Open letter to the IRV community

I want to convince you of the following 2 points.

  1. Range and Approval voting are superior to IRV=Instant Runoff Voting.
    If you are going to go to years of trouble to change the voting system, you might as well do it right.
  2. Range Voting is a better "stepping stone" than IRV is, to PR=Proportional Representation.

I then would love it if CfER and RangeVoting.org could find a way to somehow merge with us and/or pool our talents and resources, and/or to convince, say, the CVD also of thought-reform. We now give more detailed arguments, and hyperlinks you can follow to reach data backing up our points.


1. Range and Approval voting (RV & AV) are superior to IRV:

  1. AV & RV are simpler to define (e.g. shorter computer programs).
  2. Simpler for voters in the sense that, experimentally, voters have the most ballot-invalidating errors with IRV (in San Francisco 7 times as many with IRV than plurality; also supported by cross-country comparisons), followed by plurality, followed by Range, followed by Approval (i.e. Approval is the best in this respect).
  3. Less painful and cheaper to adopt. Range and Approval both work with every voting machine in the USA, right now, no modification or reprogramming required. (Admittedly, with more or less hassle depending on the voting machine type, but it always does work.) IRV does not. Also, IRV cannot be "counted in precincts" – only centralized counting is possible – necessitating a big change to procedures and a security hit because a small central conspiracy then can throw elections. (As IRV advocate Rob Richie admitted in print in 2008, no voting machine in the USA comes, out of the box, ready to handle IRV elections. Many computerized voting machines have been decertified by their states, in, e.g. California, Colorado, Ohio.)
  4. IRV is most likely to lead to tie and near-tie chad-counting lawsuit nightmare scenarios; range voting is least likely.
  5. Range is less likely to lead to massive self-reinforcing 2-party domination when adopted countrywide. Every IRV country has quickly fallen into 2PD status in all IRV seats (examples: Australia, Fiji). Plurality also almost always has led to 2-party domination (the USA being a primo example). But, in contrast, ordinary (not instant) runoff historically has usually not led to 2-party domination (example=France), perhaps for these reasons. With 2-party domination, voters effectively have only 2 choices, in which case "improved" voting systems largely aren't. (IRV is superior to plurality in the sense that it works better in "two and a half candidate" elections, but when you extend your thinking to genuine 3-candidate races – or more – IRV behaves poorly.) Also, e.g. 70% of US voters right now (July 2007) are antiwar. But it appears probable that once again they will have a 2-way choice between a pro-war and pro-war candidate. (Same thing as in 2004 and 1968.) That is a very bad and very undemocratic phenomenon and it is the fault of 2-party domination and the plurality voting system. IRV still leads to 2-party domination, still leads to "spoiler" effects, and "favorite betrayal" and the "lesser evil" phenomenon still are important under IRV (contrary to pro-IRV propaganda).
  6. A voting system fails "favorite betrayal" if a voter finds it strategically unwise to vote her true favorite top. IRV fails FB frequently. In one probabilistic model of 3-candidate elections it fails it 19.6% of the time, in another 20.2% of the time. This is serious and we believe it has a lot to do with why IRV countries become 2-party dominated (the 3rd party candidates get betrayed, because that is the strategic move for voters to make).
  7. Another reason (we conjecture) for IRV 2-party domination is that "dumb" voters unaware of the above strategy-facts, still often will automatically vote one of the two major-party contenders top and the other bottom just because, intuitively, that maximizes their vote's "impact" on the "real battle." (I claim based on my polling experience that they'd commonly do this in any rank-order voting system – Condorcet, IRV, Borda, whatever – whether it makes logical sense or not.) If enough of them do that, then it is a theorem that – with either Condorcet or IRV voting – a third-party candidate can never win. So: strategically-wise IRV voters cause 2-party domination, and "dumb" ones also cause 2-party domination.
  8. Computer simulation studies show range voting is superior to all common alternative proposals (including IRV) – very robustly to changes of simulation parameters – measured by "Bayesian Regret." This is true for both honest and strategic voters (and hon-strat mixtures) and many different utility generators, numbers of voters, numbers of candidates, and "ignorance levels." The amount of the superiority over plurality is comparable to the amount plurality is superior to "random winner" i.e. "is comparable in importance to the invention of democracy."
  9. Honeybees and at least one ant species are known to use range voting to make important collective decisions. Honeybees have carried out far more elections than either humans, or computers doing election simulations. They are thus the greatest authority on elections on the planet. Ants are claimed to be the most successful land animal ever (in terms of total biomass). They both evolved range voting independently. No nonhuman animal I know of uses any voting method besides range voting.
  10. Range voting is used very commonly to elect school valedictorians. As such it has had a tremendous amount of actual use, comparable to the amount of use of all other voting systems and probably considerably exceeding the amount of use of IRV. Range voting was also central to the governments of Ancient Sparta and Renaissance Venice, which were the two longest-lasting governments with a substantial democratic component, in world history.
  11. Range voting obeys "participation" (casting an honest vote cannot hurt you vs not voting at all) "monotonicity" (raising X in your vote cannot reduce X's winning chances) and "clone proof" (if an extra "clone of X" candidate runs, that neither hurts nor helps X's or anybody else's winning chances aside from replacement of the winner by a clone). These are important properties. IRV fails the first two (example, example). It has been proven that no voting system based on rank-order ballots can both obey cloneproofness and avoid favorite betrayal, so in that sense range voting is superior to every voting system based on rank-order ballots.
  12. IRV and ordinary runoff both sometimes fail to elect "beats-all" Condorcet winners in 3-way races. This happened in the Peru 2006, Chile 1970, and France 2007 presidential races as well as the famous "Lizard versus Wizard" Louisiana 1991 governor race. In all these cases the "centrist" candidate was eliminated in the first round and then an "extremist" won, but polls indicated the centrist would have been preferred over every opponent in a head-to-head race. Indeed, IRV tends to have a strong extremist-favoring effect as you can see from these pictures (and note observation 2 there). Also it can be proved, under a simple and fairly realistic model of strategic voter behavior, that range voting always elects Condorcet winners whenever they exist; and Range Voting under a modified definition-wording of the Condorcet property is a Condorcet method; and range voting under a similar pictorial study shows no particular bias in favor of either centrists or extremists.
    (Range voting with honest voters seems to exhibit little or no bias favoring or disfavoring extremists in such pictures, whereas under the strategic voter model in which Condorcet winners always win when they exist, range is totally unbiased and coincides with optimal voting in the 2D-picture scenarios.)
  13. IRV exhibits a tremendous number of crazy pathologies.

