
Wheel also noted (again, just before the 2016 election) Ada County, Idaho as another longstanding Republican county Hillary Clinton might flip. Ada, the largest county in Idaho (and one that casts over 100,000 votes), was eminently in the range of the flippable after 2012, having voted for Romney by 11.3%. It was still in that range after 2016, having voted for Trump by 9.2%--both of these smaller margins than Bush had won Loudoun in 2004 by. And yet, although Biden closed the gap in Ada further to 3.9%, the county stayed red for at least one more election (indeed, with Trump recovering to win a majority in it, after a plurality win in 2016).
Ada didn't get the kind of attention as other possible Biden beachheads, obviously being in a solidly red state. (There were, similarly, a number of counties with similar profiles as the 'Romney must-win counties' in various other states, but--other than Nate Cohn's mention of Somerset, NJ--they generally weren't noticed because they were either in safe red or safe blue states. Limiting ourselves to counties Obama won in '08 by between 6% and 16%, they included such noteworthy counties as San Diego, CA, Dallas, TX, and DuPage, IL [with Obama's margins in Harris, TX and Riverside, CA being much narrower--although, ironically, those two have gone on to vote Democratic in every subsequent election--and his margin in Lake, IL being much larger, approaching 20%].)
But, typically--and, as we have just seen, this was true of Obama in '08--a candidate's flipping a certain kind of county in swing states will be accompanied by his or her flipping that kind of county in non-swing states as well. If it is dramatic enough, it can begin to make what had been seen as a safe state seem endangered. After the 2020 election,
Kevin Richert wrote that the narrowing of Ada County didn't
make Ada County into a Maricopa County, Ariz., or a Clark County, Nev. —
diverse and vote-heavy urban centers that have transformed their
respective states into presidential battlegrounds. But these numbers
explain why many rural Idaho Republicans view Ada County’s politics with
a mixture of concern and disdain.
Overall, however, Richert concludes that 2020 was a 'more-of-the-same election' in Idaho, with the Republican 'secur[ing] Idaho’s four electoral votes by winning by more than 30 percentage points' '[f]or the fifth time in the past six elections'. It might not have been so easy to put it down as a 'more-of-the-same' election in Idaho if Biden had actually managed to flip Ada.
---
Of course, when one is running for president, one is not running against oneself, but against someone else. In certain respects, Biden may have underperformed Kerry, and in certain respects, he may have outshone Kerry. But, accepting the premiss that they did (roughly) similarly well in flipping counties, their opponents did very differently to one another. Bush flipped about 150 counties--which allowed him to stay just ahead of Kerry. Trump, on the other hand, flipped an order of magnitude fewer counties than this.
Much about the Trump turnover counties does feel like they are 'anti-trend'. They are nameably finite. They tend to be geographically concentrated in a certain corner of the
South (both Burke and Scotland
border South Carolina) and in the Rio Grande Valley. As has often been the case with
anti-trend counties, they were characterised in terms of their
geographical idiosyncrasy and size, for example by
'TheValuesVoter': '1
3 of the 15 had fewer than 20,000 total votes. More than half of them are small counties in south Texas.'
(Alamosa, in Colorado--the only one not in the Rio Grande Valley, in or bordering South Carolina, or in Ohio--is in the San Luis Valley, around the
headwaters of the Rio Grande.)
Now, unlike in 1988, 1992, or 2008 (the three most recent elections in
which the Republican turnover counties could be characterised as
'anti-trend'), Trump flipped two major counties in 2020: Lorain and
Mahoning, Ohio (which cast 157,768 and 119,190 votes, respectively). (And, to correct for the possible ahistoricity of the threshold of 100,000 votes, Lorain cast 0.09961% of the national vote in 2020, and Mahoning, 0.07525%, whereas Washington, PA cast 0.07467% in 2008; Woodbury, 0.04105% in 1992; and Bibb, 0.04847% in 1988.) Out
of 78 total turnover counties in 2020, Lorain was the 18th largest,
and Mahoning, the 25th largest. Lorain was in the top 25%, and Mahoning,
in the top third, by size. (That said, the larger of the two Hillary
margins in those two counties was just 3.25%, in Mahoning.)
These two also happen to be in the same state (also the only state containing
major counties in which the Democratic vote share declined in 2020);
and, as 'TheValuesVoter' points out, there is a steep dropoff in size
after Mahoning (unlike with the Kerry-McCain counties, despite that none
of them quite cast 100,000 votes). The various Trump turnover counties
don't seem particularly thematically unified, unlike the Kerry-McCain
counties.
