Why I Prefer IRV to Condorcet

by Greg Dennis // Published November 22, 2008 in gregdennis.com

If you consider yourself a supporter of Condorcet voting methods, then you've clearly put some critical thought into electoral reform, thought which has driven you to support a very good single-winner voting method. We need more people to think as seriously about alternative voting systems.

That said, I'd like to explain why I believe Condorcet methods, while very good single-winner voting systems, place a close second behind Instant Runoff Voting. As you probably know, Arrow's Impossibility Theorem proves that no voting system satisfies a reasonable set of fairness criteria, even under the assumption that voters mark their preferences honestly on the ballot. You may also be familiar with the Gibbard-Satterthwaite Theorem, which proves every system to be susceptible to some form of strategic voting. (While these theorems don't technically apply to systems that don't involve ranking, like range and approval voting, these systems do violate numerous fairness criteria nonetheless.) Choosing a voting system, therefore, requires balancing the relative pros and cons of what are ultimately imperfect options.

As I attempt to demonstrate below, IRV not only elects Condorcet winners regularly in practice, but in a competitive electoral environment, IRV may actually elect true Condorcet winners more often than Condorcet itself. Furthermore, I argue IRV is more politically feasible, prevents obscure candidates from winning, and increases the political and technological feasibility of proportional representation and other forms of preferential voting, including Condorcet itself. Even if this essay fails to convince you that IRV is superior to Condorcet, I hope it at least persuades you to support IRV campaigns where they arise in the United States.

Reason #1: IRV has a track record of electing Condorcet winners

First and foremost, IRV eliminates the most common violation of the Condorcet criterion --- the so-called "spoiler" scenario --- where the presence of a candidate with little core support causes a Condorcet winner with strong core support to lose. Under IRV, the candidate with little support is eliminated in an early round of tallying, allowing the deserving Condorcet winner to beat another front-runner in the final round. By eliminating this "spoiler" dynamic, IRV accommodates more parties and candidates who cannot yet gain enough votes to win. Having more choices from across the spectrum deepens political discussion and allows new parties to disseminate their message and grow (or not) organically over a period of years.

Admittedly, there is another situation similar to the spoiler problem --- the "center squeeze" scenario -- in which IRV may fail to elect the Condorcet winner. In this scenario, the presence of a candidate with strong core support causes a Condorcet winner with little core support to lose. Fortunately, despite the theoretical possibility of this scenario, the empirical evidence suggests that it is vanishingly rare in practice. In fact, if you look at the ballot data that is publicly available for the IRV elections held in the US (San Francisco, CA, Burlington, VT, and Pierce County, WA), for instance, you'll see that IRV and Condorcet agree on the winner every single time.

Lacking sufficient examples of real elections in which IRV has failed to elect the Condorcet winner, a few IRV critics have resorted to using top-two runoff elections in which the Condorcet winner lost as evidence of IRV's center-squeeze problem. However, top-two runoff and instant runoff are different systems that can produce different results, so this "evidence" is hardly convincing. Given the hundreds of public IRV elections that are conducted worldwide every year, if the center-squeeze scenario under IRV were so common, it would presumably be easy to find a good percentage of actual IRV elections that exhibit it. This is not to say that the center-squeeze will never happen, but it is to question why so much hand-wringing is devoted to a problem that is so rare in practice.

Conclusion: IRV will elect Condorcet winners more often than plurality, and cases in which it fails to elect them are rare.

Reason #2: IRV is less complicated than Condorcet

The one advantage that Plurality has over IRV is its simplicity. Fortunately, voters don't need to know how to tally an IRV election in order to vote in one, in the same way the average person can drive a car without understanding what's going on under the hood. Still, the simpler the system, the easier it is to explain to those who do wish to look under the hood, and the more trustworthy and transparent the system appears to the voting public. Plus, the simpler the system, the greater the political feasibility of enacting it.

The basic Condorcet idea may sound simpler than IRV at first; that is, until you realize that pairwise elections can result in a cycle and thereby fail to produce a clear winner. Any usable Condorcet method must, therefore, have a backup plan to resolve such cycles. The need for a backup plan has lead to quite complicated Condorcet systems, perhaps the most popular being Schwartz Sequential Dropping, which are ultimately more opaque and difficult to explain than IRV. The prospects of enacting such a system in the U.S. are dim.

Conclusion: The complexity of Condorcet renders it less transparent and less politically feasible than IRV.

Reason #3: IRV is less susceptible to intuitive strategies

Unlike IRV, Condorcet is susceptible to intuitive voting strategies. Condorcet suffers these strategic problems because it fails the Later-no-harm and later-no-help criteria, which means ranking additional candidates can affect whether earlier ranked candidates win or lose. As we shall see, satisfying these criteria, as IRV does, is crucial to encouraging complete, honest ranking of ballots in a competitive electoral environment.

