Perhaps no feature of our Constitution is more problematic or hard to explain than the Electoral College. It has generated results that thwart popular vote outcomes on five separate occasions. Given our intensely divided politics, and the peculiarities of our electoral system, the problem is likely to recur. In the Electoral College, each state is afforded a number of votes equal to its congressional delegation, meaning the sum of its membership in the House of Representatives, determined on a population basis, and of its Senators, fixed at two per state. (My separate post proposes how to fix the Senate, and, if adopted, those changes could readily be incorporated into this proposal as well).
Like Congress itself, the Electoral College reflects a compromise between large and small states. Because the Senate produces notable distortions in representation as compared with a population baseline, so too the does the Electoral College, albeit blunted by the hybrid feature of further incorporating population-based House representation.
Although it is most unlikely that anyone would design such a system today, the Electoral College possesses some notable attributes that must be accounted for in any attempt to modify or replace it. First, it encourages campaigning in low population states, and, second, it avoids the problems that would arise with a National Popular Vote of inviting challenges quite literally anywhere in a close race as needed to change the result. For example, in the contested 2000 George W. Bush versus Al Gore election, the Electoral College cabined the constitutionally relevant challenges to Florida. Although there were other states with voting-outcome challenges, because it had twenty-five electoral votes, only Florida’s delegation had the possibility of controlling the election outcome. The effect was to render any other counting problems within other states the electoral equivalent of harmless error.
Our electoral system has been modified since its original inception, most notably in the Twelfth Amendment. By implicitly recognizing parties, that amendment allows presidential candidates in general elections to choose running mates, thereby avoiding the original practice of shotgun weddings marrying those who ran against each other from different parties as president and vice president. The Constitution’s original architects envisioned the Electoral College as a deliberative body, and further anticipated that if it failed in selecting a President, the task would devolve to the House of Representatives, with one vote per state. Except in the 1824 election of John Quincy Adams over Andrew Jackson, who received a larger share of the popular vote, this has never occurred. Our system of direct, yet filtered, presidential elections effectively bakes in the two-party system with the consequence that one party almost invariably prevails under the Electoral College, even if occasionally that result is at odds with the other party prevailing in the popular vote. Although the formulation controls who wins, the meeting of the Electoral College is non-deliberative formality.
The most notable alternatives to the Electoral College are National Popular Vote and instant-runoff. National popular vote runs up against the two objectives previously identified: promoting campaigns in low population states and cabining counting errors. Instant runoff invites a separate set of complexities, including voter confusion and operating within a two-staged election in which, at the general election stage, we are already settled upon only two major-party candidates.
The proposal that follows, like that for fixing the Senate, operates more closely within the framework of the existing Constitution. It will not solve all problems with electioneering, including voter suppression and partisan gerrymandering, but it holds the promise of greatly improving the manner by which we select our President and Vice President. The proposal tips toward the Electoral College system, as opposed to the National Popular Vote, but it incorporates elements of each in ways that are likely to avoid the Electoral College’s most problematic results while retaining some of its continuing benefits.
Each state will continue to have the number of electoral votes equal to the full congressional delegation. (If proposal for Fixing the Senate were adopted, this would favorably modify that formula, while leaving the general scheme otherwise unaffected.) If the outcome of the Electoral College vote coincides with the outcome of the National Popular Vote, as certified, for example, by state attorneys general in finalizing the state Electoral College delegations, the Electoral College outcome is conclusive. If the Electoral College and National Popular Vote, so recorded, diverge, such that the former favors the presidential and vice presidential candidates from one party and the National Popular Vote favors the presidential and vice presidential candidates from the other party, the election is resolved by the House of Representatives subject to the following rule: as is presently the case, each state obtains a single vote, but unlike the present system, the House is permitted to select only among three top candidates for President in the primaries and caucuses from within the party that prevailed in the Electoral College. Those candidates that proceed to House resolution shall each select a running mate by a specified date sufficiently in advance of the House deliberations and voting.
The Electoral College produces two related problems for our constitutional system. First, it allows for outcomes that potentially thwart the express will of a majority of voters, as occurs when there is a divergence between the Electoral College and the National Popular Vote outcome. I say potentially because the Electoral College rules motivate campaign strategies. If the voting system were National Popular Vote, candidates would, in theory, campaign differently, and it is possible that in some instances, those who prevailed in the Electoral College but not in the popular vote, might have obtained a popular majority with an alternative campaign strategy.
Second, unlike parliamentary systems, which tend to produce more than two parties, our two-party system is baked-in. As a result, the Electoral College invites the possibility, seemingly realized in the 2016 election, of a successful candidate who is considerably distant from the electoral median voter. Within the existing system there is even a risk of producing what social choice theorists refer to as a Condorcet loser. (For an introduction to social choice theory, see here, especially chapters 13 and 14) Although we lack firm data, Donald Trump might be the first successful presidential candidate in modern history to meet that criterion.
