To the Spoilers Belong the Victors
A FairVote Innovative Analysis
Presidential elections wrongly decided due to plurality voting rules and "spoiler" candidacies since 1828: 5, possibly 7 (11%, possibly 15.5%)
Presidential elections wherein no candidate received a majority of votes since 1828: 16 (35.5%)
Plurality Leadership: The President of the United States is tasked with being the leader of the entire country, including those who did not vote for him or her. Unlike Congress, where different viewpoints are represented in one body (even if only one party happens to be in control at any given time), a president is in office because his or her side won, and the other lost.
It is because of the singular, zero-sum nature of the office that it is so crucial that the person chosen to fill the position is someone who truly is the choice of the majority of Americans; it is far more difficult to claim a popular mandate when most of the country voted for someone else. The principle of accountability is lost when a majority in fact would have rejected the "winner" in a two-person race.
Unfortunately, our current system does nothing to ensure majority support for White House occupants. Instead, states have structured their presidential elections - and indeed most of their elections - to allow candidates to win their state with a mere plurality of votes. When more than two people run for office, the system breaks down.
The result is a system that is dangerously susceptible to so-called "spoiler" candidacies, often causing candidates to prevail who are opposed by a majority of voters, and who may even hold positions that are the polar opposites of the "spoiler" candidate. Thanks to the Electoral College, this is true even with "spoilers" who only run or campaign in a handful of states. With the memory of Ralph Nader in 2000 fresh in our minds, along with the looming semi-candidacies of Michael Bloomberg and others in 2008, the spoiler issue has never been more salient or explosive. It is also nothing new.
A History of Spoiled Presidential Elections: Inspired by William Poundstone"s upcoming book Gaming the Vote: Why Elections Aren't Fair (and What We Can Do About It), we took a look at spoilers throughout presidential election history, and the numbers are quite telling. Since 1828 (about the time it became the established norm for all states to hold popular elections), the outcomes of at least five presidential elections were likely wrongly decided due to the presence of a spoiler candidate - and the number may in fact be as high as seven.
- In 1844, slave-owning James K. Polk defeated Henry Clay due to Clay losing votes in key states to the more strongly-abolitionist candidate James Birney.
- In 1848, former President Martin Van Buren split the vote with fellow Democrat Lewis Cass to elect Zachary Taylor.
- In 1884, two minor candidates took a combined 3% of the vote in a race where Grover Cleveland defeated James Blaine by a tiny margin of 0.3%.
- In 1912, Teddy Roosevelt entered the fray on the Bull Moose ticket, splitting 50.6% of the vote withfellow Republican William Howard Taft and electing Woodrow Wilson.
- In 2000, consumer advocate Ralph Nader won 181 times more votes than the decisive Florida victory margin for corporate-friendly Republican George W. Bush.
The other two cases in question are stark, if not conclusive. In 1892, a close election between Grover Cleveland and Benjamin Harrison was shaken up by Populist James Weaver, who took a sizable 8.5% of the vote while Cleveland defeated Harrison by 3%. A century later, George H. W. Bush was out of a job after billionaire Ross Perot took almost 19% of the vote. Pundits still scratch their heads over whether he handed Bill Clinton the presidency, but what"s certain is that Clinton only won a single state (Arkansas) with a majority of the vote.
Non-Majority Winners: Regardless of how one characterizes these outcomes, certainly the concept of majority rule is not being served. Even putting aside the spoiler issue, 16 out of 45 presidents have been elected with less than 50% of the vote since 1828. That means in over one third of those elections, most of the country voted for someone other than the winner. In fact, as of 2004, 14 governors were in office under the same circumstances.
All of this confusion and all of these "wrong-winner" elections are due to our first-past-the-post plurality system of electing office holders. Remember, in a three-way race, it takes as little as 34% to win, but it also means that 66% voted against you.
The Solution: If we want our elected leaders to have the political legitimacy of majority support, we need to have a system that allows for it. If we want to reap the benefits of independent candidacies -(such as a greater range of issues being discussed, a wider range of choices, and an increased level of discourse) we need a system that does not turn them into spoilers.
