Sunday, February 21, 2010

Electoral Games People Play



Electoral Games People Play

Book Review: 'Gaming the Vote: Why Elections Aren't Fair (and What We Can Do About It)' by William Poundstone

by Steven Hill,

The best solution will be the method that most liberates voters to choose the candidates they truly like without fearing that honestly ranking or rating the ones who are their second and third choices might help to defeat their favorite candidate.

The realm of electoral system design is still a fairly esoteric branch of political science in the United States--unfortunately so, since no single detail has a greater impact on the quality of representative government. The choice of an electoral system affects which candidate gets elected and all other aspects of a representative democracy, including the number of viable political parties, the quality of campaigns, voter participation levels, the role of campaign finance, legislative policy and more.

But despite the central importance of electoral system selection, the subject has never received much attention in the United States, especially compared with such topics as the role of money, term limits, voter registration and, more recently, the security of voting equipment and election administration. Debate over what system works best has been largely the province of a handful of academics. William Poundstone's Gaming the Vote: Why Elections Aren't Fair (and What We Can Do About It) is a welcome, though flawed, attempt to bring these important discussions to a popular audience.

Poundstone focuses on exploring the unintended consequences unleashed by the widespread use of the electoral system known as plurality voting. Under this system, each voter votes for one choice, and the candidate with the most votes wins, even if he or she has received less than a majority of the votes cast.

The problem with this method, as revealed both by mathematical models and by analysis of results from actual elections, is that it can lead to victory for the wrong candidate. If there are just two candidates, then a plurality is also a majority, and the method works fine. If candidate A is favored by 55 percent of the voters and B by 45 percent, nobody will disagree that A should win. But if candidate C enters the race and draws votes from some supporters of A, this hypothetical three-way race could result in A receiving only 40 percent of the votes, B still getting 45 percent, and C, the "spoiler candidate," obtaining 15 percent. Although a majority of all voters still favor A over B, B is declared elected under plurality rules.

B in fact is opposed by 55 percent of the electorate and yet has benefited from a dreaded defect of the "plurality wins all" method: the split vote, and the spoiler candidates who are blamed for causing it. Some voters soon figure out the rules of the game and, not wanting to waste their vote, begin voting for "the lesser of two evils"; others insist on their right to vote their conscience, even if doing so contributes to their least-favorite candidate winning. The voters and candidates with the most in common--the A s and C s in this example--end up bitterly divided.

The problem is a very real one, as Poundstone points out. The 2000 presidential election became the poster child for the defects of plurality voting when a center-left majority in Florida split between Al Gore and Ralph Nader, allowing Republican George W. Bush to win the state and the presidency. This split-vote dynamic has occurred in at least 11 percent of U.S. presidential elections. If cars or airliners had a failure rate that high, Poundstone pithily notes, they would be deemed "unsafe at any speed." In some U.S. Senate races in 2006, the candidacies of Libertarian Party spoilers resulted in incumbent Republicans losing their seats, allowing Democrats to take control of the chamber. So this is a nonpartisan dilemma; no major party is exempt from having votes siphoned off by a spoiler candidate.

The question at the heart of Poundstone's inquiry is, Which electoral systems will prevent spoilers and vote splitting, given that we don't want to ban independent and third-party candidates from the ballot? In the most interesting part of the book, he explores several electoral systems--approval voting, range voting, Condorcet voting, the Borda count and instant runoff voting, all of which allow voters either to rate or to rank multiple candidates. Reading descriptions of electoral systems can be about as much fun as watching paint dry. Poundstone, however, grounds his exegesis in personal stories and amusing anecdotes. He also manages to keep his account lively with an easy writing style and a sharp eye for illustrative narrative.

Poundstone does a competent job of weighing for his readers the pros and cons of the methods he highlights, but ultimately he goes awry in his analysis. The best solution to plurality voting's defects will be the method that most liberates voters to choose the candidates they truly like without fearing that honestly ranking or rating the ones who are their second and third choices might help to defeat their favorite candidate. Also, that method must do this not just in the mathematical models of the academic scholars on whom Poundstone relies as his primary sources of information but in the real world of rough-and-tumble politics, in which candidates and voters don't always behave as expected.

