‘Large state’ is a term that needs some precision. It might not seem all that important, but in 2008, George Will said that ‘Ohio’s the only large state outside the South that [Bush] carried.’ Insofar as Will is making a point about national politics, it is actually important to have a non-fuzzy definition of a ‘large state’.
Now, a number of analysts’ intuitions appear to have converged at around 15 electoral votes or so as the threshold for a ‘large state’. In 1988, Walter Manley II analysed Dukakis’ and Bush Sr’s chances in the seven contested ‘electoral-rich states’ (explicitly justifying leaving out New York and Florida on the basis of their safely Democratic and Republican statuses, respectively). The seven states he analysed, together with New York and Florida, were the nine biggest electoral prizes that year, ranging from 16-EV New Jersey to 47-EV California. (The next-largest electoral prizes after New Jersey in 1988 were Massachusetts and North Carolina, each with 13 electoral votes at the time.) In 2008, Rhodes Cook defined a ‘megastate’ as a state ‘with 15 or more electoral votes at the time of the election’. However, neither Manley nor Cook gives a justification for his definition.
Interestingly, there is a phenomenon that is directly tied to state populations, which can be used to define the threshold for a ‘large state’ (a threshold which ends up fairly close to Manley’s and Cook’s definitions). We have to start with something David Wasserman tweeted after the 2016 election: ‘If more than 50 percent of your ’16 voters lived in just nine states, you’re probably not a healthy national party.’
We can call this number — nine for Hillary Clinton in 2016 — a presidential nominee’s ‘Wasserman number’: the minimum number of states that can account for half of a nominee’s voters.
Now, there are multiple groupings of nine states that can account for half of Hillary Clinton’s 2016 voters. This is one; this is another; this is another. However, there are certain states that will be in any grouping of nine states that accounts for half of Hillary Clinton’s 2016 voters. We can call these a nominee’s ‘necessary Wasserman states’.
We can then find every winning and narrowly-losing nominee’s necessary Wasserman states, and see if there seems to be a size below which they tend not to go. (I am limiting my consideration mostly to winning nominees only, because we are trying to find something out about the country — what makes a state ‘large’ in the United States — rather than about the parties. Perhaps unsurprisingly, things can get weird in the cases of badly-losing nominees. McGovern’s smallest necessary Wasserman state was then-41-EV-New York, which, as we will see, was atypical for his party and time; George Wallace appears to have had no necessary Wasserman states at all. However, I am considering narrowly-losing nominees — that is, nominees who lost by less than 4.5% in the popular vote and by less than or equal to 23.6363…% in the Electoral College — on the basis that a narrowly-losing nominee could [often] have won on the basis of technicalities or litigation, without having had to persuade ahead of Election Day any more than he or she had already done. Examples include the butterfly ballot and the projection of Florida ahead of polls closing in the Panhandle in 2000, and contemplated challenges to both Wilson’s win in California and Hughes’ win in Minnesota in 1916.)
In 1912, the Electoral College jumped from 483 to 531, and has stayed at roughly the same size ever since. In both 1912 and 1916, Woodrow Wilson’s smallest necessary Wasserman state was then-13-EV California. However, if we look at every winning and narrowly-losing nominee’s smallest necessary Wasserman state from 1920 on, we get this:
Harding, 1920: Ohio (24)
Coolidge, 1924: Ohio (24)
Hoover, 1928: Illinois (29)
Roosevelt, 1932: California (22)
Roosevelt, 1936: California (22)
Roosevelt, 1940: California (22)
Roosevelt, 1944: California/Ohio (25)
Truman, 1948: California (25)
Dewey, 1948: California/Ohio (25)
Eisenhower, 1952: Illinois (27)
Eisenhower, 1956: Ohio (25)
Kennedy, 1960: California/Pennsylvania (32)
Nixon, 1960: Ohio (25)
Johnson, 1964: California (40)
Nixon, 1968: Illinois/Ohio (26)
Humphrey, 1968: Massachusetts (14)
Nixon, 1972: Illinois (26)
Carter, 1976: Florida (17)
Ford, 1976: Ohio (25)
Reagan, 1980: Florida (17)
Reagan, 1984: New Jersey (16)
Bush Sr, 1988: New Jersey (16)
Clinton, 1992: Michigan (18)
Clinton, 1996: Ohio (21)
Bush, 2000: North Carolina (14)
Gore, 2000: Florida (25)
Bush, 2004: Georgia/North Carolina (15)
Kerry, 2004: Ohio (20)
Obama, 2008: Illinois/Pennsylvania (21)
Obama, 2012: Ohio (18)
Romney, 2012: Florida (29)
Trump, 2016: Florida (29)
Hillary Clinton, 2016: Florida/New York (29)
(Hughes’ smallest necessary Wasserman state in 1916 was 24-EV Ohio.)
