Instant Runoff Voting is a essentially the second worst of the commonly proposed voting methods, producing a poorer social utility efficiency than Borda, Condorcet, Approval, and Range Voting, not to mention numerous other (more esoteric) voting methods.

Contrary to popular myth, IRV does not prevent spoilers. In fact, the following type of hypothetical scenario occurs about 19.7% of the time in IRV elections:

% of voters - their order of preference

36% Obama > Edwards > McCain 15% Edwards > Obama > McCain 17% Edwards > McCain > Obama 36% McCain > Edwards > Obama

With IRV, Edwards is eliminated first, and McCain wins. Yet Edwards is preferred by a huge 66% majority to both McCain and Edwards. IRV picks the wrong winner, and Obama acts as a spoiler, because had his supporters only voted for their second choice, Edwards, they would have had a better result, hence the huge incentive in IRV to vote strategically for the "most electable" candidate.

Proof of this behavior lies in the history of IRV in Australia's house of representatives, and the Irish presidential post (more of a monopoly than a duopoly, save one exception), as well as widespread use in Malta and Fiji.

The solution is the vastly superior, and simpler, Range Voting - score all the candidates and elect the winner. It is spoiler free, has a vastly higher social utility efficiency than IRV, and picks the Condorcet winner (when one exists) more often than IRV with 10 or more candidates (which you most certainly will have for the most contentious/important posts, like President).

See the official site of the Center for Range Voting for more information - http://RangeVoting.org/IRV.html

Clay Shentrup San Francisco, CA 415.240.1973

With IRV, Edwards is eliminated first, and McCain wins. Yet Edwards is preferred by a huge 66% majority to both McCain and Edwards.

Doh - I meatn to say "McCain and Obama".

Doh! I meant to say "meant". ;)

What's so especially unfortunate about the IRV issue is that so many of its proponents (i.e. FairVote.org) will tell blatant lies about it, like saying it eliminates spoilers, when in fact spoiler scenarios in 19.7% of IRV elections.

The IRV propaganda machine is at it again right here: http://youtube.com/watch?v=wqblOq8BmgM

Slowly but surely, the truth is getting out. The Libertarian Reform Caucus, for instance, already knows that IRV is deadly to third parties.

http://reformthelp.org/issues/voting/range.php

I would be interested in obtaining the 363 rank-order votes for research purposes if that is possible.

In particular I am interested in the comment above that "We definitely saw LOTS of strategic voting in our little straw poll..."

I wonder what is meant by that, more precisely.

To Clay and others opposed to IRV efforts, a few questions:

<h1>I note this language in Oregon Constitution (Art. II, Sect. 16):</h1>

Election by plurality; proportional representation

...

Provision may be made by law for the voter's direct or indirect expression of first, second or additional choices among the candidates for any office.

For an office which is filled by the election of one person it may be required by law that the person elected shall be the final choice of a majority of the electors voting for candidates for that office.

<h1>These principles may be applied by law to nominations by political parties and organizations.</h1>

1) Do you agree that this 1908 quoted language contemplates ranked-choices voting (IRV)?

2) Do you think that this language permits or describes any other form of voting (other than IRV)?

3) Do you think that there is any argument that could be made that approval or range voting satisfies the process described, in which voters may "express[] ... first, second or additional choices among candidates," which seems to call for an ordinal ranking?

The "rangers" strike again. Range voting can be potentially useful in certain surveys, but for all the huffing and puffing and reform-slamming tactics of Shentrup and Smith, I'll be stunned if it ever gets adopted anywhere in the U.S. for any election of meaning. It immediately promotes insincere voting -- just ask the Canadian figure skaters who lost out on the gold medal they should have won in the 2002 Olympic games, and that was even where judges had to vote publicly and reveal their potential strategizing. It can produce absurd results like a candidate winning who wouldn't get a single first choice vote or a candidate losing who won nearly all first choice votes.

Instant runoff voting has a LONG history of voters handling it just fine and or themselves as voters trying to be tactical.

Range voting folks want political minorites to win. I do too, but only their fair share through proportional voting methods. When we're electing a single person, I want a candidate who can command first choice support but also reach out to enough people to earn a majority against other top contenders.

In the spirit of this exchange, I also will correct a typo.. I wrote "Instant runoff voting has a LONG history of voters handling it just fine and or themselves as voters trying to be tactical." I mean to say in the latter part a long history of "tactical voters not having any advantage over sincere voters."