However, as with the Romney-Hillary counties, I didn't mark the Hillary-Trump counties as 'anti-trend'. For one thing, 2020 was a close (or close-ish, if you like) election, following another close election. The Democratic swing in 2020 was slight: 2.36%.
One thing that seems to rule the Hillary-Trump counties out from being 'anti-trend' forthwith is that Trump actually flipped a slightly larger proportion of the blue counties, than his opponent did of the red counties (which underscores just how few counties Democrats now carry, and how this must change our perception of the numbers of counties Republicans flip).
For another thing, the largest Hillary '16 margin in any county Trump flipped was a massive 32.9% (/1,034-vote) margin in Zapata, TX. This is the largest county margin any nominee has managed to roll back in any election since 2000. In addition, Hillary Clinton's 65.7% vote share in Zapata overtook Kerry's 63.4% in Knott to be the highest vote share that a turnover county had given its winner in the previous election since 1992 (when Clinton flipped Miller, GA, which had voted 68.00% for HW Bush in '88, and Pickens, GA, which had voted 67.52% for HW Bush in '88 [and had given him a > 1,000-vote plurality]--amongst others).
Now, it is true that, after the Hillary '16 margin in Zapata, there is a steep dropoff in the sizes of the Hillary '16 margins in the other Trump turnover counties. Trump flipped only four counties that voted for Hillary Clinton by at least 5% and by at least 1,000 votes:
Zapata, TX (
65.7%-
32.8%) (HRC '16 margin:
32.9%)
Jim Wells, TX (54.1%-43.8%) (HRC '16 margin: 10.3%)
Val Verde, TX (51.1%-43.3%) (HRC '16 margin: 7.8%)
Scotland, NC (52.6%-44.9%) (HRC '16 margin: 7.7%)
He also flipped an additional three counties that voted for Hillary Clinton by over 5%, but not by at least 1,000 votes:
Frio, TX (55.6%-42.2%) (HRC '16 margin: 13.2%/588 votes)
La Salle, TX (54.8%-42.4%) (HRC '16 margin: 12.4%/257 votes)
Reeves, TX (52.1%-44.5%) (HRC '16 margin: 7.6%/242 votes)
Oh, I forgot one:
Kenedy, TX (53.23%-45.16%) (HRC '16 margin: 8.07%/15 votes)
OK, so I'm not going to be counting Kenedy at all in the subsequent discussion.
In any case, adding these counties in shows that, while impressive, Zapata was an outlier--unlike DeKalb, TN for Bush in 2004. As noted above, in 2004, Bush flipped six counties outside Tennessee
(and at least five in Tennessee) that had voted for Gore in 2000 by over
1,000 votes and by more than Val Verde, TX voted for Hillary Clinton
by. (And it is not as though Trump's
largest-prior-election-Dem-margin-turnover-counties weren't
idiosyncratically geographically concentrated.) We can further say that
Bush flipped four counties that had voted for Gore by over 1,000 votes
and by more than Frio County, TX had voted for Hillary Clinton by. In
other words, even though the Hillary '16 margin in Zapata was more
impressive than the Gore '00 margin in DeKalb, TN (which nevertheless
was still impressive), the Gore margin in DeKalb was not an outlier
amongst Bush '04 turnover counties, whereas--even ignoring the size of
these counties' raw vote pluralities for Hillary Clinton--the Hillary
margin in Zapata was an outlier amongst Trump '20 turnover counties.
However, if we compare these to Biden's turnover counties, they become more impressive. Obviously, Hillary's '16 margin in Zapata was greater than Trump's '16 margin in Inyo. What is more surprising is that Hillary's '16 margin in Jim Wells--which cast more vote than Inyo in 2020--was greater than Trump's '16 margin in any Biden turnover county save Inyo. (As were--if we include them--her margins in Frio and La Salle.)
Furthermore, Hillary Clinton's vote shares, not only in Zapata, but also in Jim Wells and Scotland (and, if we count them, Frio and La Salle) were greater than Trump's '16 vote share in a single county Biden flipped. Her vote share in Reeves (with the same caveat) was higher than Trump's '16 vote share in any county Biden flipped save Talbot.
