Al Gore won the popular vote in 2000 by more than 500,000 votes. Hillary Clinton leads Donald Trump by more than 2 million. Yet she still lost the White House because Donald Trump narrowly won Wisconsin, Michigan, and Pennsylvania combined by fewer than 100,000 votes.
The electoral college is not neutral, it favors the presidential candidate who wins in the small states and/or the largest states over the candidate that is most popular. How should this problem be fixed? First we will identify where the electoral college goes bad. Then we will examine whether the leading proposed electoral college reform, the National Popular Vote Interstate Compact, remedies the problems.
One problem is that all states get two electors for free, in addition to one elector per House district, the same way that each state gets two Senators. This gives smaller states substantially more electors per voter than larger states.
A second problem is that 48 states and the District of Columbia award all of their electors to the candidate who wins the statewide vote instead of awarding electors individually by each House district vote result or collectively in proportion to the statewide vote. Although the electoral college is a federal institution, the states decide how their electors are selected. State lawmakers assign their electors this way, despite it being unfair to their own voters who voted for the non-first place candidates, because it increases the influence of the state over the final result.
When states select electors individually based on each House district result (as does Maine) then there is a different problem. The drawing of House district boundaries for partisan advantage (a.k.a. gerrymandering) biases both the state and national results.
Avoiding these problems with the electoral college is the motive behind the National Popular Vote Interstate Compact. It is an agreement among several U.S. states and the District of Columbia to award all of their electoral votes to whichever presidential candidate wins the overall popular vote. Once states totaling 270 electoral votes join the compact—which only requires passing state laws—then the next presidential election will be determined by the popular vote, not the Electoral College. As of early November 2016, 10 states and the District of Columbia have signed the compact, totaling 165 electoral votes which is over 60% of the way to 270. This approach to reforming the electoral college avoids a federal constitutional amendment that requires support from two thirds of both houses of Congress and three fifths of the states.
Under that National Popular Vote Interstate Compact, what happens if six (or more) candidates for president split the vote so that no one candidate wins more than twenty percent of the vote? The candidate who won twenty percent of the vote would receive 100% of the electors from the majoritarian subset of the states that have adopted the compact. Will the resulting president be the most popular candidate overall among the voters? The popularity ranking of the newly elected president relative to the other candidates could be anywhere from first place to last place. We do not know because the voters did not express their second (or third, etc.) preference. This is not merely a problem of lacking information, it is a problem with the election outcome. The election winner may, in fact, be the most disliked candidate overall who was elected by the 20% of the population that, to quote Hillary Clinton, define the "deplorables".
The only context where the National Popular Vote Interstate Compact avoids the problem of a candidate who is less preferred overall among the voters winning the election is the context where there are exactly two candidates. The moment a third candidate draws some votes the National Popular Vote Interstate Compact makes it possible for a candidate who is not the most popular to win the election. Our electoral college provides an imperfect mechanism for resolving this problem by requiring that the House of Representatives select the president from the three candidates with the most electoral college votes when firstly the voters, and secondly the electoral college, failed to identify a most popular candidate with their votes. The National Popular Vote Interstate Compact subverts that mechanism by always giving one candidate an automatic majority of electors even when that candidate is unpopular nationwide.
The National Popular Vote Interstate Compact is like a game of whack-a-mole. It fixes a problem in one place while enabling the same problem to recur in a different place. In the short term the interstate compact could help avoid the wrong candidate winning the electoral college. But for the longer term, fixing our 18th century method of electing people to public office will require more changes than the interstate compact would implement.
It is easy to diagnose the problems, including the problem with the National Popular Vote Interstate Compact. It is more difficult to design an optimal method for electing a single person in multi-candidate elections. There is arguably no one election method that is the best method, certainly there is no perfect method.
But it is not difficult to identify methods for electing a president that are technically better than both the existing electoral college and the National Popular Vote Interstate Compact. Unfortunately, the better methods alter the electoral college and will therefore require a constitutional amendment to implement. The integrity of our presidential elections is important. The presidency of the United States is a very powerful position. We need something better than either the existing electoral college or the National Popular Vote Interstate Compact.
