APPENDIX 4
SOME FURTHER THOUGHTS ON IRV
September 21, 2010
Is IRV a Better Alternative Than Plurality Voting?
In the US, and in the UK, Instant Runoff Voting (IRV), or the alternative vote as it is called in Britain, is touted by the Center for Voting and Democracy (FairVote) in the US and by the Electoral Reform Society in the UK as a better alternative to voting than plurality voting. Both maintain that it would be fairer to both voters and third parties.
In one sense, IRV has a clear advantage over plurality voting in that it would, to quote William Poundstone in a paper subsequent to his book "Gaming the Vote", "fix the most glaring fault of the plurality vote system--namely, third-party spoilers unfairly tipping the race from one major party to the other." Thus under IRV in the 2000 Florida presidential election, Nader's votes would have been dropped and his voters' second choices reallocated to the other major candidates, surely overwhelmingly to Gore who would have won under that system. But the same would almost certainly have been true had approval voting been used instead in Florida in the 2000 presidential election.
Outside of this, however, it is not so clear that IRV would be an improvement over plurality voting and in some crucial ways may even be clearly worse. For one thing as Gavin Thompson put it: "The Australian House of Representative has been elected using AV (our IRV) since 1918. Election outcomes have proven only slightly more proportional than in the UK..." where plurality voting is used for parliamentary elections. (see "Keeping things in proportion: how can voting systems be fairer?" in the September 2010 issue of the popular statistical magazine Significance published by the UK's Royal Statistical Society) Or as Warren Smith, the advocate of range voting, stated it in a broader context "IRV still leads to massive 2-party dominating (has in every country it has been tried in at a national
level) in IRV seats and still would, I have no doubt, in the USA."
There are some convincing clues as to why this is the case in William Poundstone's paper "A Test Drive of Voting Methods" published after his book on "Gaming the Vote." In that paper Poundstone focuses on result of work by Ka-Ping Yee simulating voting in three and four candidate elections in which the results for plurality elections and IRV turn out to be remarkably similar. Poundstone's explanation for this is that while, as we have seen, IRV may get rid of the spoiler role, it does not get rid of vote splitting in elections processes which under plurality voting does much to engender a two party system.
According to Poundstone this is "because IRV can penalize popular independents, much as the plurality vote does." Whatever the case, though, it is amply clear from the empirical record that IRV does very little if anything to open up the political process in the voting system to third parties.
Aside from this, however, IRV has a serious flaw that even plurality voting does not have as I pointed out in my June 22, 2010 piece on the Guardian Newspaper UK website on approval voting. The system lacks, in technical terms, monotonicity. That is, as the numerical example in the piece clearly demonstrates, it is possible under IRV for a candidate, set to win an election, to wind up losing despite gaining more support or first place votes in the course of a campaign.
Conversely, a candidate, set to lose an election, can wind up winning despite losing support in the course of a campaign. The upshot of this is that in such a situation, given the opacity of the system of IRV, a voter cannot know for sure what the impact of their intended vote will be. Even plurality voting with all its flaws does not experience such a difficulty since more support for a candidate, set to win, will only enhance that candidate's prospects while less support for a candidate, set to lose, will always worsen the candidate's prospects.
Proponents of IRV, like the Center for Voting and Democracy and the Election Reform Society, argue that this non-monotonicity problem is so rare as to be of no real significance. Some even suggest that such a situation might occur only once in a century. Because of the opacity of the system of IRV and because usually only results are published rather than detailed data on how they were arrived at, there is little information available on this issue. But the information that is available strongly suggests that the problem is far more serious than supporters of IRV recognize. Burlington, VT is a case in point. There have been two IRV elections held in Burlington for mayor, the latest in 2009. In that election as analysis by Warren Smith and Joe Ornstein clearly shows non-monotonicity was present.
Work that Ornstein has done indicates that in close multi-candidate elections, those for which it is claimed IRV is particularly suited, the incidence can range from 13% of such elections to 43%. The upshot of this is that in such elections, voters cannot know for sure the degree to which giving one's favorite candidate a first place vote will help or hurt that candidate. For a similar problem to take place under plurality voting one would need to have outright vote stealing in 13% to 43% of such elections. Viewed this way, it should be amply clear that the problem is not a minor one and ought to be avoided to much of the same degree that we try to avoid vote stealing in our plurality elections.