Most nations elect their presidents with two-round runoff systems based on the majority principle, but what might work best in the United States is instant runoff voting (IRV), an increasingly popular voting method that allows voters to rank their choices in order of preference and ensures a majority winner. By ranking candidates in order of preference, it becomes possible to vote one"s conscience while still lending support to a candidate with a more realistic shot at winning. With IRV, no one plays the spoiler, and one candidate always emerges with majority support.
States are empowered to adopt instant runoff voting for president, and states where this has been seriously debated range from Vermont to Alaska.
Read more about instant runoff voting at: instantrunoff.com or fairvote.org/irv. Previous editions of Innovative Analysis can be found here.
Comments currently closed for To the Spoilers Belong the Victors
-
Posted by Clay Shentrup, 2007-12-12 13:56:04 (5 years ago)
-
Jack: In an earlier comment here I stated that the more widely used 'traditional' approach to evaluation single winner election methods was to look at properties and whether methods pass or fail on these properties. Are you a subscriber to this approach? If no, please let us know how you evaluate election methods. If yes: * Which properties do you consider import and why? * Which property failures of IRV on properties that you consider important do you consider forgivable because they are minor failures.
Posted by AllAboutVoting, 2007-12-05 15:21:49 (5 years ago) -
Jack, I don't think you can say that 20% is "not likely". You talk about different "universes" with regard to Smith's simulations, but ignore that his simulations used 720 different universes, in which parameters, such as the proportion of strategic vs. honest voters, were gradually moved from 0% to 100% - and Range Voting surpassed IRV in ALL of them. In fact, IRV was one of the worst. Such a wide margin of victory across such a wide range of conceivable "universes" supports our confidence in these figures. Aside from your vague incredulity, you offer absolutely no scientific/mathematical criticism of the results, nor any alternate contradictory results. Our simulation code is open source, and freely available if you'd like to critique it, or make it more realistic. But until you do that, your non-expert hand waving does not qualify as a scientific refutation of anything. I don't care about your "faith", I care about what evidence you can actually present. So far you present nothing but unqualified opinion. And in fact, if you don't trust these utility figures, then you have no way to suppose IRV is superior to any other method. You would have to see all methods as equal in quality, and so the only distinguishing feature would be simplicity and ease of adoption. Well, in that case, I'm happy to be the devil's advocate and "agree" with you. Let's implement Approval Voting, since it's radically simpler and cheaper to adopt than IRV - and makes it safe to always vote for your favorite candidates, so would plausibly dismantle duopoly, unlike IRV.
Posted by Clay Shentrup, 2007-12-05 15:15:11 (5 years ago) -
In response to AllAboutVoting's question about simulations: Models have their applications. I don't object to the simulations, per se. Rather, my point is that they should not be confused with the real world. Put differently, an automaton with computerized preferences is not a living, breathing voter. With respect to judging the relative "merits" of voting systems: the question is a normative one. The answer, which will be subjective, depends on the metric used, which also will be subjective. Social utility efficiency is a subjective metric. It follows that IRV is not conclusively inferior to range voting. That contention is just an opinion.
Posted by Jack, 2007-12-05 03:28:48 (5 years ago) -
An aside on comment formatting: This is at least the second comment I submitted to the FairVote blog that has rendered goofily. In both cases I believe that the problem was a 'hash' symbol that rendered very strangely. It would be nice if the blog administrator either: * fixed the rendering bug * added a 'preview' button for previewing comments or * added the ability to edit a submitted comment for a few minutes after the comment is submitted
Posted by AllAboutVoting, 2007-12-04 22:29:45 (5 years ago) -
>[Clay]Plus, IRV (which FairVote vigorously advocates) >can easily elect X even when Y was preferred to X by a >huge majority. So IRV also is not ???majoritarian???. >[Jack]The criticism that IRV can ???easily??? elect X >when Y is preferred is not a fair one. That outcome >is possible but not likely. This is a disagreement about the severity of a property failure of an election method. Evidence that it is a severe failure: #15 from http://rangevoting.org/rangeVirv.html attempts to address the issue of whether this problem is significant or not. The response points out that to reliably determine whether this problem is occurring one actually needs to see each voter's ranking. Usually elections do not provide this data so usually this analysis cannot be done in a straight forward manner. The few elections / polls that RV has found do suggest that the problem is significant. (Aside: The page references a range voting exit poll with 36 voters that Clay conducted. The rankings were inferred from the candidate orderings. I do not consider this poll to involve enough voters to yield any meaningful results. I also do not consider inferred rankings from a poll using range voting to be at all meaningful for analyzing IRV votes.) Aside: Is FairVote aware of other IRV elections where the individual votes may be examined? Evidence that it is a non severe failure: ?? (Jack: Can you provide support for your statement that the 'outcome is possible but not likely'?)