A constitutional amendment focusing on federal elections provides an opportunity to populate the House of Representatives with lawmakers who more accurately represent the voters by including a provision that prevents gerrymandering when drawing House district boundaries. To create election districts without gerrymandering we can rely on automated mathematical methods designed to optimize election district compactness. This will tend to result in major changes to election district boundaries in ten or twenty year intervals. This potentially increases the turnover of elected officials without decreasing voter choice or eliminating all of the experienced lawmakers, as happens with term limits.
Four changes to the electoral college would render it substantially better: Eliminate the two statewide electors, when a candidate won a majority of votes nationally then the electoral college is dismissed and that candidate is automatically the president elect, otherwise automatically assign the electors from each state to the individual candidates proportional to the statewide vote for those candidates, and automaticaly cast each elector's vote to their assigned candidate. Two more changes that should also be considered: Give the electoral college some time, maybe a week, to try to reach a majority consensus with several more votes before the decision is turned over to the House of Representatives, and sequester the electors, like jurors are sequestered, during their deliberations to secure them from bribes or threats.
The first four changes to the electoral college would elect the nationwide vote winner while increasing the likelihood that the electoral college winner will also be the most popular candidate. The latter two changes give the electoral college an opportunity to select a nationally popular candidate when no one candidate initially won a majority of votes or electors. Without those two additional changes each elector is a particular vote and there is no need for any real people to serve as the electors. But there may be better ways to elect a popular candidate than to hand the final decision over to the electors or the House of Representatives when there is no clear winner.
For more reliably accurate election results we would need to change the method of voting and tallying votes to allow voters to approve more than one candidate, or to rank the candidates, particularly in single winner contests. This affects the electoral college. The state electors could now be individually elected by House district provided that gerrymandering is eliminated. Alternatively, the electoral college could be scrapped because it is very unlikely that there will not be a clear nationwide winner. A constitutional amendment should therefore choose between retaining our current election method or replacing it with a better method and then modify the electoral college to match the election method.
My personal favorite methods for tallying the overall preference of the voters in single winner contests are the Condorcet methods. Voters are asked to rank the candidates. A complete ranking of all candidates is best, but omitting some candidates, or even voting for only one of the candidates, is acceptable. Condorcet methods start by pairing the candidates with each other and incrementing the count for one candidate of each pair each time a voter preferred that candidate over the other candidate. The pairs are usually ordered based on a measure of how strongly the losing candidate of each pair is defeated. Any circular rankings within groups of multiple pairs, referred to as cycles (e.g. A beats B beats C beats A), are then resolved. Different Condorcet methods take different approaches to eliminating the cycles. After all of the cycles are eliminated there is one candidate that beats all of the other candidates. Two different approaches to eliminating the cycles can be tried simultaneously by iterating through all combinations of culling the cycles using those two approaches. The cycle culling that disregards the fewest voter preferences then identifies the winner.
There are also non-Condorcet methods that are better at identifying the most popular candidate than the overly simplistic method of voting for one with the candidate obtaining the most votes winning. There are technical criteria for identifying good election methods. We need to rely on mathematicians who study election methods to tell us how well the different methods comply with various criteria.
Voter behavior also must be considered when selecting an election method. It may require some experimentation and time to determine which methods actually work well in different contexts. Primary elections, where it is common for many candidates to vie to be a political party's single winner nominee, are good for experimenting with better single winner election methods. A problem with preference voting is that voters have incentive to vote strategically by ranking the candidates they do not want to win last because the polls say those candidates are among the most popular candidates instead of voting sincerely by ranking the candidates they most dislike last. Sincere voters may be disadvantaged by strategic voters but strategic voting undermines the integrity of the election result.
Eliminating gerrymandering and thoughtful electoral college reform, with or without voting and ballot tallying method reform, can be accompanied by other steps to improve the quality of our elections. The federal government should automatically register everyone to vote in federal elections when they turn 18. Voter registration and de-registration could be automatically linked to state driver's licenses, state tax returns, post office change of address applications, and death certificates. There should be regular auditing of the voter registration rolls. Ballots could be mailed to all residents. Governments could arrange free transportation to polling places and publish videos of the candidates promoting themselves on the Internet. Keeping polling places open 12 hours every day for one week, as is done in Maryland, should be the national standard. Election results should always be audited before they are finalized.