For one document comparing IRV to other reforms, focusing in particularly on qualities that make a system politically viable, see: http://www.fairvote.org/?page=1920

<h1>To answer JMG re this language in Oregon Constitution (Art. II, Sect. 16):</h1>

Election by plurality; proportional representation ... Provision may be made by law for the voter's direct or indirect expression of first, second or additional choices among the candidates for any office.

For an office which is filled by the election of one person it may be required by law that the person elected shall be the final choice of a majority of the electors voting for candidates for that office.

<h1>These principles may be applied by law to nominations by political parties and organizations.</h1>

--WDS: Thanks JMG, that is very interesting.

1) Do you agree that this 1908 quoted language contemplates ranked-choices voting (IRV)? --WDS: it plainly contemplates ranked-choices voting, but not IRV alone, because there are many voting methods involving ranked choices, and IRV is only one of them (and not a very good one, either). Also, this language does not EXCLUDE voting methods based on ratings (such as range voting) although it does not mention them explicitly. The language aims to be inclusive rather than exclusive. But it does permit Oregon to make a law which WOULD exclude range and approval voting.

2) Do you think that this language permits or describes any other form of voting (other than IRV)? --WDS: Sure. For example, Schulze beatpaths voting, the method used by Debian each year to choose their leader. RangeVoting.org currently recommends Schulze beatpaths if it is insisted that rank-order ballots be used. In this we agree with N.Tideman's recent book "Collective decisions and voting", which regards IRV as "unsupportable" but supports Schulze's method. (However I do not always agree with Tideman.)

3) Do you think that there is any argument that could be made that approval or range voting satisfies the process described, in which voters may "express[] ... first, second or additional choices among candidates," which seems to call for an ordinal ranking?

Well... there are several answers. A. range and approval do not operate with pure-ordinal rankings. B. The language "it may be required by law that the person elected shall be the final choice of a majority of the electors voting for candidates for that office"... IF it actually WERE required by law, might be argued to FORBID IRV voting.

Why? Because IRV can elect candidates who are NOT the choice of a majority of voters, when a clear such choice exists. An example is http://www.rangevoting.org/rangeVirv.html#nasty In this election, which is entirely realistic, IRV insists on electing D even though C is preferred versus every opponent (including D) by approximately a 2:1 ratio landslide supermajority. These supermajorities are based on the "final choice" of the voters of their rank order votes. So if such a law were made, which would be an OK law under the Oregon constitution, then IRV would be illegal.

The current PLURALITY voting method could also be made illegal by Oregon law for the same kind of reason - it can elect X even though Y is preferred over every opponent (including X) by a large supermajority.

This constitutional language might also permit Oregon to outlaw range voting because range voting also can elect X even when a majority of the voters say Y is their first choice (if enough regard X as a high-enough-rated second choice).

But as far as I can see Schulze beatpaths would be very constitutional in Oregon.

In my opinion the authors of the Oregon constitution were ignorant about voting methods and unwise in their choice of language. "A little knowledge is a dangerous thing."

<hr/>

Finally, concerning Rob Richie telling you to read his webpost http://www.fairvote.org/?page=1920 , Richie unfortunately neglected to mention the massive refutation of almost everything in that webpost, which you can read here: http://rangevoting.org/RichieRV.html

I hope this has been helpful. Warren D Smith http://RangeVoting.org

Rob Richie (as corrected by himself): "Instant runoff voting has a LONG history of voters handling it just fine and tactical voters not having any advantage over sincere voters."

--this remains simply false. Many counterexamples are found on the RangeVoting.org web site. Indeed the straw poll that started this entire conversation was here claimed to be yet another counterexample. The entire history of Australia is a counterexample (IRV + strategic voting led to their state of massive 2-party domination).
Both are facts Richie ignored like usual. Yawn.

What unfortunately is actually the case, is Rob Richie has a long history of shutting his eyes to evidence contrary to what is convenient for him to believe. He is aware of, because I told him about long ago, these counterexamples and of that refutation of his latest webpost. Note how he acted. It is unfortunate to see a man so opposed to the truth in a position as a "prominent reformer."

Warren D. Smith

Sticks and stones may break my bones, Warren, but .... All the mud that Warren and his zealous rangers like to throw melts under scrutiny. If Australians are being so strategic with IRV, for example, why do they like it so much and why does IRV keep being used in a whole myriad of places? His "counterexamples" are poorly argued, and he just can't grasp the limitations of his proposals.

