The Tech - Online EditionMIT's oldest and largest
newspaper & the first
newspaper published
on the web
Boston Weather: 40.0°F | Partly Cloudy


Votive Offering: How to Have Your Election Cake and Eat It Too

David Strozzi

When Californians vote in the recall election, they will be confronted with 135 candidates. Voting could prove harder than an 18.03 exam. Should a Republican vote for Schwarzenegger or McClintock? Running can be even more dangerous: why are dozens of sure losers running? Who will “cost” whom the election? Similar discussions happen in many elections, most notably the 2000 Presidential race. In particular, Ralph Nader and those who voted for him were accused of costing Gore the election.

As a Nader voter, I do not take these charges lightly. If Nader had not run, Gore would have won the election. One can speculate about what would have happened in 1992 if Perot hadn’t run, and similarly for many other races. The outcome of these elections changes if additional losing candidates run.

In many elections, a third candidate produces a “spoiler effect.” Voting for someone besides a front-runner does not affect who wins. Third-party candidates are charged with “taking votes away” from the front-runner closer to their positions and costing him the election. Would-be third-party supporters realize this and hide their sincere preference by choosing a front-runner -- the “lesser of two evils.” These aspects of plurality voting are why we have a two and not multi-party system. Plurality voting leading to two-party systems can be seen in many countries, prompting political scientists to dub this trend “Duverger’s Law.”

These problems are a damning indictment not of third parties but of plurality voting (where a voter selects one choice, and the choice with the most votes wins). Many voting systems lack these drawbacks, and we should consider using them instead. For example, the system I prefer is the Condorcet Method. Here, each voter ranks as many choices as they want to. The candidates are paired up and the rankings of the candidates are compared. This is done for every possible combination of candidates. If someone wins all her pairwise comparisons, she is shown to be preferred over every other candidate (the “Condorcet winner”), and she is elected. As a result, there is no spoiler effect: I could express my strong support for Nader and still take part in the contest between Bush and Gore.

Alas, there isn’t always a Condorcet winner. This means the voters prefer A over B, B over C, and C over A. This requires three candidates to have substantial support, and there are various methods to resolve this “cycle” and choose a winner. Condorcet does not produce this cycle, but merely reveals it, since it really exists in the electorate’s opinions.

The alternative system with the most momentum in the U.S., and currently used in Australia and Ireland, is instant runoff voting. As in Condorcet, a voter ranks her choices. If someone has a majority of the first-place votes, she wins. If not, the candidate with the fewest first-place votes is eliminated from the ballots. Each voter’s rankings are renumbered to reflect this (e.g., if they ranked the eliminated candidate first, their second choice becomes their first choice, and so on). This continues until someone has a majority.

Instant runoff voting eliminates the spoiler effect provided all but two candidates have weak support. However, a candidate who would beat two other candidates in pairwise races can lose an instant runoff election. One way this happens is if the “centrist” has the fewest first-place preferences, but many second-place preferences from supporters of each “extremist.” Either an extremist wins, or supporters of one extremist rank the centrist first so he makes it to the second round (and beats the other extremist). Either the spoiler effect or strategic voting occurs. Notice that I discuss “preferences” and not “votes” precisely because voters are forced to use strategy and not vote their sincere preferences.

Approval voting approaches Condorcet’s results in a simpler way. Each voter approves (without ranking) as many choices as she wants, and the candidate with the most approvals wins. This is almost always a Condorcet winner (if one exists). There is no spoiler effect to speak of. The downside is voters have to choose when to stop approving candidates. A simple rule, which is close to an optimal strategy, is to approve your favorite front-runner and every candidate you prefer over him. This assumes you can reliably identify front-runners, which only happens when there are strong historical precedents or accurate and stable polling data.

Our voting system “cost” Gore the 2000 election. It is also why we have only two significant parties. Other systems have been studied and used elsewhere that remove the spoiler effect and enhance the democratic process. We should adopt one of them here.

An excellent introduction to voting systems and voting reform is wiki/Voting_system.

David Strozzi is a graduate student in the Department of Physics.