Landslides

[Skill and Chance]

I agree with Muon.  Double digit national margin = landslide.  Of course, these are heavily clustered in the End WWI-End Cold War era when most people got along unusually well and had a strong sense of common identity that is anomalous compared to the rest of US history.  You can see this clearly in how many EV held out against Teddy Roosevelt even in his 19% win (Teddy and the 1912 vote splitting were the only double digit wins between the end of Reconstruction and the end of WWI).

[Skill and Chance]

The median result of an American election (in electoral votes) is about 62% of the electoral vote. That is what Obama did in 2012. If you expect the electoral-vote percentage to be distributed randomly, then percentages of the electoral vote for the winner would cluster around 62%. Margins beginning in 1900 from the narrowest to the broadest divides are

01 2000 50.4
02 1916 52.2
03 2004 53.2
04 1976 55.2
05 1968 55.9 mostly squeakers
06 1960 56.4
07 2016 56.5
08 1948 57.1
---------------
09 2012 61.7 hard to characterize
---------------
10 1900 65.3
11 1908 66.5
12 2008 67.8
13 1992 68.8 bare landslides
14 1996 70.4
15 1904 70.6
16 1924 71.9
---------------
17 1920 76.1
18 1988 79.2
19 1944 81.4
20 1912 81.9 landslides
21 1952 83.2
22 1928 83.6
23 1940 84.6
24 1956 86.1
---------------
25 1932 88.9
26 1964 90.3 big landslides
27 1980 90.9
---------------
28 1972 96.7
29 1984 97.6 gigantic landslides
30 1936 98.5

That's thirty Presidential  elections, and much has changed in technologies of transportation, communication, and computation. Campaign management has become a near-science. Public polling did not exist before Gallup invented it in 1948. Five states have been added to the Union  between 1907 and 1959. Educational standards are much higher now, with lots of small children being dragooned to toil in mines and factories around 1900 with the end of their formal education, and lots of people having graduate degrees now. The American population has expanded greatly and its distribution is very different. Women got the vote by law for the 1920 election. Blacks rarely voted in the former Confederate States until the 1960's due to sundry subterfuges, but that is over.  Life spans have increased, lengthening the number of years in which people could vote by about thirty years on the average. The legal age for voting has gone from 21 to 18, which typically adds one potential vote in a presidential election for us all. 

The Electoral College remains, and except for the small states of Maine and Nebraska (such happening recently) practically all states allocated their electoral votes on a winner-take-all basis. The population has changed greatly in 120 years. The largest states in their counts of electoral votes today (CA 55, TX 38, FL and NY 29) having had 5, 15, 36, and 4 electoral votes in 1900, respectively. The four largest states in electoral votes in 1900 had 36 (NY), 32 (PA), 24 (IL), and 23 (OH). Iowa, Missouri, and Pennsylvania have lost nearly half the electoral votes that they had in 1900.   

First of all, eight elections are close. Except for 1948 and 1968 (which involve racist secessions against the mainstream Democratic Party), those elections could have gone either way. Truman kayoed Dewey by 114 electoral votes. In a two-way split of the electoral vote, such would give a  326-211 result, which would not be a close election.

that's 116 years between the first and last election in this  group, and forty elections in times of peace and war.

By looking for gaps one can subdivide the levels of victory. The biggest gap (5.8%) now separates Reagan 1980 from the three biggest electoral blowouts. A significant gap (4.2%) separates elections of 1924 and 1920. That's a good candidate for a break in classifications.

Splitting the shares of electoral votes by obvious gaps one finds

 
01 2000 50.4
02 1916 52.2
03 2004 53.2
04 1976 55.2
05 1968 55.9 mostly squeakers
06 1960 56.4
07 2016 56.5
08 1948 57.1
---------------
09 2012 61.7 hard to characterize
---------------
10 1900 65.3
11 1908 66.5
12 2008 67.8
13 1992 68.8 bare landslides
14 1996 70.4
15 1904 70.6
16 1924 71.9
---------------
17 1920 76.1
18 1988 79.2
19 1944 81.4
20 1912 81.9 landslides
21 1952 83.2
22 1928 83.6
23 1940 84.6
24 1956 86.1
---------------
25 1932 88.9
26 1964 90.3 big landslides
27 1980 90.9
---------------
28 1972 96.7
29 1984 97.6 gigantic landslides
30 1936 98.5

Do you like my classification?  Just over two weeks from now we will have some idea of where the election of 2020 belongs. The one in its own category is Obama in 2012, which was a few thousand votes away in Florida from being in the 'close' category. Until 2012 there was a huge gap (8.2%) between mostly squeaker elections and 'bare landslides'. 
 

Awesome analysis! It's fascinating that, except for 2012, no one since 1900 received between 57.1% and 65.3% of the EV.

I guess 2020 is the 9th "squeaker" election, though it's relatively decisive as "squeakers" go.

I would favor putting the "squeaker" cutoff at 55% or 56% of the EV.  All of the top 3 were one-state victories, which is an important distinction given how elections are regulated and disputed at the state level.  Going to 55% separates them out.  Going to 56% includes some 2 state victories where the 2nd state over the line for the victor was extremely close.  The elections over 56 of the EV are in a different class (even when the victor lost or arguably lost the PV as in 2016 and 1960), because several states have to be flipped to change the outcome (3 in 1948, 3 in 1960, 3 in 2016, and 4 in 2020 for an outright win).  These victories are inherently more robust and fit better in the same class as 2012 (Romney needed to flip 4 states).  Imagine Obama lost instead of won Florida by <1% and this becomes even clearer, though 2012 still stands out as the most robust of them due to the state margins.  It also becomes clearer if you think about the likelihood that PA gets contested by the state legislature in 2020 if it had been a bit closer and Biden lost AZ and GA.