In a race with two major front-runners, one obvious strategy, known as burying, occurs when a voter who prefers one front-runner dishonestly ranks the other front-runner last. To illustrate, first consider the following honest rankings:

  46: A > B > C
  44: B > A > C
   5: C > A > B
   5: C > B > A

In this scenario, Condorcet, IRV, and Plurality all agree that A is the winner. Suppose, however, that B's voters, in an effort to manipulate the result, decide to dishonestly rank A last. While IRV and Plurality would still elect A, Condorcet methods would find a cycle, so it would be up to the particular resolution strategy to choose the winner. Now consider a case where B's supporters again bury A, but A's supporters bury B as well. Now Condorcet elects C, who according to the honest rankings is the Condorcet Loser, i.e. the absolute worst choice. IRV, being fully resistant to burying, continues to elect the true Condorcet winner A. Thus, in the face of rampant burying, IRV may, ironically, elect true Condorcet winners more often than Condorcet itself.

Particularly worrisome is that the campaigns themselves may encourage burying. It might begin with B's campaign quietly encouraging its supporters to bury A. If A's campaign gets word, they may encourage its supporters to bury B. In the end, both campaigns may engage in a kind of high-stakes game of chicken that could result in their both losing to C. Professor Burt Monroe points out this problem in his paper "Raising Turkeys" (PDF) -- in my example, the "turkey" candidate C is "raised" by supporters of A and B. Monroe invents a criterion called "Nonelection of Irrelevant Alternatives", which a voting method passes if, when the voters all act strategically in their individual self-interest, the method cannot elect the voters' least preferred choice. IRV passes this criterion but Condorcet methods do not.

In addition to burying, Condorcet may lead to increased bullet-voting, the strategy whereby the voter ranks his or her top choice and leaves the rest of the ranks blank. Bullet voting under Condorcet may increase the chances one's first choice will win. To illustrate, consider these honest rankings:

  46: A > B > C
  44: B > C > A
  10: C > A > B

Condorcet results in a cycle in this case, but if A's supporters bullet vote instead, then A wins. This isn't a bad outcome, per se, because A looks like the deserving winner and would win under IRV as well. However, because bullet voting can help one's top choice under Condorcet, bullet voting may become more prevalent. In a close three-way race, for example, this may cause a return of the spoiler problem. For example, consider these honest rankings:

  18: A > B > C
  17: A > C > B
  19: B > C > A
  14: B > A > C
  18: C > B > A
  14: C > A > B

Candidate B is the both the IRV and Condorcet winner in this case. But voters using Condorcet, aware that bullet voting may help their top choice and unable to predict the exact results in advance, may nevertheless decide to bullet vote. If they do, then Condorcet elects candidate A, even though A is the Condorcet Loser. Thus, if Condorcet necessarily leads to bullet voting in practice, then IRV may, again, elect true Condorcet winners more often than Condorcet itself.

Furthermore, any increase in bullet voting gives ammunition to those who would seek a return to Plurality. "If so few people are voting beyond the first rank, do we really need the rest?" a Plurality advocate may claim. Indeed, the prevalence of bullet voting sped the repeal of Bucklin voting (a ranked voting method, which also failed the Later-no-harm criterion, used in several state party primaries early in the 20th century). Widespread use of multiple ranks is important to the political sustainability of any preferential voting system.

As mentioned, every voting system is theoretically vulnerable to strategic manipulation, and IRV is no exception. However, strategies that help under IRV are are largely counter-intuitive and may very well backfire -- actually causing the defeat of the candidate the strategizer sought to help -- so they can't be universally applied to any election the way burying can be employed under Condorcet. Given their increase complexity and chance of backfiring, they are unlikely to be advocated by the campaigns themselves. It is not surprising, then, that there's little evidence that voters strategically manipulate IRV elections to their benefit in practice.

Conclusion: Under Condorcet, voters are more likely to engage in strategic manipulation that leads to results poorer than those delivered by IRV.

Reason #4: IRV ensures that we know where the winner stands

Imagine an upcoming election between two well-known candidates, A and B, who observers expect to both engage in mud-slinging and dirty tactics. If the election were to use a Condorcet method, an enterprising but little-known candidate C might decide to throw her hat into the race, but otherwise lay low and avoid revealing any potentially alienating policy positions throughout the campaign. With a little luck, and "anybody-but-that-other-candidate" thinking by voters, candidate C may see the following results on election night:

  49: A > C > B
  48: B > C > A
   2: C > A > B
   1: C > B > A

The supporters of A and B, feeling disdain for each other's candidate and lacking any reason to think negatively of C, place C second on their ballots. Note that this is subtly different than the "burying" discussed in Reason #3. In this case, supporters of A and B aren't dishonestly burying; rather the candidate they know nothing about truly seems more appealing at the moment to a candidate they actively dislike.

A theoretician may look at the results of this election and presume that C would win in a head-to-head match-up with either A or B, but I challenge the conclusion. Although laying low could prove a winning strategy in a Condorcet race, it would be unlikely to succeed in a head-to-head match-up where a candidate would have to make her positions known and earn a majority of support to emerge victorious. Candidate C may hold fringe positions, have a history of corruption, or have other unappealing characteristics that would be revealed under the limelight of a head-to-head race.