From this, we can observe some things. The size of the smallest Wasserman state of winning and narrowly-losing nominees appears to fluctuate as a function of time and party. From 1920 through about 1964, the smallest necessary Wasserman state for winning and narrowly-losing nominees of both parties appears to have tended to be large. Then, winning and narrowly-losing Democrats’ smallest necessary Wasserman states tended to be in the teens from 1968 through about 1992, before becoming big again, for the most part, from 1996 on. The Republicans seemed to enter the smaller era slightly after the Democrats — around 1980 — and then to enter the second bigger era slightly after the Democrats as well — around 2012.
We can also see that the smallest that any winning or narrowly-losing nominee’s smallest necessary Wasserman state has been from 1920 on, has been 14 electoral votes. This has occurred twice, for two different parties, in two different states, 32 years apart: for Humphrey in Massachusetts in 1968, and for Bush in North Carolina in 2000.
In addition, however, it is worth noting that there were three elections in the 1968–2004 period in which there were no 14-EV states at all: 1984, 1988, and 2004. In all three cases, the winning nominee’s smallest necessary Wasserman state was one (or more, in Bush’s case) of the states with the smallest value above 14 at the time: 16-EV New Jersey in 1984 and 1988, 15-EV North Carolina and Georgia in 2004.
We have now had a 531–538-EV Electoral College for over a century, and the size of winning and narrowly-losing nominees’ smallest necessary Wasserman states has exhibited a high degree of regularity over this period. Aside from California in 1912 and 1916, the smallest any winning or narrowly-losing nominee’s smallest necessary Wasserman state has been is 14 electoral votes, and this was in a period when values at or just above 14 was the norm. This can arguably be said to be the Electoral College reflecting something about the states’ sizes. On this basis, I would say that a ‘large state’ is one with 14 or more electoral votes at the time of the election in question.
What about the exceptions of California in 1912 and 1916? Well, unlike Massachusetts in 1968 and North Carolina in 2000, these were not part of a broader phenomenon — from 1920 on, the size of winning nominees’ smallest necessary Wasserman states (including for the next winning Democrat) jumped sharply, leaving California in 1912 and 1916 as a relatively isolated phenomenon.
And what about the elections prior to 1912? Well, the Electoral College’s overall size was constantly fluctuating before 1912, which would make it less likely that the size of nominees’ smallest necessary Wasserman states would tell us as much. Aside from that, however, the size of winning and narrowly-losing nominees’ smallest necessary Wasserman states themselves fluctuated wildly from 1840 through 1908, in contrast to the long-term regularity we see from 1920 on. For example, in 1848, Taylor’s smallest necessary Wasserman state was 23-EV Ohio; in 1852, Pierce’s was 11-EV Illinois; in 1856, Buchanan’s was again 23-EV Ohio. (And lest one think that this was a function of party, while Cass’s smallest necessary Wasserman state in 1848 was 12-EV Indiana, Polk’s in 1844 was 23-EV Ohio. And Whig nominee Winfield Scott’s in 1852 was 13-EV Indiana — a sharp departure from Taylor in 1848.) Similarly, in 1872, Grant’s smallest necessary Wasserman state was 21-EV Illinois; Hayes’ in 1876, 11-EV Michigan; Garfield’s in 1880, once again 21-EV Illinois. In 1904, Theodore Roosevelt’s smallest necessary Wasserman state was 23-EV Ohio; in 1908, his fellow Republican Taft’s was 14-EV Michigan.