But hey, we've got reform work to do, while Warren needs to keep working as the agent of the status quo that he essentially is.

1) Do you agree that this 1908 quoted language contemplates ranked-choices voting (IRV)?

There are lots of "ranked choice" voting methods, from instant top-two runoff, Borda count, to Condorcet methods. This language doesn't specifically say anything about IRV.

2) Do you think that this language permits or describes any other form of voting (other than IRV)?

It absolutely does, as a matter of fact.

3) Do you think that there is any argument that could be made that approval or range voting satisfies the process described, in which voters may "express[] ... first, second or additional choices among candidates," which seems to call for an ordinal ranking?

Range Voting allows for the expression of ordered preferences, plus it goes even further, allowing voters to specify not just that they like A more than B, but by how much - intensity.

Moreover, if Oregon law did forbid Range Voting, then I would simply suggest we work to overturn the law. Range Voting is too important.

If Australians are being so strategic with IRV, for example, why do they like it so much...

"Like" is a relative term. When compared to plurality voting, IRV seems "not half bad". But Warren's utility calculations show that voters typically derive significantly greater satisfaction with election results by using Range Voting. This isn't just a plug for Range Voting - Borda, Condorcet, and Approval voting typically outperform IRV too. IRV isn't even near the top of the list of good voting methods.

and why does IRV keep being used in a whole myriad of places?

It only enjoys widespread use in four countries (if Fiji even still uses it), whereas top-two runoff - which as we pointed out, tends to escape duopoly - is used in at least 27 countries that we know of. I don't know off hand of any countries that use Borda or Condorcet methods, even though these methods appear to be substantially better than IRV. So I don't think popularity is a very good indicator of the quality of a voting sytem.

Although I will point out that Australia's third parties want to get rid of IRV.

I also want to mention that Rob Richie's false claims that IRV deals better with strategy than Range Voting, have been refuted before, at length. But notice how he still makes the same arguments for a "new" audience, pretending he doesn't know any better.

It's quite like the frequent myth that IRV proponents tell that IRV prevents spoilers. They've been corrected about this, but many of them still keep saying it.

The rangers just keep going, like energizer bunnies, but it's absurd to suggest range voting's incentives for tactical voting aren't one of the decisive reasons it is going nowhere. Rangers' arguments about tactical voting hold no water at all. They tout approval voting as an example of range voting, but gloss over the fact that when it's been tried (for a private election here and there), tactical voting typically becomes dominant. That's because if one side is tactical (and it's transparent how to do so -- just cast one vote for your favorite candidate and none ofr anyone else), the other side loses. There's nothing comparable with IRV, which is why it keeps going strong in major election after election -- if tactical voting were a problem, people would get rid of it.

Third parties in Australia support IRV by a longshot over plurality voting (or other single winner methods as far as I know), but yes, they sensiblyt want proportional representation. That's what the rangers really should be focused on, given their belief in minority representation. I support proportional voting for legislative elections and wish they did too (ather than saying erroneous things like it not being constitutional for congressional elections, which is flatly, bizarrely inaccurate). But instead they support range voting for winner-take-all races so that it can allow political minorities to defeat political majorities -- and think that concept of minority rules is going to fly politically in the U.S.

Okay, enough of my little rant. Shentrup and Miller can keep swift-boating away, which is their tactic, but it's not going to stick if people look up the facts.

Rob Richie,

Your arguments about tactical voting are oblivious to the facts. Range Voting produces hugely more satisfying election results than IRV, regardless of how strategic the electorate is. Range Voting also suffers only half as much as IRV if an electorate goes from 100% honest to 100% strategic; so it's reasonable to say that IRV is twice as susceptible as Range Voting to strategy.

You have, several times now, made the case that Range Voting would suffer from strategy significantly more if one faction was much more strategic than another. Well, that has also been addressed here: http://rangevoting.org/StratHonMix.html

Your claim that "if tactical voting were a problem, people would get rid of it" is self-contradictory, because tactical voting is significantly less of a problem with Range Voting, yet you say we shouldn't adopt Range Voting because of its susceptibility to tactical voting.

As your post continues, you show further horrendous misunderstandings of election science. Many Range Voting advocates very much do want proportional representation (RangeVoting.org/PropRep.html). We also want it to be done well, which is why we offer Reweighted Range Voting (RangeVoting.org/RRV.html) as a superior alternative to Single Transferable Vote.

