I'm a wordy person, it can be a flaw. I've written an enormous amount on this subject, and learned a few things on the way.
Let me see if I can distill it down.
Why, may I ask, does the case of Ohio strike you as especially notable? I do observe that Carter's margin of victory is very thin, not much more than 10,000 votes. But so what? he clearly still led Ford. That is not what your OP question picks up on though; what seems out of line to you is that Carter's share of votes, though clearly the largest, does not comprise a majority. But again, so what? I wondered if this was a unique and rare case...but take a look at the complete rundown of all states won, using the prevailing winner take all rule,
in the 1976 election. Jump down to the chart of state by state results; it is a semi-interactive spreadsheet, and one can reorder the states by various columns, so let's order it by Carter's victory percentages first. Click on the little arrows next to the percentage signs.
Under the white "totals" row we can see Carter won not one but three states with less than 50 percent--but also that Ford trailed in every one of them. Besides Ohio, he benefited from winner take all in a state he did carry over 50 percent also in Mississippi and Wisconsin; the three states get him 43 EV altogether. His total was 297 OTL and that means yes, if he were to forfeit all three he'd be down to 254, 16 under the 270 score needed to win.
Now of course you aren't saying a candidate who fails to carry a state with over 50 percent should lose all EV (though, in lieu of your proposal to be
selectively proportional, that might be a valid rule too--strike the votes of
all states which fail to return a single majority victor, and let the race be decided only by those states where one candidate or another has a clear lead. I don't think anyone has ever proposed that solution and it doesn't strike me as a great one, but I will consider it as an alternative, having just thought it up).
No, you suggest that--in Ohio only, for some reason--the votes should be split proportionately. Well sort of--you say, "award some to Carter, some to Ford," which rather overlooks the reason such a situation is possible in the first place. If every state were a race where all voters cast votes for two, and only two, candidates, then it would always be the case that one or the other wins each one with a solid majority, unless there were an even number of voters and they split exactly evenly. Well, that is too improbable to actually happen, and the truth is just about every US Presidential election in any year you look at has some number of third party candidates, and so it is that the leading candidate in one of the two leading parties often does fall short of 50 percent, because of the third party vote.
If it makes sense to split EV between the two leading party candidates proportionally in any race, does that not imply as well we ought to count the third party candidates as well?
It can be argued, no, we shouldn't; the race should really be between the two leading parties and third party candidates do not really count. All I'm saying now is, the situation you point to in Ohio would not be possible were it not for third candidates--in this case, led by far by Eugene McCarthy. McCarthy got over 1.5 percent of Ohio's vote, a number of votes far outstripping Carter's victory margin over Ford.
As it happens I did apply several methods of proportionally allotting votes and only the one that is most generous to low-vote candidates, the Hamilton/Largest Remainder method, allots any EV to McCarthy at all, and none to the smaller ones behind him. Any other method awards Carter 13, Ford 12 of Ohio's EV.
Keeping this simple for now, focusing on the question of the 2 main parties and ignoring third candidates like McCarthy, suppose we assume for the moment that every one of Carter's 3 less than majority wins were split the same way, an even split with Carter getting the odd one left over? That gives Carter only 4 and 6 from MS and WI, and Ford gains 3 and 5 from them, and of course 12 from OH--Ford is up by 20 and Carter down by the same amount.
This leaves Carter with 277 EV and Ford with 260--Ford still loses.
But wait! Was Jimmy Carter the only candidate to benefit from getting all of the EV from states where the leading candidate did not get over half the votes?
Look at the chart again, this time sort by Ford's PV percentages state by state--we see that Ford's percentages also drop below 50, and that this happens in
seven states to Carter's 3:
Washington, Oklahoma, Iowa, California, Virginia, Maine and Oregon! It may not seem Washington counts here but actually although sources say it was "exactly" split, Ford falls short of half of all votes cast by some 30 or so. (To be sure, if we were to say that only votes cast for the major two candidates count, Ford's margin would clearly be over half--but by what logic should we discount the minor party voters? If there is any sense in making a fuss over failure to get 50 percent plus one, surely dissidents voting for third parties are relevant to the argument the candidate does not command an actual majority?)
Looking at their EV, and applying the quick and dirty rule of Carter getting half of each state's EV, rounding down to even numbers in each case, Carter would pick up 45 votes Ford got OTL.
Thus, the upshot of applying this ATL rule of special case proportionality is actually to
widen, not narrow, an OTL Electoral College outcome of 297 to 240 that was already far out of proportion to the ratio of popular votes between the two--Carter is, net, up 25 votes and Ford down the same number to give us 322 to 215!