---
There is another impressive thing about Trump's turnover counties, although this is not really a feature that argues against their being anti-trend counties. Namely, several of them had last voted Republican a very long time ago. One, Zapata, was indisputably a
'streak-breaker'. (Robert Wheel doesn't nail down how long ago a county had to have last voted for a party to be eligible to be a streak-breaker, but the county he names that voted most recently for the party it had since become traditionally opposed to was Swift, MN, which [had] last voted Republican in 1952, 64 years prior. He explicitly rules out counties that had last voted Republican in 1972 [Trumbull and Mahoning]. He doesn't name Dubuque, IA [had last voted Republican in 1956] or Mower, MN [had last voted Republican in 1960], but I'm not sure whether this is because he felt these last Republican votes were too recent, or because he didn't think of them--four counties came extremely close to being 'streak-breakers' in 2016 that he didn't name [Chesterfield, VA and Seminole, FL, both of which had last voted Democratic in 1948; and Carlton and Lake Counties, MN, which had last voted Republican in 1928 and 1932, respectively].)
In any case, Zapata had last voted Republican in 1920, so its status as a 'streak-breaker' would not really be in question. Trump carried several further counties--including Mahoning--that had last voted Republican in 1972.
Now, every major-party nominee (with one somewhat curious exception) from 2000 on has flipped at least one county that last voted for his or her party before 1972 (if a Republican) or 1964 (if a Democrat). (When thinking about truly long streaks of voting for a party on the county level, at least in the context of the last few decades, using 1972 and 1964 seems more interesting than setting an arbitrary span of time. These were the two 'civilisational chaos elections': as Rick Perlstein put it, the two elections in which to do anything but vote for the Democrat [in 1964] or for the Republican [in 1972]
'seemed to court civilizational chaos'. And unlike in most landslides, a significant proportion of the loser's support came from voters who had hitherto been typically aligned with the
other party [but who were presumably attracted to this candidate's brand of 'civilizational chaos']. Conversely, counties that voted for the loser out of sheer partisan loyalty were much, much rarer than in, say, 1936, 1956, or 1984.)
Gore, 2000: Orange, FL [last Dem vote: 1944]
Bush, 2000: Wolfe, KY [last GOP vote: never], Morgan, KY [last GOP vote: never], Greenlee, AZ [last GOP vote: never], Red Lake, MN [last GOP vote: 1920], Cottle, TX [last GOP vote: 1928]
Kerry, 2004: Alpine, CA [last Dem vote: 1936], Mono, CA [last Dem vote: 1940], Albemarle, VA [last Dem vote: 1948], Danville City, VA [last Dem vote: 1948]
Obama, 2008: Carroll, IL [last Dem vote: never], Boone, IL [last Dem vote: 1844], DuPage, IL [last Dem vote: 1852], Kane, IL [last Dem vote: 1852], Carroll, NH [last Dem vote: 1912], Waupaca, WI [last Dem vote: 1936], Gallatin, MT [last Dem vote: 1944], Teton, ID [last Dem vote: 1948], Henrico, VA [last Dem vote: 1948], Jefferson, AL [last Dem vote: 1952], Oktibbeha, MS [last Dem vote: 1956]
McCain, 2008: Floyd, KY [last GOP vote: never], Knott, KY [last GOP vote: never], Stewart, TN [last GOP vote: never], Breathitt, KY [last GOP vote: 1908], Perry, TN [last GOP vote: 1920], Logan, WV [last GOP vote: 1928]
Obama, 2012: Warren, MS [last Dem vote: 1960]
Romney, 2012: Jackson, TN [last GOP vote: 1920], Houston, TN [last GOP vote: 1928]
Hillary Clinton, 2016: Orange, CA [last Dem vote: 1936]
(Trump, in 2016, hit every mark set for him by Robert Wheel save Deer Lodge, MT, and additionally carried Dubuque, IA and Mower, MN.)
(Incidentally, even though, for elections before 2000, flipping a county that had last voted for one's party before 1964 or 1972 would have, at least potentially, involved breaking a significantly shorter streak in temporal terms than in more recent times, it didn't happen with this kind of regularity. From 1980 through 1996, the only losing nominee to flip a county that had last voted for his party before the relevant civilisational chaos election was Carter in 1980 [although Dukakis came tantalisingly close in Whitman, WA]. Neither HW Bush in '92, nor Dole in '96, even flipped any county that had last voted Republican in 1972. And even one winning nominee in this period failed to flip any counties that had last voted for his party before the relevant civilisational chaos election: the longest-ago that any county HW Bush flipped in 1988 had last voted Republican was 1972.)
In this context, Trump breaking a long Democratic winning streak in a county in 2020 isn't surprising.