This is important, because these mathematical models and electoral paradoxes can break down in real-world situations. For example, approval voting certainly has virtues, such as simplicity: The voter simply "approves" as many candidates for a given office as he or she wishes by making a mark next to their names. A running tally is kept of the vote total, and the one who gets the most votes wins. Simple.

Or at least it would be if all elections were tepid affairs, as they often are in academic and private institutions that use approval voting from time to time. In those elections, the voters usually don't care much about the outcome. But elections for public office are a different matter. There, voters and candidates often care a great deal about outcomes; they are what is known as "self-interested advocates," and that leads to a paradox.

Let's imagine that we're back in the spring of 2008 and the Democratic Party is using approval voting to nominate its presidential candidate. And let's say that I am a voter who strongly prefers Barack Obama but also likes Hillary Clinton. I can "approve" both Obama and Clinton, maximizing my voter choice. Unfortunately, however, by checking the box for Clinton, I might help her defeat Obama. Obama knows that, so he is going to tell his supporters, "Approve only me." Clinton will tell her supporters the same thing, as will the other candidates. Consequently, approval voting in the real world will have a tendency to revert back to the dreaded plurality voting.

This is not merely a hypothetical consideration. Approval voting has an at-large cousin known as bloc voting, which is the most widely used electoral system in the United States, used for thousands of local elections. And bloc voting has had a strong tendency to encourage voters to engage in single-shot or bullet voting--the approval of only one candidate instead of several.

The system Poundstone favors most is range voting: Voters approve as many candidates as they like and award each a certain number of points by rating them--on a scale of 1 to 10, for example, with 10 indicating the greatest degree of approval. The winner is the candidate with the highest average score. Range voting is basically a more sophisticated version of approval voting, which means that a rating for your second- or third-favorite candidate can contribute to the defeat of your favorite. Range voting is used on many Web sites for rating everything from the sexiest bodies to the best films, and it works fine when voters don't care greatly about the outcome. But if range voting is used for public elections, once again smart candidates will urge their supporters to vote strategically by not rating other candidates--that is, to bullet vote. So range voting also would tend to regress to plurality voting.

In short, range and approval voting sound good in theory but have serious shortcomings that become apparent when one takes into account human psychology and the blood sport of politics, with their disincentives to honest voting. Poundstone describes the problems with approval voting in some detail, but in my view he doesn't do enough to emphasize them. And for range voting, he makes the claim that "no one seems (yet) to have found anything dreadfully wrong with it." He dismisses concerns about strategic voting with either method. His discussion of these methods would have benefited if, instead of relying only on academics and their mathematical models, he had sought some input from actual politicians and political consultants, the practitioners who deal with electoral choices every day in the real world.

The other systems Poundstone discusses all involve ranking candidates. The Borda count method has voters rank all the candidates from most to least preferred by putting numbers next to their names. The rankings from every ballot are added up for each candidate, and the one with the highest score wins. This sounds easy--but, once again, a vote for your second choice can help defeat your favorite. If voters are required to rank every candidate, they can manipulate the outcome by giving the lowest rankings to the strongest rivals of their favorite, as has sometimes happened in sports polls.

The Condorcet, or pair-wise, method uses voters' rankings to conduct a head-to-head vote among all possible pairs of candidates, and the candidate who beats all the others is the winner. This approach is good at ensuring that the winner has a broad base of support, but it can also lead to a "cycle" in which no candidate wins all the head-to-head matches. The methods for breaking such a cycle are very complex, and it's hard to imagine that these complexities would be tolerated in a public election.

The voting method that best liberates voters from the paradox that support for a lesser candidate can hurt one's favorite choice is known as instant runoff voting. This system elects majority winners in a single election by allowing voters to rank their candidates first, second, third and so on, and using the rankings to simulate a series of runoff elections. Initially, all first rankings are counted, and if one candidate has a majority, she or he is declared the winner. If not, the last-place candidate is dropped from the running, and each ballot on which the eliminated candidate had been ranked first now is counted for the second-ranked candidate on that ballot. All ballots are recounted, and if at this point someone has a majority, she or he wins; if not, the candidate now in last place is eliminated and the process is repeated, round by round, until someone reaches the majority threshold.

