By Daniel B. Szyld
Published May 6th 2007 in Philadelphia Inquirer
One possible solution is a runoff election between the two top vote-getters. This voting method is standard practice in many places around the world, and in some municipalities here in the United States.
We are witnessing one such example today, in France's presidential runoff election. The usual criticisms of this option include the cost of a second poll, and that sometimes a second primary may lead to low turnout.
One solution is called Instant Runoff Voting (IRV). It is like having the election and runoff in the same election, but better.
In this method, the voter ranks the candidates first, second, third, etc. If no candidate receives 50 percent of first-place voting, the candidate with the least number of "first" votes is eliminated, and the votes are immediately reassigned according to the rankings.
One common criticism of IRV is that it is too complicated, but in fact it has been very successful in several U.S. cities, such as San Francisco; Burlington, Vt.; and Minneapolis.
For example, if a voter feels torn between Dwight Evans and Michael Nutter - as The Inquirer Editorial Board indicated in its April 29 mayoral endorsement - one can vote for Evans as first, even though the polls put him last at the moment, and rank Nutter second. If at the end of the vote, Evans is indeed last, in the "instant second round" this hypothetical voter's ballot would indicate Nutter as first.
Some other voter may rank Brady first, Evans second and Fattah third.
After the first round determines that Evans is last, this voter's ballot would indicate Brady first and Fattah second.
The process is repeated until one candidate has a majority (a true majority) of "first" votes.
In many elections, the winner receives less than 50 percent of the votes, and two other candidates who might be considered very close in their position divide the rest. One example was that of my own district in 2006, when we had precisely such a situation in the Democratic primary for state representative. Terry Graboyes stated to me during the campaign that a vote for Anne Dicker was a vote for Michael O'Brien, indicating that Dicker was the spoiler. In the end, O'Brien indeed won the primary, with 34 percent of the vote, with Dicker receiving 32 percent, and Graboyes 29 percent. If what Graboyes indicated was true, and support for both Dicker or Graboyes came from like-minded voters, their combined base was about 60 percent. Had we had IRV, it is possible that one of the two women would be the winner, with broad voter support.
If we had IRV in the presidential election, there would be more room for third-party candidates, and the concept of the spoiler would disappear. Had we had IRV in Florida in 2000, all votes cast for Ralph Nader would have been distributed in an instant runoff between Gore and Bush.
No voting system is perfectly representative, but IRV has a chance of mirroring well the voters' preferences.
Furthermore, a candidate from one party who does not have broad support in the primary may lose the general election.
All parties would benefit from IRV in their primaries, by having a stronger candidate.
When I teach an undergraduate mathematics course for nonscientists at Temple University, the topic of voting methods elicits a great deal of interest and enthusiasm. I sense that the public might similarly embrace IRV as more representative. If people feel they can vote with their consciences and not for the lesser of the two (or more) evils, there might even be higher voter participation. It is time to consider IRV for our primaries and general elections.
Daniel B. Szyld is a professor of mathematics at Temple University.