Our claim that PR is plausibly unconstitutional in US Federal and state governments, is supported by cited evidence, which you made no attempt to refute - another argument by assertion. And PR is definitely illegal for Congress, under federal law.

Your last claim, that we want political minorities to defeat majorities, is simply ludicrous. This page thoroughly disproves that claim, with graphical representations of the data: rangevoting.org/IrvExtreme.html

I wonder why you think we're "swift-boating" you, when we go to such lengths to provide large amounts of supporting evidence for our claims.

Posted by: CLAY SHENTRUP | Mar 10, 2007 8:07:12 PM

"% of voters - their order of preference

36% Obama > Edwards > McCain 15% Edwards > Obama > McCain 17% Edwards > McCain > Obama 36% McCain > Edwards > Obama"

Let's see...

36%+ 36%+ 17%+ 15%= 104%

Vote early and vote often!

Ah yes, that was a mistake. The example should have been:

"% of voters - their order of preference 34% Obama > Edwards > McCain 15% Edwards > Obama > McCain 17% Edwards > McCain > Obama 34% McCain > Edwards > Obama

Thank you for pointing that out.

Clay,

On intensely supported minorities defeating more moderately supported majorities, see your own website at: http://www.rangevoting.org/MajCrit.html

There your site gives an example of a candidate who would lose 51%-49% in our current system turning around and winning with range voting, with the comment that "in such a situation we would argue that it is good that Y [the candidate that loses 51%-49% in a one-person, one-vote election] wins 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."

Of course these kinds of reversals of majority rule could be more extreme. Systems like approval voting and range voting can allow a candidate to win who would lose 99%-1% in a plurality election -- unlikely, of course, but that extreme points to how it could happen rather frequently in a 60%-40% type race. The way it would happen often is if one side is tactical and other isn't. Tactical voting is a reality from the get-go for range voting despite all the "simulations" you can do. And IRV doesn't lead to that kind of tactical voting, with thousands of major elections providing evidence of that.

People weren't thrilled when George Bush won the 2000 presidential election despite having fewer votes than Al Gore. Range voting would make such results rather common. And you don't see that as a political problem?

'Nuff said.

Rob,

In my first post, I outlined a common type of IRV scenario in which the Condorcet winner loses: Edwards loses, even though he would win by a 66% supermajority to both Obama and McCain. So failing to pick the majority winner, in 19.7% of elections, is a major problem for IRV. You cited our majority criterion page, which explains why Range Voting also sometimes fails to elect the majority winner. The important difference is that when Range Voting does this, it isn't just some fluke - it has the benefit of picking a winner who is more broadly satisfying to the voters, instead of giving the majority their way, at the possible expense of the minority's misery. This inherent "compromise" tendency is why Range Voting tends to produce the greatest satisfaction for the greatest number, whereas IRV tends to elect extremists. In other words, if voters were about to hold an election, and had a choice to use IRV or Range Voting, they would all be wise to choose Range Voting, because it would give them a higher expected value from the election outcome.

You continue to insist that tactical voting has not been a problem for IRV, even though it has lead to two-party domination in all four countries where it has seen widespread long-term use - exactly the effect we would expect from tactical voting. Nevertheless, we gave IRV the benefit of the doubt in our one-sided tactical voting simulations, assuming that IRV would encourage 100% of voters to be perfectly honest angels, whereas 50% of Range voters would be strategic. Even after giving IRV a big head start, Range Voting came out well ahead of IRV.

I just noticed another error from an earlier post by Rob Richie. He says that the tactical way to vote under Range/Approval voting is to

..just cast one vote for your favorite candidate and none ofr anyone else.

This is incorrect. Say your favorite candidate is Ralph Nader, whom you know probably won't win - then doing this is strategically bad, because you effectively waste your vote. Whereas if you vote for Nader and also vote for the most Nader-like candidate who is "electable", you get to express your support for Nader, but still take part in the "real" election between the electable front-runners.

For more discussion on this, see:

http://rangevoting.org/RVstrat1.html http://rangevoting.org/RVstrat2.html http://rangevoting.org/RVstrat3.html http://rangevoting.org/RVstrat4.html

Correction: Above I incorrectly stated that the probability that IRV will fail to elect the Condorcet winner (when one exists - often it doesn't) is just under 20%. This is actually the probability that there will be a favorite-betrayal/lesser-evil scenario in IRV. For figures specifically pertaining to the odds of picking the Condorcet winner, see charts 2 and 14 here.