In their defeats for another term, William Howard Taft, Gerald Ford, and George H. W. Bush all flipped counties, but every county they flipped had either voted Republican eight years prior, or was participating in only its second election. (The latter was the case of the Bryan '08-Taft '12 counties in Oklahoma, which had become a state in 1907.) Hoover, of course, flipped no counties at all. In 1892--the last election before 1912 in which a president lost re-election--it would have been impossible for a county to have built up a streak of voting for one party as long as Zapata's streak of voting Democratic was by 2020.
In 1980, Carter flipped a handful of counties in Mississippi that had last voted Democratic in 1956, and one (Yazoo) that had last voted Democratic in 1952--28 years prior. Until 2020, this would have been the longest streak of voting for (or, in these cases, more against) a party on the county level that an incumbent president unsuccessfully seeking a second term would have broken (in the 20th or 21st centuries--possibly ever). However, in time terms, at least, even the 48 years ago that Mahoning (and others) had voted Republican as of 2020 dwarfs this. (Of course, these Mississippi counties had voted for Goldwater, whereas Mahoning etc. had not voted for McGovern. But 48 years is beginning to be a lot longer than 28 years; added to which, Mahoning etc. did vote for Mondale.) And in any case, Trump's flipping Zapata easily outdoes Carter's flipping Yazoo, and without the need for any qualifiers.
In fact, I have not checked, but even simply the
48 years since Mahoning etc. last voted Republican might be the longest streaks of voting for a party on the county level that an incumbent president losing re-election has broken. (There were four elections in the 19th century--since elections
'began to resemble today's system'--in which an incumbent president lost election to another term in a general election: 1828, 1840, 1888, and 1892. Of these, 1888 and 1892 are the only elections in which it would have been possible for the loser to flip a county that had last voted for his party even so much as 48 years prior [as 1828 was
when elections began to resemble today's system]. And in the case of Harrison--the loser in 1892--one would have had to accept the Whig Party as a stand-in for the Republican Party, as the Republican Party didn't exist 48 years prior to 1892. In fact, even Cleveland in 1888 couldn't break a long Republican
winning streak in a county--just a long streak of a county
not voting
Democratic.)
And in fact, even in the context of the past six elections, Trump flipping Zapata, at least, still sets him somewhat apart. This is, easily, the longest streak of a county voting for a party that an incumbent president has broken--whether he won or lost re-election--since 1996. (This means that, in defeat, Trump broke a longer streak of a party winning a county, than either Bush or Obama did in victory.) The longest-ago that any McCain '08-Obama '12 county had last voted Democratic was 1960--52 years prior. The next-longest-ago was not even in 1964--it was in 1976 (Nash, NC).
And, curiously, Bush's '04 candidacy is the one exception this century: it is the one major-party candidacy that failed to flip any counties that had last voted for its party before the relevant civilisational chaos election. Bush did flip
40 counties that had last voted Republican in 1972, 32 years prior. Given the conjunction of the number of these counties, Bush's large margins in many of them (he won twelve by over 10%), and the overall closeness of the national election, he almost certainly would have carried some of these if he had narrowly lost re-election, thereby breaking Carter's modern record for the longest streak of a county voting for a party broken by an incumbent president losing re-election (although not by much). Still, it is curious that Bush in '04 should have been the only nominee this century not to flip a county that had last voted for his party before the relevant civilisational chaos election.
Of course, Biden had his own streak-breaker counties, a fair number in fact: Johnson and Riley, KS, Maricopa, Seminole, FL, and Chesterfield and Lynchburg City, VA. (Although it may be worth noting that Hillary Clinton had already softened most of these significantly in 2016, much as Romney had Elliott, KY in 2012. Only Lynchburg City gave Trump a majority in 2016. In contrast, Mono County and Danville City had both given Bush a majority in 2000--as we have seen, Mono gave Bush a higher vote share than
any Biden turnover county gave Trump in 2016--and Orange, CA had given Romney a majority in 2012. By 2020, the idea of a Democrat carrying any of these--except, again, perhaps Lynchburg City--would not have been shocking, the way that Hillary Clinton's carrying Orange, Cobb, and Fort Bend was
to David Jarman. [And Lynchburg City would probably be too small to be sufficiently noticeable to be all that shocking, although its being home to Liberty University might have outweighed its relative smallness.])