A Condorcet winner is the candidate who, in the absence of a first-choice winner, defeats all others in direct comparisons. In a slate of candidates, none of whom have first-choice majority support, it is possible that voters cycle over the candidates, preferring Al to Becky to Carly to Al. Alternatively, it is possible that although none is a first choice candidate, one of the candidates, called a Condorcet winner, would defeat each of the others in direct comparisons, such as Becky defeating Al in one direct contest, and Becky also defeating Carly in another direct contest. This could even occur if Becky received fewer than a plurality of the votes. Failing to select a Condorcet winning candidate is far less problematic than actually selecting a Condorcet loser. A Condorcet loser is a candidate who would be defeated in a direct comparison against each of the remaining candidates, as might occur, for example, if the rest of the field is dividing majority opposition to an outlier with plurality support.
Although Trump prevailed against a slate of 16 other Republican candidates, those candidates might well have fractured his opposition, allowing several early plurality leads to solidify his position over the candidates that remained. By the time John Kasich was the only remaining Republican challenger, it might have been too late. And then, after receiving the Republican nomination, Trump prevailed in the Electoral College against Hillary Clinton, but did so despite losing the National Popular Vote by 2.8 million votes. Although we cannot be certain, it is possible that Donald Trump is the first Condorcet loser to succeed to the presidency.
My proposed fix is designed to accomplish the following objectives:
First, as with my proposal on Fixing the Senate, it is intended as sufficiently modest to resolve a major constitutional problem without the need to upend the overall constitutional structure and design. Together, these proposals are designed to blunt the perceived need for a highly risky constitutional convention.
Second, it is intended to encourage a change in the conduct of presidential primaries and general election campaigns. The visceral tendency in modern presidential history has been to campaign heavily toward the deep base on each side, and then, having secured the party nomination, to only grudgingly and occasionally move toward the center. The intuition for candidates is often that it is easier to secure two base voters than to flip a single median voter. (For a more detailed analysis, see here, pages 740-42). The result is bimodal campaigns placating a bimodal electorate. This proposal blunts those incentives by signaling a realistic possibility of House resolution if there is a divergence between the Electoral College and National Popular Vote. But unlike the present scheme, the House will be limited to those three top primary candidates for the party that prevailed in the Electoral College. This modification, having the Electoral College, even when indeterminate, conrol which party succeeds to the White House, reduces the likelihood of electing a Condorcet loser. With this regime in place in Election 2016, the House result might have been Ted Cruz or John Kasich, but almost certainly it would not have been Donald Trump. Although accounting for National Popular Vote in making this choice might, once more, suggest the search for votes everywhere, that problem is substantially blunted. The candidate who loses the Electoral College election will not personally benefit by such vote-tally challenges; the election will go to the other party, even if not to the candidate who prevailed in the Electoral College.
Third, the scheme ensures that the candidates do not disregard elections in low population states, but it also ensures that in so campaigning, candidates are more attune to the nation’s more centrist voters since a contrary popular vote risks undermining a close Electoral College victory. This is a hard balancing act, which is the point. This scheme encourages candidates who are sufficiently charismatic to appeal to the base, who are sufficiently nuanced to appreciate the complexities of actual governance, and who are more apt to build or favor broad coalitions.
Fourth, the proposal continues the high stakes of our peculiar electoral system. Our system embeds the serious promise, or risk, depending on one’s view, of policy swings depending on which party prevails. My personal belief is that by eschewing centrism, our swings have been too wide. (For my related post, see "The 'M' Word.") And yet, I also believe that a critical motivation for electoral participation in our system is the promise that victories result in meaningful policy payoffs to the side that prevails. Those promises will likely hold more value in a system that rewards the party prevailing in the Electoral College than under National Popular Vote.
Fifth, this proposal is designed to avoid having the House option kick in except in the rarest of circumstances. A central lesson of Game Theory is that combined strategies that are not played often have a major influence in encouraging those strategies that are played. (For more on this point, see here, p.577 n.17). So too here, the possibility of the House selecting, say, Ted Cruz or John Kasich in place of Donald Trump in the event that Hillary obtains more popular general election votes might well have affected how the primaries were conducted. The candidates might have played their cards differently, and it is also possible that Donald Trump might not have succeeded in securing the nomination. Either way, the revised scheme would reduce the likelihood of Condorcet-loser emerging victorious.
In my Senate post, I mentioned that no one would likely devise my solution on first principles. We live in a second-best world, but even that world can have much to commend it. I do not pretend that this fix is perfect by any stretch, but I think it is better than the risks of a convention and certainly better, by far, than the system we have.
I welcome your comments.
[Special thanks to Brandon Draper for comments on an earlier draft.]