Last post from me in this exchange. Two points:

1) Clay is defending range voting electing a losing candidate in a two-person race -- e.g., a candidate who would lose 60%-40% (or by a far more lopsided margin) under our current rules being able to defeat the candidate who has 60% support. He's welcome to try to win a campaign for a system that can easily allow such results, but count me out.

2) Clay writes about various simulations. We don't need to simulate IRV, as it's been used in thousands of significant, closely studied elections. Australia has had more IRV elections than anyone, and all Clay can say about how it's promoting tactical voting is that the major parties have won almost all the seats. That's not evidence of tactical voting have anything to do with those results. On the other hand, though, in surveys of voter satisfaction with democracy across dozens nations, Australia consistently ranks right at or near the top. See:

http://www.umich.edu/~cses/resources/results/CSESresults_SatisfactionWithDemocracy.htm

Such surveys aren't the be-all, end-all, but help explain why there is absolutely no energy or interest in Australia in going to U.S.-style plurality or French-style runoffs. Any reform energy there is toward proportional representation.

A recent paper suggested there is a real link between voter satisfaction and the kind of voter choice available in ranked-choice systems. See: http://repositories.cdlib.org/cgi/viewcontent.cgi?article=1034&context=csd

Rob,

As we have seen, IRV and Range Voting can both elect a candidate who is not the Condorcet (beats all) winner. But when Range Voting does this, it is typically a benefit, producing greater average voter satisfaction. Whereas when IRV does it, it will more than likely cause a decrease in average voter satisfaction. It seems like you realize this, and so you are choosing now to specifically look at the case of a two-candidate race, in which IRV can't exhibit its problematic non-Condorcet behavior. But even in this one isolated example, consider some important facts:

1) The expected voter satisfaction will be greater with Range Voting than with IRV. Why would a wise voter, once informed of this, want to reduce his expected satisfaction, just to ensure a "majority" winner? After all, he may not end up being in the majority when all is said and done. 2) Range Voting still allows voters to strategically maximize one option and minimize the other - so if we use Range Voting, and a solid majority wants to ensure majority rule instead of greatest societal happiness, they can do it. 3) This has been pointed out to you many times before, specifically in Warren D. Smith's response to your first "common sense" criterion, here.

e.g., a candidate who would lose 60%-40% (or by a far more lopsided margin) under our current rules being able to defeat the candidate who has 60% support. He's welcome to try to win a campaign for a system that can easily allow such results, but count me out.

Why count you out? Why would you want to be less happy just for the sake of using an antiquated and problematic voting system like IRV?

Clay writes about various simulations. We don't need to simulate IRV, as it's been used in thousands of significant, closely studied elections.

But unfortunately we can't read voters' minds, so all the elections studies in the world won't tell us certain things that we can calculate with simulations. For example, if you look at the Obama/Edwards/McCain IRV election earlier in this page, you see that the Obama fans would have been strategically justified to vote for Edwards, the centrist, so as not to risk helping McCain win. If we saw the ballot results of such an election, we would have no way to tell which Edwards votes were sincere, and which were strategic votes from Obama fans. So accurately deducing utilities from ballot results is infeasible. But the opposite process - assign random utilities to simulated voters and see how the different election methods fare - is quite feasible. This is why we don't want to use real people for these experiments.

The persistence of two-party domination in all four IRV-using nations is in fact strong evidence of strategic voting, because if we model IRV elections under honest behavior, there is no reason to expect a duopoly; whereas under the obvious strategy of top-ranking your favored front-runner, IRV does lead to two-party domination. A traditional "genuine" runoff election, on the other hand, produces slightly different strategic incentives, hence 21-23 of the 27 countries that use it have escaped duopoly.

I'm happy to hear that Australia's voters are among the happiest with their voting system, but is that really a good reason to pass up the large additional benefit they would see from using a better voting method such as Borda, Condorcet, or Range Voting? After all, one can never be too satisfied, especially when it comes to electing leaders, who affect everything from which wars are fought, to how much is done to stem global warming.

A recent paper suggested there is a real link between voter satisfaction and the kind of voter choice available in ranked-choice systems.

This paper suggests that rank-choice systems are undesirable.