---
Ultimately, neither Biden nor Trump was all that impressive in terms of their county flips in 2020. There are many counties that I would have expected either to be flipping, if he were winning or even coming close, that he did not. In broad terms, Biden's county flips resembled Kerry's in '04 (which would also have been historically unusual for a nominee successfully taking the presidency back for his party). But what Kerry did was almost enough. And Trump's county flips, in all but a couple isolated ways, were far less impressive than Bush's in '04 (which, again, enabled Bush just to keep slightly ahead of Kerry). While there are other factors that would make Trump's turnover counties unusual for 'anti-trend counties', the main factor that militates against seeing them as such is that they represented a larger proportion of the blue counties than Biden's turnover counties did of the red counties. (Indeed, the raw number of Biden's turnover counties sits just above the normal range of how numerous the anti-trend counties in any given election tend to be.)
The largest-prior-election-Dem-margin Trump '20 turnover county was traditionally Democratic (as in every election from 1996 through 2016 save 2012), but not--really--in the South. (Not, at any rate, in a way such that it could be said to reflect the movement of 'Borderer' sentiment.) Nor was it in the same broad region--what used to collectively be called the Frost Belt--in which all the Obama '12-Trump '16 counties that had voted for Obama by at least 20% in 2012 were located, indicating perhaps that Trump's appeal in the traditionally Democratic parts of this region had stalled out in 2020 (unlike Bush in 'Greater Appalachia' in 2004). (And there were--and are--certainly still red-trending 'old coalition' Democratic counties in the Frost Belt that were in the range of being flippable in 2020, but which would have made a list of largest-Hillary '16-margin Trump '20 turnover counties [even if they wouldn't quite have outdone Zapata], such as Rock, WI, St Louis County, MN, and Douglas, WI [which handed Wisconsin to Gore in 2000]. There are also counties in this region that had last voted Republican before 1972, but Trump's flipping which would have been about as expected--assuming he was winning or even coming close to winning--as the Democrat's flipping Chesterfield, Seminole, Johnson [KS], and Riley would have been [i.e., Lake and Carlton Counties, MN and Portage County, WI].)
And despite how impressive the flip of Zapata was, it sets in greater relief the near-lack of flips where there had been so many impressive ones in 2016, which, more than anything else, might explain Trump's defeat.
Conversely, Biden's flips--relative even to Kerry's--illustrate a stunning collapse in the ability of the Democracy to persuade 'Cracker Barrel voters'; and they also illustrate a somewhat surprising limitation (even relative to Hillary Clinton, before Trump had even assumed office) in the ability to keep peeling off affluent, highly-educated, traditional Republican voters.
------
In most elections, one nominee carries at least four times as many counties as the other. In some elections, one nominee carries between 0 and 3 times as many counties as the other--anything in that range seems to indicate an election roughly of a similar kind as the last election. (These would include 1936, 1956, 1996, 2004, and, of course, 1948--although in 1936, 1956, and 1996, neither nominee even so much as doubled the number of counties being flipped by the other nominee.) I haven't found any elections yet in which one nominee carried between 3 and 4 times as many counties as the other.
The proportion of the opposition party's counties that a nominee needs to flip to retake the presidency for his party seems generally to have been at least 20% or so (with 1960 having been a slight anomaly), although this doesn't guarantee it.
'Anti-trend counties' (counties flipping to the nominee flipping fewer counties when one nominee is flipping more than four times as many counties as the other nominee) have a tendency to be--in addition to being not very numerous--geographically narrow, indicating a very regionalised sub-national trend that goes against the national trend. Often they are concentrated in the South.
Nominees taking the presidency back for their party, conversely, have generally flipped counties in at least 40 states, at least until recently. They are still generally flipping counties in at least 30 states. (Kerry, almost taking the presidency back for his party, did so in just shy of half the states--24.) Anti-trend counties are usually located in under 1/3 of the states, and generally number fewer than roughly 50 (with 1960 and 1964 having been anomalies in this respect--although the 1964 anti-trend counties compensated for their relative numerosity by being highly geographically concentrated). I got the rough figure of '50' from the fact that the next-smallest number of turnover counties, after the 44 Kerry-McCain counties, is the 63 Trump-Biden counties (followed by the 64 Bush-Kerry counties, in what was a close election).
The number of turnover counties in general has declined sharply since 2000. In elections with only a slight national trend, it had not been uncommon for only a few hundred counties to switch sides altogether (e.g., 1936, 1956, 1996). In 2004, 220 counties flipped, whereas over 300 counties had flipped in 1936, 1956, and 1996 (although not in 1944). Obama in 2008 and then Trump in 2016 (and then Biden in 2020) set likely records for the fewest counties flipped by a successful out-party nominee in over a century, although I didn't gather these figures for 1968, 1952, and 1932. In 2020, fewer than 100 counties flipped overall for likely the first time in over a century. I think this is most likely due to the historically small number of counties being carried by Democrats these days. In 2008, Obama reached beyond the base of
'dynamic' counties that he was building to some degree (although not a very large degree given that the proportion of the opposition party's counties he flipped was already historically low for a successful out-party nominee), but in 2020, Biden appears to have done this to essentially no significant degree whatsoever. As for Republicans, the fewness of counties carried by Democrats simply means there aren't all that many counties available for them
to flip.
