The updated 'worse than Dukakis' states (and more)

In 2013, Cook Rhodes observed that Obama in 2012 had gotten a lower share of the vote in 19 states than Dukakis had in 1988. 

We can now find the states in which Biden got a lower vote share in 2020 than Dukakis did in 1988 as well. Before we do that, for context, these are the states in which Gore, Kerry, Obama in '08, and Hillary Clinton got a lower vote share than Dukakis. 

Every Democratic nominee in the 21st century got a substantially higher vote share than Dukakis, but Biden in 2020 and Obama in 2012 got particularly comparable vote shares (51.3% and 51.1%, respectively).

These are the states where Biden in 2020 got a lower vote share than Dukakis:

Alabama (39.86%/36.57%)
Arkansas (42.19%/34.78%)
Idaho (36.01%/33.07%)
Iowa (54.71%/44.89%)
Kansas (42.56%/41.56%)
Kentucky (43.88%/36.15%)
Louisiana (44.06%/39.85%)
Minnesota (52.91%/52.40%)
Missouri (47.85%/41.41%)
Montana (46.20%/40.55%)
Nebraska (39.20%/39.17%)
North Dakota (42.97%/31.76%)
Oklahoma (41.28%/32.29%)
South Dakota (41.28%/32.29%)
Tennessee (41.55%/37.45%)
West Virginia (52.20%/29.69%)
Wisconsin (51.41%/49.45%)
Wyoming (38.01%/26.55%)

In 2016--with a national vote share of 48.2%--Hillary Clinton got a lower vote share than Dukakis in every state Obama had gotten a lower vote share than Dukakis in in 2012 (including Texas and Utah), plus Wisconsin, Pennsylvania, Rhode Island, Ohio, Indiana, and Oregon. 

Oregon and Indiana were probably not very meaningful (neither Gore nor Kerry had fallen below Dukakis' vote share in the latter state), but the other states represented something of the best of both worlds: Hillary Clinton remained below Dukakis' level in Sun Belt states such as Texas and Utah, and fell below his level in the Frost Belt states of Wisconsin, Pennsylvania, and Ohio (and Rhode Island, a state where Kyle Kondik observed the potential beginnings of a long-term Republican trend in 2016).

In 2020, however, Texas, Utah, Pennsylvania, Ohio, and Rhode Island all fell off the list of 'worse-than-Dukakis' states. Wisconsin stayed, which should be of some encouragement to Republicans; but Wisconsin was also the only Frost Belt state that Hillary Clinton added in 2016 that Gore and/or Kerry (and, in this case) also had done worse than Dukakis in.

In many of these states, a national Trump win would likely have been correlated with Biden staying below Dukakis' percentage:

Ohio
Dukakis: 44.15%
Biden: 45.24%

Rhode Island
Dukakis: 55.64%
Biden: 59.39%

Pennsylvania
Dukakis: 48.39%
Biden: 50.01%

Even with Trump losing the election, it might have been hoped that Biden would stay below Dukakis' level in Ohio. (It would be hard to see Biden winning Pennsylvania whilst staying below Dukakis' level in that state; and it would be hard to see Biden winning without Pennsylvania, although of course mathematically it was possible.)

With Biden doing better than Dukakis in Pennsylvania, Ohio, and Texas, that left no large states (states with 14 or more electoral votes) on the list of 'worse-than-Dukakis' states. (The largest was 11-EV Tennessee.)

Wisconsin is the only state in which Biden did not do better than Dukakis, but Obama did do better than Dukakis in 2012. Even though Gore and Kerry also both did worse than Dukakis in Wisconsin, this should be encouraging to Republicans, given that Biden did nearly 3% better than Gore nationally, and slightly more than 3% better than Kerry. Also of some encouragement to Republicans is that Minnesota remained a 'worse-than-Dukakis' state, despite Biden's surprisingly convincing win there. 

Kansas and Nebraska came surprisingly close to falling off the list of worse-than-Dukakis states. (Especially Nebraska, with regard to which I should note that I am using Wikipedia's general elections page, which itself cites the FEC's official results.) Of the nine states that have not voted Democratic since 1964, both Gore and Kerry did worse than Dukakis in all nine. Obama in 2012 and Hillary Clinton did worse than Dukakis in eight of the nine--all save Alaska. Alaska fell off the list in 2008 and has never come back, and got even more atypically close in 2020, with Trump winning it by just over 10%. Utah also fell off in 2020, making it now seven of nine. Utah falling off was probably not to be unexpected--Dukakis ran in the era when Utah was consistently among the two or three brightest red states in the country (the reddest, more often than not), and polled a low 32.05%. It was, or should have been, fairly clear after 2016 that, even if it bounced back to being safely red (which it did), Utah would not bounce back to being one of the handful of the very reddest states, and this alone would have been enough to allow the Democrat to beat Dukakis' percentage assuming a two-party contest in the state. (Furthermore, Utah is a growing, Sun Belt state.)

