It's the (voting) system, man!
by Dan Keshet


In the last three U.S. presidential elections, no candidate has received a majority of the popular vote. It seems quite possible that George Bush would have won the 1992 election had the voters who voted for Perot instead chose between Bush and Clinton. Clinton was the better choice because Perot ran, but would Bush have been better if Perot had not run?

In most US elections, though, some candidate does win a majority of the votes. But in these elections, we don't know how many voters spurned their preferred candidate in favor of one they thought might win. It's possible a majority of voters preferred Perot but didn't vote for him because they "wanted their vote to count". Even polls wouldn't tell us this, though, because under our election system, most voters don't bother evaluating candidates whom they know won't win. Our election structure systematically discourages voters from entertaining debate outside of leading candidates, then makes it impossible to glean from the results what voters really wanted.

There is another unanswered question: how many candidates didn't enter races because their campaign could do damage to their own cause? John McCain likely could have won in head-to-head races with either Al Gore or George W. Bush in the 2000 elections. But had he entered, McCain also could have "siphoned" more votes from Bush than Gore, resulting in Gore's election. Our system systematically discourages popular candidates from entering.

Far from being a simple technical matter, the rules of voting form the framework in which all campaigns, candidates, and movements must work. Changes in our voting system could have profound effects on the political system. Given the movements for other aspects of electoral reform, such as new voting machines or campaign finance reform, the debate on which voting system to use is sparse.

One proposed change is moving to a system called Instant Runoff Voting (IRV). IRV is used in Ireland, Australia, and, as of 2002, San Francisco. In IRV, voters rank all candidates (e.g. 1. Ross Perot, 2. George Bush, 3. Bill Clinton...). The candidate with the fewest first-choice votes is eliminated and their votes are transferred to the next choice on their ballots. The process is repeated until some candidate wins a majority of the votes. When there are one or two strong candidates and many weak ones, this process allows voters to vote for a weak candidate, but still get a say in the head-to-head race between the strong candidates. This should let weaker candidates run without being constantly hounded that they were going to "cost the election" for another candidate.

But as soon as there are competitive elections with three or more viable candidates, IRV suffers from the same problems as our system. Bush could beat either Clinton or Perot in head-to-head races, but still be eliminated first because he received fewer first-choice votes. If we do switch to IRV, we will maintain our two-party system in some places; where we step out of it, we will step into a mess of "compromise" voting, risky candidacies, and unclear messages.

In 1785, the French mathematician Marquis du Condorcet devised a way of thinking about elections that avoids these contradictions. Imagine conducting a series of two-way elections with every pair of candidates in an election. If any candidate wins all of their two-way elections, then that candidate is the "Condorcet winner". There are many methods to choose a winner if no Condorcet winner exists.

Under a Condorcet system, John McCain could have safely entered the 2000 race knowing that he would have only affected the outcome if he could have beat the winner head-to-head. In 1992, voters could have safely ranked Ross Perot first, knowing that their preference in the head-to-head race between George Bush and Bill Clinton would have still been counted.

However, systems based on the Condorcet winner are rarely used. They are complicated, both to explain and to count the votes. If voters and candidates don't have trust in the outcome, then one of the purposes of an election is defeated.

The system I prefer for single-winner elections is Approval Voting. In this system, each voter votes yes or no for each option and whichever option gets the most votes wins. A voter who prefers Ross Perot can still choose between George Bush and Bill Clinton by voting for both Perot and Bush. It doesn't guarantee it, but this system is very likely to result in the Condorcet winner. Although strategic voting is built into the voter's choice of how many candidates to vote for, voting for your favorite candidate can never cause you to get worse results. Because of that, this system gives much cleaner information about which candidates voters really prefer.

Changing voting systems will not make politicians honest, puppies cuter, or ice cream more delicious. But if it's worth it to us to hold elections, it should be worth it to us to hold elections well.

To learn more, visit the Administration and Costs of Elections Project at

or the Voting systems page of Wikipedia:

----Beginning of inset text---------

Imagine an election for the capital of Tennessee. If voters vote based on geographic proximity, here's a chart of voter preferences (highest preference listed on top):

(Picture and chart here)

Nashville is the Condorcet winner, defeating Memphis (58-42), Chattanooga (68-32), and Knoxville (68-32) head-to-head. In IRV, Chattanooga would be eliminated first (all votes transferring to Knoxville), then Nashville (all votes transferring to Knoxville), then Memphis, resulting in victory for Knoxville. If Memphis were withdrawn from the race or if Memphis voters "compromised" by listing Nashville first, Nashville would win. In Approval voting, if voters choose between the leading two options and also vote for any options they prefer more (a simple and good strategy), Nashville would be approved the most.

(This example and picture were drawn from Wikipedia)