---
Smallest numbers of counties flipping from one major party to the other in elections from 1912 on:
1. 1932: 0 Smith '28-Hoover '32 counties
2. 1920: 1 (2?) Hughes '16-Cox '20 counties
2. 1952: 1 Dewey '48-Stevenson '52 county
4. 2000: 2 Dole '96-Gore '00 counties
5. 1976: 3 McGovern '72-Ford '76 counties
6. 1912: 6 Bryan '08-Taft '12 counties
7. 1988: 7 Mondale '84-Bush Sr '88 counties
8. 2012: 12 McCain '08-Obama '12 counties
9. 1980: 13 Ford '76-Carter '80 counties
10. 1940: 14 Landon '36-FDR '40 counties
11. 2020: 15 HRC '16-Trump '20 counties
12. 1972: 16 Nixon '68-McGovern '72 counties
13. 2016: 17 Romney '12-HRC '16 counties
14. 1944: 23 Willkie '40-FDR '44 counties
14. 1992: 23 (24?) Dukakis '88-Bush Sr '92 counties
16. 1924: 27 Cox '20-Coolidge '24 counties
17. 1968: 35 Goldwater '64-Humphrey '68 counties
18. 1984: 36 Reagan '80-Mondale '84 counties
19. 2008: 44 Kerry '04-McCain '08 counties
20. 2020: 63 Trump '16-Biden '20 counties (Biden wins presidency back for his party)
21. 2004: 64 Bush '00-Kerry '04 counties
22. 1916: 69 Taft '12-Wilson '16 counties
23. 1928: 82 Coolidge '24-Smith '28 counties
24. 1960: 97 Stevenson '56-Nixon '60 counties
* Since 1976, to be more precise. These are the largest prior-election opposition-party margins in turnover counties casting at least 100,000 votes in the six elections from 1976 through 1996:
1996
Clinton: Ocean, NJ (44.4%-34.9%) (HW Bush '92 margin: 9.5%)
Dole: San Luis Obispo, CA (38.4%-34.8%) (Clinton '92 margin: 3.6%)
1992
Clinton: Ventura, CA (61.64%-37.19%) (HW Bush '88 margin: 24.45%)
1988
Dukakis: Pierce, WA (57.85%-40.75%) (Reagan '84 margin: 17.10%)
1984
Reagan: Davidson, TN (59.1%-37.5%) (Carter '80 margin: 21.6%)
Mondale: Marin, CA (45.8%-36.2%) (Reagan '80 margin: 9.6%)
1980
Reagan: Worcester, MA (60.41%-36.89%) (Carter '76 margin: 23.52%)
Carter: Monroe, NY (55.1%-44.4%) (Ford '76 margin: 10.7%)
1976
Carter: DeKalb, GA (77.4%-22.6%) (Nixon '72 margin: 54.8%)
Ford: Washtenaw, MI (50.9%-45.6%) (McGovern '72 margin: 5.3%)
Carter did have one of the more powerful home-state appeals amongst presidential nominees. He also flipped three major counties in Florida that had voted over 70% for Nixon in 1972:
Broward, FL (72.41%-27.31%) (Nixon '72 margin: 45.1%)
Duval, FL (72.19%-27.50%) (Nixon '72 margin: 44.69%)
Hillsborough, FL (70.13%-29.71%) (Nixon '72 margin: 40.42%)
1976 is also the last time that a nominee flipped a major county that had voted more than 2/3 for the opposition party in the prior election. In his article on streak-breakers, Wheel wrote that 'As for Waukesha itself, it’s almost impossible to imagine a Democrat
winning a county that voted more than two-thirds for Mitt Romney without
third-party help.' (Waukesha cast 243,856 votes in 2012; it is naturally easier to imagine a small county flipping after voting over 2/3 for its winner in the prior election, although even this hasn't happened since 1992.) Kennedy flipped two counties casting over 100,000 votes that had voted over 2/3 for Eisenhower (Oneida and Niagara, NY, which voted 69.8% and 67.4% for Eisenhower in '56, respectively).
Comments
Post a Comment