It's hard to tell what's going on in Alaska, because Alaska doesn't report election results by borough, but something is obviously going on there that should be concerning to Republicans. Given the narrowness of the Republican path in the Electoral College, they can't afford for whatever is happening in Alaska to also happen in Kansas and Nebraska. (It's already bad enough as it is for Republicans that Nebraska's allocation of electoral votes by CD looks likely to essentially mean Nebraska's 'safe' status for Republicans is good only for four, not five, electoral votes henceforth. If Kansas did the same, the Democrat would have received one electoral vote from Kansas in 2008, 2016, and 2020.)

Conversely, every state that voted for Bill Clinton twice and has never voted Democratic again is securely on the list once again, with Biden coming closest to Dukakis' performance in Tennessee (where he still fell over 4% short). (In hindsight, it seems it should have been shocking that already in 2000, Gore was underperforming Dukakis in Kentucky.)

---

Why was Cook Rhodes comparing Obama's vote shares in given states with Dukakis' in the first place?

Why compare these particular elections? Basically because of the parallels in their historical timing. Dukakis ran in the sixth and last presidential election in an era of Republican domination that lasted from 1968 through 1988. The GOP won the White House five times in this period.

Meanwhile, Obama was reelected last fall in the sixth election in a more Democratic period that was launched in 1992. Beginning with Bill Clinton’s election that year, the Democrats have won four presidential contests and the popular vote in a fifth.

 

Obviously, one couldn't say that Republicans are in a similar period now. However, there are certain weaknesses the Republican Party showed in 1996 (and 1992) that have not been replicated since, even in 2008. In every election from 2000 through 2020, one would have been able to walk from coast to coast on red counties (and only on red counties), for example--which is now the most elections in a row that that has been true of either party. In five of the past six elections, the Republican nominee has flipped at least one county that had voted over 60% Democratic in the previous election (and given the Democrat a > 1,000-vote plurality), and the exception was not 2008, but rather 2012. In the same period, Democrats have done so only in 2008. Republicans have increased their vote share in Arkansas for seven straight elections, from 1996 through 2020, and improved their vote share in Tennessee for six straight elections, from 1996 through 2016. (National majority parties have sometimes declined continuously in states for many consecutive elections, but the Democratic declines in Arkansas and Tennessee--and West Virginia and Oklahoma--have been unusual, in that they have not been accompanied by carrying these states [except in 1996, when the Democratic vote share in all these states also went up], as discussed here.) During the same time, the most elections in which the Republican vote share has gone down consecutively in a state is four, and only in one state, Maryland, from 2008 through 2020 (and the Democratic vote share did not rise in the state in all those four elections, as both parties' vote shares fell in 2016). 

And in each of the last six elections, the Republican had a 'fornia'--a collection of states in which the party's nominee was winning by at least 20% and with at least 55.5% of the vote. The last time any nominee losing the national popular vote by over 5% (as McCain was) won even one state by over 20% and with at least 55.5% was 1964. The last time any nominee losing the national popular vote by over 5% won even one state outside the South by over 20% and with at least 55.5% was 1944 (when Dewey's fornia consisted of Kansas; McCain's included Wyoming, Utah, Idaho, and Alaska). No nominee losing the national popular vote by over 5% had done as well as, or better than, McCain's 64.78% in Wyoming, in any state outside the South, since 1856, when Frémont got 77.96% in Vermont. In the intervening time, only Bryan in 1900 and Dewey in 1944 had even broken 60% in a non-Southern state despite losing the national popular vote by over 5% (Bryan with 62.25% in Nevada, Dewey with 60.25% in Kansas). McCain did so in three non-Southern states (Wyoming, Utah, and Idaho).

This is a somewhat impressionistic thing to say, because in objective terms, 2008 was not a competitive election. But one could say that 1996 was the last election before a period in which Republicans have consistently been able to keep presidential elections competitive. 