Another thing one can notice is that nominees’ smallest necessary Wasserman states were frequently in the 20s — out of a smaller overall Electoral College. (The smallest size reached by a winning or narrowly-losing nominee’s smallest necessary Wasserman state from 1840 through 1908, however, was 11, in 1852 [Illinois, for Pierce] and in 1876 [Iowa/Michigan, for Hayes]. It was a 13-EV state [Texas/Michigan] for Cleveland in 1888; and, if one counts Cass in 1848 as narrowly-losing, it was a 12-EV state [Indiana] for him.)
We generally have less occasion to talk about ‘large states’ in the past. To the degree that we do, the Orange County Register proposed another index of a state’s being large in 2016:
As of this writing, Clinton’s lead in the popular vote nationwide was 1.7 million. Clinton’s lead in California was 3.5 million.
That means, when all the votes are counted and Clinton’s popular-vote edge is official, it probably will in a manner of speaking have hinged entirely on the results in California.
This is rare: In the Bush-Gore election, four states gave Gore vote margins larger than his 543,895-vote national margin. The last time a single state’s voting accounted for a candidate’s national vote margin was 1888, which happens to be the last election before Bush-Gore to see an Electoral College/popular vote split. Grover Cleveland received 90,596 more votes than Benjamin Harrison nationwide, and for that Cleveland could thank the 146,461-vote margin he received in Texas.
This year’s popular-vote count — nationally and in California — is a reminder that the Constitution’s authors were addressing a legitimate fear by adopting the Electoral College in part to prevent someone from becoming president by rolling up a huge margin in one big state, perhaps the candidate’s home state or a state with a narrow political interest.
This was not the Register’s focus, but the implication is that if one and only one state can account for a candidate’s national popular vote margin, it is a ‘big state’. This makes some sense — Gore’s national margin in 2000 was so small that it would have been hard for multiple states not to account for it. (And it is probably likely that if, say, only then-12-EV Massachusetts had done so, that would indicate that Gore was not doing well enough in other states to win the national popular vote in the first place.)
As the Register noted, however, this is a rare occurrence. In fact, there are only four times that a single state has accounted for a nominee’s national popular vote margin (counting all the Whig nominees in 1836 as an aggregate):
1836 (42-EV New York for van Buren; national margin 1.74%)
1884 (13-EV Texas for Cleveland; national margin 0.57%)
1888 (13-EV Texas for Cleveland; national margin 0.83%)
2016 (55-EV California for Hillary Clinton; national margin 2.09%)
The cases of 1836 and 2016 don’t tell us much (at least as far as what the threshold for a ‘large state’ is), since New York in 1836 and California in 2016 were the largest states. And because this doesn’t happen very often in general, there are long periods of time when this index alone wouldn’t tell us much about what the threshold for a large state is at all.
That said, it did happen twice in the 1880s. Cleveland’s national margin in 1884 was not much bigger than Gore’s in 2000 (0.51%). However, the margins in both 1884 and 1888 were greater than the margins in all the elections where multiple states could separately account for the winner’s national margin (1880, 1960, 2000), and 1888’s margin, in addition, was larger than the smallest margin that could not be accounted for in any one state (Nixon’s 0.70% margin over Humphrey in 1968). Furthermore, it is true that 1888 is (along with, now, 2016) one of two elections in which the Electoral College selected a winner who had lost the popular vote without the intervention of an ‘external agency’.
So, perhaps we could say that in the late 1800s, a ‘large’ state was one with 13 or more electoral votes.