The New Yorker’s Hendrik Hertzberg demolishes the arguments made against the National Popular Vote proposal by Tara Ross, author of Enlightened Democracy: The Case for the Electoral College, at a recent Cato Institute debate
(via Jon Chait):

Ms. Ross argues that [National Popular Vote] would undermine the two-party system. She says that there would be “five, six, ten Presidential candidates in elections. There’s no reason for there not to be.” As a result, she says, we would end up with a President elected with fifteen per cent (“or it might be twenty per cent, or whatever”) of the popular vote.

In reality, there is a very good “reason for there not to be.” The domination of two large, coalition-like parties is a function of the fact that there can be only one winner of a Presidential election. If it were remotely true that popular-vote elections cause parties to proliferate, then you would expect to find examples of this phenomenon. Since all fifty states elect their governors this way, there ought to be at least a couple that have, or have ever had, this problem. If the problem is a function of size—the larger the electorate, the more likely parties are to proliferate—you would expect to find such proliferation in, say, at least one of the four largest states, each of which is more populous than the entire country was in 1840. You find no such thing. It doesn’t happen in California (pop. thirty-seven million), it doesn’t happen in Wyoming (pop. half a million), and it wouldn’t happen in the United States of America (pop. three hundred million).

So that argument is merely untrue. A second argument—that N.P.V. would empower regional candidates—goes further: it is the exact opposite of the truth. Do I really need to explain why awarding a hundred per cent of a state’s electors to the plurality winner in that state favors candidates whose appeal is regional as opposed to national? “The George Wallaces of the world, which right now have basically no impact on national elections, would have a much larger voice,” she argues. No impact? In 1968, Wallace, whose appeal was regional, got 13.5 per cent of the popular vote and 46 electoral votes. In 1992, Ross Perot, whose appeal was national, got 18.9 per cent of the popular vote and zero electoral votes.

Ross’s argument displays a disturbing failure to grasp basic principles of political science. The reason that we wouldn’t observe a proliferation of candidates under a popular vote system is what’s known as Duverger’s Law, which pundits who hype third-party candidates repeatedly fail to grasp. As one textbook puts it, “In any election where a single winner is chosen by plurality vote (whoever gets the most votes wins), there is a strong tendency for serious competitors to be reduced to two because people tend to vote strategically.” In short, most voters do not want to waste their votes, and the contest is almost always winnowed to the two strongest candidates. Moving from the Electoral College to a popular vote system would not change this dynamic. Indeed, as Hertzberg notes, it might make it stronger by eliminating the possibility of regional candidates capturing a handful of electoral votes.

Update 8/2 11:03 AM: Hertzberg has posted a response to a comment by Rob below. See also Jon Chait for more on the weakness of pro-Electoral College arguments.