In terms of popular vote margin, 2008 was similar to 1996, 1992 (in the two-party vote), and 1988. Of the four, 2008 took the longest to call--it wasn't called until the West Coast closed at 11 PM in the East. 1992 took the second-longest to call, so it presents a useful comparison point. Shortly after 10 PM in the East in 1992, Bill Clinton was at 254 electoral votes. (Without California, Hawaii, Oregon, and Washington, the most Obama would have been at at any point before 11 PM was 207, as, forthwith after he was projected the winner, he was at 284.) With Clinton at 254, Tom Brokaw said 'it's just that close for Bill Clinton'. At this point, Clinton has been projected the winner in Louisiana, Arkansas, Tennessee, Kentucky, West Virginia, Missouri, Georgia, and even Colorado; at 11 PM in 2008, these are all states that either haven't been called (Colorado, Missouri), or been called for McCain (all the others). The only states that were called for Obama by 11 PM in 2008, that haven't yet been called for Clinton by a little after 10 PM in 1992, are Wisconsin and Ohio. 

Just after 11 PM in the East--at about the same time as Obama is at 284 in 2008--Clinton is at 362. Both Wisconsin and Ohio have now been called for Clinton, meaning that there are now no states that have been called for Obama by 11 PM in 2008, that had not been called for Clinton by 11 PM in 1992. (The first state to be called for Obama in 2008, that hasn't been called for Clinton at this point in 1992, is Virginia, which takes several minutes following 11 PM to call.)

It's also worth noting that, shortly after 11 PM in the East in 1992, George H. W. Bush was projected with 64 electoral votes, whereas, at around the same time in 2008, McCain was projected with 146. McCain has been projected to win the following states that were either uncalled at this same time in 1992, or had been called for Clinton: West Virginia, Kentucky, Tennessee, Georgia, Arkansas, Louisiana (had been called for Clinton), Mississippi, Kansas, Texas, North Dakota, South Dakota, and Wyoming (had not yet been called, would end up voting for HW Bush). States that have been called for HW Bush at this time, that have not been called for McCain, are Virginia and Indiana. Some of these states were surprisingly close in 1992. The future reddest-in-the-country state of Wyoming was just a 5.60% win for HW Bush. Texas was a 3.48% HW Bush win in 1992 (still a narrower margin than Trump's in 2020); South Dakota, a 3.52% win; Kansas, a 5.14% win.

So, even though there is no question that 2008 was a decisive national win, it might be worth comparing later Republican vote shares with Dole's vote shares in 1996. The problem is that 1996 was a genuine multi-party contest, and so Dole's percentage in any given state would be lower than his performance in the two-party vote would suggest. Now, from 2008 on, Republicans have been getting lower vote shares in some states than Dole got in 1996. Looking at these states would be understating the Republican 'problem', for the reason mentioned above, but it may still be worth looking at:

2008
2012
2016
2020

California has been on every map, which surely largely explains Republicans' difficulties in the past four elections in the national popular vote. (Dole got 38.21% in California.) Colorado and Virginia have been on every map save Romney's, in 2012. Interestingly, North Carolina (where Dole got 48.73%) is on none of these maps. 

Rhodes Cook's explanation for comparing 1988 and 2012 also suggests another parallel Republican comparison. Although he doesn't go as far as some have in making the 1992-2012 period out to have been a period of Democratic presidential domination, he does imply that he sees it as such at least to some degree, comparing it to the Republican period of domination from 1968 through 1988.

However, some analysts have pointed to a broader period of Republican domination at the presidential level, from 1952 (or perhaps even before) through 1988. For example, in its analysis of the 1996 election, the CQ Almanac says:

Democrats were in a similar period in 1996. They had won six presidential elections since the end of World War II, but voter approval had been grudging. Only once in the previous 50 years had the Democratic nominee exceeded 50.1 percent of the popular vote. That was Lyndon B. Johnson in 1964.

(Carter got 50.06% in 1976, which likely explains the '50.1%'.)

And Sean Trende discusses the 'underrated Eisenhower coalition':

The importance of the Eisenhower coalition is all the more apparent when one looks closely at the three Democratic majorities that emerged from 1952 to 1988. Far from being throwbacks to the New Deal era, these were either unique “deviating” elections, or antecedents of the Clinton coalition that would emerge in 1992.

...

Interestingly, under this narrative, Ronald Reagan does not represent the birth of a substantial Republican majority, but rather the beginning of its decline.

Trende explicitly dates the beginning and end of the era of Republican dominance at 1952 and 1988. However, when the CQ Almanac refers to the six presidential elections since the Second World War that the Democrats had won, 1948 is the first of those six (followed by 1960, 1964, 1976, 1992, and 1996). In 1948, Truman did not win a majority: he got 49.55%.