Somehow I think this is not the result you were expecting. I certainly didn't either, but there it is.
What happens by the way if we just strike out all these 10 states and 133 EV who could not decide to cast a majority for anyone?
Well, then, Carter loses 43 EV, and Ford, 90! (actually only 89, for one vote in Washington OTL was cast by an elector to Ronald Reagan, who was not on the ballot and received no popular votes whatsoever). This leaves Carter with 254 and Ford with 150, and if we add those together we get 404 and 203 should be sufficient to win.
If instead of applying proportionality in the limited case you suggest, what happens if we apply it consistently? The easiest way to do that is to just look at the national totals and assign the 538 EV proportionally on a nationwide basis, without regard to state. I find if I do that, Carter still wins--but only by the thinnest fraction, at 270, maybe as high as 272, EV, but never 269 or lower. Carter won the the 1976 election. He did so by very thin margins, but the outcome, given how people voted, was reasonable.
Of course, given the current Constitutional rules, we cannot aggregate state totals into national ones; each state must cast its EV separately. (Perhaps they can agree to pool them voluntarily, as has been proposed by a current movement to get a number of states controlling over 270 EV to agree to automatically elect the electors for whoever gets the PV majority nationally, thus guaranteeing that the person who does that wins (when they do; quite a few Presidents have been elected with less than 50 percent of the national PV, though only rarely with someone else having more). But the legality of this plan is under question; I don't see how it can be forbidden but there are lawyers out there threatening to invalidate it somehow or other).
It would be a pain to go through all 50 states and DC to allocate each of their popular votes proportionally and then total them up to see who wins that way.
Fortunately a fellow board member has already undertaken to do so! We can see the state by state outcomes Golfman76 calculated, and that Carter still wins 272 to 266.
I fear there may be some flaws with his methodology unfortunately, but I think if you check you'll find that by most proportional methods of allotment the figures are substantially correct, and deviate only a little bit where they are not. I got worried when I looked at California--by my figuring California should either be 22 Carter, 23 Ford, or else the odd vote goes to McCarthy by the Hamilton method. Note the effect of this is to narrow Carter's lead. I doubt correcting each state carefully can possibly flip Ford ahead--though the most generous methods of allotting EV to third parties might possibly cut Carter's total down below 270!
Anyway even if there are some fluctuations, Golfman76's work ought to be of some interest to you since it is comprehensive, from the first elections where popular vote was a relevant concept (1824) to the present day.
--------------------------
Here are some lessons learned and considerations to weigh:
1) the purpose of adopting winner-take-all rules is to make the final outcome more decisive. The more accurately we allot votes proportionally, the narrower victory margins appear in terms of EV, and we will quite often fail to achieve majorities at all, if we count third party votes.
2) If we are sticking with the winner-take all rule, there is no reason to make an exception for failure to get over 50 percent. Any of the accepted methods of assigning votes proportionally degenerate down, in the case of there being only one "vote" to assign, in this case all of a given state's EV in one throw, to simply naming whoever leads the pack the winner, and it does not matter if this is with just 2 candidates with one getting a clear majority or with 27 candidates with the leading one having a third or less of the total. There is no logic saying that some other candidate should win, after all. To be sure, the more common it is that there are in fact multiple candidates and not just two, the more illogical the idea that one winner somehow represents everyone is. To a great extent, the problem lies in the fact that some offices, such as the US Presidency, are unitary in nature and trying to split it up to represent a broad spectrum of opinion seems to doom the executive power to self-contradiction and internal gridlock. Unless we can find a way to make the powers of the Presidency something exercised by a broad committee, and have that sort of system work out well for us, or attach checks and balances on the office dependent on the magnitude of vote the single winner actually got, we are stuck with the possibility of being ruled by someone actually elected by a minority. It happened quite a lot in the USA OTL!