The opacity of IRV as a voting system is a serious problem in two senses. First, even with full disclosure of the voting data from an IRV election it is not easy for most voters to discern the monotonicity problem. Besides the difficulty is really an ex ante, not an ex post, problem for voters. That is, before deciding how to vote they need to know what impact their votes will have on influencing the outcome, not after the fact.
Second, Sir Michael Dummett, a prominent British philosopher who has done widely respected work on voting systems, has pointed out that in any voting system there are always cases where voters can gain a strategic advantage by voting tactically. This makes a great deal of sense.
Voters can best reflect their preferences in terms of the outcome if they can vote both sincerely and strategically at the same time. The distinct disadvantage of plurality voting is that it forces voters to choose to vote either sincerely or strategically, hence the wasted vote syndrome. But at least in doing this, the voter knows what the likely outcome of his or her decision on influencing the elections results can be. Thus in the 2000 Florida presidential elections Nader supporters who actually detested Gore and consequently decided to vote tactically for Bush instead would surely have been effective in that contest in terms of their preferences to block Gore.
The problem with IRV is that because of its very opacity, strategic voting on behalf of a candidate is difficult for an individual voter both to discern and implement. As a result, Samuel Merrill in his book on "Making Multi-candidate Elections More Democratic" suggests that the incidence of "manipulation," or tactical voting by individual, is low in IRV elections. But the anecdotal evidence from Australian elections seems to indicate that tactical voting by individual electors is replaced by "How To Vote Cards," lists handed out by individual parties advising on how to vote for second and third places on the IRV ballots their voters cast to favorably effect the outcome. Thus, while IRV at an individual voter's level may avoid "manipulation" on an aggregate level it may lead to collusive behavior which may be even more pernicious than what could happen under plurality voting once the field of candidates is determined.
Given these considerations, it is hard to see how one could make a very cogent case against those who would argue that between plurality voting and IRV the latter is a worse system. But in any case approval voting, a system in which in multi-candidate elections voters are allowed to give one vote each to the candidate or candidates they support with the candidate having the most votes winning, is surely superior to both. Approval voting not only gets rid of the wasted vote but it also gets rid of the spoiler role and vote splitting.
These are surely a necessary, though not sufficient condition, for putting third parties on a equal level with the two major parties. In this sense, third parties in countries with single member constituencies as we have for the British Parliament and the American Congress, ought to support approval voting over the other two systems discussed here. Even if some of the arguments in this essay may have weaknesses, the historical record of IRV in opening up political processes to third parties suggests that this is sound advice.
September 7, 2010
50% + 1 Is a Mirage
The major advantage claimed for instant runoff voting, IRV or the alternative vote in the UK, is that it always assures that, absent a tie vote, the winning candidate wins by at least a vote of 50% + 1, or a majority. Unfortunately, the very opacity of the voting system hides the fact that this claim is bogus. The problem here is that data in IRV elections on how the results were arrived at are usually not made available; only the final results are made available.
The 2009 IRV election for mayor of Burlington, VT is an exception to this because the underlying voting pattern was made available in addition to the final results. In that election, in the penultimate round of vote counting the Republican candidate had 3,297 votes, the Democrat 2,554 votes and the Progressive 2,982. Under the system the Democrat candidate was dropped and the Progressive candidate won by
4,314 votes to the Republican's 4,064 votes or a margin of 250 votes.
Had the Republican candidate been dropped instead and his voters'
second choices been reallocated, the vote would have been 4,067 for the Democrat and 3,477 for the Progressive for a margin in favor of the Democrat of 590 votes, more than twice the original margin of 250 votes. Worse yet, in neither case were the votes for the winning candidate even equal to half of the 8833 votes counted in the penultimate round where a majority, 50% +, would have required a minimum of 4417 votes. From this perspective, it is clear that the claim that IRV always leads to a majority candidate being elected is a mirage.