To win under Condorcet, a candidate need not give a good reason to vote for her; not giving voters a reason to vote against her can suffice. The latter can be accomplished by speaking in platitudes, refusing to take clear positions on issues, and generally not making one's views known to avoid alienating anyone. Indeed, a candidate who garners zero first choice support may win under Condorcet. Such a campaign strategy could be detrimental to our political discourse, leading to poorly-informed voters and depressed voter turnout.

To win IRV elections, in contrast, candidates must distinguish themselves and make their positions clear, so as to attract a significant amount of core support. While the strategy of saying little of substance may prove successful under Condorcet, such a candidate would likely be eliminated in an early round of IRV.

Conclusion: By requiring the winner to garner a level of core support, IRV ensures the winner is someone who made his or her positions clear.

Reason #5: IRV makes Condorcet feasible

Condorcet and IRV share the same hurdles any preferential voting system would face on the road to implementation. These include the need for user-friendly preferential ballots, voting machines capable of capturing these preferences, and voters comfortable with the whole idea of preferential voting. IRV and Condorcet advocates can work jointly in this regard, leaving it to the future to decide between which tabulation rules should be applied to the ranked ballots, since both systems are superior to plurality and two-round runoffs.

IRV is currently in place in many jurisdictions around the U.S. and the world, and the "center-squeeze" effect has yet to generated serious concern in practice. Though if it does, and if a jurisdiction currently using IRV wishes to experiment with Condorcet instead, there will be far fewer obstacles in its way. Maybe my concerns about strategic voting and substance-less campaigns will come true, or maybe they won't. But enacting IRV now makes it more likely that Condorcet experiment can take place in the future.

Conclusion: Enacting IRV removes technological and political hurdles to Condorcet and preferential voting generally.

Reason #6: IRV is the best stepping-stone to proportional representation

IRV, Condorcet, and all systems for electing single-winner offices (such as a mayor, governor, president) are "winner-take-all" systems. Many electoral reformers, myself included, believe no winner-take-all system is ideal for electing a legislative body, such as a city council, or state or national legislature. In those cases, some form of proportional representation would better reflect the diversity of the public. Proportional representation ensures that the majority of voters will elect the majority of the legislative seats, but also, that significant minorities will be able to elect their fair share as well, in proportion to their support among the electorate.

Thus, improving our single-winner elections is ultimately only a small step in ensuring our electoral system is competitive and fully democratic, and our ultimate goal should be some form of proportional representation for our legislative bodies. To that end, one of the great advantages of IRV as a voting system is that it is just a special case of a more general system for proportional representation called the Single Transferable Vote or "Choice Voting". Praised by activists and academics alike, Choice Voting is the only system of PR used in the world today that is preferential and the only one that can be used in both partisan and non-partisan elections.

The synergy between IRV and Choice Voting is evident by looking at electoral systems around the world. The use of IRV in Australia set the stage for that country's later adoption of Choice Voting for its upper house. Ireland, too, uses both IRV and Choice Voting for its elections. Similarly, the IRV legislation passed in Minneapolis also enacts Choice Voting for the city's At Large commissions. Due to this synergy, IRV serves as a great stepping-stone to proportional representation in the United States.

While there do theoretically exist proportional representation systems that become Condorcet in the single-winner case, such as CPO-STV, they are exceedingly complicated, and to my knowledge, have yet to be used for any election -- public or private -- in the world. Even the inventor of CPO-STV, Professor Nicolaus Tideman, does not take a position on whether the added complexity of his method is worth its "cost of manageability." Enacting such a complex system would, in my view, be politically infeasible.

Conclusion: IRV makes a good single-winner stepping-stone to a feasible system of proportional representation, but Condorcet does not.

Reason #7: IRV has political momentum

Among proposals to reform the way we elect single-winner offices, Instant Runoff Voting is by far the most popular and has gained significant traction in states and cities around the nation. Currently, at least a dozen jurisdictions in the United States use IRV or Choice Voting and at least nine implementations of IRV are pending, meaning the U.S. should have at least 21 IRV implementations by 2010, more than half of which have been adopted in the years since 2002. Given the glacial pace at which electoral reform usually moves, these numbers are a sign of impressive political momentum. Although Condorcet is used by a handful of technical organizations, it is not currently used for any public election in the world.

I welcome Condorcet voting advocates seeking to persuade voters and elected officials to support their preferred system over plurality elections and runoff elections, just as IRV advocates do. But efforts to adopt IRV are far more likely to be successful. Rightly or wrongly, the fact that Condorcet could elect a candidate who would come in dead last in a traditional American plurality or runoff election essentially makes it a political non-starter - both from the perspective of ordinary voters and of elected officials. In contrast, IRV has a widely-accepted analog: the traditional runoff election that Americans use in many elections and that most nations use around the world when electing a president. The fact that many people are familiar with runoffs and appreciate their logic helps IRV advocates fend off attacks by opponents.  Indeed, recent wins at the polls, growing support from civic leaders and editorial writers and activity in half of our state legislatures indicate IRV's viability.

Conclusion: Even if you disagree that IRV is superior to Condorcet, I hope you find IRV to be a politically-viable system worth supporting, so that we may continue to make our elections in the United States more competitive and democratic.