What sets instant runoff voting apart from the other systems is that your second or third choice can never help defeat your favorite candidate, because by the time your vote moves to your lower choices, your top choice has been eliminated from the race. So in real-world elections with self-interested voters and candidates, instant runoff is the method that best allows voters to sincerely express their preferences and overcomes the defects of plurality voting. According to experts such as Nicolaus Tideman, instant runoff voting also is less apt to result in strategic voting than is range voting.

But, as Poundstone points out, some experts find fault with instant runoff voting. In particular, they are troubled by what is known as the "winner-turns-loser paradox"--the possibility that in a close three-way contest the order of elimination can be affected by small numbers of voters who change one of their rankings, which in turn flips the result. But I find this critique unconvincing. It is based exclusively on mathematical models and not on real-world experience. Poundstone fails to mention that instant runoff voting is the only one of the methods he discusses that has a significant track record of use in public elections, going back decades in Australia, Ireland and elsewhere. There are data from thousands of elections in which to search for signs of the paradox, yet no evidence of it has so far been found.

It is important to note that, pros and cons aside, all five of the systems that Poundstone focuses on are better than plurality voting. And his book performs a public service by engaging a popular audience in these discussions. But the book has regrettable shortcomings that stem from what Poundstone left out.

Foremost among these omissions is Gaming the Vote 's failure to consider that when you select an electoral system you are also choosing a system of values. Do you want a method that will elect candidates who tend to be ideologically moderate? If so, the current plurality system is the worst, because extremists who have a strong core of supporters can win a low-plurality victory when the field is split among too many candidates. If you want to ensure that the candidate who wins will be someone who has broad-based support, then the Condorcet method works well. And instant runoff voting will produce winners who have not only a broad base of support but also a strong core of supporters. Electoral methods experts--such as Douglas Amy (the author of Behind the Ballot Box: A Citizen's Guide to Voting Systems [2000]) and staff members of the International Institute for Democracy and Electoral Assistance in Stockholm--recommend that you establish the values you are trying to instill in your democracy and then design the electoral system that will achieve those values. This perspective is a valuable foundation for the analysis of electoral methods.

Also, Poundstone limits his exploration to "winner take all" elections for a single office (governor, mayor, president or representative of a single-seat legislative district). He mostly ignores the group of methods known broadly as proportional representation, which is too bad, because an analysis of them would have shed even more light on the themes in the subtitle: "why elections aren't fair" and "what we can do about it." True, there is a certain logic to Poundstone's decision to focus on the type of election used overwhelmingly in the United States, but because of it, he never really confronts the defects and paradoxes that plague all "winner take all" electoral methods. The fact is, most elections in the United States are not even remotely competitive, with most legislative districts, indeed entire states, having become one-party "red" or "blue" strongholds in which three-fourths of the contests are won by landslide margins. But the problem that creates these noncompetitive races is the use of "winner take all" itself, not any particular variant of it. Perhaps Poundstone believes that poking at the underbelly of "winner take all" elections would be too vast an undertaking, but this omission makes his thesis less robust.

Nevertheless, for those concerned about the state of American politics, and for those who have never thought much about the role of electoral systems, Poundstone's effort is a useful introduction to the idea that we don't have to accept the flawed rules currently in use. If the United States is to bring its political system into the 21st century, it will be necessary for more Americans to ground their understanding of politics in an awareness of the impact of electoral systems.

1 comment:

  1. Steve Hill says:
    "The best solution will be the method that most liberates voters to choose the candidates they truly like without fearing that honestly ranking or rating the ones who are their second and third choices might help to defeat their favorite candidate."

    That's simply incorrect, Hill knows the reason why. Voters don't focus on supporting their favorite candidates -- they focus on supporting their favorite front-runners.

    This is explained in detail here.

    ReplyDelete