3) Although I cannot put my finger yet on a general rule why, and I am extrapolating from a single case that may be peculiar for some obscure reason or other, applying a mixed strategy, of winner take all if any candidate gets over 50 percent but proportionality should they all fail to cross this hurdle seems to have very perverse outcomes indeed, actually multiplying rather than mitigating the inflated victory margin apparent in the outcome of the Electoral College OTL. 297 to 250 implies that Carter got 19 percent more popular votes than Ford, or over 54 percent of the total--but this is false of course, he only got 50.8 percent of the total, or looking just at him versus Ford, 51.4 in that race. The EC tends to distort outcomes in favor of the victor and make their mandate appear stronger than actual public votes would suggest--indeed as we are so strongly aware at this moment, sometimes flip them! But applying selective proportionality to the 10 states only where neither secured a majority actually raises Carter's margins, even if we adopt the exceedingly crude method of just throwing out their votes completely. Indeed that makes the margin the starkest of all, with Carter winning 5:3! Perhaps if we examine all the states where no single candidate led in all years, we'd find this is a strange fluke and that sometimes the margins narrow instead. And perhaps the outcomes might flip--but that would be bad news indeed, because only 4 times in the history of US Presidential elections with popular vote as part of the machinery has the EC failed to ratify the popular vote outcome.
If it is logical to count these "no majority" states proportionally, it is logical to count all states proportionally, every time. The price we pay is that we will fail to achieve clear EV majorities more often, especially if people start to vote for third parties more now that some of their votes are being counted in the EC.
If we could also incorporate a way to arrive at a majority, such as with preference voting, so that ballots cast for low vote getters are transferred to higher vote winning candidates until someone gets a majority, then I would say we should definitely go over to a principle of proportional allotment of EV across the board, in all elections, in all states. Better would be a national proportional vote, for running it through 50 states with numbers of EV varying from 3 to 55 tends to filter out the weaker candidates and give a national total distorted from the actual proportion of votes cast.
----
I am taking some time to nevertheless apply the method of switching votes around in less than majority states, with interesting results. There does not seem to be a general rule or pattern so far, the way the procedure changes the outcome seems rather arbitrary.
In 2016, applying the method still leaves Trump the winner, but only by 277 EV. In 2008 Obama loses about 6 EV but of course still wins big. In 2000, the outcome flips! Gore gets 272 and Bush 265. I then looked at 1992 and lo and behold the simple shortcut of just looking at the 2 leading candidates and swapping votes between them based on near-even splits breaks down completely, for H. Ross Perot did very very well indeed in the PV--in the state of Maine, he got more votes than Bush did, and trailed Clinton pretty closely; all three were well below 50 percent. Applying the Largest Remainder procedure to award EV there I find Clinton would have got 2, Bush 1, Perot 1.
Since Perot's aggregate share of the PV was over 18 percent, applying Largest Remainder there (thus getting the proportional outcome for the nation as a whole) I figure Clinton would have got 232, Bush 202, Perot 102, and Libertarian Andre Marou 2. Either it would have gone to Congress (and although the Democrats controlled Congress I think perhaps there were more Republican controlled states than Democratic, thus presumably Bush would be selected) or if there were also some mechanism to transfer votes in place, it would depend on how Perot voters ranked Clinton versus Bush as second choices--obviously they could have some moderate tendency to prefer Bush and Clinton could still win with his 30 EV vote margin. A strong preference of Perot voters over Clinton would put Bush back in of course--but can we be sure of that?
Anyway the method the thread author proposes would be very painful to apply to 1992 consistently, because almost every state comes under the special proportionality rule! Clinton carried DC and his home state of Arkansas with solid majorities, but everywhere else, including Bush's most favorable state of Mississippi, no one got a majority thanks to Perot's strong showing nationally. So only Arkansas and DC follow standard rules, the notion of simply splitting the EV between the two leading candidates almost evenly breaks down badly and in at least one case if followed would award 2 EV to Perot at Bush's expense. The only just procedure is to follow a more careful proportional procedure for each and every state. Golfman76 did this of course for us, and very remarkably the split between Clinton and Bush is exactly as Largest Remainder would apportion them nationally, at 232 and 202 leaving Perot to vacuum up all the third party EV at 104.
1980 is almost as bad as 1992, with John Anderson's strong third party challenge. It is very hard and I have fallen back on a crude estimate, that suggests Reagan would lose about 112 EV versus OTL, Carter would lose an even larger portion of those he got OTL but gain a fairly large fraction of those Reagan loses, and so come out ahead by 76, with Anderson getting the rest: Reagan 377, Carter 125, Anderson 36 or so--here because Reagan, unlike Bush in 1992, did command solid majorities in many states his OTL titanic landslide is only partially eroded. Going proportionally nationally would leave Reagan barely winning with 272 or so, though Carter is clearly trounced; Golfman76 has 274, 232, 31, 1 (the last for the Libertarian Clark)
1968--again a headache due to Wallace doing quite well as a third party candidate. I am going to stop now.