Posted by AllAboutVoting, 2007-12-04 22:06:04 (5 years ago) -
>I do not share your faith in the concept of social >utility efficiency Fair enough. There are two approaches to analyzing the quality of an election method that I am familiar with. One is to define desirable qualities and to note which methods meet which properties. In my view, IRV does very poorly by this approach. This approach's main flaw is that the severity of failures is not expressed nor is there general agreement about which properties are important and which ones are unimportant. Warren's approach was to instead measure how satisfied simulated voters are by the results of elections after voting under different methods using different strategies. Measuring this by necessity involves simulated voters but the idea is that election methods that satisfy simulated voters best would also satisfy real voters. The problems with this approach is that the case must be made that the quality measure is the correct one and that the simulated elections are reasonable. I believe that Jack objects to using simulations of voter behavior in general and not to the specifics of whether simulated elections closely enough model real voters. Please correct me if I am wrong.
Posted by AllAboutVoting, 2007-12-04 21:49:19 (5 years ago) -
The criticism that IRV can "easily" elect X when Y is preferred is not a fair one. That outcome is possible but not likely. Smith's conclusions about range "maximizing social utility efficiency" are based on the output of choice theoretic computer models. In those models, preferences are artifically ascribed to simulated voters, and the "utilities" of election outcomes are somehow quantified. In certain universes, that methodology sufficiently accounts for what outcomes people will 'like.' In others, it does not. Model specification, no matter how rigorous, is not real politics. Warren Smith's admitted hard work notwithstanding, I do not share your faith in the concept of social utility efficiency or its operationalization in abstract models.
Posted by Jack, 2007-12-04 20:38:36 (5 years ago) -
A view of ???wrong-winner??? whereby it is desirable for the majority-preferred candidate to lose is out of step with convention. But today???s novel ideas can become tomorrow???s standard currency. So sally forth with Bayesian regret as a metric for election fairness! Well, it's not an issue of popularity, or contemporariness. It's an issue of mathematical proof. The majoritarian idea, that X is more utilitarian than Y if preferred to Y by a majority, is obsolete. It has been disproved, decades ago. We have looked at (and vigorously debated over) various alternative social utility functions, such as Pareto optimality (which essentially equates to Condorcet-ness). Also, the book mentioned above, Gaming the Vote, apparently supports Range Voting. Also, IRV is very non-majoritarian. See for example: http://rangevoting.org/CoreSupp.html
Posted by Clay Shentrup, 2007-12-04 14:20:16 (5 years ago) -
The "majority criterion" is flawed, since it can be mathematically proved that Y can be preferred by a majority over X, and X can still be the social utility maximizer. Condorcet ambiguities easily demonstrate this. Plus, IRV (which FairVote vigorously advocates) can easily elect X even when Y was preferred to X by a huge majority. So IRV also is not "majoritarian". The bottom line is social utility efficiency, or Bayesian regret as it can alternatively be expressed.
Posted by Clay Shentrup, 2007-12-04 14:11:42 (5 years ago) -
Back on topic to the "wrong way" elections. Discussion about this page and some fact checking attempts can be found here and here.