If we are looking at the 1952-88 period, the election that would seem to be the obvious parallel to 1988 would be 1948. However, 1948 was, as mentioned above, one of the six 'grudging' Democratic postwar wins the CQ Almanac was referring to in 1996. It was also, unlike 1988, a close election. Although Rhodes Cook doesn't directly reference Dukakis having lost badly, the fact that Dukakis lost badly makes it meaningful to see in which states later, more nationally competitive Democratic nominees have fared worse than him in. (It would mean little--to take it to an extreme--to show which states Kerry did worse than FDR in 1936 in, for example.) Cook does list the basic facts about the 1988 election ('In 1988, Dukakis carried 10 states (plus the District of Columbia), 111 electoral votes [sic], and 45.6% of the total popular vote as the Republican era was winding down'), from which I suppose one might draw one's own conclusions.

In 1944, on the other hand, Dewey lost by 7.5%, a margin of defeat comparable to Dukakis' in 1988. (And Franklin Roosevelt got 53.39%, a vote share that has not been matched or surpassed by any subsequent Democrat save Lyndon Johnson.) That said, Dewey's actual vote share in 1944 (45.89%) was actually slightly higher than his vote share in 1948 (45.07%). Both were comparable to Dukakis' (45.65%).

There were some odd features in all three elections of 1940, 1944, and 1948. (A fuller discussion of these features is here.) In short, 1948 was the most 'normal' of the three elections (although it still had some odd features). In 1940 and 1944, the Republican carried a coalition that was somewhat more atypical for his party, at the time. But there was no election in the 1932-48 period in which the Republican lost badly, but with a 'normal' coalition--except the 1932 and 1936 landslides, in which the Republicans did so badly that any comparison with them would, again, understate the Republican 'problem'. 

Basically, I have argued here for using 1940 as the comparison election, because, even though Willkie's coalition was somewhat atypical, he was doing badly enough nationally to make up for this atypicality (while still doing well enough to make the comparison worthwhile). One could also do comparisons using 1944 and 1948; the ones using 1944 would look largely similar, while the ones using 1948 would not show the same level of decline in the rural Upper Midwest (although it would show more the Republican decline in the large industrial Frost Belt states of Michigan and Pennsylvania). Here are the states in which Republican nominees have gotten a lower vote share than Willkie in 1940 in some elections: 

1952 (none)
1956 (none)
1960
1972
1976
1980
1984 (none)
1988
2000
2004
2012
2016
2020

Here are the states in which Republican nominees in those elections have gotten a lower vote share than Dewey in 1944:

1952 (none)
1956 (none)
1960
1972
1976
1980
1984 (none)
1988
2000
2004
2012
2016
2020

And finally, the states in which Republican nominees did worse than Dewey in 1948 in those elections:

1952 (none)
1956 (none)
1960
1972 (none)
1976
1980
1984
1988
2000
2004
2012
2016
2020

Somewhat counter-intuitively, one can see a Republican decline, in the Reagan years, not simply in the Northeast, but also in the rural Upper Midwest and in the industrial state of Michigan. In 1980--winning a national majority of 50.75%--Reagan underperformed both Willkie's and Dewey '44's vote share in the future Dukakis Upper Midwest states of Wisconsin, Iowa, and Minnesota, as well as in Michigan. He underperformed Dewey '48's vote share in Michigan and Pennsylvania. In 1984, Reagan managed (winning a national landslide) to outperform both Willkie and Dewey '44 everywhere (and the only state in which he underperformed Dewey '48 was Vermont, which is interesting but not unexpected). But in 1988, George H. W. Bush underperformed both Willkie and Dewey '44 throughout the Dukakis Upper Midwest (as well as in Kansas and South Dakota), and underperformed Dewey '48 in Pennsylvania (and managed to underperform Dewey '48 in Iowa as well, despite that Dewey didn't do so well in Iowa in '48). 

Many of the Upper Midwest states in which Bush did surprisingly well in 2000 and 2004 nevertheless show up as states in which he underperformed Willkie and Dewey '44: he underperformed Willkie in Minnesota and Iowa in both of his elections, and in Wisconsin in 2000; and he underperformed Dewey '44 in Iowa and Wisconsin in both of his elections, and in Minnesota in 2000. He didn't underperform Dewey '48 in any of these states in either of his elections, but did underperform him in both elections in Michigan and Pennsylvania. 

But Trump also underperformed Willkie in Minnesota, Wisconsin, and Iowa in 2016 (and again in Minnesota in 2020); and underperformed Dewey '44 in Minnesota, Wisconsin, and Iowa in 2016 (and again in both Minnesota and Wisconsin in 2020); and Dewey '48 in Michigan and Pennsylvania in both 2016 and 2020.

In other words, much of the ground recent Republicans have been making up in the Upper Midwest and Frost Belt was a) lost in the Reagan years, and b) has been made up to a relatively limited degree by both Bush and Trump. (Trump's breakthrough states, although they had all become known as Democratic strongholds by 2016, had been, in the deeper past, historically Republican states.)