2. Range Voting is a better "stepping stone" than IRV is, to PR.

Suppose your goal is to have the USA eventually adopt "Proportional Representation" (PR) multiwinner voting. I personally am not sure whether this is a good idea (summary) and it is difficult to do in the USA for real without a massive rewriting of the constitution and risks being stopped cold by the (2-party dominated) supreme court. However, ignore those concerns for our present purposes – just suppose we agree PR is a wonderful idea and ask "then what?"

My claim is that Range Voting is a better "stepping stone" than IRV is, to PR:

  1. There are PR methods that are both better and simpler than STV for large elections, and that resemble range voting instead of IRV. They are called "asset voting" and "reweighted range voting." See papers #77, 78, and 91 here. (Admittedly they are less familiar right now, but considering PR in the USA isn't going to happen for 20 years at least, that is hardly a concern.)
  2. There is plenty of evidence that the Democratic and Republican parties are opposed to PR. They have fought it actively and highly successfully in the past, e.g. organizing campaigns that got rid of it in New York City and Cincinnati. (And in Ireland, the dominant Fianna Fail party tried twice to abolish PR, both times unsuccessfully.) In these three cases they had to get rid of PR via referendum and could not do it via direct action. But only one of the USA's top two parties, acting alone and directly, is enough to block a PR constitutional amendment in the USA forever. They are motivated to stop PR. Therefore PR on a nationwide scale is simply not going to happen as long as the D & R parties reign supreme.

    Are there any then-2-party-dominated countries that have ever successfully been made to switch to PR? As far as I currently know – although my knowledge is incomplete on this question – almost all PR countries got that way by starting with a PR system (or something very similar to one) at their initial founding as a democracy and/or in their constitution (or after the collapse or overthrow of a previous government). The best exception I know is New Zealand, which switched from Plurality in single-member districts to German-style "Mixed Member Proportional" (MMP). New Zealand's top party opposed the switch but apparently did not fully realize the danger until it was too late. The population gave the third party 21% of the vote (and the fourth party 12% in the 1984 election) so evidently was highly favoring of third parties, far more so than in the USA (but this did not help those parties of course – they won two and zero seats respectively). There then was a nonbinding referendum in which 70% of the population wanted MMP and 85% wanted to get rid of the present system, a huge landslide!! The top party then saw the danger, but couldn't tell a massively publicized 85% supermajority to go to hell without presumably suffering for it big time. So they instead just made heavy propaganda efforts to kill MMP, and they nearly succeeded since then it passed by only 54% in the binding referendum. If they had realized the danger earlier, they plausibly quite likely could have and would have averted MMP.

    In view of this I believe the only way PR will ever happen nationwide in the USA is if its massive 2-party domination first is made to gradually vanish. IRV cannot and will not do that because IRV reinforces and yields 2-party domination (every IRV country so far has developed heavy 2-party domination in all IRV seats). Range Voting probably will do that because there is heavy experimental evidence that it leads to greatly enhanced relative vote counts for third-party candidates. The interesting unexpected nursery effect enhances that.

3. My fear is that IRV will lead to a permanent fossilization of 2-party rule and permanent prevention of any further election-method reform because of the following excuse: "we gave you reformers what you asked for, and now you are not satisfied with your own reform?? Morons. Oh and by the way, you third-parties are still out of power, da hahahahaha." Typical timescales for election reform are 200+ years and we cannot afford to have an inferior system for that long. So it is foolish to go with an inferior method (IRV) due to shortsighted thinking.


Return to main page