Posted by AllAboutVoting, 2007-09-08 10:31:41 (6 years ago) -
Context also matters. IRV supporters often sell IRV as a way to get rid of run off elections which cost money and tend to get poor turnout. But preference ordering during the general election may be very different in an an actual runoff where the circumstances are different: * there are now only two candidates instead of many * the press and the voters can scrutenize these candidate more carefully Consider the election of Ed Jew as San Francisco suprevisor representing the Sunset District. Quoting from the SF Chronicle: In the old days, Jew would have been forced into a runoff against Ron Dudum, the candidate who was just 53 votes behind, with 25.96 percent of the vote. Jew's belated switch from Republican to Democrat and questions about his family residence in Burlingame surely would have been significant issues in a one-on-one race against a lifelong Democrat and lifelong Sunset resident. "I always thought if Ed and I were in a runoff, I would have had a good chance," Dudum said in a interview last week.
Posted by AllAboutVoting, 2007-09-08 10:26:13 (6 years ago) -
I endorse the page that Jack references in the above comment. Also from that page is the Lomax 'Pizza example': A group of people, let's say there are three of them to make it very simple, want to buy a pizza and split it. The pizza parlor, amazingly enough, will only allow them to choose one kind of pizza. Two of the three prefer pepperoni, but the third is a Jew who cannot eat pepperoni, but, we'll say, would have no problem with cheese or any vegetarian pizza. Now, if, for some reason, the two pepperoni fans must have their pepperoni, they certainly can get it. They might have some problems, though. The Jew might balk at paying for it. Suppose that the pepperoni fans, though, can agree that mushroom would be just fine, almost as good as pepperoni. And that just happens to be the favorite of the Jew. Please give a cogent argument why the first preference of the majority should win. The point is that preference strength matters.
Posted by AllAboutVoting, 2007-09-08 10:18:22 (6 years ago) -
Range voting is advocated for its ability to produce wrong-winner elections, FYI. I cite: This illustrates the failure of the "majority criterion" by range voting. However, in such a situation we would argue that it is good that Y won and it is good that range voting found a way to evade the "tyranny of the majority." Indeed this is an advantage of range voting that all other common voting method proposals cannot match. A view of "wrong-winner" whereby it is desirable for the majority-preferred candidate to lose is out of step with convention. But today's novel ideas can become tomorrow's standard currency. So sally forth with Bayesian regret as a metric for election fairness!
Posted by Jack, 2007-09-05 21:17:31 (6 years ago) -
Warren Smith has a page at rangevoting.org about "wrong way" elections in the US and outside. Important elections whose results (probably) would have differed under different voting systems
Posted by AllAboutVoting, 2007-08-31 00:31:34 (6 years ago)
Jack, Some remarks. ..an automaton with computerized preferences is not a living, breathing voter. No one's arguing with you there. But what is your overall point here? The utility distributions used are very realistic, and some utility generators in the program actually were based on real election data. Plus, there were hundreds of election models used, and hundreds of thousands of simulated elections were averaged together to get the results for each model. Range Voting handily surpassed the common methods such as IRV in all of them. So your vague assertion that the simulations were so unrealistic that we can just write them off entirely just won't fly. With respect to judging the relative ???merits??? of voting systems: the question is a normative one. The answer, which will be subjective, depends on the metric used, which also will be subjective. Social utility efficiency is a subjective metric. Well, no. Individual preferences are subjective. Some people like candidate X and hate candidate Y, while others have fairly neutral feelings about both. Social utility efficiency is an objective measure of how well a voting method satisies voters' subjective preferences. Understanding this is essential to any serious discussion on voting methods. Now, maybe you think that the social utility is not merely the sum of individual utilities; but some other function entirely. I can relate. I have, in the past, proposed alternatives, such as: "MaxiMin" - the best candidate is the one who maximizes the minimum individual voter's utility. Problem: a single voter can effectively become a "dictator". "Utility Bar" - the social utility of candidate X is the maximum product of U, and the number of voters for whom X produces at least U personal utility. Problem: Explained here (warning: MATH) Simply put, the invalidity/inconsistency of (all) alternatives to additive utility is essentially proven.